Adds a Special Themes row to Settings > Appearance > Theme. Halloween is a
scheme on the existing data-scheme axis that is always dark: the applier and
the boot script force .dark while it is on, and the color mode control is
dimmed with a hint.
- Palette for the whole app (surfaces, dashboard, Vacay glass, Journey,
mobile chrome) in theme/special/halloween.tokens.css, loaded with index.css.
- Decor as a lazy chunk mounted by SpecialThemeLayer while the scheme is on:
fog, embers, bat flocks, a passing ghost, corner cobwebs, spiders that drop
on a thread and flee when poked, a pumpkin patch and an ember running along
the navbar. Figures are pre-rendered 3D sprite sheets.
- Nothing is drawn over a map: the Atlas gets no decor, the trip planner only
dresses its side panels. Marker clusters turn into lanterns.
- The 3D mascot plays its own Halloween clips in every scene (witch hat, a
spider on the brim, a lit pumpkin beside it); the SVG fallback wears the
same costume.
- Reduced motion stills everything, transparency off drops the fog, shared
pages never show it.
- Navbar, bottom nav, notifications, Atlas and Collections read the dark flag
off the .dark class instead of settings.dark_mode, so a forced dark theme
reaches them.
The tour list row puts the description between the name and the badges.
The badges keep the three that matter most (time, then a paddle tour's
check bench and distance, or a hike's distance and climbing) and only
those that fit whole, so a narrow sidebar shows two instead of cutting
one off. A check bench without obstacle data shows no pill in the list
or the day plan; the checklist still names the gap.
The checklist and the obstacle data sorted texts without a compare
function. They now sort by code units through one shared compareCodeUnits,
the order a bare sort gave, the same on every machine; the material data of
a lock's parts no longer orders by the server's locale, so its hash cannot
change with it. The note editor takes the focus through a ref instead of
autoFocus, a direct function reference in a map becomes a call, and the
GPX import's default fields are a named constant.
A save answered on one tour no longer clears the pending state of a save
started on another. A checklist changed here offline waits for the queue to
replay before it is read online, as a bench of such a write already did. A
saved tour the list has not loaded yet is no unsaved paddle draft, and Next
walks the rows as the list shows them, a vanished object among them.
A decision answered after another tour was opened no longer replaces that
tour's checklist. A decision made offline on a checklist already older than
the route stays marked as such, and once online a bench from a write made
offline waits for the queue to replay, then takes in the server's. While
the checklist reads again it stays in view, locked, so a panel open on the
map keeps its note. A saved tour turned into a paddle tour asks to be saved
first, a vanished obstacle stays among the rows, Next follows the list
order and keeps the focus in the panel, a note closes only once saved, the
earlier choice in review is named for screen readers too, planning a
portage is offered to editors only, and the map card scrolls within the
screen. A tooltip of several lines breaks at them. A page test waits for
the tours row instead of reading it at once.
A lock flight chained into one group of more than 30 parts broke the
checklist's contract and made every decision about it fail; an entry now
names its nearest part and the first others, a portage lists at most 30
obstacles and a checklist at most 100 own entries. Reading along a line
takes in the parts of a lock the corridor cut off, so its data basis no
longer depends on how the route runs. A line or an area stands at the
middle of its extent, so adding a node leaves it in place, and an earlier
decision is judged by where the obstacle was, not by the route point of its
passage. The cache keeps its readings without their lines, for one data
version at a time, and the data version is read without the region
boundaries.
The tour map stage and the trip planner page hook had grown past the
complexity and length the lint ratchet holds them to. Marking a spot on a
recorded paddle track moves into useTrackLinePick, and what the planner map
frames into usePlannerFocusPoints, each with its own tests. The planner's
viewport snapshot no longer reads a ref while rendering, so the count of
that warning goes down by one.
PAD-06 and PAD-09 on screen. A paddle tour open in the planner or shown as
recorded carries a checklist card: the obstacles of the data along its
route in route order with the paddler's own portages and locks among them,
each with its status, its paddling rule and the SUP advice at a lock, and
the earlier decisions the route or the data no longer carries behind a
fold. A row moves the camera onto the obstacle and the other parts of its
lock and opens its card on the map, where the decision panel sits: four
choices, the notices that stay after a choice (a ban, the SUP advice, a
portage still to plan), the date of the last check on site, a note and who
decided when. An object the map does not hold opens in a dialog instead.
Deciding waits while the route has unsaved changes, and a stale or dirty
list is dimmed, never hidden.
The tours list, the desktop day plan and the phone day plan show the bench
as one pill, the tour detail shows it with the data's date, and every paddle
tour reaches the safety note from its footer. Nothing reads as safe where
TREK knows nothing, and nothing blocks.
PAD-06 and PAD-09 on the client's data side. A tour's cached row keeps its
checklist beside its provenance, written through one cache helper by the
repository and the queue replay, which also keeps provenance an answer
lacks. The checklist is read for the bench the list shows and served from
the cache while its revision matches. Decisions made offline, or behind a
queued write of the same tour, change the cached checklist and bench at
once under a local revision and replay with their idempotency key; a route
written offline marks the checklist stale until the server judged it.
The obstacle store learns a camera focus: a click in the rail moves the
camera onto an obstacle and the other parts of its lock, the newest of such
a focus and a picked search result wins, and a selected lock lights up
whole on both maps. The checklist hook and its model hold the logic the
cards will show.
The checklist, its statuses, choices and notices, and the safety note of a
paddle tour, translated into all 27 languages. A spec keeps the checklist
strings in English and German from claiming the water is safe, cleared or
passable. The paddle strings of the last batch that Prettier had not seen
are formatted.
PAD-06 and PAD-09. The tours list carries the check bench of every paddle
tour and the detail its whole checklist, worked out of the obstacle data
along the tour's line and the decisions stored for it. The obstacle data is
read along the line run by run and kept in memory under the line, the boat
and the data version, so a list of a trip reads it once; a hike costs
nothing. While the data cannot be read the checklist says so and keeps every
decision visible.
PUT /api/trips/:tripId/tours/:placeId/obstacle-reviews stores what the
paddlers decided, each against the basis the user saw, and answers the
detail. It needs place_edit, refuses a tour that is no paddle tour with a
400, waits with a 503 while the data cannot be read, and skips an obstacle
no longer on the checklist instead of failing. A copied tour and a copied
trip take their decisions along. The MCP read tools describe the new fields;
there is no write tool, since a decision is a statement of the paddlers.
The obstacle data keeps the dates of checks on site, covers a line by
sampling it, and tells its version by its rows and regions.
PAD-06 and PAD-09 in shared. A paddle tour's checklist lists the obstacles
of the OpenStreetMap data along its route, each bound to a basis: the route
near the obstacle as samples around every passage, and the material data of
the obstacle as a hash and its position. A decision keeps that basis, so it
reads as one to check again once the route or the data changed there, and a
portage decision holds while the paddler's own portage carries round every
passage. The check bench sums the list up for the tours list and the day
plan. Nothing in it is a safety or legal clearance.
The route relation learns the passages of a route by an object and the
distance of a point to a route. The client takes the obstacle key from
shared instead of its own copy.
PaddleRegionsRepository types and writes the status through the generated PaddleRegionsStatus instead of a copy of its four values; the paint tokens, a fact key type and a minutes input stay private to their modules, the wind test pins its 502 text, and the planner page's size entry follows it down.
The strings of the paddle plan, the time estimate, the wind card, the obstacle card and notice and the admin's paddle water data, translated into the 25 locales that still lacked them, with the plural forms each language needs. The pinned count of Tours keys follows.
The time card's estimate with water, portages and locks and a lock without times leaving it incomplete, every state of the wind card with day switching and missing values that never read as calm, and the track marks with their tools, minutes, removal and saving through the repo.
How an admin adds a region from Geofabrik's list, what the card shows while it is read, the monthly refresh on the 3rd, what a failed read keeps, and the OpenStreetMap attribution under the ODbL; a record in the data is no check on site.
Thirteen client suites for the paddle obstacle work and the region admin: the detail model and card, the overlay with its outside press and Escape, the Leaflet and GL layers with the area zoom threshold and the single click handler, the route band marks, every state of the obstacle notice, the sync hook, the tiled obstacle repo, the online only region repo, and the regions panel. The wind hook for the start of a paddle tour is covered too.
PAD-05B: a click on an obstacle on the map or on the route band opens its card in the look of the via ferrata card: the kind with its glyph, the name, how it stands to the route (on it at a distance along it, near it with the link marked unclear, or only in the area), the paddling rule its tags give with canoe read before boat, a condition as written, height, grade, hours and fees, the source with the date of the data and the OpenStreetMap link. A missing rule reads as unknown, never as allowed, and a portage way recorded as private says so. Objects only in the area show from zoom 13 on, on Leaflet and GL alike; on GL the lines only near the route are dashed as on Leaflet, one click handler picks the marker over the line, and a picked marker no longer flickers. The route band marks the obstacles on and near the route by their share of the measured line. The obstacle notice also names an answer cut at its limit.
PAD-05A: the admin area gets a Paddle water data card under the addons, on desktop and phone: a search in Geofabrik's list with each region's path, the regions with their state (waiting, reading with its progress, ready with objects and data date, failed with the reason), reading again and removing after a confirmation. The list follows a read without reloading. Region changes write audit rows.
Map colours of paddle objects and portages come from theme tokens read at draw time, as the via ferratas do; the ferrata layer now shares that reader. The rail takes the planner page's parts as one bag, the map hit tests of the tour layers are one helper, and the unused noWaterData string is gone from every locale.
The PBF reader streams the extract and refuses a varint longer than ten bytes; the block offsets are read as the plain int64 the format defines. Region reads queue without loose promises: a region queued while another is read is still picked up, removing a region calls its read off, a shutdown stops the read under way and the next boot queues it again and clears the download folder. The admin list shows how far a read has come.
Geofabrik is asked only on its own servers, redirects included, and an outage of its list answers 502 instead of an unknown region. Nested ids such as us/texas are accepted.
A region being read again or after a failed read still counts with the data it holds, a region without a boundary spans its objects, and an object two overlapping regions hold is answered once. Stored geometry, tags and boundaries go through the JSON column declarations. An obstacle is related to the route both ways, so a long weir passed in its middle and a route inside an area count.
Tests cover the reader with extracts written in the test, the tag rules, the extraction and lock grouping, the geo helpers, the Geofabrik client, the region queue, the read side, both controllers, the job, the repositories and the HTTP routes; the domain gets its coverage entry.
PAD-03 portages and locks on planned tours and recorded tracks, PAD-07 paddle time with portage and lock stops, PAD-08 wind forecast for the start, PAD-05A regional obstacle import from Geofabrik extracts with admin endpoints and a monthly refresh, PAD-05B first cut of the obstacle layer on Leaflet, GL and 3D. Work in progress: checklist (PAD-06), check bench (PAD-09), admin UI, translations and tests follow.
- Paddle as a plannable tour type: one boat (kayak, canoe, SUP, other), river or
lake, a pace that a boat change never overwrites; time is distance over pace
- Hand-drawn legs beside routed ones, drawn dotted and labelled as not routed,
kept through save, reopen and edits elsewhere
- Reverse a tour and close it back to its start, each one undo step
- GPX import shows a preview first (line, measures, duplicates, type) and only
the confirmation stores the tour
- Base maps and overlays per tour type in the planner: hike keeps topo and the
via ferrata layer, paddle gets the OpenSeaMap seamark overlay
useToast handed back a new object on every render. useWandererConnection lists
toast in its load effect, so each keystroke reloaded the saved settings and
reset both fields to them.
The repository stubs of the tour service tests answer findProvenance, which getTour and createTour read since P2C-02.
Switching the Tours addon off now ends live MCP sessions like every other addon that gates a tool (MCP-GATE-PARITY-001).
copyTour refuses a line without a point instead of placing the copy at 0/0, and new tests cover the reachable paths
of the copy, the source check and the provenance of a snapshot without its import record, including a trip copy
keeping that record. On the client the shared day assignment hands the plan exactly the tour and the day, the new
visibility store is reset between tests, and the tests of the sidebar, the detail, the phone sheet and the origin
section match the views this PR built.
- P2C-02: tours record their origin (planned, GPX with file name, wanderer)
instead of reading it off their shape; the detail answers a provenance
contract and a trip copy keeps a tour's import record
- MF-03: a wanderer snapshot's source can be checked on purpose, read only,
with a confirmed reimport as a new snapshot
- PR1-18: every way of putting a tour on a day shares one set of rules
- PR1-07/08: each person can hide tours from the trip map; the list shows
each tour's line colour and the picked tour stands out on both maps
- unit suites for WandererService (every refusal and fallback), the
wanderer connection and tour source repositories, and restoring a
recorded track (GPX and wanderer imports, empty or missing track, 404s)
- client cases for a body that breaks off and the trail count fallbacks
- an import whose realtime notification fails still stands (e2e)
- the search controller passes the validated body on; the schema already
defaults q and page, so the fallbacks there could never run
- tours locale key count, entity count and NOT NULL field count follow
the wanderer tables and strings
- the shared MSW admin handlers answer /api/admin/tours-wanderer
- the mobile tour detail test checks the dialog, not the dropped heading
- the refused-origin e2e sends one request per case instead of four at once
The admin switch for wanderer moves to api/wanderer.ts and the recorded
track restore to api/recordedTrack.ts; trips.service.ts keeps dev's
formatting with only the derived-from remap on copy.
- GPX files drop onto the map or a day; the import lands on that day
- wanderer connection per user in Integrations; the workspace offers the
user's own trails only, behind an admin switch under the Tours addon
- wanderer difficulty is stored on the SAC scale (T1 to T3), backfilled
- a GPX or wanderer import can be planned as an editable tour, either in
place (the recorded track is kept and can be restored) or as a new tour
that remembers the import it came from
- rails, rows and detail dialog follow the plan sidebar's cards and badges
- admin addon shelves with more than one row fold, collapsed by default
- reverse-geocoded waypoint names are cached on the device
The tours map gets a 3D mode on the same map surface between the side panels:
a "3D" switch in the layer pill drapes the route over open Terrarium elevation
tiles (AWS Open Data), colours it by grade, plays a flyover tied to the
elevation profile in the rail, and keeps planning working by clicking and
dragging waypoints on the terrain. The corridor around the route is preloaded
before the flyover so the landscape never loads in front of the camera.
GPX imports no longer get a made-up hiking difficulty: max_hiking_difficulty
becomes nullable (new migration, waypoints kept), the import writes null and
the client shows no T badge for it. Drawn tours still default to T2.
Info hints in the planner show their tooltip only, and the discard
confirmation reads "Discard changes".
The column showed the last sign-in, which a renewing cookie leaves weeks
behind for someone using TREK daily. It now shows the later of the sign-in
and the last request. Saving an edit no longer blanks the row's avatar,
activity and online dot, and the list refreshes every 30 seconds while open.
Cloud models got the first 4,000 characters, so totals and lodging sections
further down never reached them, without a word. They now get 24,000 with the
start and the end kept and a warning when the middle is left out; a server
whose context is too small is asked once more with 8,000. A total the model
repeated on every flight of one booking stays on the first flight only, when
the document prints it once as a total, and the prompt says so too.
Both admin forms hid the field for Anthropic and the shared hook saved an
empty address, wiping a gateway configured on the server. The field shows for
every provider now and is saved as typed; switching provider empties it, so an
address meant for one never receives another provider's key.
A missing arrival time skipped the time check, so a return listed first
merged into the outbound flight. The layover is now measured from the
arrival, or from the departure when there is none, must be forward and under
a day, and a leg landing anywhere the chain has been is a return.
Journeys from the transit search are routing, not bookings, and a day of them
pushed the real bookings out of the six slots. Booked trains and buses keep
their own types and stay. The MCP tool reads the same feed.
The entry dialog fills an empty weather from the forecast, and leaving it
untouched then asked to discard changes. The suggestion is not an edit. In
development StrictMode also dropped the suggestion before it arrived.
1Password filled the signed-in user's login into a place's e-mail field, and
offered the admin's own credentials for an account being created. The fields
now carry each manager's opt-out; autocomplete=off alone is ignored.
A hotel booking reaches its place through the stay, not its own place_id, so
the card listed the booking but not its confirmation. Links to a stop, which
MCP's link_trip_file and plugins can make, were never returned by the file
list at all; it now carries linked_assignment_ids. Without files the row no
longer opens to nothing.
The hover controls took their room at rest, so a place's line and a note's
text broke early. A place now gives way only while hovered, a note's controls
fade in over the end of its text in the note's own colour, and a picked place
shows none. All three controls of a place come out together: the add-booking
ticket waited for a React hover state and came a beat late. A click no longer
leaves them showing; keyboard focus and an open menu do.
A proxy image_url whose cached file was gone answered 204, and the self-heal
got the same URL back, so the list avatar kept its broken image until a
reload while the inspector showed the photo. The image is now mounted afresh
after the refetch; a second failure still ends on the category icon.
Both side columns keep their content width, so the pill gives way: it drops
its labels to icons when they no longer fit (measured, since plugins and
languages change the width) and scrolls only after that. The username is no
longer cut at 96px.
This commit completes the reviewed Tours follow-up scope after the Tours MVP was merged into dev.
Included feature blocks:
- RC-01: identify Tours in PLAN day rows and show compact Tour facts;
- RC-06: search for a city or address in TOUR-PLANNER without hidden writes;
- RC-07: reorder Tour waypoints in the left sidebar;
- RC-08: drag Tour waypoints on Leaflet, MapLibre and Mapbox with stable viewport, history and routing behavior;
- RS-01: persist Tour descriptions and HTTPS information links with local provider classification;
- RS-02: persist breaks/additional time and calculated or manually adjusted planned Tour duration without changing scheduling semantics.
The implementation preserves:
- ordinary Place and Legacy Track behavior;
- Tours-off fallback behavior;
- existing GPX/import behavior;
- Tour assignment and deletion behavior;
- calculated routing duration as a separate value;
- existing Tour provenance fields;
- keyboard/list alternatives for waypoint editing.
Not included:
- PR1-09 trip-wide Tour summary;
- Wanderer integration;
- GPX import changes;
- new Tour types;
- Tour-to-Day drag-and-drop;
- further roadmap or Post-Release features.
Validation:
- focused RC-01/RC-06/RC-07/RC-08/RS-01/RS-02 tests passed;
- affected shared/server/client typechecks passed;
- migration and generated DB type checks passed;
- scoped ESLint passed;
- staged diff checks passed;
- owner manual tests passed for all six feature blocks.
The pre-migrate snapshot docstring said an older release refuses a database a newer TREK migrated. Only images from 5.0 on do; 4.3.x and earlier start on it without a warning. server/CLAUDE.md said anything the legacy baseline cannot map refuses the boot, which holds only above step 258.
The admin screen's model list (/api/tags) and pull (/api/pull) went out
without the Authorization header that the extraction and /api/show already
send, so an Ollama behind a Bearer proxy answered 401 and the list came back
as a 502. The controller now adds the key saved with the local provider, read
server side through LlmConfigResolver.readLocalServerAccess, and only when the
requested base URL reaches the same server root as the saved one, so another
address in the form never receives it.
A 401 or 403 now says whether the key was rejected or none is saved for that
address, and the admin form shows the server's reason instead of the axios
status line. A server without /api/tags (LM Studio, vLLM) is asked for its
OpenAI-compatible /v1/models instead, and the list body is read with a size
cap and checked before use.
Refs #2649
The TREK Places service takes an optional lang now and puts localities into
it, and names its OpenStreetMap layer rows in it with the settlement they are
in. TREK never sent one, so a Japan trip in an English UI kept reading
"福岡市博多区" and the layer rows had no place at all, only the local name.
Search, autocomplete, nearby and the details of an index place now ask in the
reader's language (TREK codes through toApiLang, anything that is no language
tag left off). A layer row with a locality reads "Sumida, Japan" like an index
row, the country named from its coordinate; without one it keeps the local
name as before. The offline area download stays without a language. REST and
MCP share the service methods, so both get it.
Refs #2592
Deleting a day said its bookings stay under Bookings without a day, which
read as if their date went too. They keep their date and stay listed,
upcoming ones included; the hint now says so.
Refs #2604
Deleting a booking also deletes the expenses linked to it, but the
question only said the booking would be deleted. A booking with a linked
expense now says so in every place that asks: the bookings list and detail,
the transport and transit dialogs, and the phone's booking cards. The MCP
delete_reservation description names the linked expenses and the stay a
hotel booking created as well.
Refs #2604
A booking saved with 18:00 to 09:00 and no end date was stored as ending
earlier on its start day without a word. The dialog now says that the end
time lies before the start time and that an end date is needed if it ends
the next day, on desktop and on the phone. It only warns and does not block
the save, since time zones can make such times correct.
Refs #2604
A booking linked to a stop has two days to go by, its own date and the
stop's. They part when the stop is moved to another day or the trip is
shifted with "keep bookings on their dates", and the plan then listed the
booking at the stop and again in the timeline of its own day. The stop now
wins wherever both are drawn: the desktop day plan and day detail, the
phone's plan tab and day sheet, and the road trip rail on both. The
booking's date is left as it is; a stop on one of the booking's own days
changes nothing, so a rental picked up at a stop still runs across its
other days.
Refs #2604
The travel stats and the upcoming bookings were only fetched when the
dashboard mounted, so "Trips" and "Days" kept their old values after a new
or duplicated trip and the sidebar still listed booking dates from before a
date change until the page was left. Every successful trip mutation now
reloads both, on desktop and mobile alike since they share the hook.
Refs #2604
Collab, packing bags, packing templates and imports, the category a place
form creates and the place imports are not kept on the device, so offline
their writes cannot wait in the queue. They failed with a bare "Error" or the
generic save message. While TREK is offline the failure now reads "This needs
a connection. Try again once TREK is back online.", on desktop and on the
phone's Collab notes; online nothing changes.
Refs #2601
The day plan wrote straight to the API, so offline a place dropped on a day
snapped back with "Error assigning place", a renamed day or a new day note
failed, and a moved stop, a leg's travel mode or who joins a visit were lost.
Visits (put on a day, taken off, reordered, moved, travellers, leg mode), the
day's name, notes text and travel mode, and the day notes (add, edit, delete,
move) now go through assignmentRepo, dayRepo and dayNoteRepo and the mutation
queue: the cached day changes at once, a visit or note made offline gets a
temporary id that the edits after it wait for, and an online write held behind
an older queued one joins the queue, since none of these carry a version
token. A new order of a day that names a visit made offline is sent with its
real id, or without it when its create failed. Desktop, the phone timeline and
note sheet all write through the store, so both are covered.
The replay reads each answer for what it names: the day answer keeps the
cached notes and visits, the participants answer lands on its visit, and a
visit the server refuses with 409 (a Tour already on that day) fails instead
of waiting as a conflict no retry can solve.
The trip's name, description, reminder and the archive flag wait in the queue
too, from the planner, the phone and the dashboard. What only the server can
do says so instead of failing: new dates, a day count, a currency or a cover,
a new trip, a copy or a delete, members, guests, share and invite links, and
a new order of the days, which moves the dates of the bookings.
Refs #2601
Every file write went straight to the API, so offline a star snapped back
without a word and moving a file to the trash failed with "Failed to delete
file". Notes, the place or booking of a file, a further link, the star, the
trash move, restore, delete for good and emptying the trash now go through
fileRepo and the mutation queue like the other trip data: the local copy
changes at once, and an online write to a file that still has an older one
queued joins the queue behind it, since file writes carry no version token.
The desktop file manager, its assign dialog and the phone's files tab,
menu, link and trash sheets all write through it.
The single file answers carry no ids of further links, so a replayed edit,
star or restore keeps the ones the local copy holds, and the link answer
({ success, links }) writes no row.
What only works online says so instead of failing: an upload shows that it
needs a connection, a place saved with newly picked files is stopped with a
hint, and removing a further link (its row id is known only to the server)
asks to try again once online.
Refs #2601
Every booking and stay write went straight to the API, so offline a new
booking failed with "Network Error" and was lost, and a status tick, the
travellers or a moved transport snapped back. Bookings (add, edit, status,
travellers, delete, day order) and stays (add, edit, delete) now go through
reservationRepo and accommodationRepo and the mutation queue like places and
costs: a temporary row offline, held behind an older queued write online,
since neither carries a version token. Desktop, the phone sheets, the day
plan, the road trip nights and the PDF all write and read through them.
The replay reads the booking from `reservation` only, so the travellers
answer ({ travelers, reservation }) lands on the booking it names. A booking
or its stay that names a place or stay created offline is sent with the real
id once that one has synced, inside create_accommodation too, so an imported
hotel reviewed offline can be linked to a place made for it.
What only works online says so instead of failing: an AirTrail import, a
booking file import, the AI retry and an attachment on a saved booking show
that they need a connection, and a save with newly picked files is stopped
with a hint, since an upload cannot wait in the queue.
Refs #2601
Every task write went straight to the API, so offline a new task failed with
"Network Error" and was lost, and a tick flipped back. Tasks now go through
todoRepo and the mutation queue like packing items: a temporary row at the
end of the list offline, the edits and the delete of it wait for its real id,
and an online write to a task that still has an older one queued joins the
queue behind it, since task writes carry no version token. A reorder offline
is queued without the tasks that have no server id yet. Desktop and phone
both write through the store, so both are covered.
Refs #2601
Every Costs write went straight to the API, so offline an expense failed with
"Unknown error" and was lost. Expenses (add, edit, delete, split, paid tick,
reorder) now go through budgetRepo and the mutation queue like places and
packing items: a temporary row offline, held behind an older queued write
online. The replay reads the expense from `item` only, so the split answer
({ members, item }) keeps the local payers and the paid answer ({ member })
never lands in the expense table. An expense linked to a place or booking
created offline is sent with the real id once that one has synced.
Payments are queued without a local row and a hint says they show once
sent; the ledger reads the balances again when the queue is through, and
offline it shows the last balances it read (new Dexie table, v10). Receipts
need an upload, so an offline save with new ones is stopped with a hint.
Refs #2601
The three planner view switches lived in sessionStorage only, so a new tab,
another browser or another device opened every trip on the plain view again.
They are now kept per trip in one user setting, planner_trip_views, validated
against a contract in @trek/shared. The server keeps only switches that are on
and only trips the user can still open, so a deleted trip or one the user left
drops out on the next write. The tab copy in sessionStorage stays and answers
until the settings have loaded.
Refs #2639
Autocomplete and the full search took their location hint only from the
open day or the whole trip, and dropped a trip spread over several cities
as too wide. Zoomed in on one city of such a trip, the search went global.
The plan map now hands its viewport to the trip store on desktop and on
the phone, and useLocationBias tries it first. A viewport wider than the
same 60 km radius, one across the antimeridian or one from another trip
is ignored, and the day and the trip follow as before. The viewport is
cleared when the plan map leaves the screen and on a trip switch.
Refs #2620
The OIDC page explains when a login by email links an account, what
email_link_blocked means and how an admin lets a blocked account link by
entering its address in the admin panel. The audit log page lists
user.email_change.
The userinfo endpoint answered email_verified: true for every account, although
nobody had confirmed the address. It now follows users.email_self_asserted, the
same mark the SSO login uses before it links an account by email.
An SSO login that matched no account by its subject took the account with the
same email address and linked the provider identity to it. That address proves
nothing when the account typed it in itself: anyone could sign up with, or
switch their profile to, a colleague's address and receive the colleague's
first SSO login. The provider switch branch (#2110) relinked a row on any email
match without even asking for email_verified, and also for another subject of
the same provider.
users.email_self_asserted remembers an address the account set itself: signup
(not the first account) and an email change in the profile settings set it, an
admin entering a new address clears it, and accounts the provider creates start
without it. Existing accounts keep linking, except those whose address moved
away from the registered one without an admin editing it. A login by email now
needs a verified address and an unmarked account, and is refused for another
subject of the same issuer, with email_link_blocked and a translated message on
the login screen. An email change in the profile settings writes a
user.email_change audit row.
Refs #2110
A guest's payer rows and recorded payments hung on the users row by ON DELETE
CASCADE, so deleting the guest wiped money that had really moved: expenses
they paid were left without a payer, settled payments reopened, and a custom
or receipt split lost the guest's share so the balances no longer summed to
zero.
Deleting a guest who paid for an expense or is part of a recorded payment is
now refused with a 409 (REST and the delete_trip_guest MCP tool, same text),
and the share dialog says to change the payer or delete the payment first.
Otherwise the guest's share is handed on: an equal split re-splits as before,
a custom amount goes to the others by largest remainder in the currency's
smallest step (members without an amount get one, so the expense stays a
custom split), and a receipt drops the guest from its lines and is split
again like the cost editor does. A legacy paid_by_user_id is cleared as on
account deletion. Account deletion goes through the same re-split. The
confirmation text in all languages and the wiki say what really happens.
Refs #2604
calculateTicketShares in the cost editor now converts the hundredths of the
new shared splitTicketLines. Same arithmetic as before: a line nobody is
ticked for is carried by everyone on the receipt and the leftover hundredths
of a line go to the lowest user ids first.
Refs #2604
The night went to the last visit of the hotel. On a day without hours the
booking seats its own stop first, and Days does not show that stop, so a rest
the traveller put mid-day was the last visit: it was held to the check-in and
the places behind it followed from there, the way the report describes.
The night is now the visit that ends the day, else the booking's stop, else
the last visit. A visit the traveller adds to the check-in day no longer moves
the night or its check-in. nightVisits in shared holds the rule, and the road
trip, calculate_roadtrip, the stored ride slot and the stop editor all read it.
Refs #2636
wiki/Languages.md has the account's language first, then the browser's.
The browser step only ran on the login page, so opening a trip with a
live session in a new browser (or after the site data was cleared) showed
English to anyone whose account had no language of its own.
The app now starts in the browser's language when nothing was chosen in
this browser, and keeps it when the account has none. Nothing is written
to the account; a language set in the account or picked in the app still
wins. The detection moved to its own module so the settings store can use
it at startup.
Refs #2604
A member without the right to archive a trip clicked Archive and read
"Failed to archive trip", while the server had answered why. The archive,
restore, delete and copy toasts now show the server's reason, as creating
and editing a trip already did; a failure without one keeps the old line.
The buttons stay where they are.
Refs #2604
The base-layer pill sits in the map's bottom-left corner on purpose. The
place inspector, a tour's detail and the day panel are centred between the
sidebars and were assumed never to reach that corner, but with both
sidebars open in a 1280 window they do, and the pill covered "Remove from
Day" and Edit. These cards now keep their left edge clear of the pill: a
narrow map moves the card right of it, a wide map keeps it centred as
before. The pill itself does not move, and the phone sheets are unchanged.
The Bookings tab showed "Book" at every width, which read as cut off. The
short form now only shows where the bar is tight; from xl up the tab shows
its full name.
Refs #2604
"Title is required" stayed on screen after the title was typed, until the
next save. Any edit in the trip form, desktop and phone, now clears the
last save's complaint; the next save checks again.
The keyboard date field always hinted DD.MM.YYYY, while it was filled and
read back in the user's own order: a Korean user saw "2026. 12. 10." under
a German hint. The hint is now the locale's own pattern (YYYY. MM. DD. in
Korean, MM/DD/YYYY in US English), and the field's name is translated.
Refs #2604
A screen reader lists a page's buttons together, and the costs views gave
it twenty "Edit" and "Delete" buttons with nothing to tell them apart.
- Edit and delete on an expense name the expense ("Edit: Dinner"), edit
and undo on a payment name who paid whom, on desktop and phone.
- A settle-up row shows only initials in its avatars, its names lived in a
tooltip; the row now carries them for assistive tech too, and its Settle
button says who pays whom and how much.
- The search field of every searchable select (currency, time zone,
category) was read out as "..."; it now says Search in the user's
language.
- The public link toggles on the phone share sheet and the journey share
toggles say whether they are on (aria-pressed), not only by colour.
- Japanese still said "you" in the settle-up lines.
Refs #2604
Several small things in the costs views, on desktop and phone alike:
- The outstanding card no longer says "You're all settled up" while a
payment is still open in Settle up.
- The custom split shows its suggested shares as a faint placeholder in
the currency's own form (30000 for won, not 30000.00), so it no longer
reads like an amount already typed.
- Amount fields take only the decimals the currency has: won and yen take
none, KWD and the other three-decimal currencies three. A switch of
currency cuts what was already typed. 1000.555 won was saved as typed,
listed as 1,001 and reopened as 1001.
- The expense sheet on the phone prints shares and sums with the same
formatter as the list, with grouping and the currency's decimals.
- A payment recorded with Settle or Settle all is dated on the viewer's
own day, and one without a date of its own is grouped by its UTC record
time read on the local calendar. In Seoul before 09:00 it landed on the
day before.
- "1 expenses need a payer" now uses the plural forms of each language.
Refs #2604
Creating the expense from a booking opened the editor on today, and the
cost a booking import writes by itself had no date at all, so it sat under
"No date". Both now take the booking's own date, else a hotel's check-in
day, else the day the booking sits on. The server sets it for every
booking write that creates a cost, so REST, MCP, plugin RPC and the import
agree.
A start time typed on a booking with no day picked no longer borrows
today's date on desktop. Today was mostly outside the trip, the server
then moved the booking onto day 1 while the list kept today. Desktop now
does what the phone already did and keeps the time without a date.
Refs #2604
The ledger nets every currency in hundredths, and the settlement handed those
hundredths out as they were. For a currency shown without decimals the panels
then rounded each balance and flow on its own: 10,000 KRW three ways read
+6,667 / -3,333 / -3,333, two flows of 3,333 sat under "You're owed 6,667",
and Settle booked 3,333.34 won.
The balances, flows and final budget lines are now allocated by largest
remainder in whole steps of the display currency (whole won or yen, unchanged
for two and three decimal ones), so the shown set adds up to zero and the flow
amount Settle sends is the whole amount on the row. The ledger, stored rows
and manual settlements are untouched. REST and the MCP settlement summary
share the same method.
Refs #2604
Moving a trip's start shifted the owner's vacay entries one row at a time,
checking each target against rows that were about to move themselves. A
block shifted by less than its own length kept its first days behind and
pushed the rest onto the weekend, even with weekends blocked.
The shift now decides every day before writing: a day whose target is a
blocked weekend or a whole company holiday is dropped, as is one whose
target already holds leave that is not moving; a half company holiday
leaves half a day. The rest move in the direction of travel inside one
transaction. The plan is the one the owner is active in, so a member of a
fused plan gets their days moved too, and a trip edit no longer creates a
vacay plan for an owner who has none. Open vacay views are notified after
the trip update commits.
The pure vacay helpers moved to vacay.helpers.ts and the trip window tests
to their own suite, so both files stay inside their size budgets.
Refs #2604
The undo wrote a new place from nine of its fields and put it on its days as a
bare stop, so notes, website, phone, hours, picture, the visit times, day notes
and travellers were gone while the toast said it was undone. It now writes every
field the place had and plans each visit again with its times, note, end of day,
route flag, travellers and leg modes, and casts the own rating again.
The server takes a booked night, its booking and expense, linked expenses, file
links, other members' ratings and an uploaded picture down with the place, and
writing the place again brings none of that back. Such a delete is no longer
offered for undo. Undo steps that name a deleted place are dropped at the delete
instead of failing later, and adding a place, moving categories and the phone
reorder and optimize now carry their place ids for that.
Taking a stop off a day and clearing a day undo with the visit details too. The
stays are read again when a deleted place had one, here and for collaborators.
Refs #2604
Someone removed from a trip stays on the expenses they paid for or shared,
but the costs views only knew the current members. They showed up as "?"
in the expense chips and the settle-up suggestions, and the balances and
final budget lists left them out, so the listed balances no longer added
up to zero.
The names now come from what the server already sends with the expenses
and the settlement, and anyone off the trip who still carries money gets a
row after the members, on desktop and on the phone. The expense form is
unchanged.
Refs #2604
The copy has nobody on it but the copier, yet it took every payer and
split row along. Its balances then had the copier owing people who cannot
see the trip, and the first save of such an expense dropped them and
shrank its total.
An expense anyone else pays for or shares now comes across as a planning
entry with its total and no payers, split or receipt lines; one that is
the copier's alone is copied whole. Assignment participants and the
legacy paid_by_user_id keep only the copier for the same reason.
Refs #2604
Removing someone from a trip leaves their payer and split rows on the
expenses they were part of. Every save ran the ids the form sends back
through the roster filter, so an unchanged save dropped them, total_price
shrank to what the rest had paid and a custom split stopped adding up.
Ids already on the expense now stay allowed, as payer or as member, and
can still be taken off it on purpose. A new expense and an id that never
was on the expense are filtered as before (#1952). The paid ticks of the
split carry over a form save instead of being reset. REST, the plugin RPC
and the MCP split checks all read the same allowed set from the service.
Refs #2604
A place whose image_url already is the place-photo proxy URL skipped the five
slot limiter and loaded its marker picture right away. A map showing a big
list, such as "All saved" in collections, asked for every picture at once.
Those downloads now wait for a slot like the provider lookups do; the entry
itself is still there immediately.
Refs #2559
The "All saved" view read every list in full at the same time, and every
collections:* websocket event started that again right away. An agent saving
places one by one over MCP sends one event per place, so an open collections
tab stacked up full reads of every list on the server.
The union now goes through the same batched loader the trip's saved-places
picker uses (four lists at a time), reloads asked for while it runs fold into
one more pass, and a list or union read that lands after the user switched to
something else is dropped instead of painting its places over the new view.
The page waits for a run of events to settle before it reloads and keeps the
selection on "All saved" like it already did on a single list. Desktop and
phone share useCollections, so both get it.
Refs #2559
API keys were split off from MCP to work without any addon, but the tab
that holds them still hid unless a photo provider, MCP, AirTrail, AI
Parsing or Dawarich was enabled. On a fresh install all of those are off,
so there was no way to create a key. The desktop tab now always shows;
the addon backed sections inside it keep their own gates. The phone layout
has no API Keys section and keeps its gate.
The wiki now says where keys are created and that no addon is needed.
Refs #2561
On an install without the TREK Places index and without a Google key the
suggestions come from Nominatim, and autocompleteNominatim asked it without
the box the client had sent, so every suggestion was ranked globally. The
search fallback on the same install dropped the bias the same way. Both now
pass it on as a viewbox with bounded=0, which ranks results near the trip
without hiding anything further away.
Refs #2620
The suggestion list under the place search showed the index's raw ISO code
("福岡市博多区, JP") and listed a place twice when the index and its
OpenStreetMap layer both carried it, which the search list after Enter
already merges. Suggestions now name the country through Intl in the
language of the search, and a layer row at the spot of an index row with a
matching name is dropped by the same rule mergeSearchResults uses, now
shared as samePlace.
Refs #2592
reservation:positions was on the ignored list, while the stop order the same
reorder sends was applied, so in other tabs and for other members the bookings
sat between the wrong stops until a reload. The event now writes the slots the
way the mover's own list did, for the day it names (day_id from REST, dayId
from the MCP tool) or as the global slot without one, and caches them offline.
A booking handed back by an edit or a reservation:updated event without day
slots keeps the ones it had, in the store and in the offline cache.
Refs #2631
The day list draws a row with a time by that time, and an untimed stop takes the
time of the timed row before it. A transit ride dropped behind a later stop was
stored there, drawn back at its time on the next render, and every untimed stop
behind the drop went with it; the next drag then stored that shuffle. A dropped
booking or timed note now lands in the slot nearest the drop the list draws it
in, so nothing else moves and what is stored is what is shown. Rides dropped
where their time allows stay on the drop, and the rule for untimed stops behind
a booking is unchanged.
The chronology check for a timed stop reads a booking's time for the day (the
arrival on the last day of a multi-day ride) instead of its departure, and the
three copies of the remove and insert step share one helper.
Refs #2631
The create and update answers, the reservation:updated broadcast and the MCP
reservation tools all hand back the single booking read, and that read never
loaded reservation_day_positions. The day list puts a booking moved on a day by
that slot before its time, and the client replaces the booking with the answer,
so a booking moved on a day jumped back to its time after every edit until the
next reload. The read now attaches day_positions in the shape the trip list
already sends.
Refs #2631
A GPX or KML import stores the track's first point as the place's position,
so the day route treated it as a single stop there and routed from the
track start to the next stop, a second road running alongside the track.
Any assigned place with a usable track now gets the gap a Tour already had:
the route runs up to the track start and continues from the track end, on
the desktop map and day list, the mobile timeline, the trip overview and the
PDF map, which now draws the track as a line. A place without a usable track
stays a plain stop and never shows as an invalid Tour.
Refs #2600
A chosen alternative or a hand-bent drive is stored as vias, and only the
road trip rail routed through them. The day route on the map, its
connectors in the day list and on the phone, and the overview of all days
asked the router for the bare stops, so they drew and measured the default
road again.
They now thread the same vias between two stops that follow each other in
the stored list, with the anchor counted the way the rail counts it, and
fold the answer back to one leg per stop pair. Only while the road trip
addon is on; avoid settings and the routing engine stay with the rail.
Refs #2580
A stop counted as the booked night whenever a stay began at its place on its
day, so a hotel on its check-in day twice (the booking's own stop and a rest
the traveller planned there) was read as two nights. Both were held to the
check-in: the rest waited for the room or was set back to it, and with a check-in
far from its arrival the rest and every stop behind it moved onto the next day's
card.
oneNightPerStay in shared keeps the night and its check-in on the last visit of
the place and turns the earlier ones into ordinary stops, still a hotel. The
browser and calculate_roadtrip both run it, the stored ride slot times only that
row by the check-in, and opening an earlier visit in road trip mode opens the
stop rather than the night.
Refs #2636
The invite notification links to /collections/:id, a list the invitee
cannot read until they accept. The phone screen had no invite UI at all
and showed "Start your first collection", and on desktop the rail with the
invites is hidden below 1080px while the empty state had no way to open it.
useCollections now recognises a route id that is a pending invite, keeps
it out of the route sync and puts it first in the invites it hands out.
The phone screen shows the pending invites as cards above the lists, and
the desktop body shows them when the person has no lists of their own,
or the one invite the route points at. Both reuse the existing invite
strings. A route id that is neither a list nor an invite (declined,
withdrawn, deleted) now drops back to /collections once the lists are in,
instead of leaving the page on a list it cannot show.
Refs #2645
The delete on a gallery tile only showed on mouse hover, so a phone or
tablet never saw it, and the phone screen had no delete at all. Where it
was reachable it deleted at once, although the server removes the photo
for good and takes it out of every entry that uses it.
The desktop tile now shows its delete on hover, on keyboard focus and
always where there is no hover, with a larger target there and an aria
label; viewers no longer get the button. The phone gallery gets the same
corner button for editors. Both ask through ConfirmDialog first, saying
that the photo also leaves the entries it is used in, and both delete
through the store action. The entry editor's photo controls follow the
same hover rule.
Refs #2593
The Leaflet journey map only drew the dashed route between entries when
fullScreen was off. Since the desktop sidebar and the mobile map plus
carousel view both pass fullScreen, the line was missing on the public
journey page (desktop and phone) and in the logged in views with the
Leaflet provider, while the GL renderer always drew it. fullScreen now
only enables wheel zoom, matching JourneyMapGL where it only sets the pitch.
Refs #2606
Phone booking and transport cards printed only the title of the linked day,
so a long title cut off in the field left no date at all. They now print the
date beside the title like the desktop card, as a span that never shrinks
while the title is truncated, and a day range takes the full width with the
time under it. The desktop card gets the same non-shrinking date.
A day without a date of its own (a trip planned without dates) now borrows
the booking's start or end date, on the desktop cards, the list, the detail
dialog, the shared page and the phone alike.
Refs #2590
The avatar menu of the mobile dashboard lost the version pill that the
desktop account menu shows under sign out. It now renders the same
appVersion from the auth store as a small footer, with the wordmark
following the active theme. Discord, help and the prerelease badge stay
desktop only.
Refs #2581
A stop with a start time but no end time went out as a VEVENT without
DTEND. Calendar clients fill in their own default length for that, mostly
an hour, so a 20 minute stop showed up as an hour. The event now ends at
start plus the place's duration_minutes when that is set and above zero,
carried over midnight like the stay markers. An explicit end time still
wins, so start and end under an hour stay as they were. The trip download,
both calendar feeds and the MCP export share this builder.
Refs #2641
Marking a region wrote its country with source 'manual', the same as ticking
the country off by hand, and removing the last region then took the country
away whatever its origin, Dawarich confirmations included. A country that
only comes in through a region is now stored as 'region', marking it visited
afterwards turns it into the user's own mark, and the cascade in unmarkRegion
only runs for a country without such a mark. Readers still see 'manual' for
those rows, so the country detail, the MCP resource and the plugin RPC answer
as before. Rows written before this change stay 'manual' and keep their
country.
REST, MCP and plugin RPC share the service. The atlas popup on desktop and
phone now asks the server whether the country is still marked before it
drops it from the map instead of guessing.
Refs #2597
The shared Costs tab grouped expenses by the stored category and printed it
as is, so the fixed keys ("accommodation", "flights") showed in English
whatever language the reader picked. It now groups through catMeta and
labels each group with costs.cat.*, the way the Costs tab does, so old
free-text labels also join their key instead of forming groups of their own.
Refs #2596
The field showed the stored place as soon as the search query was empty, so
deleting the last character snapped it back and the location could never be
removed. Clearing the field now empties the place and its coordinates, closes
the suggestions and drops any search answer that comes back afterwards. Typed
text without a picked result keeps the stored location as before. The hook is
shared, so the desktop dialog and the phone sheet both get the fix.
Refs #2615
NODE_ENV="production " passed the boot check but matched none of the
exact production comparisons, so the web UI was never served and every
page answered {"error":"Cannot GET /"}, while the banner showed a clean
"production". The boot import now trims these three variables in
process.env before anything else reads them, so Express, the plugin
children and readEnv() agree, and reports each change as a process
warning. Casing stays as it was. A padded "test" is only reported, so an
instance cannot switch to an in-memory database after an update. The
banner prints NODE_ENV as set instead of lowercasing it.
Refs #2565
A Google place URL carries two points: @lat,lng is the centre of the map
view, the !3d!4d data param is the place itself. extractCoords took the
camera first, so a pasted share link put the place around 170 m off and
reverse geocoded the wrong street. The data param is now read before @;
the order of @ and ?q=/?ll= is unchanged. REST and the resolve_maps_url
MCP tool share the resolver, so both pick this up.
Refs #2635
Moving a saved place to another list takes it out of the source list, but
updatePlace only asked for edit rights there, so an editor could empty a
shared list they may not delete from. The source now takes owner or admin,
like a delete, and the target still takes edit rights. REST and the
update_collection_place tool share the service path. The Move to list
action on desktop and phone follows the delete permission.
The tuple cast on clips.meta.json failed tsc. The clip lookup and the 3D
opt-out move into mascotClips.ts so the component file only exports the
component, the crop offset is allowed as a deliberate physical side, and
MDancingTrek leaves the format baseline.
Every empty-state scene of the TREK mascot now plays a pre-rendered 3D clip
instead of the SVG: a seamless loop per scene and theme plus a one-shot
animation on click that cross-fades in and out.
The clips are plain H.264 MP4 with stacked alpha (colour on top, the mask
below) that a small WebGL shader joins into a transparent image, so the same
file plays in every browser including iOS, decoded in hardware. The videos stay
2 px stand-ins and are drawn onto a canvas, which keeps browser video controls
off the mascot. Each clip is cropped to the area its scene uses and placed back
inside the 88:96 box, so every scene keeps the same scale.
MDancingTrek picks the clip when one exists for the scene and face and falls
back to the SVG under reduced motion, without WebGL, when autoplay is refused,
in tests, or with localStorage trek-mascot-3d set to 0.
The stays, hire cars and plugin label under a day's name now live in DayHeadPills, so DayPlanSidebar shrinks back under its size baseline. The stay fields are read through their types instead of an any cast, and the size, eslint and theme-disable baselines are lowered to the new counts.
The day head falls back to metric when no distance unit is set, VISIT_BOX is no longer exported, the day head is split into smaller parts and the booking icon is created once. The component tests now check the pill badges, the day tile and the ride cards instead of the old boxes, and both baselines are lowered to the new counts.
Both dialogs now open in the same frame as the stay and stop dialogs: an icon tile and title in the head, the content in a card, and the actions in the footer. Closing stays blocked while an import runs. In the right column, Search and Manual take the size, text and icon of the day plan's toolbar buttons.
Day heads take the day plan's layout: the number on a raised tile with the date
under it, the name beside it and its facts as pills that never wrap but give way
in steps. Rides, terminals and hire desks are cards in their type's colour with
their icon on the line and their times in the arrival column; a stop's bookings
sit on its own badge row. Every badge on the rail is one pill shape, the drive
bands are round, and the stay, stop and driving settings dialogs use the shared
dialog frame.
DAY-SVC-048 and DAY-DEL-011 pinned the old reorder, which handed a day
without a date the first slot's date; they now expect it to keep none and
the day it passes to keep its own. OIDC-SSRF-003 reads oidc.service.ts as
text and missed the token call once Prettier put a trailing comma after its
hop budget of 0; the pattern takes the comma.
The day badge fix grew DayPlanSidebar.tsx and its test past their size
baselines. The check-out, through, check-in order of the header's stays is
now dayHeadStays() with a test of its own, and the day header tests sit in
DayPlanSidebar.dayHead.test.tsx; both files drop below their old entries.
knip found them once the server lanes and the client ratchets met: re-exports
and helpers that moved into place-import, nest-rpc and auth-core but kept an
export nothing reads, the demo-denied helper the central MCP gate replaced,
and two settings keys only the selector uses. @OptionalAuth stays exported
as the documented opt-out and is marked @public. The baselines for knip and
the unused en keys drop to the merged tree.
Reordering handed out the dates by position, so moving a day added without a
date to the front gave it the first date and took the date of the last dated
day. The undated row stayed at the end and the move looked like it did
nothing, by drag and by the arrows alike. The dates now go to the dated days
only, in their new order; the rule lives in shared so the server, the
planner's optimistic list and the MCP tool apply the same one.
lint:i18n-keys only listed unused en keys as information, and it looked for
references in client/src alone, so every key the server sends (system notice
and notification channel keys in its registries) read as dead. The reference
search now also reads server/src and plugin-sdk/src; those references only
mark keys as reached and are not held to the client's key rules. The unused
keys per en domain file are counted against scripts/i18n-unused-baseline.json,
which only shrinks and fails on an entry above its count until --update
lowers it. Today's 841 unused keys in 44 files are frozen, none deleted.
Nothing looked for dead code: no knip, depcheck or ts-prune, noUnusedLocals
off. knip.jsonc configures knip over the client, server and shared workspaces
with their real entry points (main.tsx, index.ts and bootstrap, the MikroORM
config, the forked plugin host and its SDK surface, migrations and seeders
found by glob, the shared build entries with one per locale, the scripts,
the tests and the Playwright launcher). lint:knip (scripts/ci/knip-ratchet.mjs)
counts its findings per category and file against scripts/ci/knip-baseline.json,
which only shrinks and is built on the client's ratchet helpers. Today's 427
findings are frozen, among them unused server dependencies (nodemon,
@types/uuid, eslint-plugin-prettier, @mikro-orm/cli), the unused root unrun,
and imports of packages no package.json lists (leaflet.markercluster, geojson).
False positives are fixed in knip.jsonc with the reason. knip is a root
devDependency; the check runs in server-quality, the job that installs every
workspace, and knip.jsonc joins the code area of the change classifier.
The client compiles with strict: false, so nullable contract fields arrive as
if always present and implicit anys stay invisible. tsconfig.strict.json is the
same program with strictNullChecks and noImplicitAny, and lint:strict
(scripts/strict-lint.mjs) counts its errors per file against
scripts/strict-baseline.json: a file may not gain one, a new file has to be
clean, and an entry above its count fails until --update lowers it. Nothing is
converted here; today's 534 errors in 165 files are frozen. allowJs stays: the
tests import the .mjs scripts and read their JSDoc types through it.
The todo MSW handler maps a cleared priority to its default so the test
helper written in this branch starts at zero. The check runs after the
typecheck in client-quality.
The client lint ran untyped with two react-hooks rules, so nothing saw a
floating promise, an async handler where a void one is expected, a whole-store
Zustand subscription or a window event bus, all of which client/CLAUDE.md
forbids. eslint.config.mjs now lints with the project service and adds, as
warnings: no-floating-promises, no-misused-promises (void returns in JSX
attributes included), the rest of react-hooks v7's recommended set,
complexity 20, max-depth 4, max-params 5 and max-lines-per-function (120 in a
.ts function, 300 in a .tsx one, whose body also holds its markup),
no-console and no-non-null-assertion. Three local rules in
scripts/lib/eslint-rules.mjs cover what no stock rule sees: an empty
.catch() callback, use*Store() without a selector (the message points to
useShallow, since an object selector without it loops on Zustand 5), and
window.__* globals and window.dispatchEvent with an event of our own.
scripts/eslint-baseline.json takes today's count of each new rule, so the
existing code is frozen, not changed, and each new instance fails
lint:warnings. The CI step gets the 8 GB heap typed linting needs.
theme:lint only reported a total and exited 0, skipped whole files by name and
never ran in CI. It now counts per file under src/ against
scripts/theme-baseline.json, which only shrinks, like lint:rtl: the colour
literals, arbitrary colour classes and numeric font sizes it already saw, plus
palette classes (bg-gray-100, text-indigo-600, bg-white), raw text sizes
(text-sm, text-[13px]) instead of the type tiers, z-index literals above z-50
instead of a step of the layering scale, and reads of the dark_mode setting.
The file exemptions are gone; the map, PDF and brand files carry their count in
the baseline instead. theme-lint-disable markers stay valid in a comment and
are counted against scripts/theme-disable-baseline.json, like rtl-lint-disable
in lint:rtl; a new one raises its file's entry by hand, where review sees it.
The comment handling lint:rtl already had moves into scripts/lib/source.mjs so
both read source the same way.
The check runs in the client-quality job.
MapViewAuto took props: any and the Leaflet MapView destructured any, so a prop
one renderer learned and the other never heard of went unnoticed.
components/Map/mapViewContract.ts holds MapViewProps, derived from the GL
renderer's props and what the callers pass; MapView and the GL bindings in
glLazy declare satisfies ComponentType against it, and the GL renderer extends
it with its engine. The file imports types only, so check:gl-split is not
affected. Types only, no runtime change.
The types show two known gaps, recorded in the contract rather than papered
over: onMapReady is called by the GL renderer only, and callers pass pins
without trip_id (assignment and collection places) while both renderers type
their pins as Place; MapViewAuto bridges the latter with one commented cast.
contractHandler() parses the request body against the shared request schema,
answering 400 like the server's Zod pipe on a body it would refuse, and the
mock's answer against the response schema, failing on a factory that drifted.
The trips, budget, places, todo, tags and categories, day notes, packing,
reservations and assignments handlers and the vacay plan and colour updates use
it wherever shared has a schema; the offline bundle and the other vacay routes
have none yet and stay plain.
The typed bodies showed where the old handlers echoed request shapes back as
stored rows: payers and members are no longer echoed into a budget item, and
checked and is_private come back as 0/1 the way the server stores them.
resetAllStores read a hand list of seven stores, and a new store simply never
joined it. tests/helpers/storeRegistry.ts now names every Zustand store in
src/ as either reset or left alone with the reason, and a test fails on a
store that is in neither list or on an entry whose store is gone. The set of
stores that are reset is unchanged.
BudgetPanel, its summary, category table, inline edit cell, add row, per person
line, pie chart, ring avatar and the useBudgetPanel hook were replaced by the
costs panel and nothing outside their own tests imports them any more.
BudgetPanel.constants, BudgetPanel.helpers and BudgetPanelMemberChips stay in
use. The rtl, format, offline and eslint baselines drop the deleted files.
.git-blame-ignore-revs lists the formatting commit, so GitHub's blame view
and a local git blame with blame.ignoreRevsFile set attribute each line to
the change that last meant something.
A mechanical prettier --write over server/src, server/tests, server/scripts
and shared/src with the workspaces' .prettierrc; nothing but layout and
import order changes, and every file keeps the same set of imports. shared
is formatted throughout and its format baseline is empty. The server
baseline keeps the 35 files that Prettier would grow past their lint:size
limit (they need a split first, which is not a formatting change), and the
generated entity files stay as their generator writes them.
src/db/entities/ is written by scripts/generate-entities.ts, and
check:entities holds every file byte for byte to its output, which puts the
MikroORM import first where Prettier's import sorting puts it last.
Formatting them would fail that gate, and npm run format would do it on any
run. lint:format no longer scans the directory (its 133 entries leave the
format baseline) and a .prettierignore keeps the format script off it.
Following the method chain was not enough: Prettier also wraps a cast read
in parentheses (const n = ( above db.prepare(), puts a for head above its
body and nests the call in an argument list, each leaving it more than one
line below its marker. When the line under a marker holds no call, the
marker now excuses the first call further down in that statement, up to
the line ending it with a semicolon, never past a blank line and at most
eight lines down. Every file counts the same as before on today's tree and
on the formatted one.
Prettier wraps the emoji regex onto the line after its declaration and
spreads two any-cast lines in the plugin tests over several lines, which
would leave the casts outside the eslint-disable-next-line above them. The
regex directive now sits on the line right before the literal, the dev-link
controller takes never instead of any for the collaborators it ignores, and
the supervisor test reaches the private kill/onExit through one cast.
Prettier breaks a long schema into z on one line and .object( on the next,
which the count read as nothing: formatting the tool files would have taken
134 hand-written fields out of the tally. Whitespace before the dot counts
now. Four files already held such a split call, so their entries rise to
what they really hold (atlas 23, doc-sync 1, help 4, transit 16).
Prettier splits a long db.prepare('...').all() into db on one line and
.prepare( on the next, which put the call two lines below its marker and
made every parity oracle in the repository tests count as a raw statement.
The marker now also covers the method-chain lines that continue the
statement right below it, up to the first line that does not start with a
dot.
The sort-imports plugin moved bare imports along with the rest, which on
src/index.ts would have put reflect-metadata after the modules that need it
and the env validation ahead of dotenv. With importOrderSideEffects off, a
side-effect import stays in place and the imports between two of them are
sorted on their own. The three test files that were already formatted gain
the blank line the new grouping writes after reflect-metadata.
bootTestApp replaced JourneyPhotoCaptureService with a factory calling its constructor positionally, only to hand it an offline geocoder, so a dependency the service gained broke the boot of every MCP suite. The container now builds the service itself and the OsmClient's reverse lookup answers that nothing is known, which also keeps every other reverse lookup a booted suite reaches inside the process.
lint:test-mocks missed vi.mock(import('../src/config'), ...), the typed form Vitest takes, and counted mocks quoted in doc comments, which left slack a real mock could take. lint:test-new-service missed a service built under an import alias or as new (Name)(...). Both now strip comments first and recognise those forms; the baselines lose the entries that were only comments.
A response-contract-exempt comment counted any handler as covered, whatever it returned, so one comment line let a JSON handler past the ratchet. The marker now counts only on a handler taking @Res() or @Next(). The five vacay holiday proxies return the provider's JSON, so they lose their markers and are counted again until they declare a schema.
createInvite returned { error: 'Trip not found', status: 404 } and the admin controller never looked at it: POST /api/admin/invites with an unknown trip_id answered 201 with an empty body and wrote an admin.invite_create audit row for resource 'undefined'. The service throws the DomainError now, so the route answers 404 { error: 'Trip not found' } and writes neither the invite nor the audit row.
The ratchet only read *.service.ts for status literals and one branch shape in controllers and tools, so an envelope in a helper, impl or provider file, a branch written as result.error === 'not_found', and RPC handlers all went uncounted. It now counts the literals in every file under src/nest outside the HTTP layer, and comparisons, negations, optional chains and ternaries on a returned error in controllers, tools and RPC handlers. The baseline takes what the wider count finds.
src/nest-rpc, src/nest-mcp and src/mcp import domains and are imported by them, but the domain map only knew src/nest, so their edges dropped out of domainCycles and their repository injections out of foreignRepositories. Each layer now counts as one domain. The repository rule reads the syntax tree for @InjectRepository(X) and for getRepository(X) on an EntityManager, so a comment no longer counts and a foreign read through the EntityManager no longer passes. The baseline takes the cycles and the repository reaches the wider view finds; settings -> Users goes, it was a comment.
VacayMcp read the caller's username and email straight from UsersRepository, a table the auth domain owns. UserLookupService in auth-core answers exactly that, so vacay neither imports AuthModule nor injects a foreign repository, and its foreignRepositories entry goes.
sync_trip_documents and the five manual school-holiday tools carried no demo check of their own before the gate moved into the registry, so a demo account could call them. The gate now lets exactly those through, and a test pins the list. upload_trip_file keeps the gate's order: a demo account is refused before its trip access is checked, where the tool used to answer a trip it cannot see with the no-access result first.
lint:test-mocks counts vi.mock/vi.doMock of src/config, src/nest/realtime and src/db per file under tests/, lint:test-new-service counts new XService(/new XRepository( there with every class whose constructor takes nothing allowed (derived from src/ and tests/helpers/), and lint:size holds files under tests/ to 2000 lines against their own baseline. Both counters share scripts/lib/count-ratchet.mjs, which gains a prepare hook for a check that reads the tree first. The three run as steps of the existing server lint job; the CLAUDE.md files and the module guide describe the test helpers they point to.
src/services is gone, so the folder named a layer that no longer exists. The airtrail suites go to tests/unit/nest/integrations, the crypto, cookie, avatar, password policy, timezone and conflict helpers to tests/unit/nest/common, the kitinerary mapper to booking-import, the photo media suite to photos, the atlas bundle guard to tests/unit/nest and the two migration suites to tests/unit/db. Only the relative import paths change, and the format baseline follows the new paths.
createTestModule (tests/helpers/test-module.ts) builds a Nest testing module of the domain modules under test over the suite's test db, with a FakeRealtimeService and provider overrides for the collaborators a suite controls. bootTestApp (tests/helpers/test-app.ts) boots AppModule once per file, with recording realtime, fixture storage and an offline geocoder, and the MCP harness takes McpRegistryService from it inside a request context instead of hand-wiring roughly eighty services. The days, day removal and day notes service suites take their subjects from DaysModule and DayNotesModule.
A hand-built service gets a FakeRealtimeService where it took a RealtimeService, a Nest testing module overrides the provider, and a suite over buildApp() spies on the container's instance (spyOnRealtime). The 167 vi.mock factories on src/websocket are gone, RealtimeService reads the socket registry in ws-state.ts directly, and the re-export stub src/websocket.ts is deleted. The facade's own suite now checks what reaches a socket instead of which module function it called.
tests/setup.ts mocks src/config once with tests/helpers/test-config.ts. A suite that needs other values calls overrideTestConfig(), and the setup file restores the defaults after every case. The 191 per-file factories are gone; config.test.ts takes the real module through vi.unmock, and the few suites that signed tokens with their own literal secret read it from TEST_CONFIG.
Tightens the two comments the rawBodyBase64 field and the type parity note
added to runtime/plugin-sdk.ts, lowers its lint:size entry, and points the
shared guide at the client files it names by their full path.
The vacay plan, calendars, years, entries, stats and shares, and the
journey list, create, tracks, stats, share link and every success answer
get response schemas in shared and a @ResponseContract on their handler.
The holiday proxies pass the provider's answer through and are marked
exempt. The journey e2e stubs return whole rows, as the service does.
shared gains the response schemas these routes answer with (rows in their
storage spelling, as the routes have always sent them) plus the common
success and empty answers, all exported for the client to type its calls
with. Every JSON handler in the four controllers carries its
@ResponseContract; the file download, which streams, is marked exempt. The
files e2e fixture table gains the note_id column the real one has.
The decorator records the @trek/shared response schema a route answers
with. Under NODE_ENV=test it adds an interceptor that parses every response
and fails the request with a logged 500 naming the handler and the issues,
so the e2e and integration suites catch a response that drifted from its
contract; in production nothing is added to the request path and the body
is never rewritten. lint:response-contracts holds the handlers still without
one to a per-file baseline that only shrinks, and runs in server-quality.
The SDK accepted any number as apiVersion while the host refuses anything
but a positive integer no newer than its plugin API, a false green for an
author. The manifest patterns, sets and caps now live in the host's
protocol/manifest-rules.ts, and gen-plugin-facts writes them and
PLUGIN_API_VERSION into the SDK, so check:plugin-facts covers them. One
manifest corpus is judged by both validators in a server test (the SDK may
only be stricter, case by case), and a typecheck test holds the SDK's plugin
API types to the host child's copy, which gains rawBodyBase64.
RESERVATION_TYPES lists every type with its flags (transport, carrier,
rental, stay, legs, creatable through MCP), and the named subsets derive
from it in picker order. The MCP transport and booking enums, the share
view's travel filter, the road trip carrier query and the shared
CARRIER_TYPES now read it. The reservation response types type and status
as the known values or any stored string, so rows outside the catalog still
pass; requests stay as tolerant as before.
PERMISSION_ACTIONS, PermissionLevel, PermissionKey and evaluatePermission()
move to shared/src/permissions. PermissionsService decides through
evaluatePermission and re-exports the catalog for its importers, and
@RequirePermission takes a PermissionKey, so a misspelt action fails to
compile. Levels, defaults and every verdict stay as they were.
toMinor, sumMinor, splitEqualShares and currencyDecimals move to
shared/src/money. The budget service and its MCP tools call them instead of
private copies, and the share table both sides were pinned to becomes the
spec of the shared function. Stored values and every computed figure stay
the same.
Prettier on every new or moved file that was not formatted, so lint:format
holds them; the files taken off the format baseline are the ones that moved
or were formatted on the way.
The module comment no longer spells an @InjectRepository call the scanner took
for a foreign Users injection in nest/database, and journey's Users entry is
gone now that it reads the Immich flag through UserImmichRepository.
The reader left places.service with four strictNullChecks errors that a new
file may not carry. Declaring fetchPage's result as a page or a list import
error lets the 'error' check narrow it; the places and maps entries drop off
the strict baseline.
Non-null assertions become explicit guards, the two demo-gate loops collect
their answers with Promise.all, and the MapsService imports the maps tests no
longer use are gone. The src counts the refactors lowered are taken off the
warnings baseline.
McpToolGuardsService injects RealtimeService, and McpSharedModule counted on
some other module of the app having imported the global RealtimeModule.
AuthModule used to do that for every domain that imported it for the demo
check; with those imports gone, a test app assembling ShareModule alone no
longer resolved the guards. The module now names the dependency itself.
A domain injecting the repository of a table another domain owns couples
the two below the service layer, and nothing counted it. The new
foreignRepositories rule reads every @InjectRepository under src/nest,
looks the entity up in scripts/repository-owners.json (an injected entity
without an owner stops the run) and holds each domain/entity pair to the
baseline, which only shrinks: 209 pairs today.
Thirty services, controllers, MCP tools and RPC handlers outside the trips
domain injected TripsRepository for one call, findAccessible. They now
inject TripAccessService from the trip-membership leaf, in the same
constructor slot under the same field name; it answers with the same
access row (or undefined) and binds the id exactly as given, so each
caller keeps its own coercion and its own 404, 403 or tool error.
VacayMcp injected AuthService only to call getCurrentUser for the caller's
username and email before sending an invite or sharing the calendar, which
kept VacayModule importing AuthModule. UsersRepository.findUsernameEmail
answers the same two fields (and undefined for an unknown id, which the
tools already refuse as 'User not found.').
AddonsService read the transit backend through transit/transit-provider.ts
and built photo-provider endpoint paths with memories.helpers, so the addons
domain, which nearly every domain imports for its gates, depended on
transit and memories. Both are free functions over a repository or plain
strings; they move unchanged to common/, where the transit and admin sides
read them too.
The websocket gateway injected JourneyDomainService to decide who may join
a journey's book room, so the realtime domain, which nearly every domain
imports for its broadcasts, depended on journey and closed a cycle through
all of them. It now injects a JOURNEY_ACCESS port declared in realtime;
JourneyAccessModule (global, imported by AppModule beside the gateway
module) binds it to the same JourneyDomainService method.
The gateway imported hostVersion from the plugin install code to put the
version in its welcome frame, which made the realtime domain, imported by
nearly every other, depend on plugins. The running version (image VERSION
or APP_VERSION, else package.json) is app config; runningVersion() holds it
and the plugin host-compat checks call the same function.
NotificationsController, SettingsController, OidcController and the packing
template controller each imported their body DTO from admin/admin.dto.ts,
and AdminController imported the same four without using them. Those
imports tied the notifications, settings, packing and oidc domains to the
admin domain and through it to everything admin depends on. The classes
move unchanged (same names, same @trek/shared schemas) into each domain's
own dto file.
Every domain's *.rpc.ts was written against rpc-kit, rpc-params, rpc-errors
and PluginGuards, and every domain module imported PluginGuardsModule, all
of it inside src/nest/plugins, while PluginsRuntimeModule imports those same
domains so their handlers exist at boot. That made the plugins domain part
of every cycle. The kit imports no domain but the protocol tables and
permissions/addons for the guards, so it moves to src/nest-rpc beside
src/nest-mcp, with its tests, its README and its extraction rule; the
runtime module keeps its domain imports, which the boot-time coverage check
of the wire surface needs in any subset app.
JwtAuthGuard, OptionalJwtAuthGuard, CookieAuthGuard, AdminGuard, the global
and MFA guards, @CurrentUser, @Public, JWT verification and the ephemeral
token service lived in the auth domain, so every controller that checked a
session depended on the whole account module and its imports. They import
no domain themselves; under src/nest/auth-core they are shared kernel like
common/, every importer points there directly, and AtlasModule and
NotificationsModule stop importing AuthModule, which they only did for the
demo gate. lint:boundaries counts auth-core as kernel: 57 fewer cyclic
domain edges, the largest cycle shrinks from 46 to 40 domains.
Every write tool used to open with its own demo check, 196 copies across 30
tool files, most of them injecting AuthService for nothing else. The
registry now takes a toolGate (built from the container) that runs before
each registered tool handler, and TREK's gate refuses a demo account on
every tool not annotated read-only with the same canned result. The
per-tool checks, the AuthService injections, AuthService.isDemoUser and the
AuthModule imports that only fed them are gone; ESLint refuses a demo check
in a tool file, and a registry-wide test calls every write tool as the
demo account.
UserImmichRepository, UserSynologyRepository and UserAirtrailRepository hold the eighteen connection statements UsersRepository carried, on the same users table and with their statements unchanged, provided through a small module as factories over the EntityManager. ImmichService, SynologyService, AirtrailService and JourneyService inject them instead of the account repository. UserRow drops the dead immich_access_token column (the table keeps it), and a shape test pins the account repository's size.
app-settings.registry.ts lists every app_settings key the server reads by name with its shape, default, environment override and flags. The admin form's keys, its masked and encrypted handling and the public config's reads derive from it, isOidcConfigured() replaces the three issuer/client-id checks in AuthService, and OidcService resolves its values through the same rule while still requiring the client secret. Every literal getValue() read goes through readAppSetting()/resolveAppSetting(), and ESLint refuses a new one.
The login toggles' legacy fallback, the public config, the lockout guard of the settings save and the provider config each decide on their own whether OIDC is configured. These cases hold what each answers, including that only the provider config also wants the client secret, before the checks are shared.
The ICS export reads the stops and notes of all dated days in two statements instead of two per day, the journey's full read and entry list take the source trip titles in one join instead of one read per entry, and reordering a day's stops or sorting it by time writes every position in one CASE update. Results are unchanged: each day keeps its order (the id settles a tie), a missing trip still names null, and an id listed twice keeps its last position.
server/CLAUDE.md now says which API new repository code uses: Kysely for join, aggregate and set-based statements, the entity methods for single-table CRUD, no new QueryBuilder calls. The TrekRepository docstring no longer calls Kysely an escape hatch. lint:query-api counts this.qb()/createQueryBuilder() per repository file against a baseline that may only shrink and runs in the server CI job.
The same file documents the JSON column and timestamp rules from the two changes before.
decodeJson/decodeJsonResult/encodeJson read and write a JSON-in-TEXT column against the schema and fallback declared for it in db/json-columns.ts, and log a row that does not decode. reservations.metadata, the OAuth scope and redirect lists, consents, API token scopes, MFA backup codes, journey entry tags and pros/cons and poll options go through it. Readers that used to throw on a broken row (a hotel update, the ICS export, the price mirror, the OAuth token and session reads, a consent, a journey entry, a poll) now take the tolerant fallback the other readers of the same column already used.
A DbTimestamp brand, produced only by dbNow(), types the expiry writes of OAuth tokens and password reset links, and collection edits stamp updated_at the same way. parseDbTimestamp() reads the stored text as UTC and replaces the oauth-local parser; utcSuffix lives once beside it. A data migration rewrites the ISO values those writers left behind, and the OAuth session lists keep answering the expiries in ISO.
Two outputs change on purpose. collections.updated_at used to come back in two spellings: a collection never edited answered the column default (2026-01-02 03:04:05), an edited one the ISO text its writer stored (2026-01-02T03:04:05.678Z). It now answers the canonical text in every case, in the PUT and cover responses and, once the migration has run, in every read; the shared schema types it as a plain string and the client does not read it. The ISO expiries of the OAuth session lists are rebuilt from the stored second-precision text, so they end in .000Z instead of carrying the milliseconds of the write.
tsconfig.strict.json is the build config with strictNullChecks and
noImplicitAny on. lint:strict runs tsc over it and holds the errors per file
to scripts/strict-baseline.json (204 in 109 files today): a file may only
lose errors, a new file starts at none, and an entry above its count fails
until --update lowers it. An error outside any file or an unreadable tsc run
fails closed. It runs in the server job; the build itself stays on
strict: false.
shared's lint ran eslint --fix with prettier/prettier as an error, so in
lint-prettier.yml the format:check after it only ever saw files it had just
rewritten, and the server's format scripts pointed at a test/ directory that
does not exist. Now lint is check-only in both (lint:fix applies fixes), the
server scripts cover src/ and tests/, and a lint:format ratchet on the shared
library holds every file outside scripts/format-baseline.json to Prettier in
both workspaces: in the server job of test.yml and in lint-prettier.yml,
which keeps providing the lint check. Formatting leaves ESLint in shared, and
the unused eslint-plugin-prettier goes. Once the tree is reformatted in one
commit, --update empties the baselines and the check is a full prettier
--check.
Server lint ran without --max-warnings or a baseline, so the rules left on
warn could grow without limit. lint:warnings runs ESLint with the server's
own config through the shared tally, fails on any error or inline rule
config, and holds the warnings per rule for src, tests and eslint-disable'd
messages to scripts/eslint-baseline.json, which only falls. It replaces
lint:check in CI. no-await-in-loop, no-non-null-assertion and no-console come
in as warnings held at today's counts. npm run lint no longer rewrites files;
lint:fix does.
server/src/nest/README.md drops from 988 lines to under 200: the module
shape with weather as the reference, the cross-cutting pieces, the guards,
parity and the test layout, without the bridge playbook and the services
recipe that contradicted the code. The migration log moves verbatim to
server/docs/nest-migration-history.md, marked historical.
scripts/ci/doc-paths.mjs resolves every backtick path in the CLAUDE.md files,
the root README, the blueprint and the wiki against the tracked files. The
paths already missing are listed in a baseline that only shrinks, and the
changes job runs it so docs-only changes are checked too. The wiki no longer
says the Playwright flows stay out of CI.
None of their keys is read anywhere in client, server, shared or plugin-sdk,
neither literally nor through a key built from a prefix: the photo, sign-up
and stats screens they were written for are gone. All 27 locales lose the
same three files, so parity holds.
Thirteen interfaces nobody imported are gone. Trip is now the selectable row
of the generated TripsTable, so currency, is_archived and reminder_days read
as the nullable columns they are and feed_token and reminder_sent_for exist.
The two React profiler readers were one-off analysis helpers, and the i18n
migration read client/src/i18n/translations, which has been gone since the
locales moved to shared/.
Only their own tests imported them. The MCP knobs are parsed in
app-config/parsers.ts, which tests them directly, and toCamelCase was never
wired into a plugin RPC handler.
Nothing runs nodemon (dev goes through scripts/dev.mjs), the coverage provider
is istanbul, and uuid 14 ships its own types. uuid itself stays: the OIDC and
Unsplash services import it.
scripts/lib/ratchet.mjs carries the tree walk, the CRLF-safe reader and the
baseline handling the client ratchets already use, next to a warning tally and
a format comparison that take ESLint and Prettier from the calling workspace.
server lint:size now runs on it with the same messages and exit codes, and the
libraries have node:test suites that CI runs in the server job.
Every vacay refresh on desktop (opening the page, a vacay:settings or
vacay:invite update from another member) sent a second GET /trips just to
redraw the same trip dots, even when the leave-year window had not changed.
Cause: useVacayCalendarLogic keyed the trip dot reload on the year settings
object when tripsFollowWindowShape was false (the desktop grid). loadAll()
replaces that object with a deep-equal copy on every refresh, so the effect
ran again. The phone screen already keyed it on the window shape.
Fix: the trip dots always reload on the window shape (year type, start month,
start day, hire date), which covers every field the dot filter reads. The
option only existed to keep the desktop behaviour, so it is removed from the
hook and from both callers.
Red test: FE-COMP-VACAYCAL-010 (useVacayCalendarLogic.test.ts) failed on the
unfixed code with "AssertionError: expected "list" to be called 1 times, but
got 2 times".
Changed assertions:
- useVacayCalendarLogic.test.ts FE-COMP-VACAYCAL-010: was "the desktop mode
reloads trips on every new settings object" expecting tripsApi.list to be
called 2 times after a deep-equal settings replacement; now expects 1.
- FE-COMP-VACAYCAL-011 no longer passes tripsFollowWindowShape and no longer
names the phone mode; its assertions are unchanged.
When loading the users that can be invited to a fusion or given a calendar
share failed on desktop, the dialog kept showing the list from the last load,
so a user could pick someone who is no longer available and only learn it
from the failed send. The phone sheets already emptied the list and showed
their "no users available" state.
Cause: useVacayUserPicker only emptied the list on a failed load when its
clearOnLoadError option was true; the desktop VacayPersons and
VacaySharedCalendars dialogs passed false, a leftover of the old silent
catch in the desktop dialogs.
Fix: the hook always empties the list when the load fails. The option only
existed to keep the desktop behaviour, so it is removed from the hook and
from all four callers.
Red test: FE-COMP-VACAYPICK-002 (useVacayUserPicker.test.ts) failed on the
unfixed code with "AssertionError: expected [ { id: 2, username: 'bob' },
...(1) ] to deeply equal []".
Changed assertions:
- useVacayUserPicker.test.ts FE-COMP-VACAYPICK-002: was "a failed load keeps
the last list on desktop" expecting available to equal USERS; now expects
an empty list.
- FE-COMP-VACAYPICK-003: was "a failed load empties the list on the phone"
through load() with clearOnLoadError true; now drives the same expectation
(empty list) through a phone sheet reopening, since the option is gone.
Switching the from currency of the desktop dashboard currency widget quickly
could show the rate of the previous currency: when the older rate request
answered after the newer one, its answer overwrote the newer rates.
Cause: useCurrencyConverter only aborted a request still in flight in its
phone mode (abortStale true); the desktop dashboard passed false and fetched
without any stale guard.
Fix: the hook now always aborts the previous request when a new one starts
and when the widget goes away, and ignores answers of an aborted request. The
abortStale option only existed to keep the desktop without the guard, so it
is removed and both dashboards call useCurrencyConverter().
Red test: FE-COMP-FXHOOK-008 (useCurrencyConverter.test.ts) failed on the
unfixed code with "AssertionError: expected 1.1 to be 1.3".
Changed assertions:
- useCurrencyConverter.test.ts FE-COMP-FXHOOK-002: was "the desktop mode
fetches without a signal, the phone mode with one" expecting the desktop
call to be [url] only; now expects [url, { signal: AbortSignal }].
- FE-COMP-FXHOOK-003 no longer names the phone mode; its assertions are
unchanged. All tests call the hook without the removed option.
With the badge and toast options gone from usePhotoProviderConnections,
the desktop and phone cards called the hook identically and then derived
the same seven per card values (fields, config, values, badge, testing,
whether save and test have a route) line for line, which lint:dup counts
as a copied block of 25 lines in MPhotoProvidersSection.tsx.
The hook now hands each card those values through providerCard, and both
views keep only their markup. Nothing the user sees changes.
New test FE-COMP-PHOTOCONN-011 covers providerCard. No existing assertion
changed.
In the admin's default user settings, the map template dropdown listed
seven tile presets, while every user's own map settings listed nine: the
same seven plus Amap and Amap Satellite. An admin running an instance for
people in mainland China could not pick the one basemap that works well
there as the default, short of pasting its URL by hand.
The Amap presets (57eac32f9) were added to the two user preset copies and
not to the admin's separate copy; nothing in that change or later gives a
reason, and the server stores any template as a default. The later
refactor kept the gap as DEFAULT_MAP_PRESETS beside MAP_PRESETS. There is
now one MAP_PRESETS list, read by the user settings and the admin defaults
on both desktop and phone.
Regression test FE-ADMIN-DUS-029 (DefaultUserSettingsTab.test.tsx) was red
on the old code: "Unable to find role="button" and name "高德地图 (Amap)"".
Changed assertions:
- useMapSettingsForm.test.ts FE-COMP-MAPFORM-001: it expected the admin list
DEFAULT_MAP_PRESETS to hold 7 presets and the user list to add Amap on
top; it now expects the one list to hold 9 presets including both Amap
templates.
On a managed install the phone's Integrations section still showed the
Immich and Synology Photos cards and, with the addon on, the AirTrail
connection form. Each asks for the address of a server the reader runs at
home, which a managed install cannot reach, so saving or testing could only
time out. The desktop has hidden both since the managed mode landed
(#1946, 3a0604b3f), whose message says these "connect to a server the
reader runs at home, and there is no route from here to one".
That change gated the phone's AI section but missed these two, and the
user facing addon list still carries a photo provider that was enabled
before the switch, so the cards did render. The phone now applies the same
!managed gate as the desktop IntegrationsTab.
Regression test FE-MOB-SETINT-008 (MSettingsIntegrations.test.tsx) was red
on the old code: "expected document not to contain element, found <div
data-testid="photo-providers"> instead".
No existing assertion changed.
On the desktop, the Immich (or Synology) card could show "Connected" for a
server that was down. The settings page loads two answers at once: the
saved settings, whose connected flag for Immich only says that a URL and an
API key are stored, and the status route, which actually reaches the
server. Whichever landed last set the badge.
usePhotoProviderConnections had a statusRouteOwnsBadge option that only the
phone turned on. The option is gone: whenever a provider declares a status
route, only that route sets the badge, on both views. A provider without a
status route still takes the badge from its settings read.
Regression test FE-COMP-PHOTOPROVIDERS-028 (PhotoProvidersSection.test.tsx)
was red on the old code: "Unable to find an element with the text: Not
connected" (the late settings read had turned the badge to Connected).
Changed assertions:
- usePhotoProviderConnections.test.ts FE-COMP-PHOTOCONN-005: it expected the
later settings read (connected: false) to override the status route's
connected: true; it now checks that a provider without a status route
takes the badge from the settings read.
- usePhotoProviderConnections.test.ts: render() no longer takes the option,
so every call drops its true or false argument.
On the desktop, a failed photo provider connection test toasted "Could not
connect to Immich : Auth failed" with a stray space before the colon, and
"Could not connect to Immich " with a trailing space when the server gave
no reason. The phone already read "Could not connect to Immich: Auth
failed".
usePhotoProviderConnections took a testErrorGap option that put a space
between the sentence and ": reason", and the desktop passed ' '. The option
is gone, so both views join the reason directly.
Regression test FE-COMP-PHOTOPROVIDERS-027 (PhotoProvidersSection.test.tsx)
was red on the old code: "Unable to find an element with the text: Could
not connect to Immich: Auth failed" (the toast read "Could not connect to
Immich : Auth failed").
Changed assertions:
- usePhotoProviderConnections.test.ts FE-COMP-PHOTOCONN-010: the desktop
toast was 'memories.connectionError(Immich) : bad key', now
'memories.connectionError(Immich): bad key'; it also checks that a
failure without a reason toasts the bare sentence.
When the notification preferences failed to load, the matrix on both the
desktop tab and the phone screen stayed on its loading line for good and
nothing told the user anything went wrong. A failed load of the webhook and
ntfy settings was just as silent and left empty fields that looked like
nothing was saved.
Both loads in useNotificationSettings ended in .catch(() => {}) and had no
cancelled flag. Each failure now shows the common error toast, a failed
preferences load sets loadFailed so both views replace the loading line
with the error text, and both loads ignore a response that arrives after
the screen is gone.
Regression test FE-COMP-NOTIFSETTINGS-010 (useNotificationSettings.test.ts)
was red on the old code: "expected "vi.fn()" to be called 2 times, but got
0 times". FE-COMP-NOTIFSETTINGS-011 covers the cancelled flag and
FE-COMP-NOTIFICATIONS-032 the desktop view.
Changed assertions:
- MSettingsNotifications.test.tsx FE-MOB-SETNOTIF-002: a failed load used
to expect the "Loading..." line to stay; it now expects "Error" and no
loading line.
On the phone, every notification toggle sent the whole preference matrix as
it looked when the switch was tapped. With two toggles in flight the later
request could overwrite the earlier one on the server, and when the first
request failed the screen rolled that switch back while the second request
had already saved its value anyway.
The cause was the sendWholeMatrix option of useNotificationSettings, which
only the phone turned on. The server merges whatever cells it receives, so
the desktop already sent just the toggled cell. The option is gone and both
views send only the cell that changed.
Regression test FE-MOB-SETNOTIF-035 (MSettingsNotifications.test.tsx) was
red on the old code: "expected [ { trip_invite: { inapp: false,
webhook: false } }, { trip_invite: { inapp: false, webhook: true } } ] to
deeply equal [ { trip_invite: { inapp: false } }, { trip_invite:
{ webhook: true } } ]".
Changed assertions:
- useNotificationSettings.test.ts FE-COMP-NOTIFSETTINGS-004: the phone
toggle body was { trip_invite: { inapp: false, email: false } }, now
{ trip_invite: { inapp: false } } (the hook is built with the phone's
remaining option, skipMaskedToken).
- MSettingsNotifications.test.tsx FE-MOB-SETNOTIF-026: the PUT body was
{ trip_invite: { inapp: false, webhook: false } }, now
{ trip_invite: { inapp: false } }.
- MSettingsNotifications.test.tsx FE-MOB-SETNOTIF-032: the mock told the
in-app write apart by body.trip_invite.webhook === false, which only the
whole matrix carried; it now checks for the inapp key.
On the phone storage panel, Test on a mirrored backend sent the mirror
itself to POST /api/admin/storage/test. The server expands a mirror by
looking its primary and replicas up by name in the SAVED storage
config, so the phone tested the saved settings instead of what the
operator had just edited, and a replica added but not saved yet made
the whole test fail with "references unknown backend". The desktop
probes the primary and each replica with their draft options and
merges the results.
Cause: the phone called admin.test with the mirror candidate, while the
desktop called admin.testMirror with mirrorProbeTargets. The choice
lived in each shell's markup.
Fix: useStoragePanel now offers loaded.testRow(row), which tests a
plain row as it is and a mirrored row through admin.testMirror with
mirrorProbeTargets from the draft. Both shells call it.
Red test: FE-MOB-MSTOR-018 (new, MAdminStoragePanel.test.tsx) failed
with "AssertionError: expected [ 'mirror' ] to deeply equal
[ 'backups-local', 'off-box' ]". FE-ADMIN-STOR-PANEL-017 (new,
useStoragePanel.test.ts) covers testRow itself.
No existing assertion changed.
With the appearance set to auto, switching the OS between light and
dark left the TREK wordmark in the phone admin addon header in the old
colour (dark text on a dark page or the reverse) until something else
made the screen render again. The desktop header switches right away.
Cause: the phone worked the flag out from the dark_mode setting and
window.matchMedia during render, and nothing re-renders the screen
when the media query flips. The desktop uses useIsDark, which watches
the .dark class that applyAppearance() writes.
Fix: MAdminAddonManager uses useIsDark like AddonManager.
Red test: FE-MOB-AADD-037 (new, MAdminAddonManager.test.tsx) failed
with "Expected the element to have attribute: src="/text-light.svg""
after the .dark class was added under auto.
Changed assertions in tests/unit/mobile/admin/MAdminAddonManager.test.tsx:
- FE-MOB-AADD-003 (renamed from "dark mode and auto+prefers-dark swap
the wordmark" to "the dark palette swaps the wordmark"): it expected
window.matchMedia to be called with "(prefers-color-scheme: dark)";
that expectation is gone, since the screen now reads the .dark class.
The test now sets that class instead of the dark_mode setting and a
matchMedia stub, and still expects src "/text-light.svg". afterEach
removes the class again.
On the phone admin addon list, the Collab and AI Parsing tiles showed
the generic puzzle piece, while the desktop shows the people and
sparkles icons for them.
Cause: the addons table stores 'Users' for collab and 'Sparkles' for
llm_parsing, and the phone's ICON_MAP had no entry for either, so both
fell back to Puzzle. The desktop map gained them earlier; the phone map
was never updated.
Fix: the phone ICON_MAP now carries Users and Sparkles, like the
desktop one.
Red test: FE-MOB-AADD-036 (new, MAdminAddonManager.test.tsx) failed
with "expect(received).toBeInTheDocument(): received value must be an
HTMLElement or an SVGElement" for svg.lucide-users.
No existing assertion changed.
On the phone plugins panel, "Update all" fired every pending update at
the same time. Only one row can show as busy, so the other rows stayed
clickable while their own install was running, and the list reloaded
once per plugin as each update finished. The desktop runs the updates
one after another, and the hook's own comment calls the parallel way
wrong.
Cause: usePluginsAdmin kept the phone's parallel loop behind the
updateAllInParallel option, which only the phone set.
Fix: the option is gone; updateAll always waits for each update before
starting the next, on both shells.
Red test: FE-ADMIN-PLUGINS-HOOK-009 (usePluginsAdmin.test.ts) failed
with "AssertionError: expected [ [ 'a', undefined ], …(1) ] to deeply
equal [ [ 'a', undefined ] ]".
Changed assertions in src/components/Admin/plugins/usePluginsAdmin.test.ts:
- FE-ADMIN-PLUGINS-HOOK-009 (renamed from "with updateAllInParallel
every update starts at once, without a version" to "Update all starts
only the first update, without a version"): pluginUpdate calls were
[['a', undefined], ['b', undefined]], now [['a', undefined]]. It and
the mounted() helper no longer pass the removed option.
- FE-ADMIN-PLUGINS-HOOK-008: only its name lost "by default".
On the phone storage panel, when the operator accepted several category
moves and the first migration failed to start (for example the server
was busy), only that error was toasted. The other accepted moves stayed
in a hidden queue with no hint, and started on their own later whenever
the panel state happened to refresh. The desktop drops the rest of the
queue, names it in a storage.migrate.queueDropped toast and refreshes
the state, so the operator sees what did not move and saves again.
Cause: useStoragePanel only dropped the queue when the
dropQueueOnFailure option was set, and only the desktop set it.
Fix: the option is gone and useStoragePanel always drops the rest of
the queue on a failed start, toasts it and refreshes, for both shells.
Red test: FE-ADMIN-STOR-PANEL-011 (useStoragePanel.test.ts) failed with
"AssertionError: expected "vi.fn()" to be called 2 times, but got 1
times".
Changed assertions in src/components/Admin/storage/useStoragePanel.test.ts:
- FE-ADMIN-STOR-PANEL-011 (renamed from "by default a failed migration
start only toasts and keeps the rest queued"): toast.error was called
once, now twice; migrationQueue had length 1, now 0; refreshState was
not called, now it is.
- FE-ADMIN-STOR-PANEL-012: no assertion changed; it no longer passes the
removed dropQueueOnFailure option.
On the desktop admin GitHub panel, when "Load more" failed (for example
GitHub rate limiting), the whole release history was replaced by the
error card for the rest of the session: no list, no Load more button,
no way to retry. Had the user found a way to load again, the next load
would also have skipped the page that failed.
Cause: in its one page per load mode useGithubReleases advanced the
page counter whether the request worked or not and never cleared the
error, and GitHubPanel showed the error card whenever an error was set.
The phone (fillPages mode) already held the page, cleared the error on
the next good load and drew the error inline under its timeline.
Fix: the one page mode now moves the page on only after a good load and
clears the error then, like fillPages. GitHubPanel shows the error card
only when nothing has loaded yet; after that it keeps the timeline,
shows the error as a line above Load more and leaves the button there.
The page walk difference behind fillPages stays as it was.
Red tests:
- FE-COMP-GHREL-HOOK-012 (new, useGithubReleases.test.ts) failed with
"AssertionError: expected [ 1, 2, 3 ] to deeply equal [ 1, 2, 2 ]".
- FE-ADMIN-GH-017 (new, GitHubPanel.test.tsx) failed with "Unable to
find an element with the text: /v1\.\d\.0/" after the failed load.
Changed assertions in src/components/Admin/useGithubReleases.test.ts:
- FE-COMP-GHREL-HOOK-004 (renamed from "sets the error and still moves
the page on" to "sets the error and holds the page"): requested pages
were [1, 2, 3], now [1, 2, 2].
On the phone, the admin GitHub panel showed "Latest updates from
mauriceboe/TREK" and its releases, bug report, feature request and wiki
links, as well as the same three links in the settings About section,
all went to the old mauriceboe/TREK repository. The desktop panels
already use liketrek/TREK.
Cause: the phone screens kept the repository name from before the move
to the liketrek organisation.
Fix: MAdminGitHubPanel and MSettingsAbout now use liketrek/TREK, like
GitHubPanel and AboutTab. No other user facing mauriceboe/TREK link
is left in the client (the Ko-fi, Buy Me a Coffee and Docker image
names are unchanged and identical on both sides).
Red test: FE-MOB-SET-ABOUT-001 (new, MSettingsAbout.test.tsx) failed
with "Expected the element to have attribute: href=
https://github.com/liketrek/TREK/issues/new?template=bug_report.yml".
Changed assertions in tests/unit/mobile/admin/MAdminGitHubPanel.test.tsx:
- FE-MOB-AGH-001: wiki href was https://github.com/mauriceboe/TREK/wiki,
now https://github.com/liketrek/TREK/wiki
- FE-MOB-AGH-004: subtitle was "Latest updates from mauriceboe/TREK",
now "Latest updates from liketrek/TREK"; GitHub link was
https://github.com/mauriceboe/TREK/releases, now
https://github.com/liketrek/TREK/releases
When the notes of one category carried two colours (a category recolour
that failed halfway, or a note written over the API), the desktop panel
painted the category in the colour of the last such note in the list
while the phone notes tab used the first one, so the same category had
different colours on the two devices.
The phone's buildCategoryColorMap took the first colour on the stated
assumption that every note in a category has the same colour, so first
seen equals all seen; it was meant to match the desktop and only
differed where that assumption breaks. The desktop rule (since v2.6.0)
is the original. buildCategoryColorMap now keeps the last colour seen
and the desktop panel builds its map from the same function, merged
over its empty categories as before.
Red before the fix: tests/unit/mobile/trip/collabModel.test.ts
FE-MOB-CLBM-023 failed with "expected { Ideas: '#ef4444' } to deeply
equal { Ideas: '#10b981' }".
Changed assertion in tests/unit/mobile/trip/collabModel.test.ts
FE-MOB-CLBM-011 (renamed from "keeps the first colour seen per
category" to "keeps the last colour seen per category"): the expected
map was { Ideas: '#ef4444' } and is now { Ideas: '#10b981' }.
Uploading photos into a journey gallery on desktop gave no sign that
anything was happening until the toast at the end; the Upload button
stayed clickable the whole time. Up to v4.0.0 the button sat inside the
gallery and showed a spinner with "Uploading done/total" while it was
disabled. When the button moved up into the page header (onRegisterUpload)
that indicator was dropped, and the count kept being tracked with no
reader.
useGalleryUpload now also returns the progress it tracks, the gallery
hands it up through onUploadProgress into useJourneyDetail, and the
page's Upload button is disabled and shows the spinner with
journey.editor.uploadingProgress while the files go, as it did before.
Red before the fix: src/pages/JourneyDetailPage.wiring.test.tsx
FE-JRN-DETWIRE-043 failed with "expected undefined to be [Function
Mock]": the gallery was never given an onUploadProgress callback and
the button never showed a count. FE-JRN-GALLERY-021 and
FE-JRN-GALLERYUPLOAD-007 cover the gallery and the hook. No existing
assertions changed.
On the desktop journey page, the eye toggle that hides the suggested
entries flipped the view and then saved the preference. When that save
was refused (offline, a server error) the promise rejected into a click
handler nothing awaited, leaving an unhandled rejection behind.
The phone already treats this as a view preference: it keeps the local
flip until the next load and swallows the failure on purpose. The
desktop toggle now does the same.
Red before the fix: src/pages/JourneyDetailPage.wiring.test.tsx
FE-JRN-DETWIRE-042 failed with "expected vi.fn() to not be called at
all, but actually been called 1 times" on the unhandledRejection
listener, called with [Error: offline]. No existing assertions changed.
Saving a journey entry on the phone sheet while the server refused it
(a locked journey, a dropped connection) left the sheet sitting there
with no message, and the rejection went unhandled. The desktop dialog
toasted journey.settings.saveFailed or the server's reason.
The cause was the toastSaveError option of the shared useJourneyEntryForm
hook: only the desktop set it, and without it handleSave rethrew into a
click handler that nothing awaited. The option only kept the phone on
the broken path, so it is gone and both views toast and keep the form
open.
Red before the fix: src/components/Journey/useJourneyEntryForm.test.ts
FE-JRN-ENTRYFORM-021 failed because handleSave rejected with
"{ response: { data: { error: 'Locked' } } }" instead of toasting.
Changed assertions in FE-JRN-ENTRYFORM-007 (renamed to "a failed save
toasts for the desktop dialog and the sheet alike"): the sheet's save
was expected to reject (rejected true) with one toast in total; it now
resolves (rejected false) and the toast count is 2.
On the phone, attaching a file to a note, detaching one or deleting a
note with attachments left the trip's Files tab on the old list until a
reload. The desktop panel fires the local collab-files-changed event
that useTripWebSocket listens for, because the server leaves the
writer's own socket out of its broadcast; the phone notes tab never
fired it.
The event now lives in the shared useCollabNotesData hook:
uploadNoteFiles announces it once after a batch, and
announceNoteFilesChanged is what both shells call after deleting a note
or a note file. The desktop no longer fires it by hand after an upload.
Red before the fix: tests/unit/mobile/trip/MCollabNotes.test.tsx
FE-MOB-CNOTE-034 failed with "expected 0 to be 1" on the event count
after detaching a file. FE-COLLAB-NOTESDATA-009 covers the hook.
No existing assertions changed.
haversineKm in src/utils/geo.ts answered NaN for some near antipodal pairs,
for example 47.3812, -10.8827 and -47.3812001, 169.1173. Rounding pushed the
haversine term a hair past 1, and asin of a value above 1 is NaN, which then
spread into route sums and range checks.
Cause: unlike the Studio legKm and the roadtrip haversineKm in
@trek/shared, the shared client helper had no clamp before Math.asin.
Fix: clamp with Math.min(1, ...) before Math.asin. With the clamp the helper
runs exactly the operations of Studio legKm in the same order (2 * 6371 is
12742), so Studio now uses haversineKm and legKm goes, with its cases carried
into the new geo tests. The roadtrip variant multiplies its terms in another
order and gives different last bits, so it stays as it is.
Red test: src/utils/geo.test.ts FE-UTIL-GEO-002 failed on the unfixed code
with "AssertionError: expected NaN to be close to 20015.086796020572,
received difference is NaN, but expected 0.0005".
Removed tests: src/components/Studio/legKm.test.ts went with legKm. Its four
cases (zero for one point, a degree of latitude about 111.19 km, the same
both ways between 8700 and 8900 km, 0,0 to 0,180 equal to 12742 * PI / 2) now
run against haversineKm as FE-UTIL-GEO-001, 003 and 004.
poolCounts in the phone places browser model has had no caller since the
phone pool chips stopped showing counts in 8855b5581 (the FilterChip note in
MPlacesBrowser says counts are left out on the phone to save row space). Only
its own tests still ran it, and it had drifted from the desktop counts in
usePlacesSidebar, which count planned places against the day scoped planned
ids. Nothing on the phone should use it, so it goes, with its two tests.
No user facing change, so there is no red test: the helper is dead code.
Removed tests in src/mobile/screens/trip/places/placesBrowserModel.test.ts:
MOBILE-POOL-counts (expected all 3, planned 2, unplanned 1) and
FE-MOB-PMODEL-013 (expected all 2, tracks 1, planned 1, unplanned 1).
On the phone place form, pasting a "lat, lng" pair into the latitude field
filled both coordinates, but a later search pick whose place came without a
position emptied them again whenever an earlier pick had filled them. On the
desktop the pasted position stays.
Cause: the phone's paste handler wrote lat and lng with setForm directly, so
autoFilledRef still listed them as written by the previous pick, and
mergeResult clears such fields when the new pick says nothing about them. The
desktop goes through handleChange, which marks a field as typed by hand.
Fix: the phone's paste handler goes through handleChange for both halves, the
same as the desktop dialog.
Red test: tests/unit/mobile/trip/MPlaceEditSheet.test.tsx FE-MOB-PLEDIT-047
failed on the unfixed code with "expect(element).toHaveValue(35.7101)
Expected the element to have value: 35.7101 Received: (empty)".
No existing assertions changed.
On the phone place form, tapping a search suggestion emptied the address,
the coordinates, the map ids and the website that the previous pick had filled,
before the suggestion's place was even looked up. When that lookup failed, the
form was left with the new name and no position at all. The desktop form only
puts the name in and leaves the rest alone.
Cause: PlPlaceSearch handed the suggestion's name to the sheet as a search
pick (onPick({ name })), and the sheet folds every pick through mergeResult,
which clears each field the previous pick wrote when the new pick says nothing
about it.
Fix: PlPlaceSearch takes an onSuggestionName callback, the same one the
shared usePlaceSearch hook already offers, and MPlaceEditSheet only sets the
name with it, as the desktop dialog does. The resolved place still arrives
as a normal pick.
Red test: tests/unit/mobile/trip/MPlaceEditSheet.test.tsx FE-MOB-PLEDIT-046
failed on the unfixed code with "expect(element).toHaveValue(Taito)
Expected the element to have value: Taito Received: (empty)".
Changed assertions in tests/unit/mobile/trip/PlPlaceSearch.test.tsx:
FE-MOB-PLSRCH-004 expected onPick to be called first with { name: 'Louvre' }
and twice in all; it now expects onSuggestionName to be called with 'Louvre'
and onPick once. FE-MOB-PLSRCH-005, FE-MOB-PLSRCH-005b and FE-MOB-PLSRCH-024
waited for onPick to be called twice; they now wait for one call.
On the phone, a ticket split where a receipt line named someone who was not
ticked as a participant saved only the ticked people as members. Their share of
the line was dropped, the member amounts no longer added up to total_price, and
the settlement kept a difference nobody could clear. The desktop editor had
been fixed for this in #1382.
Cause: useExpenseForm only added the people named on ticket lines when the
caller passed ticketShareMembers, and only the desktop CostsPanel did. The
phone MCostSheet kept the old behaviour.
Fix: the hook always sends everyone a ticket line names, and the
ticketShareMembers option is gone, so both editors save the same members.
Red test: src/components/Budget/useExpenseForm.test.ts
FE-BUDGET-EXPFORM-021 failed on the unfixed code with
"AssertionError: expected { name: 'Lunch', ... } to match object { ... }":
member_ids was [1] and members held only user 1, without user 2's 6.
Changed assertions: useExpenseForm.test.ts FE-BUDGET-EXPFORM-017 used to
expect the phone run (ticketShareMembers false) to send
members [{ user_id: 1, amount: 6 }] and member_ids [1]; that phone half is
removed and the test now expects every editor to send members for users 1 and
2 with 6 each and member_ids [1, 2].
In the plan view the tabs that collapse and reopen the left and right
panels gave a screen reader no open or closed state and showed no focus
ring when reached with the keyboard. The same tabs in the tour planner
had both, because only there LeftPanelTab got the "announced" option,
and the plan view's right tab was a hand copy of the left one.
LeftPanelTab becomes PanelTab with a side prop and always sets
aria-expanded and the focus ring. The plan view's right tab renders it
mirrored, so every panel tab shares one component. The offsets are
written with literal left and right keys so lint:rtl still counts them.
The page shrinks by 11 lines.
Red test: FE-PAGE-PLANNER-016b in client/src/pages/TripPlannerPage.test.tsx
failed on the old page with
expect(element).toHaveAttribute("aria-expanded", "true").
FE-PAGE-PLANNER-LEFTTAB-005 covers the mirrored right tab.
Changed assertions (they asserted the missing state):
- TripPlannerPage.leftPanelTab.test.tsx FE-PAGE-PLANNER-LEFTTAB-003:
not.toHaveAttribute('aria-expanded') and not.toHaveClass
('focus-visible:outline') are now toHaveAttribute('aria-expanded',
'true') and toHaveClass('focus-visible:outline') (title updated).
- TripPlannerPage.leftPanelTab.test.tsx FE-PAGE-PLANNER-LEFTTAB-004: the
"announced" prop is dropped from both renders (title updated); the
expectations are unchanged. The file renders PanelTab instead of
LeftPanelTab.
Dropping a booking on the body of another day moved it but left the
shared drag data (window.__dragData) of an earlier place or map drag in
place, while dropping it on the end zone of the day cleared both. A later
drop with nothing of its own could then read that leftover place id.
The cross day booking branch of the day body drop now clears
window.__dragData along with dragDataRef, the same as the end zone. The
two clears are one chained assignment, so DayPlanSidebar.tsx stays at
the line count lint:size holds it to.
Red test: FE-PLANNER-DAYPLAN-236 in
client/src/components/Planner/DayPlanSidebar.dragData.test.tsx failed
with "expected { placeId: '77' } to be null".
No existing assertions changed.
The desktop road trip stop popup read its kind, length, night and
check-in answers only when it mounted. When the planner handed the open
popup another stop draft, the popup showed the new stop's name with the
previous stop's kind, length, night switch and check-in, and saving wrote
those stale answers onto the new stop.
The phone sheet already seeds the shared useStopDraftForm again for
every draft. The popup now does the same in an effect keyed on the draft,
and seeds the night switch and the check-in time from the same draft.
Red test: FE-ROADTRIP-STOPPOPUP-034 in
client/src/components/Roadtrip/RoadtripStopPopup.test.tsx failed with
expect(element).toHaveAttribute("aria-pressed", "true") on the Charging
chip after the draft was swapped.
No existing assertions changed.
The transport form stays mounted between openings. Editing a flight or a
train after a car kept the car's stops (desktop and phone) and, on the
desktop, the from and to picks of that earlier booking as well. Switching
the type to car in the same dialog showed those stale stops and places,
and saving wrote them onto the booking. The desktop also kept stale
flight and train rows when opening a car, bus or other booking.
seedTransportRoute only reset the rows a type does not show when the
phone asked for it (resetHiddenRoutes), and never reset the car stops for
a flight or a train. It now returns every row for every type, so a seed
always replaces the whole route. The option is removed; both views get
the same clean route.
Red test: FE-PLANNER-TRANSPORTFORM-008 in
client/src/components/Planner/useTransportForm.test.ts failed with
"expected [ { location: { ... }, time: '' } ] to deeply equal []".
Changed assertions (they asserted the stale rows were kept):
- useTransportForm.test.ts FE-PLANNER-TRANSPORTFORM-005: desktop
fromPick toBe(picked) is now toEqual({}) (title updated), and the
resetHiddenRoutes field is dropped from seedOptions and the phone call.
- transportEndpoints.test.ts FE-PLANNER-TRANSPORTEP-005: desktop route
keys toEqual(['waypoints']) is now empty fromPick, toPick, carStops and
trainWaypoints.
- transportEndpoints.test.ts FE-PLANNER-TRANSPORTEP-010: desktop car
waypoints toBeUndefined() is now two empty rows (plus empty train rows),
and the bus route now also carries the two empty row sets.
- transportEndpoints.test.ts: resetHiddenRoutes dropped from DESKTOP and
PHONE.
In the desktop booking and transport dialogs, clicking "Create expense"
(or editing a linked expense) before the booking had a title saved
nothing, which is right, but the wish was kept. The next ordinary save
then opened the expense editor out of nowhere. The same happened after a
save that threw.
useBookingExpenseIntent stored the wish in a ref and only cleared it when
the form read it, which the desktop forms do after a successful save. The
hook now clears the wish once the save its click started has settled,
unless a newer click replaced it in the meantime. The phone sheets read
it at the start of the save and are unaffected.
Red tests:
- FE-PLANNER-RESMODAL-107 in ReservationModal.test.tsx: "expected
"vi.fn()" to not be called at all, but actually been called 1 times".
- FE-PLANNER-BOOKINGEXPENSE-005 in useBookingExpenseIntent.test.ts:
"expected { create: true } to be null".
- FE-PLANNER-BOOKINGEXPENSE-006: "expected { editItem: { id: 9, ... } }
to be null".
No existing assertions changed.
Saving an imported booking in the desktop booking dialog when the dialog
had no upload handler but the import carried source files rejected the
save: the files were handed to the missing handler, the promise rejected
with nothing to catch it, and the expense editor the user asked for never
opened. The transport dialog already skipped the upload in that case, the
booking dialog did not.
The shared uploadBookingFiles helper now takes an optional handler and
does nothing without one, so every form gets the same guard.
Red test: FE-PLANNER-RESMODAL-106 in
client/src/components/Planner/ReservationModal.test.tsx failed with
"expected spy to be called with arguments" for onOpenExpense, plus the
unhandled rejection "TypeError: upload is not a function". The helper
case is covered by FE-PLANNER-BOOKINGFORM-008.
No existing assertions changed.
Editing a transport without metadata of its own (bus, car, taxi, ferry
without a transit itinerary) in the phone sheet wiped the expense price
shown on the booking card. The phone built the payload with
emptyMetadataAsNull and sent metadata: null, which the reservations
service reads as "clear the column", so keepMirroredPrice had nothing to
carry price and priceCurrency into. The desktop and both booking forms
send an empty object, which keeps the price.
The option only existed to keep the phone's null, so it is gone and both
views now send the rebuilt metadata object as is.
Red test: FE-MOB-TRFRM-060 in
client/tests/unit/mobile/trip/MTransportFormSheet.test.tsx failed with
"AssertionError: expected null to deeply equal {}".
RESV-SVC-043 in server/tests/unit/nest/reservations.service.test.ts shows
the server side: {} keeps the stored price, null drops it.
Changed assertions (they asserted the null):
- MTransportFormSheet.test.tsx FE-MOB-TRFRM-006: payload metadata null
is now {}.
- MTransportFormSheet.test.tsx FE-MOB-TRFRM-011: metadata toBeNull is now
toEqual({}).
- MTransportFormSheet.test.tsx FE-MOB-TRFRM-027: metadata toBeNull is now
toEqual({}).
- transportEndpoints.test.ts FE-PLANNER-TRANSPORTEP-014: phone metadata
toBeNull is now toEqual({}) (title updated to match), and the
emptyMetadataAsNull field is dropped from DESKTOP_SAVE and PHONE_SAVE.
- useTransportForm.test.ts FE-PLANNER-TRANSPORTFORM-007: metadata: null
is now metadata: {}.
The copies inside the map renderers, the location inputs, the day plan
sidebar, the trip planner page, the time picker, studio, the store
slices and a few small components are gone, and so is the markup the
two phone booking sheets, the two phone booking tabs, the desktop
booking dialogs, the vacay sheets and the storage panel drew twice.
lint:dup now holds every file at zero, and lint:size, lint:rtl and the
eslint warning counts go down with it. Nothing goes up.
The invite and the share sheet each carried the same unfolded user list. Both now render MUserPickerList with the users, the picked id and the pick handler from useVacayUserPicker. The markup is unchanged.
The phone storage panel rendered the same test result markup for regular and degenerate mirror backends. Both rows now render MStorageTestResult, with the same elements, classes and text as before.
The desktop and the phone addon manager each carried their own copy of the Immich and Synology SVGs. Both now import them from components/Admin/addons/PhotoProviderIcons, which has its own test. The rendered marks are unchanged.
The transports and bookings tabs drew the same title button, edit and
delete buttons, date and time row with the booking code, file chips and
delete confirmation sheet as two copies. Both cards now render them from
MReservationCardParts, and each tab still decides what a tap opens. The
rendered markup stays the same for both tabs.
The desktop booking and transport dialogs ended on the same notes field,
link and files row and linked costs block. Both now render
BookingNotesFilesCosts next to them, with the same markup as before.
The phone booking and transport sheets each carried their own copy of the
booking code and status row, the notes field, the traveler picker with its
avatar, and the transport sheet drew the card of a route stop twice, once
for flights and once for trains and cruises. They now render the same
components from MBookingFields next to the sheets, which take the sheet's
own translate function so every string stays what it was. The rendered
markup is unchanged.
ReservationPluginSlots in tabChrome repeated the plugin store selector
that useReservationDetailPlugins already holds for the desktop panels.
It now calls the hook, which gets its own test.
useMRoadtrip and useMRtCorridor each kept a nowMinutes helper and their
own check whether the selected day is today. Both now use localMinutes
and isLocalToday from utils/localDate.ts, which hold the same rules.
Both folds on the settings page opened with the same summary line and
turning chevron. FoldSummary now draws it and each fold passes its own
label; the Google fold keeps its status pill.
ApplyTemplateButton and the packing list header drew the same template
row in their menus. PackingTemplateOption now draws it; the header's
faded hover stays a parameter (fade).
Both previews built the same dialog head band with the file name, the
open in tab pill and the download pill. FilePreviewDialog now draws it;
the dialog width, the body class and the icon colour stay parameters,
and each preview keeps its own body.
The optimistic category reorder in budgetSlice and the budget:reordered
handler for another member's reorder grouped the items the same way in
two copies. Both now call orderBudgetByCategories from budgetOrder.ts,
so the local and the remote reorder cannot drift apart.
The country rows and the geocoded place rows repeated the same button,
hover tint and trailing chevron. A ResultRow inside AtlasCountrySearch
now draws that frame and each list passes its own leading content.
TravelElements ramped how far a leg bows by a local km helper and
roadRoute refused long legs by a second copy of it. Both now call legKm
from legKm.ts, the same formula with the same clamp.
The text inspector and the travel inspector each drew the same row of
typeface buttons and the same weight choice. Both now render FontButtons
and WeightLine from StudioTypeControls.tsx. The travel inspector keeps
casting its patches to Partial<BookElement>, so each side writes exactly
the fields it wrote before.
The three columns of the time picker each repeated the same chevron up,
value box and chevron down markup. TimeSpinnerColumn draws one, and the
button style it shares with the clear button moves to timeSpinnerStyle.
The AM/PM column keeps its narrower box, smaller type and leading gap as
props, and the hour and minute boxes keep their tabular digits.
The day body and the drop zone at the end of the list both carried their
own copy of two steps: moving a stop or note dragged in from another day,
and putting a same-day row after the day's last row. Both now live in the
plan hook as moveInFromOtherDay and dropAfterLastItem.
The differences stay where they were: only the end zone passes a
reservation id, so only it moves a booking to the end, and only it clears
the drag state after the move, which it still skips on an empty day.
useDayPlanSidebar returned about 145 names, and DayPlanSidebar destructured
the same list again, the largest copied block in the client. Most of it
was the component's own props passed straight back, plus values nothing
read any more.
The optional props' fallbacks move into withDayPlanDefaults, which the
hook and the view both read the props through, so they are written once.
The hook now returns only what it computes. State that only the view
reads (the context menu, the permission flags, the hover and expansion
states, the pending note deletion) is declared in the view, and the day
note handlers come back as one notes object. The values no one read are
gone, which also drops their unused variable warnings.
No behaviour changes: every value keeps its source and its fallback.
The plan view and the tour planner view each drew the left panel's tab in
full, and the place inspector and the tour dialog each listed the same props
twice, once beside the panels and once in the phone-width sheet.
LeftPanelTab, a small presentational component at the bottom of
TripPlannerPage.tsx, now draws the tab for both views. It stays in the page
file so the physical left and right offsets keep being counted there. The
tour planner's tab passes `announced`, which keeps its aria-expanded and its
focus ring; the plan view's tab still has neither.
The page body now builds selectedPlaceProps (what the inspector and the tour
dialog both take), placeInspectorProps and tourDetailProps once and spreads
them. What differs stays at each call site: the desktop inspector's
onOpenBooking and panel widths, the sheet's edit and delete closing the
sheet, the desktop tour dialog's non-modal focus return and widths, and
canEdit and canAssign on every tour dialog.
AirportSelect, LocationSelect and AddressInput each carried their own copy
of the open state, the keyboard highlight, the outside click that closes the
list, the floating list with its loading line and the suggestion row.
useSuggestionDropdown now holds the open state, the highlight, the outside
click and the arrow, Enter and Escape handling. SuggestionDropdown.tsx holds
the boxed input with its icon and clear button (airport and location), the
floating list (all three), the row button and the place row content
(location and address).
What still differs between the fields stays with them or is a parameter:
the icon, the placeholder, what typing does (LocationSelect also resets the
highlight), the rows the keys walk (LocationSelect walks the trip's places
below three characters), whether the loading line may show (not while
LocationSelect offers the trip's places) and how a row lines up its content
(centred for airports). Each field keeps its own query, search and pick.
MapView and MapViewGL each held the same photo loading effect: read the
cached thumbs, fetch the missing ones, listen for their arrival and batch
them into one animation frame. It now lives in useMarkerThumbs, which
also owns the thumb state and the place photos switch, and both maps
call it where the effect used to be.
The effect keeps its keys (the picture sources and the switch). It reads
the places through useEffectEvent, which drops the old
exhaustive-deps suppression in MapViewGL and the warning in MapView.
ReservationOverlay and reservationsMapbox each carried the booking type
list, the icon and arc table, the time zone aware duration and the loop
that turns a booking into waypoints, arcs and labels. All of that now
lives in transportHops.ts; each renderer keeps only its drawing.
The one difference stays a parameter: Leaflet asks for arcs that cross
the antimeridian to be copied 360 degrees over, the GL map does not
because it repeats features across world copies itself. Leaflet still
adds its transit segments on top of the shared hop.
calculateSegments and the cached legs call turned each OSRM leg into the
same connector record line by line. That record now comes from
osrmLegSegment; the legs call keeps adding its durationText on top, so
both answers stay exactly as they were.
Most features have a desktop view and a phone view, and until now each
carried its own copy of the feature's logic, so a fix on one side drifted
away from the other. With the logic moved into one shared hook per
feature, scripts/feature-pairs.json names every feature with its hooks
and the desktop and phone files that run on them.
lint:pairs fails when a listed view stops importing one of its pair's
hooks (a type only import does not count), when a listed file is gone,
when a hook sits in a view layer rather than under components, hooks,
utils or the folder of a page under pages/, and when a file under
src/mobile/screens is in no pair and not under mobileOnly with the
reason it has no twin. A new phone screen is so either wired to the
desktop feature's hook or marked as phone only on purpose. lint:dup
catches the copies that look alike; this one catches the copies written
anew.
The manifest starts with 120 features and 62 phone only screen files,
each with the reason it stays on the phone alone.
client/CLAUDE.md says how to change a feature, add one with both views
and add a phone screen.
Most features have a desktop view and a phone view, and each carried its
own copy of the feature's logic, so SonarCloud's duplication condition
was the gate a pull request touching both failed most often, and only
after a scan of twenty minutes. lint:dup finds the same copies before
the push.
It runs jscpd's detector (@jscpd/core with @jscpd/tokenizer) over the
source files under src/ at the thresholds the SonarCloud gate applies to
TypeScript, 100 tokens over at least 10 lines, and counts per file the
lines inside a block that repeats elsewhere or in the same file. Tests
and declaration files are left out as Sonar leaves them out, and .ts and
.tsx are read as one language, so a hook that repeats a view's code is a
copy too. The files are read sorted and with readText, so every machine
counts the same. scripts/dup-baseline.json holds the files that still
hold copies with their count; a file may hold no more than its entry, a
file without an entry none, and an entry above its file or for a file
that is gone fails until --update lowers it. --list prints every block
with its other side.
The engine packages are used instead of the jscpd command, which walks
the tree in file system order, skips files past its size limits without
a word and drops every result once it meets a file it has no format for.
The size, layers, format and eslint baselines only go down: files that
shrank, moved or lost a copied warning leave or lower their entries.
The offline baseline drops from 212 to 170 entries. 102 desktop and phone
views no longer import src/api/ and leave the list. 60 shared hooks take
their place, because the calls those views made moved into the one hook
both of them now import. Each of the 60 is imported only by views that
were on the list before, and every api module it imports was imported by
those views before. No call is new, and none moved behind the repo or the
mutation queue, so the offline behaviour is the same as before.
useSaveToCollection holds the store driven picker that
SaveToCollectionModal and MSaveToCollectionSheet each carried: loading
the lists and the membership for a new target, toggling the place in or
out of a list, the per list status, marking it visited everywhere and
opening the collections page. Both views keep their own markup.
The two copies did the same thing, so no option was needed. The load
effect now lists the membership query beside the target instead of
silencing the exhaustive deps rule; the query is derived from the
target alone, so it runs exactly as before.
Whether an invite is expired, used up or active, and which status label it shows, was worked out in both user views. It now lives in inviteStatus in the admin model; each view keeps its own row markup.
The All, Unplanned and Planned filter of the tours list was copied into the phone selection list. It now lives in filterTours next to the other tour presentation helpers, and both lists call it.
Both splashes carried their own copy of the four beats, the step time, the still beat and the interval that cycles them. These now live in useTripSplashBeats; each splash keeps its markup and styling. The phone follows the reduced motion setting live while the splash is up, the desktop reads it on each render; that difference stays as the liveReducedMotion option.
The phone trip sheet host wrote out by hand three pieces of planner
wiring the desktop planner already had.
useBookingExpenseEditor now holds the expense editor both planners open
over the trip: the booking request state, the receipt fallback, the base
currency, the current user and the save that reloads the budget.
TripPlannerPage and MTripSheets both read it.
applyTripCoverUpdate patches a new trip cover into the store for the
trip form on both sides.
The transport editor and the transit route change moved into
openTransportEditorWith and changeTransitRouteWith, which take the
planner's setters. usePlannerDialogs builds openTransportEditor and
changeTransitRoute on them, and the phone sheet calls the same helpers
on the planner it receives. There are no differences between the two
views to keep as options.
fileListRules now holds the type category of a file, the filter tab match
and the list of places and bookings a file is linked to (own column plus
extra links, transport or booking). useFileManager and FileManagerRow on
desktop and filesModel on the phone run on it. fileActions gains the field
save with its reload and assign error, which the desktop assign dialog and
the phone note edit both use, and the trash call with its toasts, used by
the desktop delete and the phone menu sheet. The phone sheet no longer
imports filesApi.
Differences that stay, each as an explicit option:
PHONE_FILE_FILTER_RULES sorts pdf, image and doc by type category (videos
count as images, doc is every file that is no PDF and no media, a pass
with a PDF mime type is a doc) and counts a note id of 0 as collab, while
desktop reads the mime type (images only, doc is Word, Excel and text).
DESKTOP_FILE_LINK_LIST_RULES skips an own place or booking column of 0,
the phone only a null one. PHONE_FILE_UPDATE_RULES waits for the reload
after a save, so a failed reload shows the assign error there; desktop
starts it and moves on. DESKTOP_FILE_TRASH_TOAST_RULES falls back to
English when the trash toast translation is empty.
The holiday countries, the calendars split by type, the weekend day toggle, dissolving a fusion, the leave year save and window label, the region options and the add check now live in useVacaySettings. The phone sheet still loads countries only while open.
The read only share reuses useVacayUserPicker for picking and sending, and useVacayShareActions holds the hide toggle and the removal with their error toasts for the sidebar card, the phone sheet and the phone chips.
Loading the users that can be invited, the picked user and sending the invite with its toasts now live in useVacayUserPicker. The dialog still resets and closes on success, the sheet still closes and resets on the next open and empties the list when loading fails.
The log mode and its half and comp modifiers, the entry, shared and company holiday maps, the trip dots, the day click rules, the default person and the day count format lived in both the desktop calendar and the phone screen. They now live in useVacayCalendarLogic and vacayFormat; the phone keeps reloading trip dots only on a new window shape.
The currency converter, the world clocks, the collection lists, the plugin widgets and card badges and the card date format each existed once per dashboard. They now live in useCurrencyConverter, useDashboardTimezones, useCollectionLists, useDashboardPlugins and dashboardModel; the phone keeps aborting a pending rate request through an option.
The phone sheet added the old length with a local setDate on a UTC midnight date, so in zones east of UTC a range crossing the start of summer time came out one day short. It now counts UTC calendar days like the desktop dialog.
The dialog and the phone sheet each carried the cover upload, the Unsplash search, the staged cover for a new trip and the title and range checks. They now live in useTripCoverPicker and tripFormModel; the blob url release keeps each view's timing behind an option.
The centre button resolved its target in two copies of the same hook. useCreateAction now holds it once, with the phone only branches (gallery upload, atlas search, collections) behind a flag.
Both presenters ran their own queue behind the window toast bus. The queue, the auto dismiss and the exit delay now live in useToastQueue; each presenter keeps its markup, its exit timing and its unmount handover.
useJourneyCarouselSync keeps the map on the card nearest the centre once
scrolling settles and brings a card to the centre for a marker, a day or a
tap, for both the phone journey screen and the narrow timeline of the desktop
and public pages. useGalleryUpload normalises and uploads gallery files for the
desktop gallery and the phone screen. The phone still moves the map on a tap
of another card, the desktop gallery still counts the files up while they go.
useJourneyEntryForm holds the fields, the photo queue that only goes out on
save, the gallery links, the location search and "use my location", the
verdict rows, quick capture and the save for both the editor dialog and the
entry sheet. journeyWeatherCategory moves to components/Journey/journeyWeather
so the dialog no longer reaches into mobile/. Where the two differed they still
do: the dialog starts with a blank verdict row, counts a moved point as a
change, toasts a failed save or location and fills the weather from the
forecast; the sheet edits tags and shows a refused location under the field.
useJourneySettings holds the name, cover, archive, the switches saved on the
spot, delete, unlinking a trip and removing a contributor for both the settings
dialog and the settings sheet. useJourneyTripLinking loads the linkable trips
and links one for the add trip dialog and the sheet, and useJourneyShareLink
manages the public share link for the share section and the sheet. The dialog
still closes after a new cover or an unlinked trip, the sheet still stays open.
useAtlasCountryActions marks a country or a region, unmarks a region, adds the
country or region to the bucket list and takes the country off it again for
both the desktop dialog and the phone sheet. The desktop dialog moves into its
own component in AtlasPage so the page body keeps calling useAtlas only. The
phone sheet keeps storing a marked region as visited and its own month input;
the desktop keeps its month and year picker in useAtlas.
useCollectionPlaceForm reseeds the edit form when a different place opens,
fetches a cover photo for a place without one, uploads and removes the custom
cover and saves the edits for both the detail sheet and the phone sheet. The
detail keeps editing the coordinates and keeps a late photo across an update
of the same place; the sheet holds the last place through its exit and drops
a late photo once the place object changes, as before.
useListEditor seeds the form from the target, handles the cover upload and the
Unsplash search, cleans the links and creates or updates the list for both the
dialog and the sheet. The sheet keeps holding the last list through its exit
and still hands over a list whose cover upload failed when it closes.
useTripCopyPicker loads the trips on every opening, filters them by the search,
formats the date range and turns the copy result into toasts for both the
dialog and the sheet. The dialog keeps its guard against an empty selection.
useCollectionSharing holds the invitable users lookup, the roster order, invite,
role, cancel, remove and leave for both the share dialog and the share sheet.
The dialog keeps one busy member per kind of row action, the sheet one for all.
Answering a question once, following a link after marking it read and the compact timestamp live in useNotificationItemActions. The desktop item keeps closing its panel after a navigation through the optional onClose.
The live preview, the debounced save with its flush on leave, the colour mode save and the widget toggles live in one useAppearanceEditor hook. The contrast helpers and widget groups move to appearanceModel.
Loading, saving and resetting the defaults for users and the map field handling live in one useDefaultUserSettings hook that reads the presets and style rules from mapSettingsModel. Each shell passes its own error text and preview stamp.
The tile presets, the provider and style rules and the preview place move to mapSettingsModel, the form state and save to one useMapSettingsForm hook. The phone keeps showing the normalized style before the save lands, through an option.
The preference matrix, the webhook and ntfy credentials and the channel test sends now live in one useNotificationSettings hook. The phone keeps sending the whole matrix on a toggle and skipping a token typed as the mask, through two options.
Profile and avatar, the password change, two factor setup with its backup codes and the account deletion now live in one useAccountSettings hook. Two existing differences stay as options: the toast key for a failed avatar removal and whether the backup code actions skip an empty code list.
Listing, adding with the step up and the WebAuthn ceremony, renaming and
deleting passkeys now live in one usePasskeys hook together with the
date and abort helpers. The desktop section and the phone card keep
their own markup.
The addon flags, the managed install flag and the addon reload of the integrations tab now come from one useIntegrationGates hook, and useSettings hands both shells the plugins tab flag next to the integrations one.
The OAuth presets, the endpoint snippets, the OAuth clients, sessions and API tokens with their handlers and the copy tick now live in one useMcpIntegration hook. Two existing differences stay as options: the desktop loads only while the addon is on, and the phone reads the client list back after a create or rotate where the desktop appends locally.
The declared field form with its save, the actions and their confirm state, the OAuth connect and disconnect and the activity log now live in usePluginUserSettings, usePluginOAuth and usePluginActivity. Each shell keeps its own form, confirm dialog or sheet and picker.
The provider field model, the seeding and hydration, save and test now live in one usePhotoProviderConnections hook. Two existing differences stay as options: on the phone the status route alone sets the badge, and the desktop keeps its space before the server reason in a failed test toast.
The grouped scopes, the folded groups and the select all, group and single scope toggles now live in one useScopeSelection hook. Both pickers keep their own rows.
The provider, model, key and multimodal state, the hydration from the settings store and the save with its reload on failure now live in one useLlmConnection hook. The phone keeps its picker sheet state.
Both settings shells now read their state, the save and test handlers and the canSave rule from one useAirTrailSettings hook and keep only their own markup.
useAdminUserActions now rotates the JWT secret and resets a user's passkeys for the desktop dialogs and the phone sheets. Each shell keeps its own confirm step, and the phone sheet still stays busy until the reset settles.
useAdminSettingsActions now holds the OIDC and file type saves and the five place switches, writing through the admin bag both shells already receive. trekApiModel keeps the TREK Places API fields, facts and sources the desktop card and the phone block list.
Saving the demo baseline and the bag tracking and collab feature switches now live in useAdmin. AdminPage and the phone admin screen pass them on instead of carrying their own inline copies.
useAddonManager now loads and toggles the addons and builds the provider shelves, useLlmParsingConfig holds the AI parsing form with the Ollama model list and pull, and addonModel keeps the shared types, collab sub features, labels and recommended models. Both shells keep their own markup.
The phone AI parsing form saved the base URL left over from a Local or OpenAI setup after switching to Anthropic, so Anthropic requests went to the old host. It now saves an empty base URL for Anthropic, as the desktop form does since #1777.
The phone rolled a failed photo provider toggle back by restoring the whole addon list it saw at the click, which also undid any switch flipped while the request was in flight. Photo providers now go through the same toggle as every other addon, with a per row rollback, as on desktop.
useAdminNotificationSettings now holds the SMTP save and test, the trip reminder switch and the admin webhook and ntfy save, test and token clear. The desktop tab drops its inline handlers and both views keep their markup.
useAdminNotificationMatrix now loads the matrix, picks the visible built in channels and saves each toggle with a per cell revert. The desktop panel and the phone card keep their own markup.
When a toggle on the phone matrix failed after a later toggle had already rendered, its catch put back the preferences from its own click and so also undid the later cell. The live preferences are now mirrored in a ref and a failure reverts only its own cell, as the desktop panel already does. The test hangs the first save, lets the second one through and then fails the first.
useBackupAdmin now loads the backups and the auto backup schedule and runs create, download, delete, restore and the schedule save. The phone keeps its form when the server answers without settings through the keepSettingsWhenMissing option.
usePackingTemplates now owns the template list, the expanded template with its categories and items, and every create, rename and delete. The phone keeps its generic delete error toasts through the genericDeleteErrors option.
useMcpTokensAdmin now loads both lists, revokes sessions, deletes tokens and keeps the expanded scope chips. Both panels keep their own cards and confirm dialogs.
useCategoryAdmin now holds the category list, the create and edit form, the colour presets and the delete. The phone sheet still shows a busy delete and closes itself on success through the trackDelete option.
Loading, editing, resetting and saving the permission levels now live in one usePermissionsAdmin hook next to the category and level tables. Both panels keep their own markup.
Both audit log panels now read one useAuditLog hook for the first page, refresh, load more and the row labels. The desktop table and the phone cards keep their own markup.
Both panels carried the same types, helpers and a four hundred line block of state and handlers. pluginModel now holds the types and pure helpers, usePluginsAdmin the state, loading and every action, and usePluginDetail the registry detail fetch. The phone keeps starting all updates at once from Update all and reads the detail manifest under its stricter shape; both panels keep their own markup, dialogs and sheets.
useStoragePanel holds the editor, removal and sync prompts, the save with its migrate prompt and the sequential migration queue; useBackendForm holds the backend draft, its checks and the backend it builds. The desktop keeps dropping the rest of the queue when a migration fails to start, and keeps recomputing the candidates at confirm time; the phone keeps moving the candidates its prompt opened with.
When five full pages in a row held only prereleases, the walk stopped with its page counter already one past the last page it read, so the next Load more started a page later and that page of releases never showed. The counter now ends on the last page read.
useGithubReleases holds the paged fetch, the prerelease filter, the expanded rows and the date format, and parseReleaseNotes splits a release body into the blocks both panels draw. The phone keeps its page walk past hidden prereleases through the fillPages option, its repository link and its own inline link scan; the desktop keeps one page per load.
Both panels loaded the same trips and users, built the same twelve test payloads and sent them the same way. That now lives in useDevNotifications; each panel keeps its own markup and the phone keeps its own wording of the failure toast.
The pure model imported its direction type from the flow view. The type
now lives in the model, which takes it as the direction parameter, and
the flow view re-exports it for its existing importers.
The model left the format baseline when it moved into components/Collab,
so it is now Prettier formatted. Its header no longer points at the
deleted services layer, and the colour palette no longer cites a rule
that stopped applying once the file sat beside the desktop panel. The
chat data hook says why it still returns setters nobody writes.
The shared toggle had mixed the two views. Desktop again treats any empty
column as free and finds a link record by comparing ids as they come. The
phone link sheet again treats only a null column as free and compares the
numeric value of a record id.
Starring stays per view. It is a single filesApi call with different
error handling and refresh on each side, so a shared wrapper would add a
layer without removing any logic.
The shared notes hook skipped loading and listening without a trip id,
which only the desktop panel did. That guard is now a desktop option, so
the phone tab loads and listens as before.
The shared polls hook had taken the desktop rules for both views. It now
keeps each view's logic as it was before the move. The phone again
replaces a poll with the one a live closed event carries, puts a vote
answer on the poll it voted on, and stops the countdown tick once a
deadline passed. The desktop panel again ticks until a poll is closed.
useScopePicker holds the provider's containers, binding one, and making
and binding a new one under the suggested name; the desktop dialog keeps
its search filter. docSyncModel holds the direction a lane switch leads to
and the folder name made from the trip title, which the panel and the
phone sheet each carried a copy of.
useCollabNotesData holds the note list with its first load and the live
note events, and the attachment upload both shells run after saving a
note. Writing notes, their colours and categories stays in each shell,
since the desktop panel keeps empty categories and colour batches the
phone tab does not have. A failed load still empties the desktop list only.
useCollabPolls holds the poll list with its live updates, the countdown
tick, and creating, voting, closing and deleting. The desktop panel reads
the vote totals, winner and closed state from collabModel like the phone
tab does and keeps its own countdown label. A failed load still empties the
desktop list only, and closing still takes the server poll on the phone
only.
collabModel moves from the phone tabs into components/Collab so both
shells read it, and the desktop chat drops its own copies of the time,
date separator and emoji only helpers for the model's.
useCollabChatData holds the messages, paging, live updates, the draft with
its images, and sending, deleting and reacting. The desktop hook keeps its
emoji picker, hover and delete animation, the phone tab its long press
popover; where they differed (older page errors, scrolling after a remote
delete, upload progress, the edit permission on send) the hook takes an
option.
useMediaLightbox holds the signed image address of the file on screen,
paging by button, arrow keys and swipe, and Escape to close. The desktop
lightbox keeps its own index and thumbnail strip, the phone lightbox stays
controlled by the files tab; both draw their own overlay over the hook.
useFileTrash holds the trashed files with their restore, permanent delete
and empty actions; the desktop trash view loads it on demand and the phone
sheet while it is open. fileActions holds the clipboard file picking and the
two tier link toggle (own column first, then link records) that the desktop
assign dialog and the phone link sheet both ran on their own.
The stop popup and the phone draft sheet take the kind and length a found stop opens with, the kind pick that brings its usual length along and the six length choices from useStopDraftForm and stopKinds. The desktop keeps its night booking and still starts from the draft it mounts with; the phone still seeds the answers for every draft it is handed.
The stay dialog and the phone stay sheet take their presets, the five minute step, the full day limit, the draft and the arrival and departure preview from useStayDraft. Each still seeds the draft when it opens and saves it its own way; the phone keeps whole minutes when a step moves a stored fraction, the desktop does not.
The export dialog and the phone export sheet keep their rows and run the PDF, the calendar file and the GPX download through useTripExport, which also reads the print settings and whether my plan is offered. The desktop still runs one export at a time and closes after each one; the phone still lets a calendar or GPX file run beside a PDF and stays open after the PDF.
The transport detail dialog and the phone transport sheet name the booking's day and each segment's own booking code through transportDetailModel, next to the files filter they already share. The rest of the two views reads the booking differently on purpose and stays as it is.
The booking dialog and the phone booking sheet keep their markup and run on useReservationForm: how a saved booking or an import fills the form, the end before start check, the start date and time, the place and hotel picks, the stay range and the request body with the stay a hotel booking creates come from reservationFormModel. The desktop's check-in window, its link to a stop inside the form and its imported price stay its own, behind options where the code is shared.
The transport dialog and the phone transport sheet keep their markup and run on useTransportForm: the route rows, how a saved booking or an import fills them, the day options and the request body come from transportEndpoints. Where the two shells differ today (unknown type fallback, stopover days, which rows are reset, train anchors, empty metadata, the link field) the shared code takes an option. The phone's upload helper moves to bookingFormModel next to the traveller checks both booking forms use.
costsModel moves from the phone tabs to components/Budget.
useCostsLedger holds the settlement, the totals, the filters, the CSV
export and the payments, and useSettlementForm holds the payment form.
CostsPanel and MCostsTab both run on them.
The old differences stay behind flags. The desktop also loads the
settlement on a trip change, leaves expenses nobody paid out of my share
(#2225) and writes each currency as stored in the CSV. The phone sheet
starts blank, refills when it opens and ignores a second save while one
is running.
useExpenseForm holds the payers, the splits, the ticket items and the
receipts of the expense editor in CostsPanel and MCostSheet.
Flags keep the old differences. Only the desktop sends the people named
on a ticket line as members (#1382). The phone ignores a second save
while one is running and stays busy after a successful save. The
rebalance effect uses useEffectEvent instead of an eslint-disable.
useBagCardEditor holds the editing of a bag's name and weight limit for
the desktop bag card and MBagsSheet, and bagLoadSummary computes the
weight totals for the bag sidebar, the bag modal and the phone sheet.
The desktop card follows the bag when it changes elsewhere, the phone
row resets a name it did not save.
usePackingItemActions, usePackingTemplates and usePackingImport take the
item, template and bulk import logic out of usePackingListPanel and
MPackingListTab, and the phone listsModel keeps only its priority
tokens.
When adding to a category the desktop looks for the placeholder in the
store and the phone in its rendered items. The desktop toasts on an
empty import, the phone runs one import at a time. The effects run in
the same order as before.
The to-do filters, counts, sort order and categories move from the phone
listsModel into todoListModel. useTaskForm holds the name, the fields,
the load and the save of a task for the desktop detail and new task
panes and for MTaskSheet. Each side keeps its own guard and error
wording.
useListsSubTab remembers whether the packing or the to-do sub tab was
open, for ListsContainer and MTripShell. The desktop restores any stored
value, the phone restores todo and otherwise falls back to packing.
usePlaceForm and usePlaceSearch take the form state, the prefill, the
opening hours selection, the duplicate name check, the payload and the
maps search out of PlaceFormModal. MPlaceEditSheet, PlPlaceSearch and
PlCategoryPicker run on the same hooks, and PlTimeFields uses the shared
timeCollisions.
Where the two shells behaved differently the hooks take a dialog or a
sheet variant: the coordinate pair rule, abort and console handling,
when the billing session ends, whether a pick clears the suggestions,
the empty search cue, whether a map link clears the results, and how a
new category is trimmed and guarded. The open effect and the submit flow
stay per shell, because service stops, expense intents and delete arming
differ too much to share.
useAddPlaceToCollection holds the maps search, the pick and the save
that AddPlaceToCollectionModal and MCollAddSheet each carried, with the
pure helpers in addPlaceModel.
The differences stay as variants. The desktop resets and stays open
after an add, sends links and typed coordinates and shows the error's
own text. The phone closes, takes coordinates only from the picked hit,
ignores a second search or save while one is running and shows the
fallback text.
usePlacesPool holds the category filter, the search, the multi select
and the bulk mark visited of the places pool, for usePlacesSidebar and
for MPlacesBrowser with its filter controls and sheet. When selected
places are deleted, the desktop leaves select mode once none of them is
left and the phone drops only the missing ids. The hook keeps both.
usePlacesFileImport holds the GPX, KML and KMZ import that
FileImportModal and ImpFileStep each carried. The desktop keeps one
toast and one undo label and counts only the summary's skips. The phone
keeps a toast per format, the importFiles undo label for a mix and the
top level skipped fallback.
useListImport holds the Google and Naver shared list import that
usePlacesSidebar and ImpListStep each carried. The desktop closes the
dialog and clears the link before the undo step, the phone calls onDone
after it, and the hook keeps both orders.
useSaveTripPlacesToList holds the logic that saves a selection of trip
places to a list, which SaveTripPlacesToListModal and
MPlacesSaveToCollectionSheet each carried as a copy. The two copies did
the same thing, so the hook has no variants.
The bookings model moves from the phone tabs into components/Planner/bookings, so
both shells read the same sections, traveller predicate and type colours. The
phone bookings and transports tabs had the same traveller filter and folded
sections twice; that state now lives once in useReservationListFilter, which only
those two tabs use. The desktop panel keeps its own filters in useBookingsView
because it remembers them for the session.
The status switch and the delete on a booking card stay per shell. The desktop
hook renders its ConfirmDialog and awaits the delete, the phone asks in an
MConfirmSheet and lets the delete run on its own; sharing them would change
when the question closes.
filesFor moves out of the desktop booking parts into utils, and the phone
transport sheet now reads a booking's files through it as the desktop
transport detail and booking cards do.
The desktop note dialog and the phone note sheet fill their editor with
dayNoteDraft and save through writeDayNote, so both trim the title, store an
empty detail as null and default the icon the same way. Both count the detail
line against one shared limit.
The desktop hotel picker and the phone accommodation sheet take their day
options, range changes, checkout default, places on offer and request body
from stayFormModel, and both save a stay through writeStay.
The desktop day panel and the phone day sheet load the day's forecast with
useDayForecast, rename the day with useDayRename, and read the day's
bookings, the stay labels and the display temperature from dayDetailModel.
The day plan's connector menu and the phone timeline's are built by
legModeMenuItems, and both save a leg's mode through useLegModeActions, which
reports a refused write and reloads the days the same way on both.
The desktop day plan and the phone timeline derive the place order, the note
and transport positions and the per leg positions of a reordered day from
mergedOrder, and write the moved bookings back through the same helpers.
dayRoute moves from the phone's lib into components/Planner. The desktop day
plan builds its Google Maps and CoMaps stops with dayExportStops, and both
optimizers put the reordered stops around the fixed ones with fillAroundLocked.
The day plan, the day panel, the phone timeline and the phone day sheet read
the day local weather anchor from dayWeatherAnchor, and every weather icon
comes from one map in weatherIcons.
The place inspector and the phone place sheet now run on one module for
their participants, file uploads, image picks, ratings and place updates,
and both build the save to collection target with placeToSaveTarget.
The lease heartbeat and settle failures, the log sink and rotation
fallbacks, the runtime env getters, the public API summary edges, the
admin-only permission level and the WebSocket handler rejection had no
spec, which took six domains under their coverage floor.
The session gate passes a session lookup to verifyJwtAndLoadUser, the
Immich save takes auto_upload, getAppUrl stays synchronous, the demo
session list marks the current session by the id it is handed, and the
lease no longer needs a transaction of its own, so its spec is gone.
The converted specs wrote accommodation_id as a plain id, while the
services store the legacy '<id>.0' text, and wrote read-only foreign key
twins instead of their relations. The files e2e schema gains the trip
columns the entity fills, and the idempotency spec seeds a real user.
findRow and the default category lookup asked findOne with an empty
where, which MikroORM refuses. insertRows returned a single id because
insertMany on plain data reports only the last insert id; it now inserts
row by row and returns every id. insertRow and updateRows reject instead
of throwing before the caller holds a promise.
TRIP-SVC-072 still pinned the old order, where the trip row committed
before the days were rebuilt. Both now run in one transaction, as
TRIP-SVC-018b already asserts.
All 100 dialect helper cases pass on both engines. Of 1402 repository
methods, 11 send one statement Postgres refuses as SQL and 4 send none
the probe can see; both lists may only shrink from here.
The specs written alongside the session, transaction and lifecycle work
still seeded and read with raw statements. They go through the ORM
factories now; the few that test SQL below the ORM carry a test-sql-allow
reason. The raw SQL baseline holds 177 statements in 6 files.
The session carry-over specs were written against the old close and
reinitialize calls; they now drive the database backup port. The rollback
specs read through the factories and await the helpers that became async
when the seeding moved to the ORM. Two readers that lost their await and a
duplicated import are fixed as well.
acquire() made sure a row existed and then raced on a conditional UPDATE,
two statements the registrar wrapped in a transaction of its own. lint:tx
cannot follow a repository reached through a local variable, so it
reported the method. One INSERT ... ON CONFLICT DO UPDATE ... WHERE now
takes a free lease, renews the holder's and leaves anyone else's alone,
atomically on SQLite and on Postgres, and needs no transaction.
Both tables arrived in migrations written before the generator existed,
so src/db/kysely/ did not describe them and check:db-types would fail.
Regenerated with gen:db-types.
These specs left the raw SQL baseline but still emptied tables with
`db.exec('DELETE FROM ...')` and aged a Dawarich connection with a raw
UPDATE, which the ratchet cannot see because it only counts `.prepare(`.
They now go through deleteRows and updateRows like the rest of the file.
packing_item_recipients and packing_bag_members have no entity; both
cascade from their item and bag, so the packing spec relies on that with
foreign keys on in the snapshot database.
findRow returns null for a missing row, and `row?.deleted_at` is then
undefined, which `.not.toBeNull()` accepts. The converted soft delete,
album sync and passkey counter checks would therefore pass if the row were
removed outright. They now assert the row is there first, as the raw
`.get()` reads they replace did by throwing on a missing row.
The permissions and plugin RPC fallback comments were reworded when
their raw SQL went away; this keeps the new wording in line with the
comment style used elsewhere in the converted specs.
The specs converted to the factories no longer carry raw prepare()
calls, or keep only ones marked with test-sql-allow, so their baseline
entries are dropped. Leaving them in place would let raw SQL grow back
into those files unnoticed.
Converts the raw prepare() fixtures and reads in the reservations service
spec to the ORM factories: expense links, stays, stops, endpoints, day
positions and guest flags are written through insertRow and updateRows,
and the narrow SELECT post-conditions read whole rows through findRow and
findRows, compared with toMatchObject so they still assert exactly the
columns the projection named. stopsOn keeps returning only the three
columns its toEqual assertions compare, and the helpers that now await
become async.
The one write left raw carries a test-sql-allow marker: it plants
dangling references inside the synchronous foreign_keys = OFF window,
which an async ORM write would outlive. The unused canAccessTrip and
isOwner entries are dropped from the database mock.
Converts the raw prepare() fixtures and post-condition reads in the users
repository spec to updateRows, findRow and readUser. The Immich read
helper becomes async and still returns only the three columns its toEqual
assertions compare, and the guest insert check compares its columns with
toMatchObject now that the read returns the whole row.
The raw statements that remain carry a test-sql-allow marker: the legacy
statements run as parity oracles for the repository reads, the
NOT NULL DEFAULT 0 probe on is_guest, which has to bypass the entity
default, and the unflushed-entity check, which must read the table
without an EntityManager that could flush first.
Converts the raw prepare() fixtures and reads in the places service spec
to the ORM factories: tours, waypoints, files, links, expenses, collection
places and ratings are written through insertRow and updateRows, and what
the service left behind is read back through findRow, findRows and
countRows. The track colour cases insert their seeded tracks with
insertRows in place of INSERT ... RETURNING, tracksOf and the file attach
helper become async, and the stale updated_at is computed with dbNow
instead of datetime('now', '-1 hour'). The rating reads keep comparing
only the rating column, as the narrow SELECT did.
Converts the raw prepare() fixtures in the calendar service spec to the
ORM factories: reservation times, metadata and endpoints, stays, stops and
place addresses are written through insertRow and updateRows. The stay
helpers become async, so their callers await them. The timezone cache
case seeds its 1010 bulk bookings and their endpoints with two insertRows
batches instead of a raw transaction. The unused canAccessTrip and
isOwner entries are dropped from the database mock.
Converts the raw prepare() fixtures and reads in the auth service spec to
the ORM factories: instance settings go through setAppSetting and
deleteRows, MFA, OIDC and invite state through updateRows, and the
password version, push device and MCP token checks read back through
readUser and countRows. The push device helpers become async, so every
caller now awaits them. The unused canAccessTrip and isOwner entries are
dropped from the database mock, since nothing under test imports them.
Converts the raw prepare() fixtures and reads in the admin integration
suite to the ORM factories and readers. The user-deletion case that seeds
one row per foreign key pointing at the departing user now writes each of
them through insertRow and checks what survived through findRow; the
expiry timestamps that used datetime('now', ...) are computed with dbNow.
The OAuth client in that case uses the same string for its row id and its
public client_id, so the token and consent rows reference it the same way
whichever column the relation keys on.
Converts the raw prepare() fixtures and reads in the trips integration
suite to the ORM factories and readers: archived trips, permission rows,
memberships, the copy's source rows (tagged places, assignments, to-dos,
the budget category order, private packing items) and every read of what a
request left behind go through insertRow, updateRows, findRow and findRows.
The copied place tags are counted per place through the place_tags
relation instead of a raw join.
Converts the raw prepare() fixtures and reads in the document sync service
spec to the ORM factories and readers. The row builders (bindings, sync
items, files, the connection) and the files service double write through
insertRow and updateRows and are async now, the shape the real service has;
the row readers go through findRow and findRows. Seeding that ran inside
Array.from callbacks is a plain loop, so every row is written before the
run it feeds starts.
Three reads stay raw behind allow markers: the helper that takes a time
from SQLite's own clock and the two checks that compare a stored timestamp
with CURRENT_TIMESTAMP and julianday('now'), since the clock is what they
are about.
Converts the raw prepare() fixtures and reads in the user cleanup service
spec to the ORM factories and readers: plugins and their per-user config,
journeys with their entries, photos, contributors and share links, trip
share tokens, settlements and memberships are written through insertRow and
makeShareToken, and what the deletion leaves behind is checked through
findRow and countRows. The slim-schema case keeps its single raw users
insert behind an allow marker: that database is hand-rolled and carries no
schema an entity could write to. The trek photo the journey case seeds now
takes its created_at from the column default instead of a literal 0 no
assertion reads.
Converts the raw prepare() fixtures and reads in the atlas service spec to
the ORM factories and readers: places, booking endpoints, travellers,
visited countries and regions, cached place regions and bucket list rows
go through makePlace, makeReservationEndpoint, insertRow and upsertRow, and
the region cache is read back through findRow.
The two #2527 cases keep their raw move and delete, with an allow marker:
they land inside the repository call the test intercepts, synchronously on
the call's own connection, which is the race they pin. The database mock
loses its raw canAccessTrip and isOwner, which nothing reads any more.
Converts the raw prepare() fixtures and reads in the memories access
service spec to the ORM factories and readers: shared trip photos, journey
photos and contributors, album links and the trip memberships the access
checks walk are written through insertRow. The legacy MA5 statement stays
raw behind an allow marker as the oracle its parity cases compare with, and
the database mock loses the raw canAccessTrip and isOwner nothing reads.
The reservation visibility parity harness now stages bookings, links stays
and reads the trip's days through the ORM factories and readers. The
legacy string fragments it holds the typed predicates to keep running raw
behind allow markers, since they are the fixed side of the comparison.
Converts the raw prepare() fixtures and reads in the collab service spec to
the ORM factories and readers: reactions, note attachments and the stored
note, message and vote rows go through insertRow, findRow and countRows.
The database mock loses its raw canAccessTrip and isOwner, which nothing
reads any more, and the repository parity cases keep the legacy joins they
are held to behind allow markers.
Converts the raw prepare() fixtures and reads in the places integration
suite and the trip members repository spec to the ORM factories and
readers. Tags, categories and linked expenses are seeded through makeTag,
makeCategory and makeBudgetItem; the member rows the roster cases backdate
or point at an inviter are written through insertRow and updateRows. The
legacy UNION and join statements the roster reads are held to stay raw
behind allow markers.
Converts the raw prepare() fixtures and reads in the unified memories
service spec and the admin service spec to the ORM factories and readers.
The memories spec drops the raw canAccessTrip and isOwner on its database
mock, which nothing reads any more; its two parity cases keep the legacy
joins they are held to behind allow markers and build the provider list
they bind through the reader. The admin spec's password reset helpers
(password version, token and push device counts, the stored hash) now read
through the ORM, and the audit log parity case keeps its legacy statement
as an allowed oracle.
Converts the raw prepare() fixtures and reads in the files service spec and
the dawarich sync service spec to the ORM factories and readers.
The dawarich service stub the sync spec hands in now reads and writes its
connection rows through the ORM and is async, the shape the real service
has. Writes that set accepted_hash from the row's own source_hash read the
row first and write the value back, since a native update cannot reference
another column. The files spec drops the raw canAccessTrip and isOwner on
its database mock, which nothing reads any more, and keeps the legacy
FILE_SELECT statements it is held to behind allow markers.
Converts the raw prepare() fixtures and reads in the system notices
integration suite, the packing e2e and the synology memories integration
suite to the ORM factories and readers.
The packing e2e moves from its hand-rolled tables to the migrated snapshot,
seeds its three users with the ids their session cookies are signed for,
and drops the raw canAccessTrip on its database mock, which nothing reads
any more. The synology suite now shares one ORM handle for the photo
registration helpers and the readers instead of declaring a second one
halfway down the file.
Converts the raw prepare() fixtures and reads in the demo repository spec,
the transports MCP spec and the immich integration suite to the ORM
factories and readers.
Where a case compared a narrow SELECT with toEqual, it now picks the same
columns off the row the reader returns, so the comparison stays exact
rather than loosening to a partial match. The immich album link cases seed
their photos through addTripPhoto and addAlbumLink, which register the
provider asset the same way the raw INSERT OR IGNORE did.
Converts the raw prepare() fixtures and reads in the files MCP spec and the
roadtrip e2e to the ORM factories and readers.
The files spec seeds its trip files through makeTripFile and its links
through linkFile, keeping the helper's explicit-null handling for size and
type. The roadtrip e2e keeps its pinned trip, day and place ids, which its
routes address directly, now written through insertRow and insertRows.
Converts the raw prepare() fixtures and reads in the share integration
suite and the passkey service spec to the ORM factories and readers.
The passkey credential helper takes a typed set of overrides instead of a
loose record, so a fixture can no longer hand the insert a column the
entity does not map. The two credentials the listing case backdates are
written with their created_at through insertRows, blobs as Buffers.
Converts the raw prepare() fixtures and reads in the collections MCP spec
and the in-app notification preference spec to the ORM factories and
readers. The collection, member and saved place helpers and the boolean
notification helper keep their signatures and simply resolve to the new
row's id, so each case reads as it did; the four copies of the boolean
notification INSERT collapse into one helper over makeNotification.
Converts the raw prepare() fixtures and reads in the trek photo cache and
google place photo cache services and the dawarich e2e to the ORM factories
and readers.
The place photo cache suite moves from its hand-rolled three-table fixture
to the migrated snapshot: the references it checks are now real places on a
trip and real saved places in a collection, seeded through makePlace and
makeCollectionPlace, which a table without names or owners could not hold.
The dawarich e2e drops the raw canAccessTrip on its database mock, which
nothing reads any more.
Converts the raw prepare() fixtures and reads in the plugin discovery spec,
the permissions service spec, the feeds and invite MCP spec and the day
accommodations repository spec to the ORM factories and readers.
The permission cache cases still write their override rows past
savePermissions, now through setAppSetting, so the stale-cache behaviour
they pin is unchanged. The day accommodations spec keeps the legacy
statements it holds the repository to behind allow markers.
Converts the raw prepare() fixtures and reads in the collections e2e, the
storage registry service spec, the transit MCP spec, the MCP token
repository spec and the immich memories integration suite to the ORM
factories and readers.
The collections e2e drops the raw canAccessTrip on its database mock, which
nothing reads any more. Reads that used to join trip_photos to trek_photos
in SQL resolve the photo row through the readers instead, and the orphan
trip photo of IMMICH-057 is still written with foreign keys switched off,
now through the ORM on the same connection.
Converts the raw prepare() fixtures and reads in the trip read model
service spec and the share token and audit log repository specs to the ORM
factories and readers.
The trip read model spec drops the raw canAccessTrip and isOwner it still
put on the database mock; nothing has read them since trip access moved
onto the repositories. The share token spec now inserts through
makeShareToken and reads what a write left behind through findRow, keeping
the SELECT * row and the legacy statements it is held to as allowed
oracles. The audit log spec keeps the raw row and the typeof() storage
class probe behind allow markers, since both are about what SQLite stores.
Converts the raw prepare() fixtures and reads in the plugin settings
isolation spec, the tags MCP spec, the place ratings and days repository
specs and the receipt scan e2e to the ORM factories and readers.
The receipt scan e2e writes the AI parsing addon config as the object the
entity's JSON column maps, not as a pre-serialised string, and seeds its
users and trips through makeUser and makeTrip with the ids its session
cookie is signed for. The days spec keeps the SELECT * row it compares the
ORM's rows with key for key, and its legacy statements, behind allow
markers; the place ratings spec backdates updated_at with dbNow instead of
datetime('now').
Converts the raw prepare() writes and reads in the websocket connection
suite, the web push and instance API key services, the days MCP spec and
the packing visibility and place details cache repository specs to the ORM
factories and readers.
The websocket suite reads the app's own ORM; the password version bump that
was a single relative UPDATE now reads the user and writes the next
version. The packing visibility parity keeps the legacy fragment it is
checked against behind an allow marker.
The visited countries and place regions repository specs seed their rows
through the ORM; the legacy statements they hold the repositories to stay
raw behind an allow marker, since those are the oracles the parity is
checked against.
The storage admin e2e moves from hand-rolled tables to the migrated
snapshot, seeds its two users through the factories with the ids their
session cookies are signed for, and reads the stored backends and the audit
rows through the readers instead of raw SELECTs.
Converts the raw prepare() fixtures and reads in the share and place shadow
e2e suites, the google quota, travel stats and MCP guard services, the
memories MCP spec, the admin seeder spec and the assignment participants
repository spec to the ORM factories and readers.
The place shadow e2e moves from hand-rolled tables to the migrated snapshot
with its users pinned to the ids their session cookies are signed for. The
share e2e drops the raw canAccessTrip it still put on the database mock:
nothing has read it since trip access moved onto the repositories. The
quota spec computes yesterday and the retention window in UTC, the same
calendar day SQLite's date('now') gives. The participants oracle keeps its
raw statement behind an allow marker.
Converts the raw prepare() fixtures and reads in the next set of specs to
the ORM factories and readers: the password reset, webauthn, dawarich,
budget, to-do and vacay repository specs, the journey, settings, addons and
share MCP specs, the plugin POI and operator egress specs, the feeds,
storage stats and query helper services, and the packing, collections,
day reorder, leg mode, upload context and trip access integration suites.
The notifications and google quota e2e suites now seed their users through
makeUser with pinned ids, and the notifications suite moves from hand-rolled
tables to the migrated snapshot. Legacy statements that serve as a parity
oracle and the raw rows the primary-relation spike holds toObject() against
stay raw, each with an allow marker naming why.
Converts the raw prepare() reads and writes in the share, airtrail, files,
todo, dev-link, vacay, tours, bucket list, unit-of-work and plugin guard
specs to the ORM factories and readers, so they go through the entities the
app uses instead of hand-written SQLite.
The vacay e2e moves from hand-rolled tables to the migrated snapshot, with
its user pinned to the id its session cookie is signed for. Statements that
are raw by nature keep the allow marker with the reason: the legacy oracle
of the bucket list parity spec, the PRAGMA probe of the trek_photos
migration, the unmapped probe table of the route import rollback, the
behind-the-ORM reads of the entity insert probes, and seedUser, which also
serves the hand-rolled users table of createTempDb. Comments that quoted a
prepare() call are reworded so the ratchet no longer counts them.
The case checks the text format revokeAllForClient and revokeById write.
Reading it through the entity runs DbTimestampType, which turns an epoch
integer or a Date into formatted text, so an integer write would still
pass. The two reads go back to raw statements under an allow marker and
assert the stored value is a string before matching its format.
The specs seeded through the repositories in the commits before this one
no longer carry unmarked raw statements, so their entries leave the
baseline. What stays raw in them is marked with a reason.
The journey domain suite now writes its photos, entries, contributors,
trip links, plan edits and place changes and reads back what a call left
through the ORM factories. Three small readers keep the shape a
statement's .get()/.all() handed back: one() answers undefined when no
row matches, pickOne()/pickAll() return just the named columns, and a
count stays { n }, so every toBeDefined, toBeUndefined and toEqual keeps
its meaning. The helpers that wrote or read rows (photoOnEntry, photoOn,
insertEntry, orderOf, daysOf, skeletonFor, skeletonsFor, skeletonsOf,
withGeometry) are async and awaited where they are used, and the M5d
parity seed binds its values in the old column order through columnsOf.
Kept raw and marked: the legacy statements of both repository parity
blocks (JG8, JG14, JG17, JG18, JG34 and the LEG set) and the unguarded
delete the JG119 mutation proof measures the guard against.
The DB-backed budget service suite now sets trip currencies, hangs
expenses on bookings, writes places, bookings, trip files, file links
and category order, and reads back items, members, payers, settlements
and links through the ORM factories. Its row readers (paidFlag,
personsOf, memberCount, itemRow, placeRow, settlementRow, totalOf,
rateOf, metadataOf, linksOf) and seeders (tripWithBooking, seedAudTrip)
are async now and awaited where they are read; projected comparisons map
the same keys off the found row.
The database mock drops its dead trip access stand-ins. Kept raw and
marked: the legacy BG10 to BG13 and settlement statements the repository
reads are held to, and the currency switch BUDGET-SVC-DB-066 makes from
inside the stubbed rate fetch.
The places repository suite now edits its fixture places, writes tags,
ratings, tours, regions, stops and tracks, and reads back what a write
left through the ORM factories. The datetime('now', ...) offsets the old
statements computed in SQL are the same UTC second-resolution strings,
computed by sqliteNow, and the full-column fixture of PLACEREPO-050 goes
through insertRow with every column it named.
Kept raw and marked: the legacy getPlaceWithTags composition, the listForTrip
statement builder and every other legacy projection the typed reads are
compared against, plus the UPDATE in PLACEREPO-013 that must land behind
the ORM's back.
The reservations repository suite now shapes its bookings, trips, stops,
stays, legs and day positions through the ORM factories, takes its day
ids from findRows and reads back the AirTrail and review columns a write
left with findRow.
Every legacy statement the typed reads are compared against stays raw
and marked as the oracle, the calendar, upcoming, share and public API
projections among them.
The trips repository suite now edits its fixture trips and reads back
what a write left through the ORM factories (setTrip, tripRow,
findRow), and promotes its admin through updateRows.
The parity statements stay raw and marked: canAccessTrip, the owner
check, the viewer, active trip, feed token, travellers, sync and share
projections are the oracles the repository reads are held to. Also kept
raw: the SELECT * snapshot toObject has to reproduce, and the UPDATE in
TRIPREPO-022 that must land behind the ORM's back.
The days service suite writes bookings, legs, stays, vias, boundaries
and day edits through the ORM factories and reads them back with
findRow/findRows/countRows. The day, booking and stay readers the cases
compare against (dayRows, reservationRow, staySpan, stopsOn, viasOn,
boundaries) are async now and awaited where they are read; a reader of a
row that is gone fails the case as the old cast did.
The database mock drops its getPlaceWithTags, canAccessTrip and isOwner
stand-ins, which the module stopped exporting when those reads moved
onto the repositories.
The packing service suite writes its items, bags and templates and reads
back what a call left through the ORM factories. Template items are read
category by category instead of through a join, and the trip access
stand-ins the database mock still carried go, as the module no longer
exports them.
The repository parity block keeps its legacy statements raw, marked: the
VISIBLE_TO_ACTOR fragment, the assignee join, the correlated item count,
the category assignees and the export read are the oracles the converted
reads are held to.
The days integration suite, the OIDC service suite and the share service
suite write their fixtures and read back their effects through the ORM
factories: invite tokens through makeInviteToken, user column edits
through updateRows, bookings, stays, packing items and messages through
insertRow on their entities. Two more database mocks drop their dead
canAccessTrip/isOwner stand-ins.
The OIDC identity-linking cases keep their raw SELECT * of the user row,
marked: they assert that no column of the table moved, including any the
entity might not map.
The days and tours e2e suites, the assignments service suite and the
journey share suite build their fixtures through the ORM factories and
read back through findRow/findRows/countRows. Fixtures that pin ids (the
trip, day and places a session cookie or a route targets) keep them.
Readers that the old suites called synchronously (order, dayOrder,
viaAnchors, insertJourneyPhoto) are async and awaited, Promise.all where
a case compares several at once.
Left raw and marked: the legacy statements the parity cases hold the
repository reads to, the COLLATE NOCASE lookups that prove the token
match is exact, and total_changes(), which counts writes on the
connection itself.
The collab and budget e2e suites, the MCP integration suite, the budget
MCP tools and the OAuth token repository suite seed and read through the
ORM factories. The collab e2e suite moves off its hand-rolled CREATE
TABLE set onto the migrated snapshot, the way the factories guide asks of
a suite that converts: an entity reads and writes every mapped column,
which a trimmed table does not have. Its users are seeded with the ids
the session cookies are signed for.
The recursive chain statement in the token suite stays raw, marked as the
legacy oracle collectChainIds is held to.
The atlas and notes MCP tool suites, day removal, the doc sync config
service and the plugin runtime conversion suite build their fixtures and
read back what a call left through the ORM factories. Helpers that used
to answer synchronously off the handle (dayRows, range, the reservation
and boundary readers) are async now and every caller awaits them; a
reader that answered undefined for a gone row still does, so the
toBeUndefined/toBeDefined checks keep their meaning.
The one statement left raw rewinds sqlite_sequence, the engine's own
autoincrement table, which no entity maps; it is marked.
Reminder jobs, user profile, trip members, Dawarich suggestions, the doc
sync controller, the plugin registry and the plugin repository parity
suite now write and read their rows through the ORM factories. The doc
sync controller's itemsForTrip stand-in answers through the two
repository reads DocSyncService makes instead of a copy of their SQL, and
one more database mock drops its dead canAccessTrip/isOwner stand-ins.
What stays raw is marked: the legacy statements the parity cases are
held to, and the places stub in the Dawarich suggestions suite, which
runs inside the service's open transaction and has to write on the
connection that transaction holds.
Eleven more suites (atlas and settings e2e, the bucket list, shadow pick
and Dawarich suggestion repositories, the budget and collab MCP tools,
notification preferences, VAPID keys, school holidays and tours) write
and read their rows through the ORM factories. Time-relative fixtures
that leaned on SQLite's date('now', ...) build the same text in JS with
dbNow(), and helpers that returned undefined for a missing row keep doing
so where an assertion depends on it. The legacy statements the parity
cases compare against stay raw, marked as oracles.
Service, MCP, integration and e2e suites with up to a dozen raw
statements now seed and read through the factories. A buildApp() suite
uses the app's own ORM, a partial-module e2e or a unit suite a
createTestOrm() or sharedTestOrm() handle on its database. Reads that
used to come back undefined now come back null, and the assertions say
so; where a projected SELECT was compared whole, the same columns are
picked off the row.
Two reads inside the scheduler's delivery callback stay synchronous raw
statements, marked, so they keep seeing the row at the moment of
delivery. Two more database mocks lose a canAccessTrip/isOwner stand-in
the module no longer exports.
The repository suites now write their fixtures and read back what a call
left behind through the factories and readers, not hand-written SQLite on
the better-sqlite3 handle. What stays raw is marked with test-sql-allow:
the legacy statements the parity cases hold a repository read to, rows
read as the table stores them (a BLOB, an integer the entity maps to a
boolean), and a write that has to land on the connection an open
transaction holds.
Cases that pinned a "raw UPDATE" between two repository reads now make
that write through a context of their own, which the repository's
identity map knows as little about, so they still prove the same thing.
The suites with one to three raw statements left now seed and read their
rows through the ORM factories instead of hand-written SQLite on the
better-sqlite3 handle. A legacy statement that serves as a parity oracle,
the schema introspection and the bare backup file keep their raw SQL and
say why with a test-sql-allow marker. Two database mocks still answered a
getPlaceWithTags the module no longer exports; that dead branch goes too,
and the budget seam suite stubs the reservations repository directly
instead of a fake prepare().
The TRIP-SVC-069 reservation seed was a raw INSERT carrying a
test-sql-allow marker, justified as reproducing the legacy raw binding of
a number into the TEXT accommodation_id. legacyBoundIntegerText already
writes that exact "<id>.0" shape and the sibling copy tests seed the same
column with it through insertRow, so nothing about this statement needs
to stay raw. The seed now goes through the factory with the same column
values, and the assertion still pins the legacy shape.
The seeding rounds above took the converted specs down to few or no
raw statements, so their entries in the ratchet drop to the new counts,
and the files that reached zero leave the baseline. A file that is
converted can no longer grow its raw SQL back unnoticed.
Two specs move to the exemptions instead, because raw SQL is their
subject: the migration step that reseats booked nights is tested on rows
seeded over the raw connection before the step runs, and the place
region trigger is pinned by raw updates on databases the migrations
alone built.
The two largest service suites seeded and inspected rows with raw
statements. They now write through the ORM factories and read through
small local readers that return a row the way the SELECT did (only the
selected columns, undefined when there is none), so every assertion keeps
its exact meaning. The trips database mock loses its dead canAccessTrip,
isOwner and getPlaceWithTags stubs.
Raw statements stay, marked, where they are the subject: the legacy day
grid rebuild that the fuzz case runs as its oracle inside a synchronous
transaction it rolls back, the reservation seed that reproduces the
legacy REAL binding into the TEXT stay link, and the legacy entry read
model the vacay suite compares against.
reservations.accommodation_id is TEXT, and every writer stores a stay id
in the shape the legacy statement bound it (`14.0`, see
legacyBoundIntegerText). The converted specs wrote and matched the bare
`14` instead, so a lookup of a booking the service had linked would miss
it and a seeded link would not look like one the app writes. They now go
through legacyBoundIntegerText, the same helper the services use.
The settings, notification, road trip plan, collection and accommodation
service suites wrote their fixtures and read back their results with raw
statements against the shared test handle. They now seed and inspect
through the ORM factories and readers on the suite's shared test ORM,
with small local helpers where a suite repeats the same read.
The collection and accommodation database mocks drop their dead
canAccessTrip, isOwner and getPlaceWithTags stubs: db/database exports
none of them since the repositories took them over.
Raw statements stay, marked, where they are the subject: the legacy
read model oracle in the collections suite, the SELECT * that proves a
saved place has no itinerary columns, and the member count taken
synchronously while the delete broadcast fires.
The service suites for mail, notification channels and transports,
idempotency, invites, tokens, OAuth, push subscriptions, the version
check, budget settlement maths, addons, place shadow picks, tour GPX
import, the place region repair, journey stats, day notes and storage
jobs seeded and inspected their rows with raw statements. They now go
through the ORM factories and readers over the suite's shared test ORM.
The database mocks of day notes and OAuth also lose their dead
canAccessTrip and isOwner stubs: db/database exports neither since the
repositories took them over.
Raw statements stay, marked, where they are the subject: the backup
file the boot restore reads before any ORM exists, the full raw row a
day note is checked against, sqlite_master, the synchronous probe that
counts tours while the GPX import may still hold its transaction, and
the edits the region repair case lands mid-pass, past the ORM's
connection queue.
The demo seed parity case, the transaction yield probe and the test
UnitOfWork check read their row counts and rows back with raw
statements; they now use the ORM readers.
The statements these files are about stay raw and are marked: the
dirty read and write the transaction yield probe measures on the shared
handle, and the reset-admin.js case, whose database file has no ORM
bound to it.
The repository suites for assignments, notifications, document sync,
tags, photo metadata, vacay, day notes, notice dismissals, idempotency
keys, settings, travellers, invites, providers, hidden regions, road
trip tracks and the collection role check seeded their fixtures and
read back what a write left behind with raw statements. Fixtures now go
through the ORM factories and insertRow/updateRows, and the state reads
through findRow/findRows/countRows.
What stays raw is raw on purpose and carries a test-sql-allow marker:
the legacy statements every parity case compares a repository against,
the full raw rows used as such an oracle, PRAGMA reads, typeof() and
date() storage checks, and the out-of-band writes the identity map
freshness cases are about, which have to bypass every EntityManager.
The plugin service, instance and user settings, OAuth, audit, route
profile, host state and request context suites seeded plugin rows,
settings fields, actions, tokens and audit entries with raw statements
and read them back the same way. They now go through the ORM factories
and readers over the suite's own test ORM.
Two raw statements stay, marked: the PRAGMA table_info read that proves
the actions scope column exists, and the audit insert that reproduces
the pre-conversion appendAudit byte for byte for the hash chain parity
check. The instance settings suite's stubbed isActive now follows the
status the test flips through markActive instead of a synchronous read.
The buildApp() suites for auth, bookings, budget, journeys, assignments,
passkeys, profiles, OAuth, the AirTrail tick and the plugin boot order,
plus the storage restore suite, seeded settings, addons and fixture rows
and read back what a request left behind through raw better-sqlite3
statements. They now write and read through the ORM factories and
readers, against the app's own MikroORM (or the suite's test ORM where
one already existed).
The account deletion case keeps every one of its foreign key checks; a
row that used to read back as undefined now reads back as null.
The e2e suites for assignments, places, bookings, trips, the public API,
doc sync, addons, transit, to-dos, admin, invite links, school holidays,
categories, feeds and plugin search seeded and inspected their database
with raw better-sqlite3 statements. They now go through the ORM
factories and readers, the way the repositories write and read.
Six of them (public API, transit, to-dos, admin, invite links,
categories) still built a handful of hand-rolled tables. They move to
the migrated snapshot, since an entity reads and writes every mapped
column; where the real schema keeps usernames unique, the seeded users
get distinct names and keep the ones the assertions read back. The dead
canAccessTrip stubs left in two database mocks are dropped: nothing in
src imports it from db/database any more.
The MCP tool suites seeded trips, places, bookings, packing, vacay and
settings rows and read back what a tool call left behind through raw
better-sqlite3 statements on the sync handle. They now go through the
ORM factories and readers in tests/helpers/factories, so the rows are
written and read the way the repositories do it.
Every assertion keeps its meaning: a missing row that read back as
undefined now reads back as null, and a partial SELECT is replaced by
the same fields picked off the full row.
The server suite seeds and reads its database through some 5,500 raw
better-sqlite3 statements on the sync handle. They restate the schema by
hand, skip the repositories the app goes through and run on SQLite only,
so nothing in the suite could ever exercise the dialect layer on another
engine.
tests/helpers/factories/ adds typed factories per domain (users, trips and
days, places, assignments, bookings, budget, packing, to-dos, files,
journeys, tours, collections, notifications, settings, plugins, tokens,
atlas, vacay, photos, collab) and generic readers (findRow, findRows,
countRows, updateRows, deleteRows) that go through MikroORM in a request
context of their own, against the app's ORM or a createTestOrm handle.
They refuse a write to a read-only foreign-key twin, which MikroORM would
otherwise drop without a word. The README there is the conversion guide.
The categories integration suite and the tags and calendar-feed e2e suites
are converted in full as worked examples; the tags suite also moves from
hand-rolled tables to the migrated snapshot.
lint:test-sql (scripts/test-sql-ratchet.mjs) holds the raw statements left
to a per-file baseline: growth and stale entries fail, files whose subject
is raw SQL are exempt with a reason, and a single statement can be allowed
with a test-sql-allow comment. CI runs it next to lint:boundaries.
A password change ends every session of the user and re-issues one for
the device it was made on, handed out as the session cookie. A caller
that authenticated with a Bearer token never reads that cookie, so the
new session sat in the session list as a live device nobody held until
it expired.
The controller now asks for the replacement session only when the
request came with the session cookie, the same check the sliding
renewal makes. The response body is unchanged.
A session renewed from a token issued before sessions were tracked has
an id derived from that token, and the renewal inserts it only when no
row by that id exists. The nightly purge deleted revoked rows straight
away, so the next request with the still valid old token inserted the
session again as an active one and got a fresh tracked token with a new
lifetime. A logout, an ended session or a sign out of the others held
only until the purge ran.
The purge now deletes expired rows only, and a revoked row stays until
its own expiry. A derived row is recorded with an expiry no earlier
than the old token's, so by the time it goes the old token is refused
by its signature check anyway.
The session e2e suite drove login, logout, the session routes, a
password change and renewal, but nothing showed end to end that a
password reset, an MFA disable, account deletion or an admin action
makes the old cookie stop working, the purge job had only a unit test
with a stubbed service, and reset-admin.js had no test of its session
handling at all.
sessions.e2e.test.ts now also resets a password through the mailed
token (both devices get 401, every row is ended, the new password signs
in), turns two-factor off from one device (the other gets 401, this one
stays), deletes the account (the rows go with it) and runs
SessionPurgeJob.tick against the real table (ended and expired rows go,
the live session keeps working). The admin paths need AdminModule and
with it most of the app, so session-revocation.e2e.test.ts boots
buildApp(): an admin setting a password, clearing two-factor and
deleting a member each sign both of the member's devices out and leave
the admin signed in, while an edit that leaves the password alone signs
nobody out. reset-admin-sessions.test.ts runs the script against a
database with session rows (only the reset account's live ones end) and
against one without the table.
The token in the sign-in bodies was declared deprecated and "read by
nothing of ours", but the web client still decided an SSO sign-in by
reading data.token from GET /api/auth/oidc/exchange, an endpoint the
deprecation notice left out. Removing the field later, as documented,
would have broken SSO sign-in.
The exchange now answers { success: true, token } and the login hook
signs in when the response is OK and success is true, so it never reads
token. The field stays for API clients and is now on the deprecation
list in the API docs, the controller comments, server/CLAUDE.md and the
wiki (Security-Hardening, OIDC-SSO). Adding a field to the body breaks
no existing client.
The sessions domain owns user_sessions, yet AuthService injected
UserSessionsRepository next to SessionsService and AuthModule registered
the entity in its own forFeature, only so verifyJwtToken could hand the
repository to verifyJwtAndLoadUser.
verifyJwtAndLoadUser now takes a SessionLookup, the two repository
methods the session check uses (findActive, touchLastSeen).
SessionsService answers both, so AuthService passes itself through the
sessions domain's API and drops the repository and the forFeature entry.
The guards keep handing in the repository from their own entity
manager, which satisfies the same type. auth.service.ts lost lines, so
its size baseline goes down with it.
A token without a jti past half its life is renewed on every request
its cookie authenticates, and each renewal minted a session with a
random id. A first page load sends many requests at once, so one browser
ended up with several active sessions; only the last cookie survived,
and the rest showed up in GET /api/auth/sessions as separate devices
until they expired, up to 30 days for remember me.
The interceptor now hands the token to SessionsService.renew, which
derives the session id from it (a SHA-256 of the token, laid out as a
version 8 UUID so it passes the same id check) and inserts the row only
if it is not there yet. Every renewal of the same token lands on the
same session; if that session was ended in the meantime nothing is
renewed, just as for a tracked token. A caller that passes no token gets
the old behaviour.
On a demo instance every visitor signs in as the one shared demo
account, so its sessions are other people's browsers. The session
routes had no demo check, unlike the password, MFA and account
changes, so any visitor could read the User-Agent of everyone else
signed in and sign them all out with revoke-others.
For the shared demo account in demo mode, GET /api/auth/sessions now
lists only the session the request came with, and DELETE
/api/auth/sessions/:id and POST /api/auth/sessions/revoke-others answer
403 "Sessions cannot be ended in demo mode." without ending anything.
Logout still ends the visitor's own session. Other accounts on a demo
instance keep the full routes. No MCP tool covers sessions, so there is
no parallel tool to change.
The demo sign-in now records the device's User-Agent on its session
like every other sign-in, which the session list needs to tell the
visitor's own entry apart.
Both swap the whole database file. The demo reset copies a baseline
saved at first seed, before anyone logged in, so it brings back an empty
user_sessions table; a backup restore brings the backup's own rows, or
none for a backup from before sessions were tracked. A session token
whose row is missing is refused, so the hourly reset signed every demo
visitor and the demo admin out, and the admin who ran a restore got a
401 on the next request. Before session tracking only the password
version applied, and those tokens survived.
Both paths now read the active rows with their owner's email before the
swap and put them back after the reopen, in a fresh fork, for every
account the new file still holds under the same id and email. A row the
new file already has is left as it is, so a session that ended there
stays ended, and a session whose id now names somebody else (a backup
from another install) is dropped instead of handed to them. The password
version check still refuses a token the restored account no longer
matches. Reading or writing them is best effort and never fails the
reset or the restore.
A session token was a signed JWT and nothing else. Logout only cleared
the cookie, a 30-day "remember me" token stayed valid wherever a copy of
it had ended up, and nobody could sign out one device short of changing
the password or rotating the JWT secret for everyone.
Every session token now carries an id (jti) naming a row in the new
user_sessions table (id, user_id, created_at, last_seen_at, expires_at,
revoked_at, user_agent cut to 256 characters; no address is kept). The
rows live behind UserSessionsRepository and the new sessions domain,
whose SessionsService is the one issuer: AuthService.generateToken and
the OIDC sign-in both go through it, and sliding renewal re-signs a
tracked token under the same id and moves its expiry. verifyJwtAndLoadUser
takes the repository and refuses a token whose session is revoked,
expired or gone, refreshing last_seen_at at most every five minutes.
A token without a jti was issued before this change; it passes until its
own expiry (or a password change), so the upgrade signs nobody out, and
renewal turns it into a tracked session.
Sessions end on logout (the one the request names, verified even when
expired), on a password change (all, the device gets a new one), a
password reset, an admin setting a password, reset-admin.js, an MFA
disable (all but the current one), an admin clearing a user's MFA, and
with the account. GET /api/auth/sessions lists the active sessions,
DELETE /api/auth/sessions/:id ends one and POST
/api/auth/sessions/revoke-others ends all but the current, with their
contracts in @trek/shared and two new audit actions. There is no UI for
them yet. MCP exposes no account or session tools, so there is no
parallel tool to change; an MCP call made with a session JWT as bearer
goes through the same check.
The token in the login, register, demo-login, MFA and passkey bodies
stays for API clients but is marked deprecated in the API docs and in
the client type, which never read it.
A nightly session-purge job on CronRegistrarService deletes the expired
and revoked rows. The migration is 042500. The sessions domain gets its
own coverage entry at the 80 floor until a full run measures it.
confirm() created the venue's place row in its own transaction and
broadcast place:created before the booking and its cost were written in
a second one. When that second write failed, the item was skipped but
the trip kept a place no booking pointed at, and clients had already
been told about it.
Per item, the venue and endpoint geocoding, the day lookups and the
cost's frozen rate still run first, outside any transaction. The place
row, the booking and its linked cost then go into one transaction
through the new ReservationsService.createWithCostInTx, the core that
createWithCost itself now wraps. place:created, reservation:created,
accommodation:created, the stay mirror and the cost events are sent
only after the commit. Each item keeps its own transaction, so a
failing item still leaves the others imported; confirm() is tagged
@txIndependent for that and leaves the lint:tx baseline.
PUT /api/integrations/memories/immich/settings saved the connection through
one service call and then wrote auto_upload through a second one from
the controller, each committed on its own. A failure between them left
a new connection with the old upload choice.
saveImmichSettings takes the auto_upload flag and writes it with the
connection (or the disconnect) inside one transaction, after the SSRF
check that does the DNS lookup. An absent or non-boolean flag still
keeps the stored choice, a refused URL still writes nothing, and the
response is unchanged. The controller's write sequence leaves the
lint:tx baseline, and the MemoriesService facade loses the
immichSetAutoUpload pass-through it no longer needs. No MCP tool writes
these settings.
TripsService.update rebased the budget onto the new currency in its own
transaction and only then wrote the trip row in a second one. When the
day grid rebuild, the booking resync or the leave shift threw after the
rebase had committed, budget items and settlements stayed pinned and
re-frozen against a currency the trip never took (#1543).
rebaseTripCurrency is split into prepareCurrencyRebase, which fetches
the rates before any transaction opens, and applyCurrencyRebase, which
reads the outgoing currency and writes the rebase. update() fetches
first and hands the plan to updateTrip, which applies it inside the
transaction that writes the row, the leave shift and the day grid.
updateTrip called without a plan (the plugin RPC host) still never
rebases, and rebaseTripCurrency keeps working as before.
The updateTrip docstring no longer claims the row is written outside
the day regeneration transaction. TripsService.update leaves the lint:tx
baseline, and the size baselines shrink with the files.
createWithCost and updateWithCost run the booking and its cost in one
transaction, but the cost sync still caught and logged its own errors
inside it. A price that could not be written therefore committed the
booking edit and left the linked expense on its old figure; only the
category re-file path actually rolled back.
The sync no longer swallows a failing cost write. Both WithCost methods
are its only callers, so the error now reaches their transaction and
the booking and its cost commit or fail together. A booking import
item whose cost cannot be written is skipped like any other failing
item instead of landing without its cost.
uploadGalleryPhotos and addPhoto read MAX(sort_order) inside the transaction
that inserts, but the only gallery test ran its uploads one after another and
would pass without that. These cases run several uploads, and several photos
added to one entry, at once and expect consecutive positions in the gallery
and on the entry, as JOURNEY-SVC-030b does for entries.
syncBudgetOnUpdate created the expense and then set its reservation_id in a
second statement, inside a try that logs and swallows the failure. Running
inside updateWithCost, a failed back-link committed an expense that was not
linked to its booking. The expense is now created through
linkBudgetItemToReservation, which writes the link on the insert, the same
way syncBudgetOnCreate does.
update_reservation and update_transport accept a type but still called the
bare ReservationsService.update. Only updateWithCost re-files a linked expense
that still carries the category derived from the old type, so changing a
booking from flight to train over MCP left its expense filed under flights
while the same change over REST or RPC moved it. Both tools now call
updateWithCost without a price entry, and send the cost events and the
accommodation change after the commit, as REST does.
erasePluginUserData deleted the host-side plugin rows and then queued the
plugin erasures as separate commits, and deleteGuest ran it before its own
transaction. When the guest delete failed, the guest kept its account but had
lost its plugin rows, with erasures queued for a user that still exists. The
erasure is now one transaction, and deleteGuest runs it inside the same
transaction as the budget re-split and the user delete, so it becomes a
savepoint of that write. Only DB writes and a synchronous readdir run inside.
Confirming an import created the booking, then froze the FX rate and created
the linked cost in a second commit, so a failed cost write left the booking
without its expense. REST, MCP and the plugin RPC already go through
ReservationsService.createWithCost. The import now builds the cost entry and
freezes its rate first, outside the write since that can fetch rates, then
calls createWithCost and announces the cost after the commit. A rate that
cannot be frozen keeps the booking without its cost, as before. The place a
venue becomes still commits on its own, with endpoint geocoding between it
and the booking, so confirm stays in the lint:tx baseline at 4.
The import created the booking and wrote its external_* link as two commits.
A failed link left the booking in the trip while the import reported the
flight as skipped. The flights are fetched before the loop, so nothing between
the two writes waits on the network: each flight or joined connection now goes
in with its link in one transaction, and the method is marked as one
transaction per imported booking instead of being held in the lint:tx baseline.
The rule that a multi-statement write runs in uow.transactional was
written down but nothing held it. scripts/tx-writes.mjs reads every class
under src/ with the TypeScript parser and counts, per method, the writes
it issues outside a transactional callback: repository calls that build
an INSERT, UPDATE or DELETE, calls to methods that write (on the class
itself or on a constructor-injected one), and a transaction holding
writes as one. A write in a loop counts twice, of two branches only the
larger counts, local helpers count where they are called, and a helper
whose every call sits inside a transaction is not reported. --explain
lists the line of every call that counted.
Today's 102 reported methods sit in scripts/tx-writes-baseline.json and
may only shrink; most are syncs with network I/O between their writes,
plugin lifecycle steps and adapters calling two service writes. A new
method fails CI, and an entry above what its method holds fails until
--update lowers it, the same ratchet as lint:size and lint:boundaries.
Two JSDoc tags mark the deliberate cases with their reason in the tag:
@txStandalone for a write that stands on its own whatever the caller does
(the audit row, notification sends, the booking price mirror, the
assignment journey refresh, the quota retention purge), @txIndependent
for writes complete one by one on purpose (the OAuth token retention
sweep, the notification claim and its release). Runs in CI after
lint:boundaries; documented in server/CLAUDE.md, the root CLAUDE.md and
the wiki's dev setup page.
Days, todo items, packing items and bags, packing template categories and
items, collections, collection labels and road-trip vias took their next
position as MAX(...) + 1 in one statement and inserted in the next, with
any other request free to run in between: two adds at once read the same
MAX and landed on the same position (the road-trip test pinned exactly
that duplicate). The read now sits in the transaction that inserts, and
the transaction holds the connection, so the second add waits and takes
the next position. A collection label's limit and name check move into
the same transaction, so two labels of one name added at once end in one
label and a 409.
Each site gets a test that adds several rows at once and reads back one
position each.
A sweep of the services for write sequences that committed step by step,
each one leaving half a change behind when the next statement failed:
- accommodations: the linked bookings' times moved after the stay's own
transaction had committed
- airports: the boot backfill wrote a flight's two endpoints one at a
time, and a flight left with only the first was never picked up again;
each flight is its own transaction now
- auth: the admin settings form and a password reset (retire the old
link, issue the new one) each land whole
- oauth: a refresh revoked the old pair before issuing the new one, so a
failure in between logged the client out; rotating a client secret and
revoking its tokens go together too
- oidc: the identity link, the role from the claims and the avatar of a
returning user are one write
- addons, notification preferences: a settings save lands whole
- places: create with its tags, update, and the GPX/KML imports with
their track colours
- trips: the trip row, the shifted leave entries and the rebuilt day grid
- vacay: a first visit provisions the plan in one transaction with a
second look against a concurrent first visit; updatePlan fetches the
public holidays between two transactions instead of inside its writes
- memories: a photo and its trip link, and the deletes with the orphan
sweep behind them
AddonsService, AirportsService and OauthService take the UnitOfWork for
this. Network I/O stays outside every transaction. Comments that only
narrated the migration plan make room under the lint:size entries, which
drop with them.
Uninstall deleted the plugin row and then up to fourteen tables of rows
about it one by one. A failure halfway left grants, scheduled tasks or
per-user settings behind for a later plugin that reuses the id to
inherit. The rows now go in one transaction, and the data directory is
removed after the commit instead of before the rows that describe it.
A registry install records the discovered row, its provenance, the
pinned author key and the cleared update block together, so a failure
never leaves a registered plugin without its pinned key. Starting an
OAuth connect drops the stale state and stores the fresh one together,
and a disconnect drops tokens and state together; PluginOAuthService
takes the UnitOfWork for that. The scheduler's per-task claims stay
separate on purpose and say so for the transaction gate.
createJourney inserted the journey, then its owner row, then linked each
trip with its skeleton entries and set the cover, every step committed on
its own: a failure halfway left a journey nobody owned. It now runs as
one transaction and broadcasts the linked trips after the commit.
addTripToJourney links inside a transaction the same way.
The photo paths wrapped only the gallery insert and left the entry link
and the skeleton promotion outside; the photo row, the gallery row, the
link and the promotion are one write now. removeTripFromJourney,
updateEntry, updatePhoto, setPhotoProvider, the photo deletes and the
place hooks write in one transaction each.
createEntry and uploadGalleryPhotos read MAX(sort_order) inside the
transaction that inserts, so two entries or uploads at once take one
position each. reconcileTripSkeletons and the photo capture helpers are
tagged as deliberately split for the transaction gate that follows.
A booking's create and update ran their own transaction, and the linked
cost sync ran afterwards from the REST controller and the plugin RPC
handler, so a failing cost write left the booking edited and its expense
on the old category or price. MCP linked the cost in a second call of its
own the same way.
ReservationsService.createWithCost and updateWithCost now run the
booking write and the cost sync in one transaction and hand the cost
events back, so REST, MCP and the plugin RPC call the same method and
broadcast only after the commit, in the order they did before. The cost
sync keeps swallowing its own failures the way the REST route always
has, so a cost that cannot be written still leaves the booking; MCP now
behaves like REST there instead of failing the tool after the booking
was already saved.
The service sheds its inventory markers (RS1..RS54, plan history) to
stay under its lint:size entry.
The time budget for an exec script was read after every statement, the
last one included. A single statement without args, or a script whose
closing COMMIT crossed the budget, had already run and committed, yet the
call threw a time budget error. A plugin that retries on error then wrote
the same rows a second time.
The script now reads the clock before each statement after the first, so
it still stops before the next statement once the time is up and rolls
back a transaction it opened, but a script whose statements have all run
is reported as done. query() likewise reads the clock only once another
row has come back, so a result read to its end is returned.
The ctx.db doc in the SDK said a single statement was not timed, which
contradicted the query budget. It now names what is really untimed: an
exec with bound args and migrate. README and wiki describe the script
stopping before its next statement.
The time budget covered query() and tx() only. exec() without bound args
handed the plugin's script to Database#exec, which runs any number of
statements in one call with nothing checked between them, so a long
script held the host loop as long as it liked.
An exec script now runs one statement at a time. sql-script.ts cuts it
after every semicolon outside a string, a quoted identifier or a
comment; a piece SQLite reports as incomplete input (a trigger body cut
at its inner semicolon) is joined with the next until it prepares. The
clock is read after every statement, and past the budget the call
throws. As with Database#exec, the statements before that point stay
applied, and an error is SQLite's own; a transaction the script opened
itself is rolled back so the connection is not left inside it.
exec() with bound args is one statement and stays as it was. A single
statement cannot be bounded at all on the host thread, a write such as
INSERT ... SELECT over a cross join included, and the SDK README and the
wiki now say so next to the cross-join read. migrate() keeps running
unbudgeted on purpose: a migration stopped part of the way is rolled back
and fails the same way on every later start, leaving the plugin unable to
update its schema.
The ctx.db rows of the SDK README and the plugin wiki pages describe the
budget for scripts, and the lines touched there and in plugin-sdk's db
comment lose their em dashes.
The output contract cut only the top level of an entity result. A derived
key's value went out untouched, so the children that are whole rows still
carried every column: a day's notes_items are full day_notes entities and
a reservation's endpoints are SELECT e.* on reservation_endpoints. About
fifty methods were also left as readModel and passed through whole, many
of them backed by SELECT * reads (day_accommodations, day_notes, todos,
tags, packing bags, journal entries, collab, collections, vacay plans).
A new column on any of those tables reached every plugin by default,
which is what the contract was meant to stop and what the SDK README and
the wiki already claimed it did.
PLUGIN_ENTITY_NESTED now names, per entity, the derived keys that hold
rows of another entity, and the shaper cuts those rows down to their own
field list, recursively. dayNote and reservationEndpoint are the child
tables that were whole rows; day.assignments and place.tags are held to
the assignment and tag lists too, although their reads already name
their columns. The remaining derived children (travelers, participants,
ratings, budget members, payers and receipts, packing recipients,
attachments, voters, bag members) are built field by field, which the
contract comment now states as the rule for a derived key without a
nested entity.
The whole-row read models get entity contracts: accommodation, dayNote,
todo, tag, category, packingBag, fileLink, collabNote, collabPoll,
collabMessage, journey, journalEntry, bucketItem and collection, and
vacayPlan with its holiday calendars. collections.listMine,
collections.get and vacay.mine return envelopes around rows; an entity
with a null table describes those, so their rows are reached through the
same nesting. Every list holds exactly the keys each read returned
before, so no plugin loses a field. What stays readModel is built field
by field from named columns, and each such entry says what it returns:
setBagMembers, atlas.visited, collections.savePlace and copyToTrip, the
two vacay toggles and journal.addEntryPhoto.
The schema test now covers every child and new table, and calls each
newly shaped method on the real services, comparing nested rows key for
key so the only field allowed to disappear is a withheld one. A further
case proves the nested rows it compares are not empty. gen:plugin-facts
publishes PLUGIN_ENTITY_NESTED to the SDK, and the SDK README, the plugin
wiki pages and server/CLAUDE.md describe the nested rows and list the new
entities.
A plugin's own database runs on the host's event loop, and the only bounds
were a regex denylist, the 256 MB quota and a 100,000 row cap. A select
whose every row is expensive, or a long scan under the cap, could hold the
loop for as long as it liked.
query() and tx() now get a wall-clock budget of 2 s, read between the rows
a statement yields and between the statements of a batch. Past it the call
throws, and a tx rolls back whole like any other failure inside it. The
row cap and the budget are overridable per PluginDataDb for tests; the
defaults are what production runs.
better-sqlite3 has no interrupt or progress handler, so a statement that
yields nothing until it is done, such as an aggregate over a cross join,
still runs to its first row. Only moving plugin SQL off the host thread can
bound that; the SDK README and the plugin wiki state the budget and this
limit.
The entity reads of the plugin RPC surface handed plugins whatever their
query produced, mostly SELECT t.*, with credentials stripped by a denylist
(withoutFeedToken). Every column a migration added reached every plugin
holding the read grant, and renaming one was an unversioned break of a
published API.
protocol/output-contract.ts now states, for every wire method, what its
result is: rows of one of the nine entities the SDK types (trip, place,
day, reservation, packingItem, tripFile, budgetItem, assignment, user), a
value the host builds itself, or a read model passed on as the REST route
returns it. The router applies it to each result on the way out, so an
entity row keeps exactly the fields plugins received when the list was
drawn, in the same order and with the same values, minus the trip's
feed_token (listMine, create and update used to carry it as null). The
table is keyed by every method, so a new one does not compile without an
entry.
output-contract.schema.test.ts holds the lists to the migrated schema: a
new column on a whole-row table fails until it is published or withheld,
and each entity method is called on the real services to prove the
contract drops nothing but the withheld fields.
gen:plugin-facts publishes the lists to the SDK as PLUGIN_ENTITY_FIELDS
and PLUGIN_METHOD_RESULT, and the SDK's entity interfaces are type-checked
against them. Place.day_id, which no read ever delivered, is deprecated
rather than removed. KNOWN_ADDONS is generated from ADDON_IDS as well,
which adds the dawarich, roadtrip and tours ids the hand-kept copy lacked.
The SDK goes to 1.9.0, and its README and the plugin wiki pages describe
what a ctx call returns.
A handler behind express.json() is reached from the request stream's 'end'
event, long after the correlation middleware returned, so it looked as if
every write lost its id. It does not: raw-body and multer bind their
callbacks with AsyncResource at the point they are called, which is inside
the request's correlation. Nothing tested that, though, because every
REQID case was a GET. REQID-007 posts a JSON body and REQID-008 one that
arrives in several chunks, and both expect the handler to see the id the
response carries. The middleware comment now says why this holds.
The ownership rule says a boot-stable value that only Nest providers read
belongs to a registerAs token, yet TransitService froze TRANSIT_API_URL in a
module-top readEnv() const and KitineraryExtractorService read
KITINERARY_EXTRACTOR_PATH and PATH through readEnv() in its probe. Both now
come from tokens: transitConfig and kitineraryConfig, derived in
boot-derive.ts and dropped from deriveIntegrations, so readEnv() no longer
reads those variables at all. TransitModule and KitineraryExtractorModule
import AppConfigModule, which keeps them buildable on their own in tests.
Behaviour is unchanged: the same defaults, the same trailing-slash strip and
PATH split, now frozen per built app instead of at import. The app-config
README names each boot-stable variable that stays on readEnv() together with
the reader the container cannot inject into.
The env reference parity test forced TREK_MANAGED, MAPBOX_ACCESS_TOKEN,
CARTO_API_KEY and PLACES_API_BASE into server/.env.example with how-to text.
They are managed hosting only and stay out of the public docs, the same way
TREK_DB_FILE and APP_VERSION do, so they move to NOT_OPERATOR_SETTINGS with
that reason and their lines leave the self-hoster reference again. The
variables themselves and their behaviour are unchanged.
The 65 s default sat above nginx's 60 s but below the 90 s Traefik keeps an
idle upstream connection, so a default Traefik deployment could still reuse a
socket Node had just closed and answer that request with a 502. 95 s clears
both, and the headers timeout one second above it stays well under Node's
300 s request timeout and the 290000 cap the schema accepts.
Also moves the stray RawEnv import in boot-derive.ts up to the other import.
The process-state ports were meant to let a store shared between
processes replace the in-memory one, but they could not. The readers
outside the container (the broadcast functions, the backup restore's
permissions-cache flush, the pending-code sweep, hand-built consumers)
went straight to the module-level in-memory instances, so an overridden
provider was used by the gateway and the services while the broadcasts
and the restore flush kept using the old store. And every port was
synchronous, while a store in the database or on the network is async,
so plugging one in would have changed every signature anyway.
RateLimitStore, PendingCodeStore, OidcFlowStore and
PermissionsCacheStore now return promises, and RateLimitService,
OauthService, OidcService, PermissionsService and their callers await
them. The in-memory stores still decide synchronously, so who gets a
429, a code or a cached level is unchanged. RoomRegistry stays
synchronous: membership holds live sockets and broadcast() is a
synchronous call every domain makes, so several processes need a
fan-out in front of broadcast(), not a different place for the sets.
The process-wide ports get a ProcessStoreSlot. PermissionsModule,
OauthModule and RealtimeGatewayModule install the store the container
resolved and release it on shutdown, and the free functions, the sweep
and hand-built consumers read the slot, so overriding the provider
changes every reader. Tests provide fake stores and check that the
broadcasts, the restore flush and the sweep use them.
Two callers needed more than an await. The transit MCP tools counted in
a RateLimitService of their own that no provider could reach; they now
take the one RateLimitModule provides (their mcp_transit_* buckets never
overlap the REST ones). The link preview claimed its in-flight slot only
after the budget check, which now yields, so simultaneous askers would
each have been charged; it now claims the slot first, and an asker that
joined someone out of budget asks again on its own budget.
The unused putPendingCode/takePendingCode helpers are gone.
The data-directory change said Nest classes inject DataPathsService, but
only AdminService did. StorageRegistryService still read the built-in
uploads and backups roots and the scratch dir from module constants
frozen at import, and StorageJobsService read the scratch dir the same
way.
The registry now takes DataPathsService through its constructor (a
hand-built one resolves the same layout), and the jobs service asks the
registry for its temp dir. The plain helper modules keep calling
resolveDataPaths() because they run without a container or are shared
with code that does: config.ts, the database path, index.ts, the logger,
the backup functions and the plugin trees, which the out-of-process
supervisor reads too. The app-config README now lists both sides, and
the nest README, server/CLAUDE.md and the provider's own comment say the
same.
The keep-alive change moved HTTP_KEEP_ALIVE_TIMEOUT_MS onto the httpConfig
token, but the app-config README still listed only TRUST_PROXY and
HSTS_INCLUDE_SUBDOMAINS for it and left the variable out of the
boot-stable list. Also drops the dashes from two lines the recent
scheduling and entity changes rewrote.
A tick whose lease another process held returned from runLeased without
running, and traceEntry still wrote `cron <name> ok`, so both processes
logged the tick as run and the per-tick line could not tell which
replica did the work. runLeased now says whether the tick ran, and
traceEntry takes a `skipped` option that logs such a call at debug as
`cron <name> skipped <ms>: lease held by another process`.
Taking the lease is two statements (make sure the row exists, then the
conditional UPDATE that decides the race), so it now runs inside
UnitOfWork.transactional like every other multi-statement write.
Rate-limit counters, OAuth pending codes, OIDC login states and codes,
the permissions cache and the WebSocket rooms were Maps on module or
instance scope. That is right while TREK runs as one process, but nothing
could be swapped without editing every consumer.
Each now sits behind a port with an in-memory implementation that keeps
the old behaviour exactly: RateLimitStore, PendingCodeStore,
OidcFlowStore, PermissionsCacheStore and RoomRegistry. The owning module
provides the implementation with the same lifetime the state had
(process-wide for the pending codes, the permissions cache and the rooms,
per instance for the rate limits and the OIDC flows), the consumers take
it through their constructor, and a hand-built consumer falls back to the
same in-memory store, so tests and the free-function paths (the backup
restore's cache flush, the broadcasts) see what they saw before. A store
shared between processes can replace a provider later without touching
the services.
Three small gaps in how the process behaves at its edges:
- An unhandled rejection or uncaught exception crashed the process with
the stack on stderr only. index.ts now logs it to trek.log and runs the
same orderly shutdown a SIGTERM gets, exiting with 1; a second fatal
error during that shutdown exits at once.
- /api/health/ready kept answering 200 while a SIGTERM was draining the
process, so an orchestrator sent traffic for another probe period. A
shutdown now marks the new ReadinessService first (and Nest's own
shutdown hook marks it too), and the probe answers 503 from then on.
- Node closes an idle keep-alive connection after 5 seconds while reverse
proxies keep upstream connections for about a minute, which shows up as
intermittent 502s. The server now keeps them for 65 seconds, configurable
through HTTP_KEEP_ALIVE_TIMEOUT_MS (owned by the httpConfig token), with
the headers timeout one second above.
waitForCompletion only kept a tick from overlapping itself inside one
process. Two processes on the same database (a rolling update's overlap,
a second replica, a dev container pointed at the same file) both ran
every job: two auto-backups, double syncs and retention sweeps, and
reminder runs racing each other.
A new migration adds scheduler_leases, and CronRegistrarService now runs
every registered tick through SchedulerLeasesRepository.acquire: one
conditional UPDATE moves the job's row to this process when it is free
(no holder, a lapsed holder, or this process already), so exactly one
process wins. A running tick renews its lease every 20 seconds, a crashed
holder frees the job after a minute, and a finished tick keeps the lease
for 30 seconds so a peer whose timer fires late does not run the same
tick again. The per-row reminder claims stay as they are; jobs need no
change of their own.
A 5xx stack and its access-log line were two unrelated lines, the
exception filter printed the stack with console.error and no request
context, and the MCP, WebSocket, plugin RPC and cron entry points logged
nothing per call. A self-hosted bug report could not be traced from a
failure back to the call that caused it.
Every HTTP request now runs under an id held in AsyncLocalStorage: a
well-formed incoming X-Request-Id is kept, anything else is replaced by a
UUID, and the response carries it in X-Request-Id. The logger tags every
line written inside a unit of work with `[http <id>]`. The exception
filter leaves a server-side failure on res.locals and the access log
writes the request line and the stack as one entry under that id; without
an access log watching the response it still prints the error itself.
MCP tool, resource and prompt calls (through a new `around` seam on the
registry), WebSocket messages, plugin RPC dispatches and cron ticks each
run through traceEntry(), which gives the call its own id (remembering the
outer one as its parent) and logs one line for it: debug on success, like
the HTTP access log, and warn or error with the reason on failure. The
call's result and timing are unchanged, a failed cron tick is logged once
rather than twice, and a WebSocket handler rejection no longer reaches
rxjs's asynchronous rethrow.
Every log line ran existsSync, statSync and appendFileSync on the event
loop before it returned. On a slow data volume (an SMB or NFS bind mount
on a NAS) that stalled every request that logged, and rotation renamed
files without checking whether another process sharing data/logs had
rotated already.
The file half of the logger is now BufferedLogFile: a line is queued in
memory and the queue goes out with one appendFile per batch, through a
single serial chain, so no append runs while files are being renamed. The
files and the rotation are the same as before (trek.log rotates to .1 at
10 MB, five files in total), but the size is re-read before a rotation
and a file another process already rotated is left alone. A stalled disk
costs dropped lines with a count instead of unbounded memory, a failed
write is reported instead of thrown, and index.ts writes whatever is
still queued synchronously when the process exits.
Eight registerAs tokens existed while every consumer read the same values
through readEnv() instead, so the boot-stable layer the repo rules describe
was unused and nothing said which path owned a variable. FORCE_HTTPS, for
instance, was frozen by the middleware through the http token and read live
by the cookie code at the same time.
A variable is now owned by exactly one side. TRUST_PROXY and
HSTS_INCLUDE_SUBDOMAINS belong to httpConfig, which buildApp() already
injects into the pre-init middleware, and TREK_PLACE_PHOTO_DIR to the new
storageConfig, which the storage registry now injects instead of reading it
through RuntimeEnvService. Their derive functions moved to boot-derive.ts,
which deriveAll() never calls. The six tokens nothing injected are gone:
their values are needed before the container exists (config.ts, the
database, src/mcp, the SSRF guard) and stay on readEnv().
config-ownership.test.ts records the keys each side reads through a Proxy
and fails on a variable read through both, on a token nothing injects and
on a variable the schema does not validate. env-reference.test.ts holds
server/.env.example to the schema, which had drifted by 34 variables (SMTP,
WebAuthn, plugins, MCP session tuning, TREK_MANAGED and others); they are
documented now. jwt-verify.ts no longer cites a middleware file that is
gone.
The data and uploads directories were derived independently in about ten
files, each counting `..` hops from its own `__dirname`, and the file
logger used `process.cwd()` instead. Moving a file or starting the server
from another directory quietly moved where secrets, backups or logs went,
and the backup code named `data/travel.db` directly, so with TREK_DB_FILE
set a backup archived and a restore replaced a file the connection never
opened.
`resolveDataPaths()` in app-config now anchors data/, uploads/, backups,
tmp, logs, the JWT secret and encryption key files and the database file
(TREK_DB_FILE when set, else data/travel.db) on the server package root.
Code that runs before the container calls it directly; Nest classes inject
the new `DataPathsService`. Every previous anchor uses it, and the resolved
paths are unchanged for the image and for `npm run dev`.
The Postgres probe only ratcheted the methods it managed to call. A method
it could not build arguments for, one that threw before its first
statement and one that timed out were counted and logged but held to
nothing, so a new repository method with a callback parameter could land
SQLite-only SQL unseen, and the part of the code the probe reaches could
shrink without anyone noticing.
pg-probe-baseline.json gains an `uncovered` list of those methods. A
method outside it that goes unmeasured fails the run, and so does an
entry whose method is measured now or gone, so the list only shrinks.
It is seeded from the same first measurement as `failing`; a baseline
without the key reads as unmeasured.
The failing count also counted every refusal, including data errors the
sample values cause (a "probe" string in a date column, SQLSTATE 22007),
so renaming a parameter could fail the gate on a change that never
touched the SQL. Only classes 42 and 0A count now; the rest is reported
in the log and the step summary without being held.
The docs said the probe measures every repository statement. They now
say it measures what each method reaches with sample arguments against
empty tables, and that the methods it cannot reach are held to the list.
The probe shipped with scripts/pg-probe-baseline.json at `failing: null`,
and an unmeasured baseline only reported. Every repository statement
Postgres refused, new SQLite-only SQL included, still passed the job;
only the dialect helper cases could fail it, so the ratchet was off until
some later commit seeded the file.
An unmeasured baseline now fails the run outside --update, and the log,
step summary and annotation say which artifact file to commit. The job
stays red until the first measurement lands, so the gate cannot reach dev
holding nothing. A run that called no repository method (an unreadable
tsconfig, a moved repositories directory) fails as well, also under
--update, so a broken plan can neither pass nor seed an empty baseline.
Nothing in CI ran a statement against anything but SQLite, so the claim
that the dialect layer keeps the repositories portable was never measured
and new SQLite-only SQL could land unseen. The new postgres-probe job
(server/scripts/pg-probe.ts, npm run probe:pg) does that against a
postgres:17 service container:
- it derives a Postgres schema from the entities with MikroORM's schema
generator, keeping TEXT timestamps, mapping the SQLite-only defaults and
the NOCASE collation, and leaving foreign keys out;
- it runs every dialect helper on an in-memory SQLite and on Postgres
against the same expectation, and every case has to pass;
- it calls every public repository method once with arguments the
TypeScript checker builds from the parameter types, through a Postgres
driver that runs each statement alone in a rolled-back transaction with
a statement timeout and records whether Postgres accepted it.
The statements Postgres refuses are held per method to
scripts/pg-probe-baseline.json, which may only shrink. The baseline starts
unmeasured (failing: null): the job reports, passes and uploads the first
measurement as the pg-probe-baseline artifact, together with a JSON report,
so the commit that seeds it and every later commit that lowers it needs no
local Postgres. The probe refuses a database holding tables it did not
create.
@mikro-orm/postgresql joins the server's devDependencies at the 7.2.1 the
other MikroORM packages resolve to; the image's production install leaves
it out. The job needs the preflight gates like every test job, runs only
for the code area, stops after ten minutes and is part of ci-ok.
The dialect layer promised that a second engine would be a branch in
src/db/dialect/ and no sweep through the repositories, but every helper
only knew SQLite and threw for anything else. Each one now has a Postgres
branch, and the helpers whose text both engines read alike (COALESCE,
LOWER, TRIM, substr, ||, the aggregates, the CASE shapes) share one
spelling behind isKnownPlatform.
The Postgres branches keep the storage model SQLite has: the clock helpers
render UTC as the same YYYY-MM-DD HH:MM:SS text (to_char over CURRENT_TIMESTAMP
AT TIME ZONE 'UTC'), GLOB becomes an anchored regular expression, the integer
cast truncates through numeric so '14.0' still reads as 14, JULIANDAY
distance comes from the epoch difference, SQLite's null-safe IS becomes IS
NOT DISTINCT FROM, and NOCASE names a nondeterministic ICU collation a
Postgres schema creates. Bound values on the Kysely side are cast to text
where pg would otherwise send an untyped parameter the planner cannot
resolve.
The SQLite branches are untouched, so nothing changes for a running
install. The Postgres text is pinned in sql-functions.postgres.test.ts.
sql-functions.ts sat at 1007 lines, held there by the size baseline, so it
had no room for the Postgres branch every helper is meant to grow. The
Kysely-expression twins are a separate family already (a Kysely Expression
and a MikroORM RawQueryFragment do not convert into each other), so they
move to src/db/dialect/kysely-functions.ts, and the platform check plus the
fail-closed unsupported() move to platform.ts, which both files share.
No helper changes behaviour or spelling. lint:dialect now skips the whole
src/db/dialect/ directory instead of the one file, and the size baseline
drops its sql-functions.ts entry.
The datetime rule only matched a literal holding a call such as
datetime('now'). The query builder spells the same function as
eb.fn('datetime', [...]), where the literal is just the name, so that form
slipped through while strftime, json_extract and glob are caught bare. A
literal that is exactly datetime, julianday or unixepoch now counts as well,
the same way the glob rule treats a bare glob. The only such call lives in the
dialect layer, which is not scanned, so the baseline does not change.
seedDocumentProviders() had no caller left: the migration that creates the
tables and DocumentProviderSeeder both import only the DOCUMENT_PROVIDERS and
DOCUMENT_PROVIDER_FIELDS constants. Its two INSERT OR IGNORE statements were
the only SQLite-only spellings in the file and had been baselined rather than
removed. Deleting it lets the file leave the dialect baseline (13 hits down to
11) and the better-sqlite3 import allowlist in eslint.config.mjs, since nothing
in it touches the driver any more. DOCUMENT_PROVIDER_SEED_IDS goes with it, it
had no importer either.
The repositories already lean on the query builder, the generated table
types and sql-functions.ts, but the rules behind that were only implied.
server/CLAUDE.md now states them: which builder call replaces each SQLite
spelling, which helper covers time, strings and casts, where PRAGMA may live,
how a new migration is treated, and that the generated types describe the
SQLite storage model.
It also says plainly that MySQL and MariaDB are out of scope. The
repositories rely on INSERT ... ON CONFLICT and RETURNING, which MySQL lacks,
so Postgres is the second engine the code is prepared for.
ESLint keeps raw SQL out of the repositories, but nothing looked at the rest
of src/: the entities, the domain services, the src/db lifecycle files or a
migration added from now on. scripts/dialect-lint.mjs (npm run lint:dialect)
reads every string and template literal under src/, skipping comments, and
counts INSERT OR IGNORE/REPLACE, datetime(), strftime, json_extract, GLOB,
last_insert_rowid, AUTOINCREMENT, PRAGMA outside the lifecycle files, ||
concatenation in SQL and .insertId reads. scripts/dialect-baseline.json
holds today's thirteen hits; the counts may only go down, like lint:size.
The dialect layer is not scanned, and neither are the migrations that have
shipped.
The four seeders that still spelled INSERT OR IGNORE (addons, photo and
document providers, the schema_version marker) are Kysely inserts that do
nothing on conflict now, so a re-run still leaves an operator's rows alone.
They leave the SQL-string exemption list in eslint.config.mjs and their pins
in the selector test.
Kysely fills InsertResult.insertId from SQLite's last_insert_rowid. On
Postgres it is undefined, so the sixteen repository inserts that returned
Number(result.insertId) would hand NaN ids to their services without an
error. They now add .returning('id') and read the id from the row, which
SQLite (3.35 and later, bundled by better-sqlite3) and Postgres both answer
with the same value.
DemoRepository ran its example-trip seed as nine SQL strings through
connection.execute, including INSERT OR IGNORE and a bare CURRENT_TIMESTAMP.
Those are Kysely inserts on the generated table types now: ON CONFLICT
(trip_id, user_id) DO NOTHING for the membership row, the dialect helper for
the timestamp, RETURNING id where the caller needs it. With no SQL text left
it drops off the ESLint exemption list and its pin.
ESLint now refuses any read of insertId in src/db/repositories and
src/db/seeders, so the pattern cannot come back.
The repositories typed their Kysely statements against 159 table interfaces
written by hand, one or more per repository. Nothing compared them with the
schema: a migration that renamed a column or changed its type left them
stale while tsc stayed green, and several had already drifted
(reservations.accommodation_id is TEXT, not INTEGER).
scripts/generate-db-types.ts migrates a throwaway database with the real
migrations, reads every table back and writes one interface per table into
src/db/kysely/ plus the DB map in db.ts. Every hand-written interface is now
Pick<DB, ...> of that type, and npm run check:db-types fails when the
committed files no longer match the migrations.
A column an insert may leave out (rowid, DEFAULT, nullable) is
InsertOptional<T>, declared as ColumnType<T, T | void, T>. The server
compiles without strictNullChecks, where Kysely's own Generated<T> still
demands the column; that is why the old code kept separate insert-only
interfaces. The ids callers pass as number | string are cast to the column
type at the bind, so every statement binds the same value as before.
Version 6 only adds the driving settings store; dropping the area cache
is version 7. The case name claimed both, so it now names only the new
store and a comment explains where the empty area cache it still checks
comes from.
The skip of newer-format queued writes only exists from this release on,
so a rollback to an older release still replays them. On a rolled-back
install the skipped rows also stay pending with no retry or discard path
and hold back later writes to their entity, leaving Clear cache as the
only way out. Say both in the wiki and in client/CLAUDE.md instead of
promising that a rollback is always safe.
Each version() of the offline database runs once, on a user's device, when a
new bundle opens what an older one wrote. None of them ran in CI: the v3 blob
cache backfill and the v8 area-fingerprint reset had never been exercised
against an older database.
The new test builds the database as each previous version left it in
fake-indexeddb, seeds a row into every table it had, opens it with the current
schema and checks that the rows survived and that the version did what it is
for. A guard reads the declared versions from offlineDb.ts and fails on a
duplicate, on a version without an upgrade case, or on stores the history
fixture does not reproduce. TrekOfflineDb is exported so the test can open a
named database with it.
A queued write outlives the bundle that wrote it. After a deploy a tab can
still run the previous bundle, and an install can be rolled back, so a row
written by a newer build may be replayed by an older one that does not know
its shape.
Each row enqueue() writes now carries schemaVersion (MUTATION_SCHEMA_VERSION)
and the UI build that queued it. Both fields are optional, so rows queued
before this read as format 1 and replay as before. A flush leaves a row from a
newer format in the queue without sending, marking or dropping it, and holds
back the later writes to the same entity as it does behind a parked write, so
they keep their order until a build that knows the format replays them.
The demo seed runs in the schema bootstrap, at boot and again after every
reopen. Since the baseline save moved out of the seed into DemoResetJob,
only the boot read the seed's mark. A restore onto a demo instance with no
baseline, whose archive held no admin trips, seeded the example trips
during the reopen and then saved nothing, so the hourly reset kept logging
"No baseline found"; the mark also stayed set for the rest of the process.
Before the move, that path saved the baseline from inside the seed.
DatabaseLifecycle now runs listeners after a reopen's schema bootstrap,
and DemoResetJob registers one in demo mode that does the same check as at
boot: save through the backup port when the seed just ran and no baseline
exists. Each check clears the mark. A failed save is logged, never turned
into a failed reopen. AdminModule imports DatabaseLifecycleModule from the
shared database kernel for it.
The DB paragraph said MaintenanceRepository is never built with `new` in
src/, but MaintenanceModule's factory does exactly that. Say what is
actually true: the module's factory is the one place that builds it, and
every consumer injects it.
DatabaseBackupModule stays @Global, but its comment gave a vague reason.
The real one is concrete: the backup domain already reaches admin through
auth and settings, so importing the module from AdminModule would add a
domain cycle that lint:boundaries rejects. The module comment now says
that and points at the e2e harness that imports it by hand, and
AdminModule notes where its DATABASE_BACKUP comes from.
SqliteDatabaseBackup.replace reopens the connection in a finally block, so
it runs even when the swap throws. A reopen that failed there was stored
and then dropped: the swap's exception propagated past the return that
would have carried it. The demo reset caught that swap error, logged
"Failed to restore baseline" and returned normally, leaving the instance
without a database connection and nothing in the log to say so. Before
the port, the same branch awaited reinitialize() and its rejection reached
the tick's log line.
replace now throws a DatabaseConnectionLostError when both steps fail. It
carries the swap error and has the reopen failure as its cause, and its
message tells the operator to restart. The demo reset lets that error
through to DemoResetJob's tick, which logs it, and still logs and skips a
failed swap whose reopen succeeded. A restore that hits it answers the
same 500 as any other thrown restore error, with both failures in the
server log.
Moving the first baseline save out of the boot-time seed into DemoResetJob
also changed when it happened. The seed saved it only on the boot that put
the example trips in; a later boot returned early and saved nothing, so an
instance with data but no baseline logged "No baseline found, skipping" on
every hourly reset. The job instead saved one on any demo boot that found
no travel-baseline.db, freezing whatever the database held at that moment,
user edits included, as the target the hourly reset rolls back to.
The seed now leaves a mark when it seeds the example trips, and the job
saves the first baseline only when that mark is set and no baseline
exists, which is the old trigger. As before, the save does not depend on
the cron registrar being enabled. The demo mode wiki page says so.
Backup, restore and the demo reset each handled the SQLite file
themselves: backup.impl closed and reinitialized the connection, built
MaintenanceRepository with new, and assumed the database was
data/travel.db. That tied the whole chain to one local file and hid two
bugs.
They now go through DATABASE_BACKUP, a port declared in the shared
database kernel (snapshot, verify, keep a copy before a restore, replace).
SqliteDatabaseBackup implements it with what the code did before:
VACUUM INTO for a consistent copy, and a close, copy, sidecar drop,
rename and reopen through DatabaseLifecycle for a restore. BackupService
injects it and hands it to backup.impl. The demo reset and the baseline
save use the same port; the first baseline is saved by DemoResetJob once
the app is up, because the boot-time seed runs before anything can be
injected. Another engine can plug in a logical dump behind the port.
Fixes on the way:
- The backup and the restore ignored TREK_DB_FILE. The backup looked for
data/travel.db and, with the database elsewhere, silently left it out
of the archive; the restore wrote data/travel.db and never replaced
the live database. Both now use DatabaseLifecycle.file, the path the
connection runs on, and the pre-restore copy sits next to it.
- A failed VACUUM INTO fell back to archiving the live WAL-mode file
without a word, the torn copy the snapshot exists to prevent. The
backup now fails: nothing is archived or committed, the existing
backups are untouched, a partial snapshot is removed, and the log says
why. A missing database file and a failed checkpoint are logged too.
MaintenanceRepository was built with new over an injected EntityManager
in the readiness probe and in UserCleanupService, outside the container.
MaintenanceModule now provides it, as a factory over the same
context-resolving EntityManager Nest injects everywhere, and both inject
it. The class itself stays free of Nest decorators, like the rest of
src/db, so tests keep building one over their own EntityManager.
No behaviour changes: the probe still pings and the erasure still runs
inside the account deletion's transaction.
db/database.ts opened the connection as a side effect of being imported,
so the moment anything reached it the database was open. index.ts had to
defer loading the app until a first-boot restore was done, and nothing
in the container decided when the connection opened or closed.
The module now opens the connection on first use and opens nothing at
import. A connection closed on purpose (a restore, the shutdown) stays
closed until it is reopened, as before. The new DatabaseLifecycle
provider owns the decisions: buildApp() opens through it, which also
binds the ORM to later swaps (the hook attachOrm used to register) and
runs the schema bootstrap in the same order as before, and the shutdown
closes through it. Its file property is the one resolved database path.
The module keeps every export, so the test suites that replace it with
vi.mock keep working; buildDbMock gains the new openDb.
_flushRun holds the running pass or null. Testing it as a boolean is
correct but reads like a forgotten await, and SonarCloud counts it as a
reliability bug (S6544). The explicit comparison says what is meant.
en now reads "Upload {n} photos", but ru, uk, fr and sv still said the
photos had already been uploaded, and the new ru and uk plural forms
copied that wording. ca, es and it used a status noun ("upload of",
"loading of") instead of the action, and ko put the counter in front of
the noun. All locales now name the same action, with each plural form
matching its Intl.PluralRules category.
NotificationsService.send() now reports how many deliveries it attempted
and how many went out, so the reminder jobs can tell a failed send from a
sent one. WPDISP-008 and WPDISP-011 asserted that the send resolved to
undefined, which only meant "no error" and broke on the new result. They
now pin the counts: the in-app row and the webhook go out, and a push the
broken key pair cannot send is skipped rather than counted as failed.
CI lints the client with lint:warnings and runs the size, boundary, layering,
offline, skip, format, i18n key and open contract checks, but the dev setup
page still called the client lint:check the CI name and its pre-push list
stopped at typecheck and lint. Add the new scripts to the command tables and
spell out the server, client and shared quality steps from test.yml, so a
contributor who follows the page passes the same gates CI does.
A run Dependabot triggers only sees Dependabot secrets, and its PRs come
from a branch in this repo, so the fork guard let s3-contract start with an
empty MINIO_LICENSE_KEY. AIStor then never became healthy and ci-ok went red
on every grouped bump PR. s3-contract and scan now also skip when Dependabot
is the actor; ci-ok already counts a skip as passed.
changed-areas skipped every gate for a change that touched only Markdown
outside wiki/. server/tests/fixtures/wiki/ is Markdown too, but wiki.test.ts
parses it, so a PR that only edited those fixtures passed ci-ok without the
test that reads them ever running.
The image tag rows already say the next release is 5.0.0 and that 4.x gets
no minor tag, while the rollback, email case and APP_VERSION notes still
pointed at a 4.4 that will never be published. Use 5.0 throughout so the
pages agree with each other and with docker.yml.
Both server ratchets only logged an entry the tree had moved past. A
file that shrank could grow back to its old size entry unseen, a
deleted file left its allowance to the next file at its path, and a
boundary violation that was fixed stayed allowed until someone ran
--update. The client ratchets now fail on such entries, so the server
ones do the same.
lint:size fails on an entry above its file's size or for a file that is
gone, and lint:boundaries on an entry that no longer occurs; both tell
to run --update, which still only lowers or drops. Both baselines match
the tree today.
lint:size, lint:layers, lint:rtl (both baselines), lint:skips and
lint:warnings only failed a count that passed its entry. A count that
fell below its entry was logged and nothing more, so a file that shrank
could grow back to its old entry unseen, and the entry of a deleted
file stayed behind for the next file at that path. lint:format and
lint:offline already fail on such a stale entry, and the docs said
these counts may only shrink.
Each of them now fails on an entry that lowerCounts would lower or
drop, naming its count or saying the file is gone, and tells to run
--update, which still only lowers. staleCounts and reportStale in
lib/ratchet.mjs hold the shared part. Every baseline matches the tree
today, so none needed lowering. The client CLAUDE.md says so, and that
CI lints through lint:warnings rather than lint:check.
The planner hook tests spy on the store's actions for the visit times,
and the note now goes through a store action as well. Asserting on that
action instead of on the raw note route keeps these tests about the hook
and keeps the file inside its lint:size entry, which the response stub
the route needed had pushed three lines past.
flush() returned at once when this tab already had a flush going, and
that includes one still waiting for another tab's Web Lock. The socket's
reconnect hook chains a refetch onto flush(), so a reconnect landing in
that window rehydrated the trip while a replay was still on its way, the
race the lock wait was meant to close. sendSoon and the online, periodic
and visibility triggers make such an overlap common.
flush() now keeps the running flush as a promise. A second call still
asks it for one more pass and then gets that same promise back, so it
resolves only once every pass, including the one it asked for, is done.
The pass itself moves to _flushPass, and the repeat runs as a loop
instead of a nested flush() call. The client CLAUDE.md sync bullet now
says so, and loses its dash.
The place dialog stopped reloading the days after a save, so the new
visit note only reached the Zustand store. Nothing wrote it to Dexie,
the tab never sees its own assignment:updated echo, and the time save
before it cached the server row with the old note. A trip reopened
offline then showed the old note until the next full sync.
The note now goes through assignmentRepo.setNotes, which caches the
saved row on its day online and queues the write behind any other
write of the same visit offline, like setTimes does. A new store
action applies it, so the page hook no longer patches the store by
hand. usePlaceEdits drops its direct API import and leaves the
offline baseline.
Each gate came with its npm script but ran nowhere. The server job now
runs lint:size, lint:boundaries and shared's contracts:open after the
shared build; the client job runs lint:warnings in place of lint:check
(it fails on errors the same way and also holds the warning count per
rule), then lint:format, lint:layers, lint:offline, lint:skips and
lint:i18n-keys. All of them are steps of jobs ci-ok already needs.
Both size baselines are taken on the merged branch. Four server files
grew while the other changes landed (the claimed reminder, the email
uniqueness error and the generator rule for an expression index) and
four client files by two or three lines with the plural forms; one
client file shrank and is lowered.
The image tag tables showed mauriceboe/trek:4.3 as the minor tag example,
"published for releases after 4.3.3". The next release is 5.0.0, so no
4.3 tag will ever be pushed, and a self-hoster copying the example into
compose gets "manifest unknown". The example is now 5.0, and the row says
4.x releases have no minor tag and are pinned by major or full version.
Both release workflows refused a commit if any ci-ok check run on it was
not green. Every workflow run is its own check suite, so a red run stays
on the commit after a later one passed: a flaky job on the dev push, or a
push run cancelled by a manual run on the same ref (ci-ok runs with
always() and fails on the cancelled jobs). Running Tests & SonarQube
again, as the error told the maintainer to do, added a green suite but
never cleared the red one, and the message did not say which run was
red.
The gate now lives in scripts/ci/ci-ok-verdict.mjs. It fetches every
attempt (filter=all), keeps the newest attempt per check suite, refuses
while one is still going and requires the run that finished last to be
green. It prints the link of every run it counted, and an older red run
it overrode shows up as a notice. A response it cannot read, one cut off
by pagination or a run without a suite, conclusion or completion time
fails the release instead of passing it. Tests run with test:ci-scripts.
The changes job listed the diff with rename detection on, so a file moved
out of an area showed up only under its new path. Moving server code or a
chart template under docs/ turned every area off and ci-ok passed on a
change that removed source from the build. The diff now runs with
--no-renames, and a test reads the exact command out of test.yml and runs
it against a scratch repository with a move in it.
The code area also missed files the test suites read from disk outside
the packages. config-templates.test.ts pins docker-compose.yml and the
Helm chart to the code defaults, help.test.ts serves wiki pages and
images, and the client's help registry test checks its links against the
wiki headings. A chart, compose or wiki change now runs the tests, and a
wiki change also builds the image, which ships wiki/ as the in-app help.
Markdown is still ignored everywhere except under wiki/.
docker.yml built and published whatever `main` pointed to when the build
and merge jobs started, not the commit it had just bumped and tagged, so
anything that landed in between shipped under a version that did not
contain it. build, merge and release-helm now check out the bump commit,
the way docker-dev.yml already pinned its prerelease builds. The push of
that commit fast-forwards main from the commit the run started on, so a
race fails the release instead of skipping over it.
Nothing tied a release to a green CI run. Both release workflows now
refuse a commit unless every ci-ok check run on it has finished green,
and docker.yml refuses to run on anything but main.
bump=auto fell back to a patch whenever no prerelease was in flight,
which would have published the ORM release, the largest change the
project has made, as 4.3.4. auto now only finalizes a prerelease above
the last stable tag and otherwise stops, asking for patch, minor or
major. The inputs are unchanged. A patch that adds migrations gets a
warning, not a refusal, since patch releases have carried them before.
Stable releases also publish :MAJOR.MINOR, so an install can take the
fixes of one minor line without being moved onto the next on day one.
The tag tables in the wiki list it and the prerelease tag, and the root
CLAUDE.md describes the release and prerelease flow.
security.yml, lint-prettier.yml and the release workflows still ran
actions built for Node 20 (checkout and setup-node v4, github-script v7,
the docker actions v3/v5, create-github-app-token v1), which GitHub now
forces onto Node 24 with a deprecation warning on every run. They move
to the Node 24 majors test.yml already uses (download-artifact to v7,
since v6 is still Node 20). Third-party actions, several of which hold
the Docker Hub and GitHub App secrets, are pinned by commit with the tag
in a comment. ubuntu-latest moves to a new Ubuntu release on 10-19, so
every job names ubuntu-24.04 and the switch happens in a PR. Every
setup-node reads .nvmrc; publish-plugin-sdk leaves 22, which it ran
under a comment claiming it matched the server.
security.yml gains an npm audit job over the lockfile, on PRs and pushes
into dev as well as main. It covers what the image installs and the
plugin SDK, takes seconds, and would have flagged the brace-expansion
regression the ORM squash brought in the day it landed. Scout keeps its
image scan on main. Neither becomes a required check: a CVE published
overnight would turn every open PR red, including the ones that cannot
fix it. The scout job now also skips PRs into dev, which the trigger
change would otherwise hand it.
.github/dependabot.yml adds weekly version updates into dev for the root
workspace, plugin-sdk, the actions and the Docker base images, minor and
patch grouped per ecosystem, majors one PR each. The Node major in the
Dockerfile is left to a hand-made change together with .nvmrc.
The ruleset required only `lint` (shared's eslint and prettier) and
`check-target`. Types, lint, coverage, end-to-end and the S3 contract
were advisory, and the three health PRs merged with red checks. Nothing
on a PR into dev built the Dockerfile or rendered the Helm chart either,
which is how #2610 shipped a Dockerfile no build could parse.
ci-ok needs every gate, test, image and chart job and is meant to be the
one check the ruleset requires. It runs with always(), fails when a job
it needs failed or was cancelled, and passes when they passed or were
skipped. The Sonar scan stays outside: its quality gate reports on the
PR by itself, and an outage there should not block merges.
A required check cannot sit behind a workflow-level path filter: on a
docs-only PR the workflow never starts and the check never reports. The
filter moves into a `changes` job. It diffs the PR's merge commit against
its base (or a push against the previous head) and scripts/ci/
changed-areas.mjs maps the files to code, image and deploy areas. It
fails closed: no usable diff, an empty one or a path it cannot read
cleanly runs everything. Its tests run in the same job.
docker-image builds the Dockerfile with buildx (layer cache written by
pushes only, read by PRs), starts it with the same hardening as
docker-compose.yml, waits up to three minutes for /api/health/ready,
checks the VERSION file and that / serves the client, and prints the
container log. deploy-manifests lints the chart, renders it with the
default values and with each secret, persistence and ingress mode,
validates the output with kubeconform, and runs `docker compose config`
on every compose file. helm and kubeconform are pinned by version and
sha256.
While the file is open: every job runs on ubuntu-24.04 instead of the
moving ubuntu-latest, reads its Node version from .nvmrc (the plugin-sdk
jobs ran 22 under a comment saying it matched the server, which runs
24), download-artifact moves from v6, still on Node 20, to v7, and the
Sonar action is pinned by commit.
Nothing said which Node TREK runs on. The image is node:24, part of CI
ran 22 under comments saying 22 matched the server runtime, and the dev
guide said "22+". A bare-metal or dev install on an unsupported Node
installed without a word and then failed at runtime: isomorphic-dompurify
needs ^22.22.2 or ^24.15.0, the plugin jail's --permission flag is new
in 22.x.
.nvmrc now holds the major the Docker image runs, for nvm/fnm and for
every workflow to read. engines in the root and each workspace states
the range the runtime dependencies accept, the same one
isomorphic-dompurify declares. Without engine-strict npm only warns
outside it, so no existing install stops installing.
The image scan has been red since 10-06 on five advisories in four
runtime packages: proxy-addr 2.0.7 (critical, IP spoofing through an
IPv4-mapped IPv6 trust subnet), brace-expansion 2.1.4, source-map-js
1.2.1 and @modelcontextprotocol/sdk 1.30.0. brace-expansion had been
fixed once already; the ORM squash regenerated the lockfile and took it
back to 2.1.4 without anything noticing.
The overrides block in server/package.json was meant to hold such floors,
but npm reads overrides from the root package.json only, so none of them
ever applied: @nestjs/platform-ws kept its own ws 8.21.2 next to the
8.21.3 the block asked for. The floors move to the root, together with
new ones for every package this lifts, so the next lockfile regeneration
cannot quietly undo them. picomatch ^4 is dropped instead of moved: at
the root it would force micromatch's 2.x copy across a major. js-yaml is
lifted only inside its 5.x line, the one @nestjs/swagger pins.
The SDK goes to 1.31.0, the first fixed release; it stays hoisted, which
the CJS path aliases in server/tsconfig.json depend on. dompurify,
fast-uri, ip-address and js-yaml come along because npm audit flags the
same lockfile for them and each fix is inside the range already used.
@mikro-orm/postgresql (with pg), marked and topojson-client are dropped:
nothing in src, tests or scripts imports them, and the ORM config names
only the SQLite driver. The Postgres driver comes back with a Postgres
job that exercises it.
npm audit --omit=dev now reports only maplibre-gl, whose fix is the
6.x major tracked separately.
The count-string gate only knew {count} and {n}, the params t() picks a
plural form by. A quantity named anything else kept one form in every
language and passed: Russian read "Все {days} дней" for 2 days and
"{objects} объектов" for 3 objects, Greek "{minutes} λεπτά" for 1 minute.
Twelve such en strings become plural groups with their forms in all 27
locales, the quantity renamed to {count}: the trip length limit, the
chat image limit, the API key limit, the storage move prompt and its
reclaimable line, the Dawarich wish rule and suggestion duration, the
road trip lateness warnings, and the road trip echo and window texts.
Their callers pass the count as a number; the storage panels used to
pass it as a string, which no form is picked by. The reclaimable line
loses its dash for a comma in every locale while it is rewritten.
i18n-parity now treats the quantity names in QUANTITY_PARAMS ({days},
{minutes}, {objects}, {max} and the like) like {count}. The four strings
where the number is followed by nothing but a unit symbol (the Dawarich
durations, the upload size limit) go on NOT_PLURAL with that reason.
A template key was read as a pattern in which every interpolation is a
wildcard, so t(`roadtrip.window.${a ? 'onRoute' : 'onRoutTypo'}`) passed
as long as some other roadtrip.window key existed, and the typo rendered
as raw text.
An interpolation that can only produce string literals (a ternary chain
of them, in parentheses or not) is now expanded, and each resulting key is
checked like a literal one. Thirteen call sites become literal keys that
way. Where data is involved, a literal fallback (`${role ?? 'editor'}`)
is still checked as a key, with the template kept as a pattern under
its own allow-list site. Literals in a condition (`phase === 'start'`)
are never taken for results.
Narrow templates over data still pass without an allow-list entry. The
script header now says plainly what that leaves unchecked: a runtime
value with no key under a narrow prefix, which needs the value's type.
Two holes let English through the ratchet unseen.
A plural form en has no key for (Russian .few and .many, Arabic .two)
was compared with en's value for that same key, which never exists, so a
locale could copy en's text into exactly the forms a new plural group
needs and the count stayed put. Such a form is now compared with every
form en spells out for the group. A form en does declare keeps the exact
comparison: Italian "{count} file" is the Italian plural, not a copy of
en's "{count} file".
The // same-as-en marker excused any value in a Latin-script locale, so
one comment waved a whole copied sentence through. It is honoured only
on a value holding a single plain word, the case shared/CLAUDE.md
already describes. No locale carries the marker today, so nothing that
passed before fails now; the counts are unchanged.
The panel fires a test photos_shared event with count: '5', which reads
like the string counts the plural work removed elsewhere. Notification
params are Record<string, string> from the server through the in-app
store, the real sender passes String(added), and resolveTemplate reads a
numeric string as the plural count, so the test event matches what
production sends.
lint:i18n-keys let any template key through as long as its fixed parts
matched one en key, and each interpolation could stand for any run of
characters, dots included. `settings.${mode}` therefore reached all 574
settings keys and `help.guide.${id}.${part}` nearly 2000, so a wrong
suffix behind such a prefix could never fail, and none of the 109
templates had been looked at.
An interpolation now stands for one key segment, and a template passes
by itself only when it reaches between 1 and MAX_IMPLICIT_MATCHES (30)
en keys, the size of an enum such as the docsync error codes. A wider
one fails unless DYNAMIC_ALLOWED has an entry for that file and template
saying what bounds the value. The twelve wide templates in use today
get one each: the help key builders (checked by the registry test),
literal tuples beside the call, typed unions, the generated plugin
permissions and the seeded photo provider fields. An entry whose file no
longer uses the template, or whose template has become narrow enough to
pass alone, fails, so the list stays short and current.
Listing all 109 templates instead was considered and left out: the
narrow ones would only restate their own prefix, and the bound already
keeps a mistake inside one small namespace.
The ratchet excused any value equal to en that held at most one plain
word, so "Settings", "Appearance", "Readability" or "Users" copied into
another locale were never counted, Japanese and Arabic included. Buttons,
tabs and section titles are mostly one word, so a feature could still add
26 English copies of its labels and stay green.
An identical value is now excused only when nothing ordinary is left
after dropping placeholders, markup, addresses, a listed set of product
names (BRAND_NAMES, which now holds one-word names such as Atlas, Immich
and Mapbox) and unit symbols (UNITS). A word that really is the locale's
own, German "Status" or French "Transport", carries a `// same-as-en`
marker on its declaration, which the catalogue reader now reports. The
marker is honoured only in Latin-script locales: a Latin word left in
Japanese or Russian is English whatever it says. Every locale folder
must be on LATIN_LOCALES or NON_LATIN_LOCALES, so a new language cannot
pass the rule unclassified.
The baseline is re-recorded from the narrower rule: the unmarked copies
go from 1252 to 3238, the marked fallbacks stay at 3157, and no entry
went down. This is the measured state under the new rule, not a raised
gate; translating, or marking a word that is genuinely the same, lowers
it with --update as before.
Parity proves a locale has every key, not that the key holds the
locale's language: copying the en string passes it, so a feature could
add 26 English copies and stay green. Nothing counted them either. Today
the locales hold 3157 strings marked `// en-fallback` and 1252 unmarked
values identical to en.
scripts/i18n-untranslated.mjs counts both per locale and file and holds
them at scripts/i18n-untranslated-baseline.json, recorded from today's
tree. i18n-parity --strict runs it, so the existing CI step enforces it.
The two counts are separate on purpose: dropping a marker without
translating moves the string from one count to the other and still
fails. --update lowers the baseline to the current counts and never
raises an entry. A missing or malformed baseline, a locale lacking one
of en's files or a value the reader cannot parse is an error, not a
pass.
An identical value is not counted when it reads the same in every
language: after dropping placeholders, markup, URLs, e-mail addresses,
paths and a short list of multi-word names (Google Maps, Home
Assistant), at most one plain word may be left, where acronyms, single
letters and names with an inner capital are not plain words. That
excuses "{count} km", "GPX", "OAuth" and "Budget" and keeps "Source
code" counted.
The plural parity check only knows the groups en already declares, so a
"{count} ..." string added as a plain key kept one form in every
language: Polish read "Usuń 1 spakowanych", Czech "Odstranit 1
hotových" and Greek "1 λεπτά". 44 such en keys become plural groups with their
forms in all 27 locales: the collection, packing, day plan and offline
banner counts, the relative times, the audit log, the tour and transit
minutes, the share and photo hints. Labels like "Notes: {count}" get
their forms too, so a language can still inflect them later. The
dashboard's hand-rolled activeOne/activeMany pair, unused, becomes one
group, dashboard.subtitle.active.
i18n-parity now refuses an en string with {count} or {n} that is no
plural group. NOT_PLURAL lists the eleven where the number is never a
quantity (a day number, a step, a waypoint, the equator multiplier),
each with its reason, and an entry that stops matching fails as well.
Callers that worked around the engine:
- Copy to trip switched between two keys on placeIds.length > 1;
collections.copyN now carries the one-place wording as its `one` form,
where the language's `one` is the number 1 alone.
- The plugin dependency toasts said "addon(s)" and "plugin(s)"; they are
groups now and the panels pass the list length as the count.
- The dismissed suggestions count went in as a string and the comp days
as a formatted string, so no form was picked; both pass numbers.
BudgetPanel titled its add button with budget.addCategory (since April)
and the trip form labelled its cover remove button with common.remove
(since August). Neither key was ever in en, and t() returns an unknown
key unchanged, so both rendered as the raw key in every language. Both
are added to all 27 locales.
Locale parity compares the locales with en, never en with the code, so
nothing could catch this. client/scripts/i18n-keys.mjs (npm run
lint:i18n-keys) now resolves every literal key src/ names against en:
the first argument of t(), tHtml() and tr() (both sides of a conditional
included), <TransHtml html>, the fallback of translateApiError and any
*Key property or prop. A template key, a prefix with the rest appended
and the help registry's key builders pass when their fixed parts match
an en key; the one template with no fixed part is on an allow-list that
names the keys it resolves to, and an entry nothing uses fails. With
--unused it lists the en keys nothing reaches, as information only.
en is read with shared/scripts/i18n-catalog.mjs, a reader for the
locale tables as text that the parity tooling will share. It refuses a
value that is not one string literal instead of skipping the key.
skipIf and runIf were left out as run-time skips, which they are with a
real condition. With the literal true (skipIf) or false (runIf) they
switch a test off for good and gave a free way past the ratchet, while
the same file already counted Playwright's test.skip(true, 'why') for
that reason. Both now count as skips; nothing in the client uses them
today, so the counts stay as they are.
A block comment such as /* eslint @typescript-eslint/no-explicit-any: off */
turns a rule off for its file without leaving a message behind, neither
among the warnings nor among the suppressed ones, so a new file could
switch any warning rule off and the ratchet stayed green. eslint-disable
comments were already counted under "suppressed"; this was the way
around them.
lintClient now reads the block comments of every file ESLint linted from
the TypeScript parser, so the same text in a string, a template or a
regex does not count, and the tally reports each config comment as an
error. Nothing in the client uses one today, so the check is zero
tolerance. The tests baseline drops by one no-unused-vars warning that
is gone.
The size check walked src/ and tests/ only, while lint:format and
lint:skips already cover e2e/. The help center fixtures under
e2e/help/fixtures.ts hold 2291 wrapped lines, past the 2000 a test may
reach, and could grow without bound. e2e/ is now measured with the test
limit; fixtures.ts enters the baseline at its current size, every other
file there is within the limit.
A file's layer was the first path segment under src/, so src/types.ts,
the shared types module every layer imports as '../types', got the
layer "types.ts" and no rule at all, while the imports into it already
resolved to "types". It could import a component or a page unnoticed.
A file directly in src/ is now the layer of its name without the
extension, for the importer and the import alike. No file changes its
count: types.ts imports only @trek/shared today.
OAuthAuthorizePage reloads the user on mount and useDashboard's retry
reloads it too. The real loadUser then registers the background sync
triggers, a 30 s heartbeat and window listeners that nothing in either
spec unregisters, so they outlive the tests the same way they did in the
account specs before 27db5dd5e. Both specs now stub the triggers, which
neither is about.
The check only looked at src/ and tests/, so a new spec under e2e/ or a
new script under scripts/ could land unformatted, and the ratchet scripts
themselves did (fixed in the previous commit). It now walks e2e/ and
scripts/ too and takes .mjs alongside .ts, .tsx and .css. The 58 files
there that were unformatted before this branch (the e2e specs and help
guides, the older build and lint scripts) join the baseline, which keeps
only shrinking; JSON baselines stay out of scope.
The checker scripts were written without semicolons, against the
semi: true in .prettierrc, so every one of them failed the formatting
rule lint:format enforces on src/ and tests/. This is prettier --write
over the scripts this branch added or rewrote, nothing else.
help/ holds the help center's screen registry and vacay/ a few date
helpers; both are plain data and code like utils/, yet without an entry in
RULES they could import any view unseen. Neither does today, so ruling
them costs nothing in the baseline. managed/ stays free and the comment now
says why: it is the slot an operator fills with whole screens at build time.
A skip only counted when its first argument was a string or template
literal; anything else was taken for a Playwright run-time condition. So
it.skip(name, fn) in a table loop, describe.skip(TITLE, fn) or
test.skip(true, 'why') switched tests off without the baseline noticing,
and test.fixme was not matched at all.
The condition form is now recognised in Playwright specs under e2e/ only,
and even there a call counts when it declares a test (a body as second
argument), is called bare inside a test or group, or has the literal true
as its condition. Under vitest every .skip counts. fixme is a skip.
Nothing in the tree uses these shapes today, so the baseline stays at 35.
Every ratchet command hands its check to runCli, which sets the exit
code from the check's result and turns a failure of the check itself
(a missing or broken baseline, a directory that is not there) into a
printed FAIL and exit 1. The test runs runCli in a separate node
process, as CI does, and also pins that an unexpected exception still
ends the run with a non-zero code.
login, completeMfaLogin and register start tripSyncManager.syncAll()
and the system notice fetch without waiting for them, and every
sign-in, loadUser included, registers the sync triggers with their
30 s heartbeat. #2613 stubbed the sync in the push spec after it was
caught logging while the worker shut down ("Closing rpc while
onUserConsoleLog was pending"); the same shape was left in the other
specs that sign in for real.
authStore.test.ts ran the real sync on every successful sign-in and
logged DatabaseClosedError from it, and its notice fetch went past MSW
to the network and logged a failed fetch, both after the tests that
started them. The sync is now stubbed there, and the login, MFA and
register tests assert that it was started. The notice store is reset
before each test, since FE-STORE-AUTH-024 left a stubbed fetch behind.
LoginPage.test.tsx signs in through the form, and AccountTab.test.tsx
and MSettingsAccount.test.tsx reload the user after an MFA change; they
now stub the sync triggers (and the login spec the trip sync), which
none of them is about.
GET /api/system-notices/active gets a default MSW handler returning no
notices, so a login or a mounted notice host anywhere in the suite
settles that fetch in the test instead of on the network. The notice
store's own tests still override it.
JourneyDetailPage.test.tsx has carried 35 describe.skip blocks since
April, for contributors, trip linking and the map view, features that
were reworked rather than removed. Nothing stops the next skip, and
nothing stops a stray .only from passing CI while the rest of its file
never runs.
lint:skips reads every vitest file under src/ and tests/ and every
Playwright spec under e2e/ from its syntax tree. A .only on any runner
fails the check, always. Skipped and todo tests (it.skip, describe.skip,
test.todo, the x-shorthands, skip.each) are counted per file against
scripts/skip-baseline.json, which only lowers. A skip that decides at
run time is not a declaration and is not counted: skipIf, runIf,
Playwright's test.skip(condition, 'why') and a bare test.skip() inside a
test.
eslint.config.mjs keeps most rules at warn so that CI stayed green on a
codebase that was never linted, and CI runs eslint without
--max-warnings. Warnings therefore only ever accumulated: 1218 today,
among them 194 unused variables and 144 missing effect dependencies in
app code, the kind behind the stale closures #2610 chased.
lint:warnings runs ESLint over the client with its own config and counts
the warnings per rule, separately for app code and for tests, so the
headroom from fixing a test cannot be spent in app code. Messages that
an eslint-disable comment silenced are counted as a third group, so an
inline disable is no way past the ratchet either. No count may pass its
entry in scripts/eslint-baseline.json, and --update only lowers. Any
error, a parse failure included, fails the check too, so it can replace
the plain lint:check step in CI instead of running ESLint twice.
The offline core only works when the views go through it: component,
feature hook, store, repo, then api or Dexie, with writes through the
mutation queue. A view that imports src/api/ skips the cache, the
optimistic write and the idempotent replay. client/CLAUDE.md said so
and noted that no ratchet held the line, and the number of view files
importing api/client grew from 161 to 190 in two months.
lint:offline lists the 213 files under components/, mobile/, pages/
and hooks/ that import anything from src/api/ for more than types
(api/client, the websocket, the per-domain modules) in
scripts/offline-baseline.json. No other view file may, and a listed
file that stops leaves the list through --update, so the next file at
that path does not inherit the exemption. Type-only imports do not
count: they carry no request.
client/CLAUDE.md describes the rule next to the data-flow section,
including that an online-only domain puts its calls in a store or repo.
Components reach up into pages/ for models and helpers (28 imports in
25 files: the journey detail helpers, the collections, atlas and
dashboard models) and sideways into mobile/ for phone hooks and sheets
(36 in 30 files), and a few stores and a repo import from components/.
Nothing stopped the next one, and every such import ties two trees
together that should move and split independently.
lint:layers reads each source file's imports from its syntax tree
(multi-line imports, re-exports, import() and type imports included,
nothing inside comments or strings) and checks them against a small
rule table: a component imports neither pages/ nor mobile/, a shared
hook no view, nothing under the views (store, repo, sync, api, db,
utils and the other helpers) any view, a repo no store, and the Dexie
schema none of the layers above it. mobile/ and pages/ stay free, since
the phone screens reuse the page hooks by design.
The 80 imports that break a rule today are counted per file in
scripts/layers-baseline.json, and a file may hold no more than its
entry. Type imports count: a type the lower layer needs belongs below
it. Tests are not checked.
client/package.json has had a format:check script, but no workflow runs
it, and 2144 of the 2171 .ts, .tsx and .css files under src/ and tests/
differ from what .prettierrc asks for. Without a formatter gate the line
count in lint:size could also be gamed, and every new file picked its
own layout.
Formatting everything in one commit would bury the history, so lint:format
lists today's unformatted files in scripts/format-baseline.json and only
lets the list shrink: a file outside it, any new file in particular, must
be formatted, and a listed file that has been formatted must leave the
list (--update takes it off and never adds one). A stale entry fails the
check, since it would let the next file at that path in unformatted.
The check runs Prettier through its API with the config that applies to
each file and reads the text with LF line ends, so a Windows checkout
agrees with CI about endOfLine. A file Prettier cannot parse counts as
unformatted, and a missing config stops the run.
rtl-lint compared its own path with argv[1] to decide whether it was
the command being run. Started through a symlink or with a path that
differs in its form, it printed nothing and exited 0. The check now
lives in scripts/lib/rtl.mjs and the command file always runs it.
A broken rtl-baseline.json read as empty, so --update would have
written {} over it. It now stops the run with its reason, like
size-baseline.json.
The rtl-lint-disable marker skipped a line wherever it appeared, inside
a string literal too, and nothing bounded how often it was used. It is
now honoured only inside a comment, and the marked lines are counted
per file against scripts/rtl-disable-baseline.json. Today that is the
four map overlay positions in PlanMapStage.tsx. A side that is geometry
on purpose still gets its marker, but its entry is raised by hand, where
review sees it, instead of the marker quietly replacing the logical
forms.
lint:size counted newlines, so a file could stay under its entry by
joining lines, and the trip planner page already carried one JSX element
on a single 3760 character line. A line now counts once per 120
characters it spans, the printWidth in .prettierrc, so formatted code
measures what it shows and packed code measures what it holds.
The check also covered only .ts and .tsx sources. Stylesheets now share
the 1000 line limit, and tests get a limit of their own at 2000: they
mirror the component they cover and carry fixtures, but the largest are
past 5000 lines and growing with the god components.
The baseline is measured again with the new rule and keyed by the path
from client/, since it now spans src/ and tests/. Entries are higher in
number than before because the metric changed, not because a file grew;
every file is held at exactly what it holds today. New entries are the
ten largest tests, index.css and studio.css, the generated
countryShapes.ts (305 lines of 2600 characters each) and
JourneyDetailPageEntryEditor.tsx, which was just under the old count.
The scanning, baseline reading and lowering move to scripts/lib/ so the
next ratchets share them. A baseline that is missing or misshapen, a
directory that is not there and a scan that finds nothing now fail the
run instead of reading as empty, and the command file always runs its
check: there is no entry point guard left that could skip it.
The open-request-schemas header pointed at scripts/open-request-schemas.spec.ts,
which does not exist; the spec lives in src/. The bracket-notation comment in
the server ESLint config kept a dash that the earlier cleanup on this branch
missed.
The foreign key index check took every index from PRAGMA index_list as a
usable prefix, partial ones included. An index like
CREATE INDEX ix ON c(p_id) WHERE state = 'open' would satisfy FKIDX-001,
yet the cascade lookup WHERE p_id = ? still scans the table, which is the
regression the check exists to catch. A partial index now counts only when
its predicate is IS NOT NULL on the key's own columns, the one form SQLite
can prove from p_id = ?. Nothing in today's schema depends on a partial
index alone, so the result on the migrated snapshot is unchanged.
contracts:open in shared only counts exports named *RequestSchema, but
the name is a convention: 37 body DTOs wrap a schema called *BodySchema,
*InputSchema or plain *Schema, and tags.dto.ts declares its schemas
locally. pluginInstanceConfigUpdateSchema (a record of unknown behind
PluginConfigDto), createAlbumLinkSchema with five open shapes and a dozen
more were never counted, and a new open body under one of those names
passed both checks.
tests/unit/contracts/dto-open-shapes.test.ts walks the other way round.
It finds every class that extends createZodDto under src/, imports it,
and counts the open shapes in the schema the Zod pipe validates with
shared's walk. Each class is held to scripts/dto-open-shapes-baseline.json,
61 DTOs with 101 open shapes today, and a new DTO may hold none. It fails
closed: a createZodDto call that is not an exported class, two DTO
classes with one name, an empty tree or an unreadable baseline stop it.
It runs with the unit tests; npm run contracts:dto-open runs it on its
own and --update only lowers the baseline.
The five DTO classes declared inside roadtrip controllers are exported so
the walk can reach their schemas. The selector test picks up prettier's
wrapping of the cases added before.
openShapes() skipped every schema node it had already visited, so one
z.unknown() instance reused across four fields counted once, and a
baselined schema could gain any number of open fields that share a node
without its count moving. z.map(), z.set() and z.custom() were not walked
at all, so an open value inside them counted nothing.
An open spot now counts once per path it sits at, and the visited set is
only a guard on the ancestor chain, so a recursive z.lazy() still
terminates. The same kind at the same path is listed once, which keeps the
two sides of an intersection from counting a shared field twice. Map keys
and values and set values are walked, and z.custom()/z.instanceof() count
as a custom open shape. The request schemas hold the same 75 open shapes
under the new walk, so the baseline does not change; the spec covers
reuse, recursion and the new types.
MaintenanceRepository, DemoRepository and six seeders are exempt from the
SQL string selectors as whole files, so new SQL text added to any of them
passed lint unseen, and nothing noticed when a file no longer needed its
exemption. The comment promised a list that can only shrink, but nothing
held it to that.
The selector test now works out which files under src/db/repositories and
src/db/seeders the config really exempts and compares that set with a
pinned table, then runs the SQL selectors over each exempt file's source
and compares the hits with the count pinned for it (23 today). One more
statement fails, one fewer asks for the pin to come down, and a file at
zero has to leave the ignores. Per-line disables would have done the same
job but add suppressions the codebase is meant to lose.
The outbound-timeout selectors flagged safeFetch(url, undefined,
{ timeoutMs }) and safeFetch(url, { headers }, { timeoutMs }), although
timeoutMs is the deadline over the whole exchange that server/CLAUDE.md
tells callers to pass. At the same time an init spelled null or void 0
passed unseen, so a guarded call could still go without any timeout.
A call whose third argument is an object literal with a timeoutMs key is
now exempt, and undefined, null and void 0 (also behind an as cast) all
count as no init. The cast forms pin the inner node type, because esquery
compares String(value) and a missing name would otherwise read as
'undefined' and flag every cast init. The selector test covers both
sides.
The repository rule that keeps dialect SQL in sql-functions.ts only
matched raw() and the sql tag, so a SQL string handed straight to the
connection, em.getConnection().execute('INSERT OR IGNORE ...'), passed
unseen, and the seeders sat outside every repository rule. A new block
refuses a string or template whose text starts like a statement as the
first argument of execute, run, all or get in src/db/repositories and
src/db/seeders; a result mode such as qb.execute('run') stays legal.
The files that already hold such SQL are named one by one so the list
can only shrink: MaintenanceRepository (PRAGMA, VACUUM INTO),
DemoRepository and six of the seven seeders.
The selector groups are named once in eslint.config.mjs, and the
selector test now also checks the new block per path, its exceptions,
and that raw() stays forbidden in repositories.
journey-gallery-order.ts exported raw SQLite ORDER BY text from
src/nest that no production code read any more since the gallery reads
moved onto JourneyPhotosRepository. Two parity tests still compare the
Kysely rebuild against it, so it moves to tests/helpers as their oracle.
The ESLint rule from #2610 that requires a signal on outbound fetches
only matched the platform fetch. The guarded wrappers hand it a spread
init, which the rule trusts, so safeFetch(url, undefined, options) passed
lint with no timeout, and the dispatcher left undici's five-minute
default in place. Nothing capped the body either. Two callers did exactly
that on hosts TREK does not control: the Synology thumbnail read (a
user-configured NAS) and the place enrichment photo download (a URL out
of third-party place data), both buffering the whole answer.
The wrappers now carry defaults. A call without a signal waits 30 s for
the response headers; the body stays under undici's idle timeout, so a
proxied video that pauses with its player is not cut off. timeoutMs adds
a deadline over the whole exchange and joins the caller's signal, once
for the whole redirect chain. Every body is capped at 64 MB, refused up
front when the declared length is larger and failed mid-read otherwise;
callers pass a tighter maxBytes, or null where the body is streamed on
(pipeAsset, the doc-sync transfers that count their own bytes, the
Ollama pull).
The two call sites get a signal and an 8 MB cap. The lint selectors now
cover safeFetch, safeFetchFollow and safeFetchAdminConfigured as well,
including an explicit undefined init, and are named once so the
repository block restates the same list. A new test asks ESLint for the
options it applies to a path and runs them over snippets, so a selector
that stops matching fails a test instead of passing silently.
coverage-thresholds.test.ts imported vitest.config.ts, which pulled the
config into tsconfig.tests.json's program; that program compiles as
CommonJS and rejects the config's import.meta, so typecheck:tests
failed. The test now loads the config at runtime. The other new gate
tests get prettier's formatting.
google-quota, managed, mcp-shared, mcp-transport, public-api,
receipt-scan, route-usage, scheduling, school-holidays, tokens and tours
had no entry in the per-domain ratchet, so they sat in the pooled
src/nest/** catch-all, where a domain can lose twenty points behind the
headroom of the files beside it. That includes the cron registrar, the
MCP HTTP transport, the public API and the new tours domain.
Each is pinned at floor(measured) minus 1 as scripts/coverage-thresholds.mjs
prints it, measured over the unit, integration and e2e suites that load
the domain. A full run can only measure higher. mcp-shared (59 branches)
and route-usage (66 branches, 77 functions) measure below the 80 floor
and are pinned where they are; managed holds an empty module.
A new unit test fails when a folder under src/nest has no entry, or an
entry names a folder that is gone, so the next domain cannot land in the
catch-all again.
Migration20200101042200 wrote 86 indexes by hand so that deleting a trip
or a user stops scanning every child table while it holds the single
connection. Nothing kept it that way: the next table with an unindexed
trip_id would bring the scans back, and the entity parity tests compare
index names, not whether a key is covered.
The test reads PRAGMA foreign_key_list off the migrated test database
and requires the columns of each foreign key to lead some index or the
primary key. Today every one does, so the allow-list for deliberate
exceptions starts empty, and a second case keeps that list honest. A
third builds a small schema to prove the check fails on an unindexed
key, a composite one, and an index that holds the column second.
The server's boot gate only checks that a write route has a Zod DTO, not
that the schema constrains anything, and 47 of the 213 exported request
schemas hold an open shape: a loose object, a record of unknown, or an
unknown field. shared/CLAUDE.md already says to narrow those rather than
add new ones, but nothing counted them.
scripts/open-request-schemas.mjs walks every exported *RequestSchema and
counts its open shapes against a baseline of today's 75. A new request
schema may hold none and an existing one no more than its entry;
--update only lowers. The CLI checks the built dist/ for CI, and
src/open-request-schemas.spec.ts runs the same walk over src/ so the
shared test run fails as well. A missing or malformed baseline, or a
dist/ without a single request schema, is an error.
Module boundaries were held by review alone. storage.module.ts documents
a cycle that only surfaced at runtime as "imports[1] is undefined", three
modules exist just to break cycles, and nothing stopped the next feature
from reaching into another domain's helpers.
scripts/import-boundaries.mjs parses every import under src/ with the
TypeScript parser and checks five rules against a baseline of today's
violations: load-time file cycles (none today once type-only and lazy
imports and the two-way MikroORM entity relations are left out), new
dependencies between domains that already depend on each other (262
pairs, one tangle around plugins and auth), shared-kernel files importing
a domain (8), imports of another domain's internals (69) and src/db
importing src/nest (5). A domain's public surface is what the code
already treats that way: modules, services, guards, decorators, types,
constants, the audit logger, the plugin RPC kit and the addon gates.
New violations fail, --update only drops entries, and an unreadable
baseline, a missing src/ or an import it cannot resolve stops the run.
The client has had lint:size since #2612, the server had nothing, and
the server is where the biggest files live: maps.service.ts holds 3231
lines and grew by 200 in a month, and nine more services are past 1100.
scripts/size-lint.mjs carries the client's semantics over to src/ and
scripts/: a file may hold 1000 lines, the 23 that already hold more are
pinned at today's length in scripts/size-baseline.json, and --update
only ever lowers an entry.
A line past prettier's 120 columns counts once per 120 columns it spans,
so joining lines does not make room under the limit. The check fails
closed: a missing or malformed baseline or a missing src/ stops it
instead of passing with nothing checked.
The planner's two time edits of a visit called the time route straight:
saving the place dialog opened from a visit, and removing a visit's time
before a reorder. They skipped the visit repo's check for an older write
to the same visit still in the queue, so with such a write parked or
backing off the new time landed first and Try again (or the next flush)
replayed the older time over it, with no version token to refuse it.
Both now use the store's setAssignmentTimes, which writes through
assignmentRepo.setTimes, queues behind the older write when there is one
and keeps the optimistic update and rollback the other time edits have.
The dialog no longer reloads the days afterwards, since the store already
holds the new time and a reload would show the server's older one while
the new one waits in the queue; a changed day note is set on the visit in
place instead. The retry docs drop the exception they made for these two
paths.
An online edit that has to queue behind an older write to the same entity
starts a flush so it goes out right after that write. When the older write
was already in flight, flush() returned at once on the in-tab flag, and the
running pass had read its pending rows before the new one existed. The edit
then sat in the queue until the next trigger (the 30 second tick, a
visibility change or a reconnect), so collaborators saw it late and a
refresh in that window showed the older server value.
A flush called while a pass runs now asks that pass for one more round,
which it starts once it is done, next to the existing "mine wins" re-run.
Each pass clears the request when it starts, so the extra rounds stay
bounded by the calls made during a pass.
Holding later queued writes behind a parked one only ordered writes that
were in the queue. Online, the repos send an edit straight to the server, so
an edit made while an older write to the same visit, tour or driving settings
was backing off or parked landed first, and Try again (or the next flush)
then replayed the older write over it. Those writes carry no version token,
so nothing refused the stale replay.
The repos of visits, day clears, tours and driving settings now ask the
queue before an online write. While an older write to the same entity is
pending, in flight, parked or a conflict, the new write is queued too, with
the same optimistic Dexie write as offline, and a flush is started so it
goes out as soon as the one ahead of it does. When the driving settings left
the cache since the older write, they are read in first.
Places and packing items keep their direct online path: they send the base
token, so a stale replay is refused with a 409 and becomes a conflict. What
this does not cover, and the docs now say so: a collaborator's change to a
token-less entity is overwritten by the retried write, as by any replayed
offline write of those resources, and the planner's two time edits that call
the API directly are not held. The wording of the parking commit also named
budget items, which never go through the queue.
Keeping archived trips stopped the sync from wiping them, but left nothing
that ever took one off the device. The trip sync never refreshes an archived
trip, and the 7-day rule and the per-trip switch only looked at the active
list, so an archived trip stayed until Clear cache however long ago it ended.
The stale and switched-off rules now also run over the archived trips that
are stored here, with the same eviction as an active trip: its rows go, its
parked changes stay. The offline settings list the stored trips the server's
list leaves out, so an archived trip gets a switch the user can turn off;
before, the switches came from the active list only and could not reach it.
The cross-tab lock was taken with ifAvailable, so a tab that found another
tab replaying returned from flush() at once. Its callers read that as "the
queue is sent": the online trigger then re-seeds every trip from the server
and rehydrates the open ones, and the socket refetches after its pre-reconnect
flush. All tabs see the online event together, so the tabs that lost the race
laid the server's rows from before the replay over the shared Dexie, and over
the optimistic edits still on their way, while the winning tab was mid-flush.
The lock is now requested without ifAvailable. A tab that finds it held waits,
then runs its own pass, which finds nothing or what the holder left behind,
so flush() resolves only after the replay and the re-seed chained onto it
reads the state after it. The holder's requests are bounded by the client
timeout, so the wait is too.
Every open tab shares the account's queue and runs its own flush
triggers, and the _flushing flag only guards its own tab. Two tabs read
the same pending rows and both replayed them, so correctness rested on
the server's replay cache answering the second request, and each tab ran
the temp id remap and the Dexie write a second time.
The flush now takes a Web Lock named after the account's database, with
ifAvailable, so a tab that finds it taken skips the run the way a busy
flush in the same tab always has; the holder picks the rows up. The lock
is released with the tab, so a tab killed mid-flush cannot strand it.
Where the API is missing or refuses the request, the flush runs as
before under the flag alone.
Parking a write the server keeps failing on unblocked its trip, so a later
queued edit of the same budget item, visit, tour or driving settings went
through while the older one waited. Try again then replayed the older
write last and it silently overwrote the newer one: only places and
packing items send a token the server could refuse it with. Before the
parking change a 5xx stopped the whole flush, so this order was never at
risk.
The flush now holds back every pending write whose entity has an older
write parked as failed or waiting as a conflict, and a write that parks
during the pass holds back the rest of that pass too. Other entities and
other trips keep syncing. Try again therefore replays the parked write
first and the held ones after it, Keep mine does the same for a conflict,
and Discard or Keep theirs flush right away so the held writes go out
without waiting for the next trigger. Writes queued against an offline
create that got parked now wait for it instead of failing as an
unresolved temp id.
The trip sync clears every cached trip the server no longer lists, but it
read only GET /trips, which leaves archived trips out. Every sync (login,
reconnect, the periodic run) therefore wiped every archived trip from the
device, pinned ones included, and with it their parked failed changes. A
trip now only counts as gone when it is missing from both the active and
the archived list, and when the archived list cannot be read nothing is
treated as gone that round.
clearTripData no longer deletes 'failed' rows. Stale eviction, a trip
switched off in the storage settings and a trip that is really gone all
go through it, and none of them is the user deciding to give up a change.
A parked change of a deleted trip, or one the user was removed from, now
stays in the queue: Settings > Offline counts it under failed changes and
offers Try again (the server answers 403/404, so it parks again) or
Discard. Dropping it silently would lose the only copy of an edit the
user may still want to look at or redo elsewhere, and keeping it costs
nothing more than a counter the user can clear in one tap.
The rollback section promised that an older image refuses a migrated
database and names the pre-migrate copy. Only images from 4.4 on carry
that check. Going back to 4.3.x, the most common rollback today, gives
neither the refusal nor the pointer, so the copy has to be restored by
hand before the old image starts.
The backfill in Migration20200101042100 marks reminders as sent by UTC
date('now'), while the reworked job counts days in the app TZ, which read
like a seam that sends twice or skips on upgrade day. It is not: the
backfill stands in for the job before it, which matched start_date
against UTC date('now') at 9 AM app time, so UTC is the clock those
reminders went out by, and with one run per UTC day nothing it marks
was left unsent. The only effect is its strict "<": the upgrade day's
own reminders stay unmarked because the boot may precede that day's
run, so a boot after it reminds such a trip once more. One extra
reminder at most and never a lost one, so the shipped migration stays
untouched and the reasoning now sits on the job.
The todo claim's docstring writes its dedup bound as SQL.
Login, password reset and the sign-up checks all match emails
case-insensitively, but the only constraint was the column's own
UNIQUE, which SQLite compares case-sensitively. Two accounts whose
addresses differ only in case (legacy rows, two sign-ups at once) left a
login to resolve to whichever row came first, and LOWER(email) lookups
had no index to use.
A migration adds a unique index on lower(email). An install that
already holds such a pair must still start, so the migration then
leaves the index out, logs how many addresses are shared and by which
user ids (not the addresses), and records email_case_index_pending.
Since a migration runs once, db/email-case-index.ts runs on every boot
while that flag is set: it repeats the warning until the accounts are
told apart, then creates the index and clears the flag. The app-level
checks keep guarding meanwhile.
A write that passes the up-front check and still loses the race is now
refused by the database. The users repository turns that unique
violation into UserIdentityTakenError, and sign-up, admin create and
edit, and the profile update answer it with the same 409 their checks
give, instead of a 500 or an unhandled rejection (ADMIN-SVC-094 pinned
the old behaviour and now pins the 409).
The entity generator could not see an expression index (PRAGMA
index_info names its column NULL), so PARITY-009c found a db index with
no entity declaration. Rule 14 carries such an index into the entity as
a named entry with its own statement; Users.entity.ts is regenerated.
The trip and todo reminder ticks sent with `.catch(() => {})` and then
marked the reminder sent whatever happened. A send that threw (a DB error
resolving the recipients) or whose every channel failed (SMTP down, no
in-app) was recorded as delivered and never retried, without a log line.
And because the mark came after the send, two processes running the same
tick (a rolling update overlap, a stray container on the same volume)
both sent.
Each reminder is now claimed first with one conditional UPDATE through
its repository: reminder_sent_for is set only while it differs from the
start date (and the start date is still the one read), reminded_at only
while it is empty or older than the 20 hour dedup window. Only the
process whose update lands sends. If the send throws, or it attempted
in-app and channels and none went out, the claim is given back, the
failure is logged, and the next tick tries again until the trip starts
or the todo falls out of its window. A send nobody wanted (everyone
opted out) stays marked, since a retry would reach no one.
For that, NotificationsService.send() now reports what it attempted and
what was delivered; a channel that rejects or answers false counts as
not delivered. Existing callers ignore the result.
The legacy-upgrade suites stopped at legacy-v242, while v4.3.3, the
release every 4.x install will come from, left schema_version 244. The
most common upgrade path only ever ran against synthetic step maps.
legacy-v244.sql is that release's own runner booted once on an empty
file, plus synthetic rows across the main tables: users (one with a
mixed-case email), trips with a member, days, places, assignments with
the canary times, a booked stay, a flight with endpoints, budget items
with splits, packing, todos, notes, file metadata, tags, a journey with
an entry and a notification. scripts/build-legacy-fixture-v244.mjs
rebuilds it byte for byte from the tag: it archives v4.3.3 into a temp
dir outside the repo, boots that runner, seeds, pins every timestamp,
VACUUMs and dumps.
The suite runs the shared upgrade assertions and checks that every row
of those tables comes through with its values, that the boot copied the
database first and the copy still is the 4.3.3 schema, and that the
reminder backfill marks only the reminders whose day has passed.
Migrations only go forward, and since #2610 an older image refuses a
database a newer one migrated. Going back after a bad update therefore
needs a copy from before it, and the only copies were backups the admin
takes by hand or auto-backup, which is off by default. The first boot
on 4.4 migrates every 4.3.x install through the legacy baseline and the
new migrations without either.
Before applying pending migrations to a file database that holds
anything, the boot now writes a VACUUM INTO copy next to it, named
pre-migrate-<schema>-<time>.db (legacy-<schema_version> for an install
the positional runner left, otherwise the last recorded migration), and
keeps the newest TREK_DB_PRE_MIGRATE_SNAPSHOT_KEEP (3). The copy is
written under a temporary name and renamed, so a crash never leaves a
truncated file that looks good. If it cannot be written the boot refuses
before anything is migrated; TREK_DB_PRE_MIGRATE_SNAPSHOT=false is the
documented way past that. The newer-database refusal names the copy
taken right before the newer release migrated past this one.
A fresh install, a restart with nothing pending, an in-memory database
and a restore's re-bootstrap (the archive is the copy) take none, and
NODE_ENV=test leaves it off unless a suite sets the switch.
A successful login starts tripSyncManager.syncAll() without waiting for
it. In authStore.push.test.ts that sync was real and could still be
logging when the worker shut down, which vitest reports as an unhandled
"Closing rpc while onUserConsoleLog was pending" and fails the coverage
job although every test passed. The spec is about push, so the sync is
stubbed.
Since the ORM migration the scan of every push to dev stops with "the
Node.js process running out of memory (heap size limit 4288 MB)" and
fails the job, while PR scans, which only analyse changed files, stay
under it. The hosted runner has 16 GB.
c6b6b83d3 wrote a real line break where `printf '%s\n'` belongs, so the
RUN instruction ended inside the quote and Docker read the next line as an
instruction of its own: "dockerfile parse error on line 135: unknown
instruction: '". Every image build from dev fails on it. PR CI builds no
image, and the scout job on the release PR was already red for another
reason, so nothing showed it.
useTripPlanner.ts (3095 lines, now 718) and RoadtripSidebar.tsx (2287, now
694) are back under the limit and drop out of the baseline, and
TripPlannerPage.tsx is held at its new 1081 lines.
Saving a reservation or a transport, deleting a booking, finding or
creating the place an imported stay belongs to, the review that walks
imported items through their editors with its state, and the bridge
that picks up an import or a scanned receipt from the background tasks
move out of useTripPlanner into useBookingEdits.
useTripPlanner calls the hook where the save handlers used to sit,
after the day add, delete and clear wiring and before the booking
detail, so the bridge still runs between the rail snapshot and the
splash and the planner returns the same object. The review state only
moves later in the render, which no effect observes. The two existing
loose record types in the save handlers move along unchanged.
Putting a place on a day, moving a stop to another day, taking one off,
reordering a day with the vias it carries along, renaming a day and
reordering the days move out of useTripPlanner into useDayPlanEdits.
None of it runs an effect, and useTripPlanner calls the hook where the
day plan's edits used to sit, ahead of the day add, delete and clear
wiring, so the planner returns the same object and behaves exactly as
before.
The right click on the map, the POI marker and the hit on the drive with
the check-out days a night can end on, the service stop added by hand
with the legs it can go on, the popup handing over to the full form with
its duplicate check, and the place editor move out of useTripPlanner into
usePlaceFormOpeners.
None of it runs an effect, and useTripPlanner calls the hook right after
the place edits, where the place editor used to sit, so the planner
returns the same object and behaves exactly as before. The note on adding
a POI with its day now sits directly above the handler it describes.
The place delete question with the booked nights it warns about, telling a
tour from a place, saving the place form with the day it was added from
and the vias that stop pushes along, deleting one place, a tour or several
places, and moving places to another category move out of useTripPlanner
into usePlaceEdits.
None of it runs an effect, and useTripPlanner calls the hook where the
place form's save used to sit, ahead of everything that asks whether a
place is a tour, so the planner returns the same object and behaves
exactly as before.
Working out which stop a point on the drive belongs behind, where a place
chosen by hand goes, adding, dragging and removing vias, asking for and
accepting a refuel stop, and a corridor hit dropped onto the route move
out of useTripPlanner into useRoadtripDriveShaping.
None of it runs an effect, and useTripPlanner calls the hook where the via
anchoring used to sit, so the planner returns the same object and behaves
exactly as before. The note on placing a via now sits directly above the
function it describes.
The other ways of driving one leg, the road the pointer is on, the
overlays and the stretch of map they cover, the automatic day points and
their markers, the point the map is asked to bring into view, the rail
snapshot a choice is checked against, and asking for and choosing an
alternative move out of useTripPlanner into useRoadtripAlternatives.
useTripPlanner calls the hook where the picker used to sit, after the
road trip feed and the stop edits, so the highlight reset, the close gate
and the rail snapshot still run in the same order and the planner returns
the same object. The note on what the map frames now sits directly above
the value it describes.
Saving the stop popup as a pause or as a night, the stay release that
asks before a booked night is dropped, the kind, fill and stay of a stop,
moving a stop within its day or onto another one from the rail, ending
the day at a stop, and the stay dialog with the call that opens it move
out of useTripPlanner into useRoadtripStopEdits.
None of it runs an effect, and useTripPlanner calls the hook where the
stop edits used to sit, so the planner returns the same object and
behaves exactly as before.
The road trip mode and feed flags, the driving preferences and the limit
save, the day boundaries with their reset and drag controls, the vias,
the refuel search, the routes, the corridor search, the followed track,
the via counts and the via index helpers, the days folded in the rail and
the map layers that follow those folds move out of useTripPlanner into
useRoadtripFeed.
useTripPlanner calls the hook where the feed used to sit, right after the
trip overview, so the preference load, the boundary and via hooks, the
routing round and the corridor search still run in the same order and the
planner returns the same object. The helpers, the folds and the boundary
controls are plain memos and callbacks, so computing them earlier changes
nothing.
The list filters the map shares with the sidebars, the declutter of
collapsed days, the compact and pending marker flags, the Tracks filter
fallback, and the selected day's order badges and fit list move out of
useTripPlanner into useMapPlaces.
useTripPlanner calls the hook where the marker set used to sit, right
after the trip's websocket, so the store subscriptions and the Tracks
fallback still run in the same order and the planner returns the same
object. The order badges and the fit list are plain memos, so computing
them earlier changes nothing.
The day route toggle and its travel profile, the whole-trip overview, the
map lock, the recorded Dawarich trail, the fit key and the effect that
fits each trip once move out of useTripPlanner into usePlannerMapView.
A new test calls every toggle and the automatic route, which no test
reached before.
useTripPlanner calls the hook where that code used to sit, so the
Dawarich fetch and the first fit still run in the same order and the
planner returns the same object. handleSelectDay now lists the map lock
ref, the phone ref and setFitKey it receives from the hook; they never
change identity, so nothing behaves differently.
The place form, the service stop form, the stop popup, the stay release
question, the trip and member dialogs, the booking, transport and transit
editors, the importers and the booking detail move out of useTripPlanner
into usePlannerDialogs, together with the ?create= and ?tab= effects and
the AirTrail sync. plannerTypes.ts holds the option types the planner's
sub-hooks share.
useTripPlanner calls the new hook where that code used to sit, so every
effect still runs in the same order and the planner returns the same
object. The state setters it now receives from the hook are added to the
dependency lists that use them; they never change identity, so nothing
behaves differently.
TripSummary leaves RoadtripSidebar.tsx for its own module and is
imported back by name; Unit and splitValue go with it unexported.
The day card's header becomes DayHeader in the same module. Its body is
the header element the card drew inline, together with the reads only
the header used: the language, the distance unit, the clock format,
the missed ride and the day colour. The card keeps its translation and
passes the day, the fold state and the track details straight through.
DayHeader returns the bare header with nothing around it, so the markup
is unchanged, and it sits at module scope so it never remounts.
The code moves as it was otherwise, with the dashes in the moved
comments rewritten as colons and commas, and the imports only these
two used leave the sidebar, which is down to about 700 lines. Nothing
renders differently.
New tests render both on their own: the summary's figures split from
their units, narrow and still loading; the day's facts and its fold
control; a day over the driving limit, roads it could not avoid, a
missed flight, the track badge with and without its name, and the
wash in the day's own colour.
ServiceStop, RouteViaStop and Stop leave RoadtripSidebar.tsx for their
own module and are imported back by name; effectiveFill goes with them
unexported, and the imports only they used leave the sidebar.
The badge row under a stop's name was written out twice, once in each
stop shape. It is now the local StopBadgeRow, which both render in its
place. It returns the same span with nothing around it, so the markup
is unchanged, and it reads the fill setting and the vehicle the two
shapes used to read themselves. Those are plain store reads, so moving
them one component down changes nothing either.
The code moves as it was otherwise. The note on the stop kind table now
sits at the lookup it describes, the comment on the two settings says
where they are read now, and the dashes in the moved comments become
colons, commas and full stops; each theme-lint-disable stays on the
line of its style. Nothing renders differently.
New tests render the rows on their own: a plugin halt with and without
its name and stay, the kind picker on a pause and on a numbered stop by
click, Enter and Space, the drive findings and the walk from the road
on both shapes, including the range and unknown findings the rail never
draws, and the fill badge with the traveller's default.
TerminalStop, BookendStop, RideBlock, BookingChips and UndatedRides
leave RoadtripSidebar.tsx for their own module and are imported back by
name. RIDE_DASH, DiscRow, CodeChip, TimetableClock and CheckInBadge go
with them unexported, since only these rows use them, and the imports
only they used leave the sidebar. The code moves as it was; the one
change is that UndatedRides spells out the type of its booking callback
instead of reading it off the sidebar's props, so the new module never
imports from the sidebar. Nothing renders differently.
New tests render each row on its own: a departure with its clock and
check-in, a hire desk, a booked night that is left after check-out, a
ride that is caught, missed or has nothing on its timetable, the chips
a reader may and may not open, and the rides on no day.
DriveBand and SpillBlock leave RoadtripSidebar.tsx for their own module
and are imported back by name. MODE_ICON goes with them, unexported,
since only the drive band reads it, and the imports only these two used
leave the sidebar. The code moves as it was, with the dashes in its
comments rewritten as colons, commas and full stops. Nothing renders
differently.
New tests render both on their own: every travel mode's sign, the
alternatives button and its pressed state, a hop, a ride with and
without its booking and timetable, and a night drive that is automatic,
has no leg yet or misses its figures.
RefuelBand leaves RoadtripSidebar.tsx for its own module and is
imported back by name; the imports only it used go with it. The code
moves as it was, with the dashes in its comments rewritten as colons
and parentheses. DriveBand's note, which the band used to separate from
its function, now sits directly above DriveBand. Nothing renders
differently.
New tests render the band on its own: idle, searching, offers with and
without the right to add one, each empty answer, and the battery words
for an electric car, which no existing test reached.
OffRoadBadge, FillBadge, NightCheckIn, StayBadge, DriveFindingBadge,
DayCarry, Arrival and LateBadge leave RoadtripSidebar.tsx for their own
module and are imported back by name. The code moves as it was. The doc
blocks that sat stranded above other functions now stand above the
badge they describe, StayBadge's two notes as one, and the dashes in
the moved comments become colons and commas. Nothing renders
differently.
New tests render each badge on its own, including the walk from the
road and the range and unknown arms of the drive finding, which the
rail never reaches.
RAIL_GRID, RAIL_DASH, DISC, STAT_LABEL and DAY_BADGE leave
RoadtripSidebar.tsx for a plain module beside it, so the parts the
sidebar is split into next can share them without importing from the
sidebar itself. Values and comments move as they were, except that the
dash in the DISC note becomes a colon; the two style objects are typed
as CSSProperties from a type import. Nothing renders differently.
The plan tab's MapView with the route alternatives bar, the overview and
trail stack, the POI and compass cluster, the phone compass and the phone
POI portal move to pages/tripPlanner/PlanMapStage.tsx. It returns the same
children as a fragment in the same place, so the DOM and the portal order
stay as they were. Its props are picked from the planner hook's own types.
The four physical sides that moved are map geometry and carry an
rtl-lint-disable marker on their line, so the page's rtl baseline drops
from 17 to 13. Two moved comments lose a dash.
ListsContainer, the packing and to-do subtabs of the lists tab, moves
unchanged to pages/tripPlanner/ListsContainer.tsx, taking along the
imports only it used. The page renders it exactly as before.
The lazyWithRetry declarations of the planner page move to
pages/tripPlanner/plannerLazy.ts, a plain module so Fast Refresh keeps
treating the page as a component boundary, and LazyPanel moves to
pages/tripPlanner/LazyPanel.tsx. The page imports both. Comments travel
with their code; two of them lose a dash. Nothing renders or loads
differently.
lint:size counts the lines of every source file under client/src, tests
aside. A file may hold 1000 lines; the 24 files that are longer today are
listed in scripts/size-baseline.json with their current length and fail
the check when they grow. --update lowers an entry once a file shrinks
and never raises one, so a long file is split by concern instead of
growing. A baseline that cannot be read or parsed stops the check with
its reason, and --update leaves it as it is. It runs in the client job
next to the reading direction check.
Finishes the plural work the previous wip commit started. t() picks a
count string's form by the language's own Intl.PluralRules, and every
locale now spells out the forms its language reaches with whole
numbers, about 160 count strings in all: one, few and many in Russian,
Ukrainian and Polish, one and few in Czech and Slovak, five forms in
Arabic, one in the others, none in Japanese, Korean, Chinese, Thai,
Vietnamese and Indonesian. Each language was translated and then
reviewed on its own. English no longer reads "1 files", "1 updates
available" or "photo(s)".
Count labels shown next to a number now take that number too: the
atlas and admin statistics, the trip cards, journey and shared trip
pills, packing templates, receipts, storage usage, the Studio export
note. A heading that shared its key with a counted label keeps a key of
its own (the boarding pass heading, the Bookings tab), so languages that
put a noun after a number in the singular keep a plural heading. The
plugin hosts chip passed its count through .replace() after t() and
never got a form; three toasts sent the count as text. Mail and push
texts for shared photos and suppressed storage failures pick their form
in code through pluralForm.
A whole number sent as text counts too, which is how in-app
notification params arrive, and the plural rule follows the language of
the strings on screen while another one still loads.
The parity check knows which forms each language needs and refuses
missing and impossible ones. It also reads keys that contain a colon
now: it skipped them, which hid 58 missing plugin permission texts in
Catalan and 35 in Vietnamese, both translated. The client parity test
covers every language of the registry instead of a hand-kept list of 19,
and the placeholder spec lets a form spell its number out only where
its category is that one number. Brazilian Portuguese error screens
were in Breton, and an Arabic dashboard label held Hebrew text.
Migration20200101042300_every_tour_type_the_contract_names runs after
the legacy steps like the other post-legacy migrations, but was missing
from the test's list of migrations without a step, so LEGACYMAP-001
failed.
The iPad case could no longer fail on its own regression: a draggable
row behind the long-press bridge is what tablets get by design since
#1616, and nothing in the case swiped. What turned swipes into drags in
#1432 was the drag-drop-touch polyfill, so the case now asserts that a
tablet never loads it.
Two help-media guides waited for "1 files" and "2 flight(s) imported",
which now read "1 file" and "2 flights imported".
On a phone in Arabic every system notice, the release notes included,
opened as an empty sheet: its three-slot strip is shifted to show the
middle slot, and in a right-to-left row that shift moved all three out
of view. The shifts point the other way in a right-to-left page now,
which the swipe handling already assumed.
The budget's total amount kept its parts in order but sat on the
opposite edge from its label, because the whole row was made left to
right. Only the amount itself is now. The phone chat's bubble tails
were still physical corners while the desktop ones follow the row; both
use logical corners now.
lint:rtl takes comments from the TypeScript parser instead of scanning
lines, so a "/*" inside a string (accept="image/*") no longer hides the
lines after it.
Three gaps in the notice from the previous commit, found in review.
A page served from the network and claimed later by its own build's
worker counted as already taken over, so Reload went to that worker,
got the old build again and spent the session's reload: no notice and
no new build until the next start. Only a worker that replaced the one
that served the page skips the handover now, as at launch.
A reload that came before the new worker could take over (a slow
precache, a failed update, an install gone redundant) also spent that
guard, so the launch behind it could no longer finish the handover with
its own reload. That reload leaves the guard alone now.
An install whose server never reports the bundle's version (a hand-set
APP_VERSION) showed the notice in every session, and Reload could not
clear it. The store keeps the first version a page hears of, from the
app config at launch, and the notice only asks when a later report
differs: a different build at launch is the launch handover's to fix.
The server's wire-envelope test pins the welcome frame's fields and now
expects the version beside the socket id. The wiki's update page says
what an open tab shows after an update.
computeTourMetrics indexed points[i - 1] and points[i], which the shared
package's noUncheckedIndexedAccess reads as possibly undefined, so the
shared typecheck in CI failed since the metrics moved there. It walks the
points pairwise from the first one now; the numbers are the same.
Both only ran on a developer's machine, and both had stopped telling the
truth. The end-to-end flows now run in CI in Chromium and in WebKit (the
iPad and iPhone cases need the engine every iOS browser uses), with the
browsers cached on the lockfile and the traces uploaded when a flow
fails. A second job builds the client and fails when one chunk carries
both map engines.
Making them pass meant fixing what had rotted. The setup closed every
first-run notice but the release notes, whose X is called Close, so the
notes covered the dashboard and every flow after login timed out on a
click. The iPad case still expected undraggable place rows; tablets drag
on purpose since #1616, behind the long-press bridge, so it now checks
what #1432 is about, a swipe that scrolls and raises no drop overlay,
with locators for the scroller and the Plan tab that match today's
markup. The iPhone case read the body scroll lock through
window.scrollBy, which moves an overflow: hidden viewport in every
engine; it now reads the lock where it takes effect and checks the page
keeps its place.
A tab kept open across a deploy ran the previous build until its next
start. The server now names its version in the realtime welcome frame,
which every open client reads again when it reconnects after the
restart a deploy brings. When that version is not the build the page
runs, a notice offers to reload. The reload goes through the same
service worker handover a launch uses: it waits for the new worker to
take over, or for a short timeout, and records the version only when
the takeover happened.
The notice reads the error boundary's texts for the same situation, so
every language has it already. It is one component for both shells,
sits above the offline pill, stays away offline and after a reload for
that version that did not bring it, and can be closed for the session.
Arabic runs the app right to left, but margins, paddings, insets, text
alignment, borders and corner radii were written for the left side, so
icons sat against the wrong edge of their labels, indents and badges
pointed the wrong way and amount columns aligned left. They now use the
logical forms (ms-, ps-, start-, text-start, marginInlineStart,
insetInlineEnd and their CSS equivalents), which are the very same sides
in a left to right language. Geometry stays physical on purpose: maps,
positions measured from the pointer or a bounding box, time axes,
sliders, transforms and print output.
Amounts drawn as several spans keep their order in a right to left row,
the time picker keeps hours before minutes, and two switches whose thumb
slid out of its track in Arabic now move the right way.
lint:rtl counts what is still physical per file against a baseline that
may only shrink (1512 uses in 370 files before, 319 in 113 now), and CI
runs it, so new code reaches for the logical forms.
Unfinished, continued in the next session: t() resolves key.one/.few/...
by the language's own plural rule, the parity check knows which forms a
language needs, 16 singular/plural key pairs are merged in every locale
and every count === 1 branch in the client uses the plural form. Still
open: the forms each locale needs for the remaining count strings, the
two new labels in every locale, and 35 tests that expect the old English
wording.
Each zone's VTIMEZONE carried one fixed offset, so a client that reads it
literally showed every event after a change to or from summer time an hour
off. It now lists every offset change in the calendar years its events
fall in, each as its own observance after the offset in force before the
first one. A zone without changes, or a span with no readable date, keeps
the single fixed offset.
The phone shell was chosen by width alone, and most phones are wider than
767px in landscape, so turning one swapped the whole route tree for the
desktop planner and an open sheet, its input and the map position were
lost. The phone query now also covers a touch screen under 500px tall,
which no tablet is. It lives in one place: the shell, the planner and todo
hooks, the warnings banner, dialog focus and the stylesheets read it, and
six components that measured the window once on mount now follow it.
The public instance allows about one request a second per client, and the
Tours elevation questions went out right behind the route request with no
pause, so they could be refused with a 429 and leave the tour without the
elevation it needs to be saved. Every request to that host now waits for
its turn, across callers; an operator's own Valhalla is not paced.
A tourRepo replaces the direct toursApi calls: reads fall back to a new
Dexie tours table, and a tour saved or edited offline is written there and
queued like a place, under one temporary id for the place and its facet.
The queue moves both to the real id on replay and drops both when the
server refuses the create. Trip sync caches each trip's tours while the
addon is on, so a tour no longer reads as a plain place offline.
The tour metrics move to shared so the offline copy shows the server's
numbers, and useTourPlaceIds keeps one Set per change as intended.
tour_types now holds every key the shared contract accepts, all but hike
disabled, and create and edit turn a disabled type away with a 400 instead
of failing on the foreign key. tours.tour_group_id is dropped: nothing ever
wrote or read it. difficulty and wanderer_ref stay for the Wanderer import.
Updates the wiki for the changes in this branch: TREK refuses an
ENCRYPTION_KEY that disagrees with the key file, the rotation script
covers the AI Parsing key and moves the file along, a restore refuses a
newer backup and keeps the replaced database, the chart's startup and
readiness probes, PORT and extraEnv, retrying or discarding failed
offline changes, trip reminders catching up, and why a Portainer
recreate used to keep reporting the old version.
On the phone a flick of the sheet's top edge, a tap on the scrim or
Escape closed the booking, transport and place forms and threw away
whatever had been typed, without a word; the desktop dialogs have asked
since #2253. The question moves into a shared hook that DialogShell and
MSheet both use, and the three sheets pass their form state to it.
The PDF export on desktop and phone, the language switch, the holiday
regions and the PDF's map image load a chunk on demand and catch the
failure themselves. After a deploy the old page asks for chunk names the
new build no longer has, so they showed their own error on every retry
(or, for the language, silently stayed in the old one) instead of
reloading onto the new build the way an uncaught chunk error does.
importChunk reloads once for a missing chunk and still hands the caller
its error.
The shared context menu closed only on a click elsewhere: Escape did
nothing, the arrow keys did nothing, and a scroll left the fixed menu
hanging over content that had moved on. It now takes focus on its first
item, moves with the arrow keys, and closes on Escape, scroll and
resize. rules-of-hooks is an error again; its one hit was a test helper
named like a hook.
One try around the holiday fetch and the deletes that follow it caught
both, so a database error while removing vacation days that fall on a
public holiday was logged as nothing and left the days standing. Only
the fetch is tolerated now. Also drops two files long gone from the
Sonar coverage exclusions.
86 foreign-key columns had no index. SQLite does not add one by itself,
so every cascading delete of a trip or a user scanned each such child
table in full, inside the transaction, while every other request waited
on the one connection. One migration adds them all, and the entities are
regenerated to match.
No job had timeout-minutes, so a hung step held a runner for GitHub's
six-hour default. Test and housekeeping jobs stop after 45 minutes, the
image builds after two hours.
The list read marked tour places with tour_place_id, but the single-place
read behind place:created and place:updated did not, and the phone's
places browser kept a second tour list of its own that no realtime event
updated. A tour another member made, or one imported from GPX, showed as
an ordinary place on the phone until the screen remounted, and every
tour did offline. The single read carries the mark now, the browser asks
the planner, and the tour id set is memoised instead of rebuilt on every
render.
A note's attachment moved to the trash from the file manager kept
showing on the note; the hydration read now skips it, as the chat one
already did. The MCP session sweep logged a line every minute while any
session was open; it logs only when it cleaned something. The compose
file caps the container log, and three test mocks that pointed at
modules long gone are removed.
Every refresh writes a new oauth_tokens row and nothing ever deleted the
old ones, so an MCP client refreshing hourly grew the table, and every
backup with it, without end. A nightly job now deletes tokens whose
refresh token expired more than 30 days ago, leaves first, and keeps any
row a live chain still points at for replay detection. Also types the
backup's per-entry store flag.
Changing PORT left the Dockerfile and compose healthchecks, the chart's
containerPort, Service target and probes on 3000, so the container went
unhealthy and Kubernetes restarted it in a loop. All of them follow PORT
now. The chart gains a startup probe (ten minutes for a first boot, a
long migration or a restore at boot, which liveness used to kill after
about 105 s), a readiness probe on the new /api/health/ready that asks
the database, extraEnv and extraEnvFrom for the settings the chart does
not list, slots for the four secrets it had none for, and the
DEFAULT_LANGUAGE and MCP_MAX_SESSION_PER_USER values it documented but
never set. The Dockerfile no longer hides a failed npm ci behind a ';',
pins gosu, and the language lists name all 27 locales.
Resetting a locked-out or compromised admin left every session on that
account valid, because the script did not move password_version on the
way every password change in the app does. It also looked the address
up case-sensitively and created a second account when the stored one
had capitals.
Both locales had their notification texts, but the index never imported
them, so their users got English mails and the parity gate passed
because the files existed. The locale map is now typed against the
supported languages, so the next one left out fails to compile.
TREK's codes 'br' (Brazilian Portuguese) and 'gr' (Greek) went straight
into <html lang> and into ten Intl calls. 'br' is Breton, so Brazilian
users got Breton month names and screen readers announced the page as
Breton; 'gr' is no language at all, so Greek users got the default
locale. getIntlLanguage maps both, and every one of those sites goes
through it.
Logging out deleted this user's offline database straight away, so a
change made offline and not yet sent was lost without a word. With a
connection, the logout now flushes the queue first, for at most five
seconds.
The trip sync and every repo list only ever upserted, so a place a
collaborator deleted while this device was offline came back on the
next offline read, and a trip the user had been removed from stayed on
the device. A sync and a full list now replace the trip's rows, keeping
only the ones created offline that have not synced yet, and a cached
trip the server no longer lists is cleared.
A queued change the server answered with a 5xx was put back at the head
of the line and the whole flush stopped, every 30 seconds, forever: one
bad payload held every later change of every trip. Now only its own
trip waits, with gaps doubling from 30 s, and after eight failures it is
parked as failed. A DELETE answered 404 counts as done instead of
showing as a failed sync. Settings > Offline offers to try parked
changes again or discard them.
A connect still waiting for its ws token when the user logged out went on
to open a socket authenticated as that user, which the next login on the
same browser then reused. And a socket in its reconnect backoff waited up
to 30 s after the network came back or the tab became visible. Both
events now reconnect right away.
The desktop day panel stays mounted across a day switch, so a forecast
or the stays still loading for the previous day overwrote the new day's
once they arrived. The location search in the transport form showed
results for text the user had already changed, or refilled the list
after a pick. Both now drop an answer for a request that no longer
matches.
CronJob ran a tick on schedule even when the previous one had not
finished, so an hourly auto-backup of a large library could overlap
itself, and a rejected tick only reached cron's own console line. Jobs
now wait for completion and log failures through the app logger. The
backup also stores photos, videos and PDFs instead of deflating
already-compressed files at level 9.
Overpass, the Wikipedia photo lookup (which holds one of five shared
photo slots), the plugin download and registry, and the two API key
checks had no signal, so a provider that took the connection and went
quiet held the request for undici's 300 s default. ESLint now refuses a
fetch() whose init carries no signal.
The reminder crons run at 09:00 in the instance's TZ but computed today
in UTC, so east of Greenwich a trip reminder arrived a day late. A day
the job did not run also lost that day's trip reminders for good: the
match was 'starts exactly N days from today'. Trips now record the start
date a reminder went out for, and the job sends every reminder that is
due and not yet sent. The active-trip redirect and the upcoming
reservations use the instance's date and clock as well.
Filling in a missing end date and keeping a trip's length when its start
moved both shifted a local-time Date, so a trip crossing a DST change
lost or gained a day. Both now use addIsoDays. The one-date-makes-a-week
default moves from the REST controller into TripsService.create, so the
MCP create_trip tool no longer stores a half-dated trip, and the trip and
its days are written in one transaction.
That branch numbered its own steps after the upstream step it was cut
from, so a database at schema_version 243 to 246 (or 216 to 218) with the
Tours tables had four upstream migrations marked as run without ever
running them. The baseline now cuts such a version back to the last
upstream step, and the Tours migration adds the difficulty cap to a tours
table an earlier branch stage created without it.
Login and the OIDC fallback match emails case-insensitively, but the
password reset lookup and the admin create and edit checks compared
exactly. An account stored as Foo@x.com never got a reset mail, and an
admin could create bob@x next to Bob@x, after which login picked
whichever row came first. The signup collision check and the first-user
count also move inside the signup transaction.
budget_settlements.created_by_user_id has no ON DELETE action and the
account cleanup never cleared it, so deleting such a user failed on the
foreign key and rolled the whole deletion back. The cleanup now clears
the recorder; the payment itself belongs to the two members and stays.
Rolling an image back, or restoring a backup a newer release made, ran
older code against a schema it does not know and failed later, far from
the cause. Boot now refuses a database that records migrations this
build does not ship, and the restore check refuses such a backup while
the current database is still in place. A restore also keeps the
database it replaces as data/pre-restore-<time>.db.
An env value different from data/.encryption_key used to overwrite the
file on boot, destroying the only copy of the key the stored secrets were
encrypted with. The server now refuses to start and names the rotation
script, which re-encrypts the AI Parsing addon key as well and moves the
key file to the new key. Under test the suite's fixed key no longer
overwrites the developer's key file.
Portainer's recreate and a copied run config carry the previous image's
APP_VERSION into a container built from the new one, so the server kept
announcing the old release and clients never handed over to the new
bundle. The Dockerfile now writes the version into server/VERSION, which
the server reads before APP_VERSION, and prerelease bundles take their
version from the same build argument.
dev now runs on MikroORM, so the positional migrations, seeds.ts and
DatabaseService the Tours server side was built on are gone. Behaviour
is unchanged; persistence moves onto dev's architecture:
- one MikroORM migration for tour_types, tours and tour_waypoints,
idempotent so instances that ran the old positional steps keep
their data; the hike tour type is seeded inside it, the addon row
in the AddonSeeder
- entities and repositories for the three tables, tours domain on
repositories and UnitOfWork
- tour facet joins in places, assignments and days re-applied on the
repository code
Fixes found while porting:
- GPX import checks addon, trip access and permission in the handler
instead of guards, so a refused upload gets a 403/404 instead of a
connection reset
- copying a trip copies its tours and waypoints
- a tour cannot be moved onto a day that already has it
- updateTour checks existence inside its transaction
Adds tours e2e tests and rewrites the tours unit tests onto the
migrated test database.
* base entities
* update deps
* fix date as string
* rename entities
* migrations + seeders
* feat(db): DbTimestampType keeps DATETIME columns as the CURRENT_TIMESTAMP text
* feat(db): discover entities from an explicit list instead of a glob
* test(db): createTestOrm binds an ORM to a suite's own SQLite handle
* feat(db): dialect helper is the one place SQL functions are spelled
* test(db): dialect helper refuses malformed column references and day counts
* feat(db): snake_case entity shape and the first three repositories (Days, DayNotes, Trips)
* fix(db): repositories mirror the statement's column set and guard their row types
* feat(db): request-context and unit-of-work helpers for the ORM, with concurrency tests
* test(server): integration suites boot on the migrated schema snapshot
* test(server): snapshot harness keeps the legacy id semantics
* fix(test): route MikroORM's dynamic imports through vitest from the setup file, not the ORM config
* test(db): coverage ratchet entries for the ORM foundation
* test(server): websocket suites boot on the migrated schema snapshot
* test(server): coverage's uncovered-file pass goes through the SWC transform
* fix(db): dialect helpers dispatch on the platform; lint gate for the repository layer
* test(server): ORM suites open the schema snapshot; RESET_TABLES derived from the schema
* chore(server): promise safety net and sweep tooling for the async migration
* refactor(realtime, systemNotices, addons, airports, audit, categories, permissions, query-helpers, tags, transit, trip-membership): async DB-touching methods; transactions through UnitOfWork
* fix(server): observe every promise the async sweep left unobserved (timer, handshake, plugin-runtime test)
* test(doc-sync): unit suites await the methods the async sweep made async
* refactor(atlas, school-holidays, settings): async DB-touching methods; transactions through UnitOfWork
* chore(server): lint rule refusing a Promise used as a plain value, and the misses it found
* test(server): suites await the methods Task 1's forced edits made async
* refactor(budget, common, day-notes, days): async DB-touching methods; transactions through UnitOfWork
* test(oauth): controller suite awaits the async service
* chore(server): db-call-graph --unawaited lists calls to async methods whose result is used without await
* refactor(llm-parse, mcp-shared, oauth, place-shadow, route-usage, storage, todo, trip-invite): async DB-touching methods; transactions through UnitOfWork
* refactor(plugins): async DB-touching methods; transactions through UnitOfWork
* chore(server): db-call-graph --sync stops its transitive walk at an async callee
* chore(plugins): drop gate-only asyncs; observe the scheduler's rejection; async event-sink stub
* refactor(notifications, place-photos, backup): async DB-touching methods; transactions through UnitOfWork
* refactor(auth, files): async DB-touching methods; transactions through UnitOfWork
* test(server): suites await the methods earlier sweep groups made async
* refactor(tokens): async DB-touching methods; the auth token handlers await them
* test(files): upload filters fail closed on an excluded or unresolvable type
* refactor(collab, collections, maps): async DB-touching methods; transactions through UnitOfWork
* refactor(memories, photos, packing, share): async DB-touching methods; transactions through UnitOfWork
* chore(server): lint rule also refuses a Promise handed to expect() without resolves/rejects
* refactor(vacay, admin, doc-sync): async DB-touching methods; transactions through UnitOfWork
* fix(vacay): type the loop accumulators the async sweep introduced
* refactor(journey, oidc, place-enrichment, tokens, trip-members): async DB-touching methods; transactions through UnitOfWork
* fix(server): the journey hooks are awaited by their callers, not detached
* refactor(assignments, public-api, accommodations, places): async DB-touching methods; transactions through UnitOfWork
* refactor(trips): rewrite AccommodationsService.createAccommodation .bind alias
Recipe R4 trap named in the Task 12 brief: a .bind alias of an
AssignmentsService/AccommodationsService method is typed any and hides a
missing await from tsc and the promise lint rules. createAccommodation
belongs to this task's group, so rewritten as an arrow forwarding to the
method instead of a bound reference.
* refactor(assignments): drop the unused Trip type alias
* refactor(reservations, booking-import, calendar, integrations, roadtrip): async DB-touching methods; transactions through UnitOfWork
* refactor(roadtrip, reservations): return await on transactional; document the single-flight promises
* refactor(trip-read-model, trips, feeds): async DB-touching methods; transactions through UnitOfWork
* refactor(server): remove DatabaseService.transaction — every transaction runs through UnitOfWork
* refactor(trips, feeds): drop added return annotations; no double await on the feed guard
* fix(server): phase 1 review follow-ups — awaited returns, logged upload-filter rejections, covered handshake/notice branches, call-graph gate in CI, dead test imports
* fix(server): upload filters answer multer once on a throwing callback; positive call-graph gate test
* test(db): entity metadata must match the migrated schema; fix the four drifts it found
* feat(db): primary-relation entities proven (BudgetCategoryOrder, VacayUserSettings)
* test(db): parity gate also holds declared pivot entities; per-field column types; missing-set ratchets both ways
* test(db): numeric ids for the primary-relation spike
* feat(db): re-runnable entity generator encoding the snake_case/DbTimestampType rules
* fix(db): generator guard rails — validation cannot mask hidden()/index loss; typed JSON; loud failures
* fix(db): generator keeps NOT NULL boolean defaults; fails if a JSON interface import survives
* refactor(db): every entity in the snake_case shape; document-sync entities added; one repository per entity
* fix(db): no NaN class-field initialisers; entity generator --check gate in CI; parity normaliser keeps literal case
* chore(db): entity --check reports stray files; class-field census is a ratchet; migrate-encryption typechecked
* feat(db): AppSettings and Users repositories; forFeature wiring pattern
* refactor(permissions): permission flags through AppSettingsRepository
* fix(server): boot-time settings read runs inside an explicit ORM request context
* refactor(audit): audit log writes through AuditLogRepository
* fix(db): PK reads refresh past the identity map; typed owned-lookup generics; e2e ORM module mirrors production config
* docs(db): owned-lookup call-site convention; refresh side effect; e2e ORM module spread notes
* fix(audit): audit rows are native inserts, not a unit-of-work flush; closed test ORMs
* fix(server): non-HTTP entrypoints run inside an ORM request context; permission loads fail closed on context errors
* refactor(categories, tags): repositories replace raw SQL
* refactor(addons): addon catalog and photo providers through repositories
* fix(db): generator captures inline UNIQUE constraints
SqliteSchemaHelper.getIndexes() drops every unnamed sqlite_autoindex_
entry before entity-generator ever sees it, so a table-level
UNIQUE(...) constraint with no separate named CREATE UNIQUE INDEX
(settings' UNIQUE(user_id, key) among ~40 others) never reached the
generated entity's uniques: metadata. Adds a general
RULE9_addImplicitUniqueConstraints (skips a column set identical to
the table's own primary key — AppSettings.key/Addons.id noise) fed by
a PRAGMA index_list/index_info scan, and regenerates every entity.
* chore(addons): well-known gate wraps only the addon read in a request context; drop unused test factories
* fix(platform): well-known gate tolerates a synchronous addon read from the context helper
* fix(db): PK lookups inside patch refresh past the identity map; corrected identity-map notes
* refactor(settings): user and instance settings through repositories
SettingsRepository (getForUser/upsertForUser/getOne/deleteForUser) replaces
the per-user settings table's raw SQL; SettingsService drops DatabaseService
entirely. instance-api-keys.ts's readInstanceApiKey/writeInstanceApiKey/
resolveApiKey take an explicit AppSettingsRepository/UsersRepository instead
of a raw connection.
An earlier version resolved those two repositories itself from the active
MikroORM request context, so none of its six callers needed to change. That
worked in production and e2e (always inside a real request) but broke
service construction in plain unit tests across every calling domain
(addons, auth x2, maps, transit, unsplash - 185 failures on a full run),
since none of them wrap a request context. Each of those six services now
injects its own AppSettingsRepository/UsersRepository and passes them
through explicitly, matching how every other converted repository in this
migration is consumed; their own SQL is untouched (outside Plan 3a).
* test(server): harness constructors follow the Task 5 repository signatures; one test ORM per handle in the plugin host
* fix(db): generator names relation properties in captured UNIQUE constraints
* fix(settings): drop dead wiring and the speculative deleteForUser; strict row assertions
* fix(server): plan 3a review follow-ups — fail closed without an ORM at every choke point; seam and D5 ratchet tests; dead method and DDL copies removed
* fix(server): plan 3a re-review follow-ups — plugin boot activation and the airports backfill inside a request context; the supervisor answers the child before failing closed; BOOT-SWEEP-001 drives the real boot hooks; ws adapter seam pinned
* feat(db): generated entities declare their repository type; test casts dropped
* fix(server): plan 3a re-review 2 follow-ups — absent-ORM boot branch pinned; airports boot hook never rejects; BOOT-SWEEP-001 asserts the seven routes; duplicate comment removed
* feat(server): pre-init photo route and WS handshake run in a request context; InviteTokensRepository
* refactor(auth): users table through UsersRepository; JWT and MFA lookups repository-backed
* refactor(tokens): MCP and API tokens through McpTokensRepository
* refactor(auth): passkeys and registration invites through repositories
* fix(db): row-out repository reads bypass the identity map — stale write-back over nativeUpdate at flush; insertUser via em.insert; settings round-trip pinned
* fix(server): non-numeric route ids answer the legacy not-found, never NaN (tokens, categories, tags); mint path back to two statements; identity-map regression pinned
* refactor(oauth): OAuth server tables through repositories
OauthClientsRepository/OauthTokensRepository/OauthConsentsRepository
replace OauthService's raw SQL (OA1-OA33). revokeChain's recursive CTE
and the oauth_clients join (its FK actually targets client_id, not the
row id the generated relation assumes) go through em.getKysely();
OA19's chain revoke is one nativeUpdate($in); OA12's consent grant is
em.upsert on the (client, user) unique key. revokeAllForUser built now
for Task 5. Every row-out read is disableIdentityMap: true; inserts go
through em.insert.
* fix(auth): non-numeric passkey and invite ids answer the legacy 404, never NaN; invite join type pinned
* fix(db): generator emits the referenced column for non-PK foreign keys; OAuth client joins through the relation; FK parity un-skipped
* refactor(db): TrekRepository base — every repository path validates the request context and reads bypass the identity map by default
* refactor(auth): AuthService and account cleanup through repositories
* refactor(oidc): OIDC identities and settings through repositories
* fix(auth): invite validation keeps expires_at null on the wire; password transaction scopes and reset-token burn pinned
* fix(server): plan 3b close-out — case folding on both sides for identifier lookups (non-ASCII login regression); coverage ratchet restored; repository scaffold and instanceof ratchet on TrekRepository; D-shape tests load-bearing
* refactor(server): trip-access primitives async at every caller
* refactor(trips): trip access through TripsRepository; demo seed in a request context; dialect helpers for the trips cluster
* refactor(days): days and day notes through repositories
* refactor(assignments): day assignments through repositories
* refactor(server): trip-membership, place-shadow, place-enrichment, place-photos and query helpers through repositories
* refactor(trip-members): membership, guests and the trip read model through repositories
* refactor(places): places, tags, ratings and place images through repositories
* fix(server): place-shadow source counts ordered in SQL; identity-map ratchets load-bearing; photo-cache race pinned
* refactor(places): import paths through repositories
* refactor(trips): trips through repositories
* fix(server): trip ids parsed once at the gate for places and assignments; NUL-safe value quoting on the SQLite platform; places parity tests completed
* refactor(trips): trip copy through repositories for the owned tables
* fix(server): plan 3c close-out — non-finite ids bound as NULL, trip id parsed once at every places/days/ws gate
Platform: NulSafeSqlitePlatform now renders a bound NaN as NULL and
+/-Infinity as 9e999/-9e999, matching better-sqlite3's own legacy binding
(rule 22 extension) — the program-wide fail-safe for any caller that still
passes an unguarded Number(untrusted) into a converted statement.
PlacesService.verifyTripAccess and RealtimeGateway.handleJoin stop doing
that: verifyTripAccess parses once with toRowId and threads the parsed id
through create/import/bulk/rating/export.gpx (deleting the seven
toRowId(tripId) ?? -1 sites), and handleJoin gets a Number.isFinite guard.
DaysService gets the same treatment across list/create/getDay/reorder/
insert/resyncAccommodationDays, closing the gap where a hex-spelled trip id
could renumber a real trip's days while the raw-bind inverted-accommodation
guard silently matched nothing.
DatabaseService.getTripTitle is deleted; PlacesService injects
TripsRepository directly and calls getTitle itself.
Template/cheap-item cleanup: DayAssignmentsRepository.listForTimeSort uses
$castTo instead of an unknown bridge; the duplicated TagRow interface in
Places.repository.ts is now imported from Tags.repository.ts;
AssignmentWithPlaceRow extends the guarded DayAssignmentRow instead of
restating its columns; PlaceShadowPicksRepository.totals drops its
unreachable ??-fallback branch; Places.repository.ts's updatePlace types
the NOT NULL name column as string; trip-members.service.ts's
transferOwnership threads one parsed trip id through all three writes and
corrects two stale "raw-bind seam" docstrings that were actually Number()
seams; a dead canAccessTrip mock is removed from reservations.e2e.test.ts.
Tests: the 23 new no-explicit-any warnings in TRIP-SVC-069 are typed
(lint back to 2416); HTTP and WS regression coverage for the non-finite
and non-canonical id shapes across places/days routes, mutation-proved;
TRIP-JOIN-005 asserts the exact concurrent-join settlement instead of
>= 1; a new tests/unit/db/transaction-yield.test.ts pins rule 24's
measured transaction-yield/dirty-read/nested-BEGIN behaviour.
* test(server): plan 3c close-out — copy-source full-row parity, Kysely rollback proofs, identity-map relabels, trips e2e on the snapshot DB
* fix(server): plan 3c close-out — days create/insert answer the guard 404 for non-canonical trip ids instead of 500; exact settlement and cross-request yield ratchets
* feat(server): plan 3d groundwork — dialect helpers, visibility predicate, feed and cron context ratchets
* refactor(roadtrip): roadtrip tables through repositories
* refactor(reservations): reservations, endpoints and travelers through repositories
* refactor(feeds): ICS feed tokens and anonymous feeds through repositories
* refactor(accommodations): stays through repositories inside the existing transactions
* test(server): plan 3d — MCP test helper constructors follow the accommodations and days services
* refactor(trips): trip copy through repositories for the roadtrip, accommodation and reservation tables
* refactor(calendar): calendar reads and reservation visibility through one predicate
* feat(server): plan 3e groundwork — packing visibility predicate, memories access-gate audit
* fix(server): plan 3d close-out — legacy text shape for stored accommodation ids restores the date-change restamp; reservation ids parsed once at the gate; parity tests for the reservation repositories
* refactor(photos): trek photos, memories access and cache sweeps through repositories
* refactor(files): trip files and file links through repositories, with transactions added
* test(server): plan 3e — hand-built FilesService, FilesRpc and AllowedFileTypesService calls follow the repository constructors
* refactor(memories): unified, immich and synology photo sync through repositories
* refactor(collab): notes, polls, messages and links through repositories
* refactor(budget): budget items, members, payers and settlements through repositories
* refactor(todo): todo items through repositories
* refactor(packing): packing items, bags and templates through repositories
* feat(server): plan 3f groundwork — systemNotices DI shape, TRAVELER_OWNS predicate, ntfy/storage-health specs, composite-PK pins
* refactor(school-holidays): country and region catalog through repositories
* refactor(notifications): in-app notifications and channel preferences through repositories
* refactor(notifications): reminder crons, mailer, ntfy and webhook transports through repositories
* fix(server): plan 3e close-out — repository registrations restore the boot; packing assignee and bag updates parse and no-op as legacy; coverage and scoping ratchets for the budget, packing and collab repositories
* refactor(vacay): vacation plans, entries, shares and holiday calendars through repositories
* refactor(system-notices): notice dismissals through a Nest provider and repository
* refactor(atlas): visited countries, regions and bucket list through repositories
* feat(server): plan 3g groundwork — GALLERY_CHRONOLOGICAL_ORDER Kysely ordering, unixEpochToIsoKysely helper, composite-PK DDL pins, JB5/R7 confirmations
* refactor(journey): journey and trip access control, CRUD and the trip-sync engine through repositories
* feat(server): plan 3h groundwork — date-offset dialect helpers, entity/repository fitness confirmations, pre-init and encryption pattern confirmations, survivor re-grep
* fix(server): plan 3f close-out — ntfy operator-scope test on the live send path; vacay date-offset and shift-order parity; todo reminder window; coverage and parity ratchets for the atlas and vacay repositories
* refactor(journey): journey stats, entries, photos, contributors and suggestions through repositories
* refactor(journey): public share links, TREK Studio books and the journal facades through repositories
* refactor(collections): access model, hydration, export/import and list CRUD through repositories
* refactor(server): journey-table survivors in trips, photos, memories, auth-cleanup and plugin-contributions through repositories
* refactor(integrations): dawarich connection, sync, tracks and review-flow through repositories
* refactor(collections): saved places, copy-to-trip, labels and invitations through repositories
* refactor(integrations,booking-import,airports,maps,route-usage): airtrail, booking import, airport backfill, maps cache and route-usage counters through repositories
* refactor(doc-sync): connections, reconciler and controller-SQL relocation through repositories
* test(integrations): add AirTrail saveSettings tx-asymmetry ratchet (Plan 3h Task 4 addendum)
* fix(server): plan 3g close-out — transactional book first-save, skeleton reconcile and share-link create; unknown-key entry patches are no-ops again; parity and authorization pins for the journey repositories
* feat(server): plan 3i groundwork — AuditLogRepository's reserved methods, the demo connection-swap spike, repository-fitness confirmations
* feat(server): plan 3j groundwork — scheduler onto CronRegistrarService, SDK-contract confirmation, repository fitness
* refactor(storage): backend/category config and usage stats through AppSettingsRepository; fix the seed-file test race and the shared-replica migration-flip gap
* refactor(share): public share links, the pre-init photo fallback, and collection-places survivors through repositories
* feat(server): plan 3j — the withCamelCase utility, unwired, pending a contract decision
* refactor(admin): user CRUD, stats, audit log, version check and addon config through repositories
* test(admin): pin createUser's pre-existing non-transactional uniqueness-check race (AD2-4)
* refactor(plugins): the trip-write role gate through a repository, in isolation
* refactor(demo,backup): seed and reset through DemoRepository, live-connection backup statements through MaintenanceRepository, isDemoUserId through an injected DemoService
* fix(server): plan 3h close-out — new-value-wins COALESCE restores stay/reservation propagation; doc-sync state filter truthiness; request-context and parity ratchets for share, doc-sync and the integrations
* refactor(plugins): runtime lifecycle, config CRUD and the uninstall cascade through repositories
* fix(server): plan 3i close-out — DemoModule imported where its service is consumed (e2e restored); admin addons list selects the provider relation; saveDemoBaseline awaited; backup tests on the maintenance repository
* refactor(plugins): outbound OAuth tokens and per-user plugin settings through repositories
* refactor(plugins): install-time gates, the capability-audit hash chain and the registry through repositories
* refactor(plugins): contribution-hook controllers, the small plugin controllers and the plugins-table survivors through repositories
* test(server): plan 4 — collapse mcp tool test harnesses onto the ORM snapshot
* test(server): plan 4 — collapse nest service test harnesses onto the ORM snapshot
* fix(server): plan 3j close-out — serialized audit-chain appends, atomic metadata and scheduler limits, rollback and race proofs for the plugin repositories
* refactor(server): plan 4 — convert the program's orphaned raw-SQL callers
* fix(server): plan 4 — un-skip PARITY-009 (FK deleteRule and index-name parity)
* refactor(server): plan 4 — inline canAccessTrip/isOwner off DatabaseService (trip-domain services)
* refactor(server): plan 4 — inline canAccessTrip/isOwner off DatabaseService (plugins + MCP surfaces)
* fix(server): plan 4 — RULE12 paren-count scanner ignores string literals
The entity generator's RULE12_fixGarbledCheckExpressions CHECK-expression
paren-balance scanner counted every `(`/`)` character regardless of
whether it fell inside a single-quoted SQL string literal. A CHECK
expression like `status IN ('confirmed)', 'tentative')` has a literal `)`
inside its first string value; the scanner would hit depth 0 at that
in-string character and truncate the expression, losing its own real
closing paren. The scanner now tracks single-quote string state (handling
the doubled `''` SQLite escape) and skips parens while inside one.
Latent-bug fix — no shipped entity's CHECK expression has hit this shape
yet (3d's ledger: "nothing affected today").
* refactor(server): plan 4 — unify the trip-access predicate's Kysely copies
TripsRepository.findForViewer/listForUser/activeTrip each hand-wrote the
identical trip-access boolean (t.user_id = ? OR m.user_id IS NOT NULL)
through the Kysely expression builder. Task 7's security review (M1)
evaluated a shared helper and set it aside for time, falling back to the
TRIPREPO-041 cross-method parity test as a regression guard instead.
Extracted the boolean into one generic helper, tripAccessExpr
(_shared/trip-access.ts), following the reservation-visibility.ts
publicReservationExpr precedent: fully generic over <DB, TB> since the
predicate has no correlated subquery, so it composes across
TripSelectKyselyDB and ActiveTripKyselyDB without needing a narrower
fixed-alias shape. The LEFT JOIN itself stays hand-written at each call
site — tripSelectQuery's own docstring already explains why a shared join
helper can't be typed across these differently-joined outer queries.
feeds.service.ts and memories-access.service.ts no longer duplicate this
predicate at all (Task 2's facade-inline sweep already converted both onto
TripsRepository.findAccessible/getFeedTokenIfReachable/
listReachableActiveTrips, which read through the canonical MikroORM
QueryBuilder form, accessibleTripsQuery) — the remaining duplication this
task closes is the three Kysely copies inside this file.
TRIPREPO-041 (64/64 in the file) stays green as the unification's own
regression guard.
* fix(server): plan 4 — trek_photo_cache_meta.cache_key NOT NULL
The column was a bare TEXT PRIMARY KEY, which SQLite does not implicitly
make NOT NULL (unlike an INTEGER PRIMARY KEY rowid alias). Every writer
(TrekPhotoCacheMetaRepository.upsertMeta, the table's only INSERT path)
has always bound a real string — 3e's R5 confirmed this. Appends a new
migration rebuilding the table with cache_key NOT NULL (SQLite has no
ALTER COLUMN, so this follows the same rebuild recipe
Migration20200101020000 uses in the opposite direction) and updates the
entity/generated metadata to match. Schema tightening, not a behaviour
change.
* refactor(server): plan 4 — retire the safely-convertible string-form where conditions (rule 23)
Places.repository.ts's existsInTrip/findInTrip/findRaw/reclaimInputs/
listByTripAndIds and Trips.repository.ts's findPublicForShare all took a
raw `.where('col = ?', [v])` string condition even though their id/trip_id
parameters were already plain `number` (or, for findPublicForShare,
narrowed to `number` here after confirming its one caller,
share.service.ts#getSharedTripData, only ever passes a number sourced
from a Kysely-typed share_tokens row, never an unconverted URL param) —
converted all six to MikroORM's typed `.where({...})` filter object.
The remaining raw string-form conditions in src/db/repositories/** are
NOT converted here: they are the `number | string` raw-bind seam
findAccessible/isOwner's own docstrings describe (trip_id/id threaded
through unconverted from req.params), which structurally requires the
deferred guard-level single toRowId parse to land first — narrowing any
of those methods' signatures now would either break the '0x10'/'007' id
parity those docstrings warn about, or duplicate the deferred item's own
work ahead of it landing.
* refactor(server): plan 4 — trek-photos.repository.ts finishes its rename
3e Task 6 renamed the CLASS (TrekPhotosRepository -> TrekPhotoRegistrationService,
since the name collided with the generated ORM TrekPhotos.repository.ts) but
left the FILE and its test file's names unchanged. Renames both to
trek-photo-registration.service.ts / .test.ts.
The old path is kept as a thin re-export stub (the src/websocket.ts shape)
rather than repointed at every one of its ~20 importers in this pass: several
of those files sit in two sibling tasks' exclusively-owned windows on this
shared branch (Task 3's budget/packing/todo/accommodations/places ripple,
reached through tests/helpers/plugin-host.ts and mcp-test-controllers.ts;
Task 5c's held-back tests/unit/services/{trekPhotoMedia,conflictUpdate}.test.ts)
and are off limits here. Flagged for a follow-up once those tasks land: delete
the stub and repoint every remaining importer at ./trek-photo-registration.service.
* fix(server): plan 4 — plugin discovery's delete+reinsert pairs are atomic
install/discovery.ts#upsert refreshed a plugin's actions and settings-field
descriptors as two independent delete-then-reinsert pairs (DI5-DI8) with no
transaction spanning either pair, let alone both: a crash between a delete
and its own reinsert left that plugin with an EMPTY actions or
settings-field set until the next discovery run, instead of its previous
still-valid rows. Wraps both pairs in one uow.transactional, so a discovery
refresh for a plugin's descriptors is now all-or-nothing.
Deliberate behaviour change, named for the user per the task brief.
DiscoveryRepos grows a UnitOfWork; PluginRegistryService (the other
production caller, registry.service.ts) picks it up the same @Optional()
way PluginRuntimeService already does, with the same guard-throw. Updates
the two hand-constructed test call sites that actually exercise discovery
(discovery.test.ts, registry.test.ts) to pass a real UnitOfWork.
* refactor(server): plan 4 — facade inline sweep part B (rosterUserIds/getPlaceWithTags)
Inline DatabaseService.rosterUserIds onto TripMembersRepository.rosterUserIds
directly in budget.service.ts, budget.mcp.ts and todo.service.ts, and
DatabaseService.getPlaceWithTags onto PlacesRepository.findWithTagsAndRatings
in accommodations.service.ts (placesRepo was already injected there).
BudgetService/BudgetMcp/TodoService collapse their constructors, dropping
the now-fully-unused DatabaseService param (budget.module.ts/todo.module.ts
register TripMembers on forFeature). PackingService keeps DatabaseService as
a vestigial param behind an optional TripMembersRepository with a fallback
in tripRosterIds: an in-flight Task 5c file
(tests/unit/services/conflictUpdate.test.ts) hand-constructs PackingService
positionally without the new repository, so its arity cannot change this
task; the fallback keeps that file's DatabaseService.rosterUserIds spy
working until Task 4 drops the param with the class itself.
* test(server): plan 4 — constructor ripple for the rosterUserIds/getPlaceWithTags inline
Hand-built BudgetService/BudgetMcp/TodoService call sites drop the removed
DatabaseService constructor arg (budgetRepoArgs() now appends a
TripMembersRepository so most spread call sites need no other change);
budget.service.test.ts's rosterHas() helper and svc() factory move from a
spied DatabaseService to a stubbed TripMembersRepository. The
AccommodationsService helper (tests/helpers/accommodations-service.ts) and
its two direct callers drop the same arg. PackingService call sites are
untouched (its new param is optional, so every existing hand-built instance
still resolves rosterUserIds through the DatabaseService fallback).
* fix(server): plan 4 — thread the discovery UnitOfWork through PluginRuntimeService too
Complements the just-landed discovery atomicity fix (DiscoveryRepos.uow):
PluginRuntimeService's own discoveryRepos getter still built the bundle
without a uow field. Threads its already-injected, @Optional() uow through
so its own onApplicationBootstrap/reload/dev-link discovery calls also get
the transactional DI5-DI8 refresh, not just PluginRegistryService's.
No throwing guard here, unlike PluginRegistryService's: several hand-built
partial-DI-graph test instances of this class (boot-registry-order.test.ts,
in a sibling task's held file window) construct it with uow omitted and
exercise onApplicationBootstrap's discovery call directly. Discovery must
never refuse outright just because a test built a partial graph, so it
degrades to the pre-existing untransacted sequence when uow is absent
(DiscoveryRepos.uow's own docstring covers this), the same "boot must
never block app init" defensiveness this file already applies elsewhere.
* fix(server): plan 4 — guard-level single parse: verify + correct the overclaiming verifyTripAccess docstring (L-2)
TripAccessGuard/TripOwnerGuard already parse :tripId exactly once
(Number(), guarded by Number.isFinite) and hand the resolved TripAccess
row to the handler via request[TRIP_REQUEST_KEY], TripOwnerGuard reusing
TripAccessGuard's own lookup when both are on the same route rather than
re-querying — verified by reading both guards directly, no change needed
there.
PlacesService.verifyTripAccess's own docstring separately claimed its
returned `tid` gets "returned... for every downstream call to reuse
(never re-parsing)" (3c's Task 9 review, L-2, carried to Plan 4). Reading
every call site (not assumed) shows this is false: `.tid` has exactly one
reader (PlacesController.requireTrip itself), which discards it — none of
requireTrip's 13 callers read `.tid` off the trip it returns; each passes
the original raw tripId string on to the write/read method it calls next,
which toRowId-parses it again itself. That second parse cannot be
eliminated by this gate alone: the same service methods are called
directly by places.mcp.ts too, an entry point with no requireTrip gate at
all (its own numeric tripId is re-stringified, String(tripId), to call
them), so each method owns its own id validation regardless of entry
point. Threading `tid` all the way through would mean giving every one of
~15 methods a second, number-only call shape for the REST path alone to
use — a real refactor, out of this docstring fix's own scope. Corrected
the docstring to say so plainly instead of repeating the false claim.
* refactor(server): plan 4 — relocate JS7/10/13/14 off the journey-share fallback stub
JourneyShareTokensRepository's own JS7/JS10/JS13/JS14 Kysely reads were a
3g Task 3 fallback (Task 2's JourneyPhotosRepository/JourneyEntriesRepository/
JourneyEntryPhotosRepository were still mid-flight). Now that those
repositories are stable, relocate: JS7/JS10 (photo/asset validation joins)
onto JourneyPhotosRepository (already injected into JourneyShareService as
photosRepo, no constructor change); JS13 (public entry list) onto
JourneyEntriesRepository.listPublicEntries; JS14 (public per-entry photo
read) onto JourneyEntryPhotosRepository.listForPublicJourney, deduped
against that repository's own JP_COLUMNS select list instead of a
hand-duplicated copy. JourneyShareService gains two new @InjectRepository
params (both already registered in journey-domain.module.ts's forFeature),
rippled through mcp-test-controllers.ts and journey-share.service.test.ts
(P04/P05 parity tests now pin the new homes).
* refactor(server): plan 4 — document unreachable ?? default arms after single-row aggregates
3f Task 7's fix wave flagged a cluster of `row?.x ?? default` arms after
unqualified, ungrouped aggregate reads (MAX/COUNT/COALESCE(SUM,0)) as
"same class as OauthClients:147" but left them unexamined per-site. Decided
per site: document, don't delete — TypeScript still types
executeTakeFirst()'s result as `| undefined` even though SQL guarantees
exactly one row for these unqualified aggregates, so the optional chain
can't be removed without a non-null assertion (worse, not safer). Six
sites: JourneyEntries.repository.ts (maxSortOrderForDate JG42,
countStatsPlaces JG67), JourneyEntryPhotos.repository.ts
(maxSortOrderForEntry JG91), JourneyPhotos.repository.ts (maxSortOrder
JG88/JG99), VacayEntries.repository.ts (sumFraction VC1, sumCompFraction
VC2). No behavior change — comments only.
* refactor(server): plan 4 — unify the two user-data-erasure enqueue paths
UserCleanupService.erasePluginUserData (the account-deletion transaction's
own UC2/UC3 enqueue) and PluginRuntimeService.enqueueUserErasure (the
emitUserDeleted sink's post-commit enqueue) independently walked the
identical SELECT id, permissions FROM plugins + JSON.parse + hook:user-data
filter + INSERT OR IGNORE — and both genuinely run for every account
deletion, since every emitUserDeleted call site (auth.service.ts,
admin.service.ts, trip-members.service.ts's guest removal) fires
immediately after deleteUserCompletely.
Extracted the shared filter into enqueueHookUserDataErasures
(nest/plugins/user-erasure-enqueue.ts) and pointed both call sites at it.
UC2/UC3 also come off DatabaseService onto PluginsRepository/
PluginUserErasureQueueRepository in the same change (they were the last
raw statements standing in the way of sharing one implementation) — this
narrows the Plan 3b Task 5 "stays raw, owned by Plan 3j" ruling to UC1
only (plugin_user_config/plugin_oauth_tokens/plugin_oauth_state), which
still stays raw, unaffected. UserCleanupService's orphan-data-directory
scan (a plugin uninstalled with data retained) has no equivalent on the
enqueueUserErasure side and stays there, now via the same repository.
Both enqueue calls still run per deletion (INSERT OR IGNORE keeps this
idempotent) — this unifies the IMPLEMENTATION, not the call count; not
touching the emitUserDeleted event-sink wiring itself was a deliberate,
smaller-blast-radius choice for this task.
* refactor(server): plan 4 — narrow TripMembersRepository.remove/TripsRepository.setOwner to number
Both accepted trip_id: number | string per 3c's carry note (their "stays
loose for other callers" reasoning), pending the guard-level single parse
landing first. Re-checked every caller directly: setOwner's one production
caller (trip-members.service.ts#transferOwnership) already passes trip.id,
a real number resolved earlier in the same transaction; remove's callers
are trips.rpc.ts (num(), already a validated number), trips.mcp.ts (Zod-
typed tool input, already a number) and TripMembersController.removeMember
(the one genuine raw-string caller, :id off the route with no class-level
TripAccessGuard on this controller by design).
TripMembersController.removeMember now parses :id once (toRowId, not
Number() — rule 15's NaN-into-SQL trap) and threads the parsed number to
both its own access check and removeMember, instead of passing the raw
string through twice. TripMembersService.removeMember and both repository
methods narrow to number accordingly; setOwner's WHERE also moves off its
raw-bind escape hatch onto nativeUpdate now that trip_id is typed, and
remove's trip_id half moves onto a typed .where({ trip: trip_id }) filter
(user_id keeps its own raw bind — a different, still-live NaN-safety seam,
unrelated to this narrowing).
TRIPREPO-018 and TMEMREPO-023, which pinned the retired string-bind seam
on setOwner/remove, no longer typecheck against the narrowed signatures
and are removed with an explanatory comment in their place.
* refactor(server): plan 4 — U6 gate-level id parsing for todo item ids
Rule 21's carry from 3e Task 8's review: TodoController.update/.remove
passed the raw :id route string all the way to TodoItemsRepository,
relying on SQLite's column-affinity CAST to match it instead of a typed
filter. Parse :id ONCE at the controller (toRowId, matching Task 8a's
TripAccessGuard/removeMember shape), 404 on a non-canonical id, and thread
the number down through TodoService.updateItem/deleteItem into
TodoItemsRepository.findInTrip/existsInTrip/update/deleteById, all
narrowed from number | string to number. todo.rpc.ts's num()-derived
todoId no longer needs its String() wrapper. tripId stays string | number
— a separate, already-accepted carry (23c292f2f). Two new e2e tests pin
the non-numeric-id 404 at the gate.
* test(server): plan 4 — full-key parity tests for ShareTokensRepository (3h L4)
3h L4's carry: ShareTokensRepository (SH1-SH6/SH17/UC6) had zero tests.
17 cases pinning every read against the legacy statement run raw on the
same connection — findRawByTrip/findTokenByTrip with every nullable
column both null and set, findValidByToken/findTripAndShareMapByToken/
findTripIdByToken's shared token+expiry predicate (valid, future-expiry,
past-expiry, unknown token), plus insertNew/updateFlagsByTrip/deleteByTrip/
deleteByCreator's write correctness (deleteByCreator's cross-trip erasure
kept distinct from deleteByTrip's single-trip scope, per the repository's
own docstring). The listPublicForShare family (the other half of 3h L4)
is a separate, larger carry — not in this file.
* test(server): plan 4 — full-key parity for Reservations share/public-api reads (3d carry)
Task 8b-2 item 1 (partial): the 8 share.service.ts/public-api.service.ts
reads on ReservationsRepository flagged by the 3d Task 7 review as having
no repository-level toEqual(<legacy raw>) coverage (listEndpointsForShare,
listDayPositionsForShare, listPublicForShare, listPublicAccommodationsForShare,
listScheduledForPublicApi, listUnscheduledForPublicApi,
listAccommodationsForPublicApi, listUnplannedPlacesForPublicApi). 12 new
cases against the legacy statement run raw on the same connection.
* test(server): plan 4 — concurrency pins for boundaries-upsert and token-mint races (3d carry)
Task 8b-2 item 2: R7's two remaining un-transacted-statement races flagged
by the 3d Task 7 review (L4) but never pinned — RoadtripDayBoundariesService
.save's belongs-check-then-upsert, and FeedsService.generateTripToken's
read-then-write. Both tests pin today's actual outcome (an atomic ON
CONFLICT upsert survives cleanly; the token mint has a last-write-wins
loser) with the same concurrent-request-shape Promise.all/allSettled
pattern roadtrip.service.test.ts's existing R7 vias pin uses. Not a fix.
* test(server): plan 4 — full-key parity for Days/DayAssignments/Places roadtrip reads (3d carry)
Task 8b-2 item 1 (remainder): the 3 non-share/public-api reads from the 3d
Task 7 review's missing-parity list — DaysRepository.listPlanDays (RPL1),
DayAssignmentsRepository.listRoadtripVisits (RPL2, the correlated-subquery
LEFT JOIN), and PlacesRepository.isTrackInTrip (RT13, the review's
"findTrackInTrip" — the actual method name). Completes the 11-file carry
together with 434113937's 8 share/public-api reads.
* test(server): plan 4 — dedupe overlapping listPlanDays/isTrackInTrip parity blocks
235f03665 landed two versions of the same DaysRepository.listPlanDays and
PlacesRepository.isTrackInTrip parity coverage side by side (a merge
artifact from concurrent work on the same carry item). Keeps one block per
method, drops the duplicate describe/it blocks and their colliding test ids.
* test(server): plan 4 — budget calc suite onto real repositories (3e carry)
* test(server): plan 4 — at-rest encryption pins for AirTrail/doc-sync/webhook secrets (3h L3)
Task 8b-2 item 4: the three credential columns the 3h Task 7 review flagged
as having no at-rest pin — users.airtrail_api_key (AirtrailService#saveSettings),
document_connections.secrets (DocSyncConfigService's whole-secret-map
column), and trip_document_links.webhook_secret (createLink's per-binding
mint). Each test saves through the real service, reads the raw column via
better-sqlite3, asserts it is enc:v1:-prefixed ciphertext that never
contains the plaintext, and asserts the service's own decrypt path rounds
it back to the original value.
* test(server): plan 4 — at-rest encryption pins for AirTrail/doc-sync secrets (3h L3 carry)
Task 8b-2 item 4: the 3h Task 7 review's L3 finding ("users.airtrail_api_key
has no at-rest pin", "document_connections.secrets ... has no enc:v1: check",
"trip_document_links.webhook_secret - no at-rest pin"). Three targets, each
proven by reading the raw column and asserting the stored value is
enc:v1:-prefixed ciphertext that round-trips, never the plaintext:
UsersRepository.setAirtrailSettingsWithKey/getAirtrailConnRow at the
repository layer, AirtrailService.saveSettings/getAirtrailCredentials at the
service layer, and DocSyncConfigService's connection secrets map plus
createLink's webhook secret.
* test(server): plan 4 — drop duplicate AirTrail encryption repository test
Task 8b-2 item 4 follow-up: 6965869dd already covers
UsersRepository.setAirtrailSettingsWithKey/getAirtrailConnRow's at-rest
ciphertext at the service layer (airtrail.service.test.ts's
AIRTRAIL-SVC-ENC-001/002) plus the doc-sync secrets/webhook pins
(DOCSYNC-ENC-001/002); this repository-level file duplicated the same
AirTrail assertion a second time.
* test(server): plan 4 — atlas read-model parity remainder (3f L6 carry)
Task 8b-2 item 3 (atlas half): full-key toEqual(<legacy raw>) coverage for
the atlas read models the 3f Task 7 review's L6 finding listed as missing
(minus AT1/TRAVELER_OWNS-consumer, already landed) — AT2 (Places.
listForTripIds), AT3/AT28 (PlaceRegions.listCountryCodesForPlaceIds/
listForPlaceIds), AT5/AT10/AT43 (VisitedCountries.listCodesForUser/
listForUser), AT21/AT24/AT25 (VisitedRegions.listForUser/
listRegionCodesForCountry/findCountryCode), AT22 (HiddenRegions.listForUser),
AT23/AT40/AT44 (PlaceRegions.listDistinctRegionCodesForCountryAndPlaces/
countPlacesByCountryForTrip/listVisitedCountryCodesForUser), AT30/33/34
(BucketList.listForUser/findById/findForUser), AT41 (Places.
listAddressesForUser), AT42 (Trips.countTripsAndDaysForUser). New standalone
files for HiddenRegions/VisitedCountries/VisitedRegions/PlaceRegions, which
had none. AT6/45/46 (the TRAVELER_OWNS endpoint reads) are proven on ids
only by the existing consumer test — full-row parity for those three is
NOT done here (carried forward, same as the review's own open item).
* test(server): plan 4 — vacay read-model parity remainder (3f L6 carry)
Task 8b-2 item 3 (vacay half): full-key toEqual(<legacy raw>) coverage for
the vacay read models the 3f Task 7 review's L6 finding listed as missing
(minus VC81/82, already landed) — VacayPlans.findByOwner/findById (the
getPlanData/getStats composite's own plan read), VacayYears.listForPlan
(listYears), VacayCompanyHolidays.listForPlan (the company-holidays list),
VacayHolidayCalendars.listForPlan/findById (holiday-calendar reads),
VacayUserSettings.findForUser (user year settings), VacayUserYears.
findForYear (getStats rows, including the NULL vacation_days case the
review named explicitly), and VacayPlanMembers.listAvailableForFusion/
listPendingForPlan/listPendingForUser (the fusion and share pickers). Seven
new standalone files — only VacayShares had a repository-level test before
this. getSharedCalendars itself is a service-level composite over these
same reads plus VacayCompanyHolidays.listDatesForRange (already exercised
elsewhere); its own dedicated test is NOT added here.
* test(server): plan 4 — full-key parity for PackingItemsRepository.listPublicForShare (3h L4)
Task 8b-4a: no repository-level toEqual(<legacy raw>) coverage existed for
the share.service.ts SH13 read. Seeds Common items with every nullable
column both null and set, plus Personal and Shared-with-people items on
both sides of the is_private/owner/recipient predicate, and proves only
the Common tier surfaces.
* test(server): plan 4 — full-key parity for BudgetItemsRepository.listPublicForShare (3h L4)
Task 8b-4a: no repository-level toEqual(<legacy raw>) coverage existed for
the share.service.ts SH14 read. Seeds two items — every nullable column
set and every nullable column null — plus a foreign-trip item, and proves
the ORDER BY category ASC shape (unlike BG72's listAllForTrip, which has
none).
* test(server): plan 4 — full-key parity for CollabMessagesRepository.listPublicForShare (3h L4)
Task 8b-4a: no repository-level toEqual(<legacy raw>) coverage existed for
the share.service.ts SH16 read. Seeds a root message, a reply (reply_to
set), a deleted message (must be excluded) and a foreign-trip message, and
proves the narrower username/avatar-only join shape (no reply_text/
reply_username, unlike joinedQuery).
* test(server): plan 4 — full-key parity for PlacesRepository.listPublicForShare (3h L4)
Task 8b-4a: no repository-level toEqual(<legacy raw>) coverage existed for
the share.service.ts SH12 read. Seeds a categorised place with every
nullable column set and an uncategorised place with every nullable column
null, plus a foreign-trip place, and proves the named 20-column allow-list
never leaks an owner-only column (reservation_status, google_place_id,
…).
* test(server): plan 4 — full-key parity for DayAssignmentsRepository.listPublicForShare (3h L4)
Task 8b-4a: no repository-level toEqual(<legacy raw>) coverage existed for
the SH9 read. Seeds assignments across multiple days (one excluded day_id
never leaks in), every nullable place/assignment column both null and
set, and proves the six owner-only columns SharePublicAssignmentRow omits
(google_place_id, google_ftid, osm_id, amap_poi_id, stop_type,
fill_percent) never appear on the row at all.
* test(server): plan 4 — collapse remaining test harnesses; triage legacy migration tests
* test(server): plan 4 — todo controller test expects the gate-parsed item id (U6 follow-up)
* fix(server): add the missing numbered migration for the booked-night reseat step (#242)
reseatBookedNights (legacy step 242) was never invoked from any shipped
MikroORM migration -- only from the retired db/migrations.ts array --
so an install upgrading through the numbered chain never got the
one-time reorder. Adds Migration20200101040300, which calls the frozen
function through the native better-sqlite3 handle so it runs inside
the migration's own transaction.
orm-driver.ts's BoundSqliteConnection now also captures that handle
onto the field SqliteConnection.getNativeClient() reads, which its
createKyselyDialect() override previously left unset -- needed for
this migration (and any future one) to reach it in production, not
just under the bare test driver.
RESEAT-009 moves off the legacy createTables+runMigrations builder
onto the real Migrator via tests/helpers/migration-step.ts, plus a
new fresh-install guard (RESEAT-011).
* test(server): plan 4 — swap 4 of 5 straggler nest test harnesses off legacy schema builder
day-notes.service.test.ts, public-api-scopes.test.ts,
registration-invites.service.test.ts, trip-invite.service.test.ts onto
createSnapshotTestDb() via the async vi.mock idiom (5a/5b/5c precedent),
dropping their own hand-rolled createTables()+runMigrations() harness.
trip-invite.service.test.ts and registration-invites.service.test.ts
also drop their now-redundant DatabaseService wrapper (its .connection
was always just the same test db handle). expect( assertions unchanged.
user-cleanup.service.test.ts held back: its constructor call was
updated to match UserCleanupService's new EntityManager-based shape,
but deleteUserCompletely() now deadlocks against a real UnitOfWork
(reproduced in isolation, independent of this test file -- see the
report), which looks like a genuine bug in erasePluginUserData's new
MaintenanceRepository(em) call versus UnitOfWork.transactional(), not
a test-harness problem.
* test(server): plan 4 — move auto-backup.test.ts and demo-reset-path.test.ts off createTestDb()
Both build a schema db while 'fs'/'node:fs' are mocked (wholesale in
auto-backup.test.ts, selectively spied in demo-reset-path.test.ts) for
their own archiver/filesystem assertions, which broke
createSnapshotTestDb()'s read of the schema snapshot through
schema-snapshot.ts's own fs import.
auto-backup.test.ts: fs is mocked wholesale, so the snapshot db is now
built via a small helper that points fsMock's three calls at the real
filesystem (fetched through vi.importActual, bypassing the mock) just
long enough to build the db, before the test installs its own
existsSync stub -- never touching schema-snapshot.ts itself.
demo-reset-path.test.ts: fs.existsSync is selectively spied to answer
only for the baseline path, so the db is now built before that spy is
installed instead of after.
* test(server): plan 4 — retire the legacy-runner-only migration hygiene checks
migration-hygiene.test.ts deleted outright: every block in it (the
destructive-op allowlist, the empty-catch guard, the full-chain smoke)
scans or runs the legacy db/migrations.ts array being deleted this
plan, never the numbered src/db/migrations/*.ts chain, so there is
nothing in it to carry forward.
leg-mode-incoming.test.ts: deleted its first describe block (the
legacy-runner column-add + replay-safety smoke, also against
db/migrations.ts); the read-path parity block, already on the modern
createSnapshotTestDb() harness, is untouched.
* chore(server): delete the legacy schema/migration/seed builders
MikroORM's migrations are the only schema source now, including for
tests: src/db/schema.ts, src/db/migrations.ts and src/db/seeds.ts are
gone, along with tests/unit/db/schema-parity.test.ts (the parity gate
that guarded them, already red on 8a's cache_key NOT NULL) and
createTestDb() + its now-dead imports out of tests/helpers/test-db.ts
(its last two callers moved onto createSnapshotTestDb() this plan).
Their three entries drop from eslint.config.mjs's better-sqlite3
allow-list -- the files they exempted no longer exist.
Ruling 7's frozen files (migration-utils.ts, reseat-booked-nights.ts,
document-provider-seed.ts) and every numbered migration in
src/db/migrations/ are untouched.
Precondition grep (server/tests + server/src, from the task brief) is
now empty:
grep -rlE "from ['\"].*db/schema['\"]|from ['\"].*db/migrations['\"]|from ['\"].*db/seeds['\"]" server/tests server/src --include="*.ts"
* docs: plan 4 — the repository layer is the only data access path
DatabaseService/DatabaseModule are gone; queries go through repositories
(server/src/db/repositories/) and UnitOfWork (server/src/nest/database/
unit-of-work.ts) for transactions. MikroORM's migrations are the only
schema source, including for tests — the hand-kept schema.ts/migrations.ts/
seeds.ts test builders and their parity test are deleted. Documents the
db/database.ts (connection only) / db/orm-driver.ts (bound driver) /
UnitOfWork / repository shape, withRequestContext for non-HTTP entrypoints,
the better-sqlite3 ESLint restriction, and the entity generator (gen:entities/
check:entities) as the schema-parity guard. Corrects two stale nest/README.md
mentions (realtime.gateway.ts's now-repository constructor, the BE-Phase 1
trip-access bullet) and marks the legacy src/services/ migration recipe as a
historical changelog rather than current guidance.
* refactor(server): plan 4 — drop the 13 vestigial DatabaseService params/getters
Removes the now-dead DatabaseService injection (or private get db() getter)
from addons.service.ts, trips.service.ts, trip-invite.mcp.ts,
place-photo-cache.service.ts, trek-photo-registration.service.ts,
categories.mcp.ts, journey-domain.service.ts, plugins.service.ts,
plugin-runtime.service.ts, plugin-mcp-tools.service.ts, places.service.ts
and packing.service.ts — every one of them was fully repository-backed
already (Task 2/3's facade inline sweep), just still carrying the param.
PackingService's TripMembersRepository is required now (the
this.db.rosterUserIds fallback is gone with it) and PluginRuntimeService's
uow is no longer @Optional() — both moved ahead of the optional
registry?/hostFactory? params, since TypeScript refuses a required
parameter after an optional one; install/discovery.ts's DiscoveryRepos.uow
is required to match, and its upsert() always runs the descriptor
refresh inside one transaction instead of falling back to an
un-transacted sequence.
user-cleanup.service.ts's UC1 (the one genuinely live raw site, Plan 3b's
"stays raw" plugin-table trio) moves off DatabaseService onto a new
MaintenanceRepository.deletePluginUserData method — the rule-4-permitted
connection.execute() home, built from a directly-injected EntityManager.
Threads the active UnitOfWork.transactional's transaction context through
execute()'s ctx param: without it, a raw connection.execute() call issued
from inside deleteUserCompletely's open transaction contends for the same
single-connection checkout the transaction already holds and deadlocks
(never released, since the transaction is itself awaiting the call).
tests/unit/db/transaction-yield.test.ts's probe B already pinned this
exact hazard class for a raw better-sqlite3 statement; this is the same
mechanism reached through the ORM's connection pool instead.
nest/photos/trek-photos.repository.ts (the Plan 4 Task 8a re-export shim)
is deleted, with every importer repointed to trek-photo-registration.service
directly. The ~7 files that only imported the TripAccess/PlaceWithTags
type re-export from database.service.ts now import them from
Trips.repository.ts/Places.repository.ts directly, the actual single source.
* refactor(server): plan 4 — delete DatabaseService, DatabaseModule and database.tokens.ts
Grep gate confirmed zero src/ importers of DatabaseService left outside
these three files (Task 1/2/3's earlier conversions + this task's own
vestigial-param sweep). Deletes database.service.ts, database.module.ts
and database.tokens.ts, drops DatabaseModule from app.module.ts's imports
and the five other modules that still listed it
(accommodations/calendar/doc-sync/feeds/trip-members) — none of them had
a live DatabaseService injection left, so the import was already dead.
OrmModule is the module that survives DatabaseModule's deletion, per the
Phase 0 ledger's own ruling; untouched.
db/database.ts's `db` Proxy export itself stays, deliberately: ~190 test
files' vi.mock('…/db/database', …) factories return a { db: <mock
handle>, … } shape, and their own `import { db as testDb } from
'…/db/database'` is typed against this file's real exports regardless of
the runtime mock — vi.mock swaps the module at runtime, but tsc resolves
the import statically. Deleting the export would fail typecheck:tests
across the whole suite for a test-harness idiom this task's file set
doesn't own (Track B's territory). src/ has zero remaining importers, so
the ESLint no-restricted-imports gate this task also shrinks still holds
this file to its permanent allow-list entry either way.
* test(server): plan 4 — repoint the test suite off the deleted DatabaseService/DatabaseModule
Mechanical follow-through of the previous two commits across ~110 test
files:
- The 41 tests/e2e/*.e2e.test.ts suites composed DatabaseModule into their
partial Test.createTestingModule graphs; the import and the array entry
are dropped (nothing in any composed domain module injected
DatabaseService any more, so the entry was already dead weight).
- The shared test helpers (mcp-test-controllers.ts, plugin-host.ts,
accommodations-service.ts, notifications.ts, test-addons.ts,
test-uow.ts) each held a DatabaseService/DatabaseService-shaped param
used only for its raw better-sqlite3 handle; every one now takes that
handle directly, and every constructor call site the helpers make
(PackingService, PlacesService, JourneyDomainService,
PlacePhotoCacheService, TrekPhotoRegistrationService, TripsService,
UserCleanupService, AddonsService, PluginsService, PluginRuntimeService)
drops the removed param and, where PluginRuntimeService's uow is
involved, reorders it ahead of the optional registry?/hostFactory?
params to match the real constructor.
- ~60 further unit/integration test files hand-constructed one or more of
those same services directly; same fix, file by file. Four of them
(memories-access.service.test.ts, places.directions.test.ts,
places.service.test.ts, unified-memories.service.test.ts) also drop a
vi.spyOn(DatabaseService.prototype | instance, 'canAccessTrip' | 'isOwner'
| 'rosterUserIds' | 'getPlaceWithTags') spy left over from when those
methods delegated through DatabaseService — dead now that the services
read the owning repository directly.
- database-service.test.ts and database-service.delegation.test.ts tested
DatabaseService directly and are deleted with it.
- tests/unit/db/transaction-yield.test.ts's probe B, which pinned a raw
DatabaseService statement's dirty-read behaviour inside an open
transaction (rule 24), now issues the same raw better-sqlite3
statement directly against the shared test handle — the exact shape
DatabaseService.get/run used to wrap (this.conn.prepare(sql).get/run(),
bypassing MikroORM/Kysely entirely) — so the hazard it documents is
unchanged. Verified still green.
No expect(...) assertion changed in any file; only construction/import
plumbing. typecheck, typecheck:tests and the full touched-file run
(101 files / 2668 tests) are green, plus the compiled boot
(bootstrap.test.ts + boot-sweeps-request-context.test.ts) and one real
e2e boot file (trips.e2e.test.ts).
tests/unit/nest/user-cleanup.service.test.ts is deliberately NOT in this
commit — see the next one.
* test(server): plan 4 — finish user-cleanup.service.test.ts's harness swap; regression-test the UC1 transactional deadlock fix
This file arrived already modified and unstaged in the shared tree —
Task 6 (the parallel legacy-schema-deletion task) had swapped its DB
harness onto createSnapshotTestDb() and updated the UserCleanupService
construction to this task's new EntityManager-first shape while landing
its own commits. Finished here: USER-CLEANUP-005 ("survives a slim
schema without the plugin tables") passed a fake `{} as unknown as
EntityManager`, which doesn't reach the real UC1 path any more (the
previous raw DatabaseService.run swallowed everything in a try/catch
regardless); it now builds a real EntityManager bound to the slim
handle via sharedTestOrm(slim), so the per-statement query against the
missing plugin_user_config/plugin_oauth_tokens/plugin_oauth_state
tables is what's actually being proven caught, not a construction-time
throw.
USER-CLEANUP-006 through -010 (deleteUserCompletely, all of them
exercising UC1 inside the real uow.transactional wrap) are the
regression test for the previous commit's deadlock fix in
user-cleanup.service.ts/MaintenanceRepository.ts: before that fix, every
one of these hung and timed out. Green now, verified directly.
* chore(server): plan 4 — shrink the better-sqlite3 ESLint allow-list
Per ruling 6: database.service.ts's entry comes off (the file is
deleted). The five plugin entries (plugin-route-normalize.ts,
plugin-host-state.ts, install/discovery.ts, settings-defaults.ts,
signature-status.ts) come off too — re-verified none of the five import
better-sqlite3 or DatabaseService any more (3j's own conversion). The
src/demo/** blanket entry was already gone (3i's own task landed it).
reseat-booked-nights.ts and document-provider-seed.ts stay, with the
`// TODO(plan 4)` comment corrected: they're frozen forever (ruling 7),
invoked from inside numbered, already-shipped MikroORM migrations, never
Plan 4's or any future plan's to touch.
Permanent allow-list after this commit: src/db/database.ts,
src/db/orm-driver.ts, src/db/durability.ts (the driver homes),
src/nest/plugins/host/plugin-data.service.ts (3j's R-out-of-scope — a
plugin's own sqlite file is never TREK's schema), src/nest/backup/backup.impl.ts
(3i's R1, exactly the untrusted-upload-file probe), src/db/reseat-booked-nights.ts
and src/db/document-provider-seed.ts (frozen forever). src/db/{schema,migrations,seeds}.ts
entries were already removed by Task 6 when it deleted those files.
0 new lint errors; warning count unchanged at 2123 (well under the 2247
lean-gate ceiling) — every file the shrink touches was already fully
converted.
* refactor(server): plan 4 — U6 gate-level id parsing for budget item/settlement ids
Rule 21's 3e carry, same discipline 218f75d0f applied to todo: every
:id/:settlementId/:userId route param in BudgetController is parsed ONCE at
the gate with toRowId, matching the legacy affinity-seam "not found" outcome
for a malformed id on each route (404 'Budget item not found'/'Settlement
not found', except toggleMemberPaid — the one legacy handler that never
404'd an unknown item/member either, so a malformed id there skips the write
and keeps the same 200 { member } + Number(id)/Number(userId) broadcast
shape it always had).
The parsed numbers thread down through BudgetService's update/updateMembers/
setPayers/toggleMemberPaid/remove/updateSettlement/deleteSettlement (and
their updateBudgetItem/deleteBudgetItem/setItemPayers/getBudgetItem/
applySettlementUpdate/getSettlement internals) into BudgetItemsRepository's
getCurrency/findInTrip/existsInTrip/findForDelete/findById/update/deleteById
and BudgetSettlementsRepository's findWithUsers/findGuard/update/deleteById,
all narrowed from number | string to number; trip_id stays loose, a
separate, already-accepted carry. costs.rpc.ts's num()-derived itemId no
longer needs its String() wrapper on the update/delete calls; budget.mcp.ts
was already Zod-typed to number throughout. Two new e2e tests pin the
non-numeric-id 404 at the gate; controller/service unit tests cover the
narrowed call args and the toggleMemberPaid malformed-id parity case.
* refactor(server): plan 4 — U6 gate-level id parsing for packing item/bag/template ids
Rule 21's 3e carry, same discipline 218f75d0f applied to todo: every
:id/:bagId/:templateId/:userId route param in PackingController is parsed
ONCE at the gate with toRowId, matching the legacy affinity-seam "not found"
outcome for a malformed id on the routes with a pre-ORM Express precedent
(update/remove -> 'Item not found', updateBag/deleteBag/setBagMembers ->
'Bag not found', applyTemplate -> 'Template not found or empty').
#858's sharing/contributors routes (setSharing/clone/addContributor/
removeContributor) are native Nest code with no pre-ORM precedent (landed in
7eabf6066, after the migration) — a malformed id there just 404s the same
way an unknown one already does. removeContributor's :userId used to reach
contributorsRepo.deleteOne through a bare Number.parseInt(), the exact
NaN-into-SQL trap row-id.ts documents; it now parses through toRowId too.
The parsed numbers thread down through PackingService's updateItem/
setItemSharing/addContributor/removeContributor/cloneItem/deleteItem/
getItemPrivacy/updateBag/deleteBag/setBagMembers/applyTemplate, all narrowed
from number | string to number; tripId stays loose, a separate,
already-accepted carry. PackingItemsRepository's getPrivacy/findVisibleInTrip/
update/updateSharing/delete/findById, PackingBagsRepository's findInTrip/
update/findWithAssignee/delete and PackingTemplateItemsRepository.listForApply
narrow the same way. packing.rpc.ts's num()-derived itemId/bagId no longer
need their String() wrappers on the update/delete/updateBag/deleteBag/
setBagMembers calls; packing.mcp.ts was already Zod-typed to number
throughout. Two new e2e tests pin the non-numeric-id 404 at the gate;
controller/rpc unit tests cover the narrowed call args and the
removeContributor malformed-userId parity case.
AdminPackingTemplatesController (/api/admin/packing-templates) is out of
scope — a separate admin-gated surface with its own string-typed template
CRUD, not part of this trip-scoped controller's :id chain.
* test(server): plan 4 review — three pins catch up with the gate-parsed ids and the cache_key NOT NULL (M1–M3)
* fix(server): boot baselines legacy-runner installs before running MikroORM migrations (B1)
An install the retired positional runner migrated has a schema_version row
and no mikro_orm_migrations rows, so its first boot on this branch replayed
all 244 migrations over an already-migrated schema: eight refuse the boot,
and the replay-safe ones overwrite user data (step 26 rewrites every
day_assignments.assignment_time).
runSchemaBootstrap now records the baseline plus every migration whose
"Legacy migration step N" docstring is <= schema_version as executed, in one
transaction through the migrator's own storage, before the ordinary run
applies the rest. The step map is derived from the discovered migrations'
source, never hand-kept, and anything unmappable (a newer or malformed
schema_version, a duplicated, missing or out-of-order step) refuses the boot.
* test(server): boot the app on three legacy-runner fixtures to pin the B1 baseline
SQL dumps of databases the retired runner itself built: schema_version 205
(main at fa406b708), 242 (417d0445b, the runner's last commit) and 82
(v2.9.14, so seven of the eight non-replay-safe migrations still lie ahead
and run for the first time). Each gets canary rows: user-set assignment
times and a booked night.
buildApp() on each asserts the baseline and the applied rest in order, the
assignment times untouched, the booked-night reseat run exactly once on 205
and 82 and not at all on 242, schema_version kept, no half-built *_new
table, a schema identical to a fresh install's, and an idempotent re-run.
* refactor(server): demo-reset reads the live path off getRawConnection, not the db Proxy (m1)
demo-reset.ts was the one src/ consumer of the db Proxy left, reached
through a runtime require that a grep on `from` misses. It now imports
closeDb/getRawConnection/reinitialize statically, the way backup.impl.ts
does, so the Proxy genuinely has no src/ importer (the claim in database.ts
and server/CLAUDE.md now holds). The path test mocks the module with vi.mock
instead of teaching Node's require loader to answer for a .ts file.
* test(server): renumber the duplicate ASSIGNREPO/PLACEREPO test ids Plan 4 introduced (m3)
ASSIGNREPO-031/032 (the legacy-statement parity pair) become 034/035, and
the later PLACEREPO-030..038 block becomes 053..061, the next free ids, so
each id names one test again. No test body changes.
* refactor(server): derive the two restated public row types instead of copying their columns (m4)
JourneyPublicEntryPhotoRow restated JourneyPublicGalleryRow's nineteen
columns with only the parent key changed; it is now that row with
journey_id swapped for entry_id. PublicApiAssignedPlaceRow and its Kysely
table shape restated PlaceRow's columns; both now Pick them off PlaceRow.
Types only. The review's CPD method over Plan 4's new src lines drops from
about 3.4% to 0.64%.
* test(server): full-key parity for TripsRepository.findPublicForShare and VacaySharesRepository.listAvailableForShare (m5)
Neither read had any test reference. Each is now pinned with
toEqual(<legacy statement run raw on the same connection>), the
listPublicForShare family shape: SH7 on a filled and a NULL-heavy trip plus
a missing id (TRIPREPO-067/068), and VC90 over one world exercising every
exclusion (caller, guest, existing share, plan owner, accepted member) and
a missing plan.
* fix(server): the legacy baseline runs createTables() instead of marking it, atomically with the run (N1)
The legacy runner ran the running release's createTables() on every boot
before its pending steps, and about 30 tables exist only there. Marking the
baseline migration as executed left installs older than those tables (below
legacy step 68) failing with "no such table: photo_provider_fields" and, since
the rows were committed before the run, stuck half-baselined for every later
boot.
Only steps 1..N are recorded now, so the baseline's IF NOT EXISTS DDL runs
first, as the legacy boot did. The rows are logged inside the same
transaction as migrator.up({ transaction }), so a failed first boot records
nothing and the next boot baselines again.
A v66 fixture (the runner at 5be2e9b26) boots: 66 recorded, 178 applied,
canaries kept. Its two schema differences from a fresh install
(users.immich_access_token, reservations.accommodation_id INTEGER) come
from the legacy chain itself and are pinned as carried.
* fix(deps): sync the lockfiles with the bumped manifests and follow lucide's icon renames
feat/orm raised react to ^19.3.0 and several plugin-sdk dependencies without
regenerating the lockfiles, so npm ci failed on every CI job. The root
overrides now pin react 19.3.0, both lockfiles are regenerated with npm 11,
and the tests that select lucide icons by their old class names use the
current ones. Two tests cover the Immich prefs reads and the boot restore
branch that the coverage thresholds were missing.
* ci: give the server lint step an 8 GB heap for the type-aware rules
* test(server): keep the no-promise-as-value rule tester out of single-run mode on CI
With CI=true typescript-eslint infers single-run mode, builds the program
from the fixture file on disk and ignores the code each case passes in, so
all 24 invalid cases reported nothing on the runner. Both testers now set
disallowAutomaticSingleRunInference.
* refactor(server): name the new migrations after the first seven words of their purpose, like the rest
* fix(server): keep tsx a production dependency so image key rotation can load it
* test(server): re-pin the repositories branch floor at the measured 89.83%
* test(planner): keep the sidebar fixtures on past dates so the jump to today never fires in jsdom
* test(server): keep console output for failing tests only and teach Nest's logger LOG_LEVEL
* test(client): keep console output for failing tests only and answer the health probe
* test(client): drop the health handler, it races the admin page's first load
* chore(sonar): assert explicitly in two e2e cases, Number.NaN, and ignore S2187 on the legacy-upgrade fixtures
---------
Co-authored-by: Maurice <mauriceboe@icloud.com>
The Tours surfaces now use the same building blocks as the rest of the
planner: tinted head bands, eyebrow labels, white field boxes and pills,
place-style list rows and the inspector frame for the map detail.
- Tour detail on the map uses the place inspector's frame, head band,
labelled metric fields, collapsible track colour and footer buttons
- Tour rows in the Plan sidebar and the Tours tab look like place rows,
with the facts as pills and a "+" that opens the trip's days as a menu
- Tours tab rails get the planner head bands, editor fields and
waypoint rows in the day card style; notices share one look
- Map hint shortened to one line and the empty state reads as a single
sentence in all locales
- New "tours" mascot scene with a pack, a trekking pole, hills and a
trail, used by the Tours empty states
PlacesSidebarHeader uses the UNCATEGORIZED constant, placesBrowserModel stops
re-exporting matchesCategoryFilter (its test imports it from utils/placesFilter),
and the orphan comment in the en places strings joins the places.filters key.
The panel button is named "Filters" like the map control, its badge counts
kinds of filter through countActivePlacesFilters (the pool, shown as chips
above it, stays out), and picking a rating floor clears the selection as
picking a pool does.
The header claimed the lists and the map always agree. They apply the same
predicate to the same values, but the map narrows "planned" to the open day
(#2024) and leaves collapsed days' stops off.
MChip only showed its state through colour. A pressable option renders
aria-pressed from active; the places filter's Show and rating floor chips use
it, and the tests now check the chosen chip by its pressed state.
MapTogglePill always set aria-pressed, so the phone map's Filters control read
as a toggle although it opens a sheet. An opensDialog option drops aria-pressed
for aria-haspopup="dialog"; the other map toggles are unchanged.
The fallback from "Tracks" to "All" once the last track disappears lived in
the desktop sidebar and the phone places browser, neither of which is mounted
while the phone map is in front: the map could sit empty under a filter no
control offers any more. useTripPlanner now owns it, once, for every screen.
Map-Features gets a "Filtering the places on the map" section: the markers
follow the places list's Show, category and rating filters, and on a phone
the Filters button of the plan map opens the same choices in a sheet.
Places-and-Search says the rating floor narrows the map markers too.
The sheet opens over the phone map, and with the default bar glass the map's
place names showed through the pool chips, the rating floors and the category
rows, which made them hard to read at 390px wide. It now uses the opaque sheet
material, like the other menus that open over busy content.
On a phone the only way to narrow the trip's pins was to leave the map for
the places list, and the rating floor could not be set at all. The phone map
now has a round "Filters" control at the foot of its right-hand stack, badged
with how many filters are on, opening a sheet with the pool (all / unplanned /
planned / tracks), the minimum rating (the desktop list's floors), the trip's
categories including "no category", and a reset. It is hidden on the road trip
stage, which draws its own pins.
The places browser's filter panel gains the same rating floor; its category
rows move into MPlacesFilterControls so the panel and the sheet share one
copy. Everything writes the trip store, so list and map stay in step.
MapTogglePill takes an optional count badge.
The minimum-rating filter lived in the desktop sidebar's local state, so it
only ever narrowed the desktop list: the map markers and the phone's places
list ignored it. It now sits in the trip store beside the pool and category
filters, and one module (utils/placesFilter) holds the pool, category and
rating matchers that the desktop list, the phone list and mapPlaces all use,
instead of three hand-copied versions of the same checks.
overflow-wrap:anywhere alone lets a long URL wrap inside a flex row; break-words
is overridden by it and word-break:break-word is deprecated. MarkdownText already
gets the rule from .collab-note-md, so its call sites drop the duplicate.
- Make the “Manage preferences” text in notification emails clickable.
- Link directly to `/settings?tab=notifications`.
- Preserve the localized link text and existing email styling.
- Add a regression test for the generated preferences URL.
A flight or train with several legs comes out of the calendar feed as one event per leg, titled with its route and carrying the leg's own departure, arrival and segment reference, so the layover shows as the gap between them. A booking with a leg missing its departure time stays one event, as before. (#2389)
Admin > Notifications gets a Defaults for users matrix: each event and channel starts On, Off (a user may still turn it on) or Blocked (off for everyone). A user's own choice is kept against the default, and a value equal to it is stored as no row, so a later change of the default reaches everyone who never chose. Blocked cells show as locked in the user settings on desktop and phone, and the server ignores any attempt to turn one on. (#1536)
Admin > Settings > Google options gets a daily limit with today's count beside it. Every call through the Places and Routes clients is counted per UTC day; once the limit is reached the Google key reads as absent until the next day, so search, details and photos fall back to OpenStreetMap the way a keyless install does, and the admin sees that the day is used up. (#1582)
Place photos, link previews, webhooks and the integrations fetch through the guard's own dispatcher and so ignored the outbound proxy, which left a server that can only reach the internet through one without them. They now read HTTP_PROXY, HTTPS_PROXY and NO_PROXY themselves and connect through the proxy, after the guard has vetted the target. The Helm chart passes the three variables through. (#1754)
The 50 MB limit for files, booking attachments and note attachments becomes FILE_UPLOAD_LIMIT_MB, 50 by default. The server hands it to the client through the app config, so the upload dialogs refuse a file over it up front and say what the limit is; on desktop a file over it no longer vanishes without a word. (#1364)
While a trip is running, today's chip on the phone carries a ring and a dot, and a button beside the rail jumps back to today from wherever it was scrolled, in the plan and on the map. It stays in place, dimmed, while today is open, so the rail never shifts under the thumb. (#2392)
The phone's collection place sheet gets the Navigation button the desktop detail and the trip's place sheet already have, first among its actions: one offered app opens straight away, more open the same picker. (#2091)
The handles on the day plan and the places panel only listened to the mouse, so on an iPad they did nothing, and in the portrait band they were left out. They now follow a finger as well, show a grip on touch screens, keep the page and the map from taking the gesture, and stay in the narrow layout within the room the map needs. As separators they also move with the arrow keys. (#1012)
The password policy moves to @trek/shared, where the server keeps enforcing it unchanged and every form that sets a password now runs it before sending: the forced change after the first login, registration, the reset link, the account settings on desktop and phone, and creating or editing a user in admin. Under each new password a checklist ticks off length, upper and lower case, a number and a special character, and a refusal names the actual reason, a too common or too repetitive password included. (#1200)
Every dialog, the shared frame and the older modal alike, moves the focus in when it opens: on a desktop into the first text field, on a phone onto the panel so no keyboard pops up. Tab wraps around inside instead of running through the page behind, and closing hands the focus back to what opened it. Fields that open a list on focus are skipped, and the older modal's close button got a label and a tooltip. (#1302)
A booked stay's pill in the day head gets a second target that opens the booking; the name still opens the place. The place panel of a hotel, on the desktop and on the phone, lists the stay booked for it next to the bookings pinned to the stop. (#2363)
Under the payers, a Split line lists every member of the expense with their share and a tick once they marked it paid, so checking that the right people are in no longer means opening each expense. More than six fold into a +N chip with the rest in its tooltip. (#1763)
The notes of a flight, train or rental print under its booking code, like the notes of a place. A Transport notes switch next to Page break per day leaves them out and is remembered; it only shows when a transport has a note. (#1571)
The search box of the places list matched only name and address. It now also reads what was written about a place, through one helper the desktop list, its filter counts and the phone's browser share. (#2252)
The image carries org.opencontainers.image.source, url, documentation, licenses, title, description and version; the release and pre-release workflows add the commit and the build time, and the multi-arch index gets source, version and licenses as annotations. (#1498)
Providers like Pocket ID send the full name in 'name', which became 'JaneDoe'. OIDC_USERNAME_CLAIM=preferred_username builds the username from that claim instead. Unset, or with the claim empty, the old order applies (name, preferred_username, the email's local part), and it is only read when an account is created. (#1677)
The counterpart of last_trip for a homepage widget: the nearest trip that has not started yet, with its countries and the whole days until it starts, or null when nothing is planned. A trip under way stays last_trip, so the two never name the same trip. (#2542)
The public journey page drew the old "Powered by" block above the shared
frame's footer, so the credit showed twice; only the shared footer is left.
Its tests read the new frame: the title in the hero, the counts in the hero
pills, the loading skeleton and the cover as an image.
The shared trip tests look for markers inside the map, since the unplanned
card names the same places. The place form clicks the search button by name,
now that a picked place adds the nearby button to that row. The journey
editors send the photo order as one PUT, so their tests follow that request,
and a refused order rolls back as a whole. Payments carry their note, and a
saved mobile entry carries its draft flag.
nodemailer goes to 10.0.13 and brace-expansion to 2.1.7 for the advisories
Docker Scout reports (CVE-2026-102276, CVE-2026-102278, GHSA-v53p-9fqp-m79j).
A double tap counted every lift of a finger, the end of a swipe included, so
two quick swipes zoomed in. Only a touch that stays in place is a tap now.
The client tests follow the fields and arguments the qol sweep added.
The hand-written e2e schemas gain the new columns and place_regions, and the
unit assertions take the new arguments and fields. The Public API page names
GET /api/auth/app-config for the running version.
Small requests from the discussions, each built end to end: planner and
map (cycling mode, date first headings, clear a day, stay times per day,
off-route stops, map lock, compact unplanned markers, pending stays,
cable car, transit quick picks), places (location filter, category from
OSM, open in maps, geo: link, image from an attachment, enrichment on
file import, places near a pin, place name language, Amap first in
China), lists and costs (packed counter, inherited sharing, Markdown
to-dos, payment notes, bookings CSV, my plan PDF), sharing (unplanned
places, travel only, without photos), journey (photo zoom, drafts,
photo order, trip suggestion, day jump), atlas (region counter, region
bucket list, places per country), vacay half company holidays, studio
copy and paste plus the book's own layouts, and RESTORE_FROM_BACKUP for
a first start restore.
Sonar counts a promise that is neither awaited nor caught as a reliability bug, and it reads every such call in a touched file as new. Each of these already reports its own failure (toast, rollback or the sheet's catch), so they are marked as fire and forget rather than changed.
The overview holds its camera frame across content edits and only publishes a
fresh one on activation or on a trip switch. Activation is covered; the trip
switch branch was not, so a regression there would leave the map framed on
the previous trip. The new case keeps days and stops identical and changes
only the trip id, so the switch alone has to produce the new frame.
A departure or arrival row wrote its zone out ("Europe/Berlin") next to the
phase label, and on the narrow day plan that ate the width the route needed,
so "BER → HAM" ended in an ellipsis. The zone is now a small globe after the
time that names it on hover, the time pill carries only the time, and the
route gets its room back.
The phone booking and transport sheets collect picked files and upload them
after the save, but only when the record was new. Editing an existing booking
showed the picked files in the sheet, then dropped them on Update without a
word. The upload now runs after every save that comes back with a record,
through one helper both sheets share. A save that fails uploads nothing.
Closes#2534
The quick picks of the public transit search listed the day's places first
and the trip's stays after them, then cut the list at eight. On a day with
eight or more located stops the hotel was always cut, so it could not be
picked as the start or the end of a connection.
The stay the day wakes up in and the one it ends in now come first and are
never cut. The day's places and the trip's other stays follow, capped at
eight as before.
Closes#2538
With ALLOWED_ORIGINS set, every request carrying an Origin that was not on
the list was failed with an unhandled error, which the exception filter
answered with a 500. A browser sends Origin on every POST and WebSocket
upgrade, including same-origin ones, so opening TREK under any address the
list did not name (the LAN IP, a second hostname, http instead of https)
broke the login with "Internal Server Error" while another browser that
used the listed address worked.
A request whose Origin is the host it was sent to is now treated as
same-origin and always passes, for REST and the /ws upgrade alike. A
foreign origin is still refused, now with a 403 "Not allowed by CORS" and a
warning in the log naming the origin to add.
Closes#2543
Booking a night writes the hotel onto its check-in day as a stop, seated first
or at its check-in, so the drive reaches it even without hotel bookends. With
"Start and end each day at your stay" on, the evening row already ends that day
at the same hotel, so the road trip drove check-out, tonight's hotel, the day's
places and back to the hotel again. Days hides that stop and draws the stay at
the day's edges only.
seatNightBookends now passes over the stop tonight's booking wrote whenever it
seats that stay as the day's evening, in the browser and in calculate_roadtrip
alike. The stops keep their stored index, a hotel the traveller placed stays,
and the booking's stop stays while it is the day's last one or no evening is
seated. With the switch off nothing changes.
Closes#2546
The opening camera of the public share map was computed for the browser
window, while the map lives in a card a fraction of that size. Leaflet then
rounded the fractional zoom up, and the first fit skipped its correction
because it believed the map had already opened framed. Only picking a day and
going back to the whole trip ran a real fit.
The card is now measured before the map mounts and the opening frame is
computed for it, on a whole zoom level, with the same padding and zoom cap the
later fits use. A card without layout falls back to the first fit.
Closes#2549
The link chip held its title on one line with an ellipsis, and nothing in
the UI showed the rest: the tooltip names the action, not the link, and a
phone or the PWA has no hover at all. The title now wraps and the chip grows
in height, the host drops to its own line under a wrapped title, and the
chip keeps rounded corners instead of pill ends once it is taller.
Closes#2550
The day pills, Total Cost at the foot of the plan and the same figures in
the PDF added up the price column of every planned place. No form sets
that price since expenses replaced it, and nothing in Costs changes it,
so deleting an expense left the plan's total where it was, and a place
planned on two days counted twice.
With the Costs addon on, the plan now adds up the expenses: each one once,
on the day its booking starts, else the first day its place is planned,
else its date, and in the total either way. A foreign expense counts at
the rate frozen when it was entered and a refund nets against its day,
so Total Cost matches Total trip spend in Costs. With Costs off, a
place's own price still counts, once per place.
Closes#2551
Sonar counts a promise that is neither awaited nor caught as a reliability bug. Every one of these already reports its own failure (toast or optimistic rollback), so they are marked as fire and forget rather than changed.
A switch at the end of the list's tab row shows every list's items in name
order, so checking whether something is already on the list no longer means
reading all of it. The order follows the reader's language (accents beside
their letter, "Shirt 2" before "Shirt 10") and the empty list stand-in stays
at the bottom.
Sorting only changes what is shown: the manual order stays stored and comes
back with a second click. While sorted the drag grips step aside, the box
keeps its column, and the choice is remembered per browser.
Every guide of the four screens is recorded again with the new dialogs.
The guide scripts follow the new markup, and a review of the pictures
fixed what they taught wrong: the Kyoto list holds Kyoto places with
their photos, the invite and share steps show the picked user, the shared
calendar step shows one shared with you, the Dawarich and wishlist steps
frame the whole dialog, and the create journey guide links a trip with
places.
The export dialog was the last popup on the old card: it is now the shared
dialog frame with its options as rows, and it takes Escape for itself so
the key no longer reaches Studio behind it. Its stylesheet goes.
The Atlas country card printed the flag emoji, which Windows draws as the
two letters of the country code; it uses the shared CountryFlag now.
Every popup of the four pages moves to the shared dialog frame the trip
uses: head band with its icon, the scrolling body, the footer with its
actions, and ConfirmDialog for the questions.
- Vacay: settings, removing a year (now says so when it fails), the
fusion request, inviting a person, a person's colour and sharing the
calendar.
- Collections: new and edited lists, adding a place, importing a file and
a trip's places, sharing, labels and assigning them, copying to a trip,
moving or duplicating, saving a trip's places and deleting a list. The
stylesheet of the old import view goes.
- Atlas: the country popup with the flag in its head, its choices as rows
and the wishlist step; Dawarich's Atlas dialog and its accept dialog on
the desktop.
- Journey: creating a journey, its settings, linking a trip, inviting a
contributor, the entry editor (discarding asks in a dialog) and the
photo provider picker; the entry card's menu is the shared one.
The phone keeps its own sheets: where a dialog is shared with it
(Dawarich, the contributor invite, the provider picker) only the desktop
frame changes.
The redesigned share page reads collapsedDays, toggleDay and pickDayOnMap
from its hook; the map test's stand-in left them out and every case there
failed on the missing set.
The page a share link opens is rebuilt from top bar to footer: the
planner's light top bar with the language picker, the trip as a hero
made of its cover, a tab bar that stays in reach, the days as the
planner's cards beside a map that stays in view, bookings as the
booking cards, packing and costs as cards with their progress and
shares, the chat as bubbles, and empty states for sections the owner
shared with nothing in them yet.
The map picks its day in a select with a step back and forth instead
of a row of chips a three-week trip did not fit into.
Share opens on the shared dialog frame: the trip's name and the number of
people in the head band, members and guests on the left, the public link
and the invite link as two cards on the right. Rows carry their actions as
icon buttons named by their tooltips, and removing a member, leaving,
handing over the trip or removing a guest asks in the app's own dialog
instead of the browser's. The invite link can no longer be created twice by
a double click. The phone, which opens the same dialog from its More sheet,
keeps the sheet it had; both run on one set of hooks.
Save to list has the place's name in the head band, Mark visited as a pill
beside it, the lists as rows with a tick, and View at its foot.
The public link's hint loses its dash in every language, and the help
guides and wiki pages of the Share dialog show the new look.
The wiki pictures of the trip are taken again with the fixtures the help
guides use, so the Bookings tab shows a booking of every kind, Files has
documents and What's Next has something ahead. New pictures: the map's
hover card, the Subscribe to calendar dialog, the PDF preview, the To-Do
view, and the Transports tab's own timeline and booking detail. Day
selection in the screenshot run clicks the day's number, now that the
middle of a day header is the booked night.
Home, Real-Time Collaboration, Road Trip, Map Features and the bookings
page are brought in line with the rewritten pages, and sixteen assets no
page uses any more (mostly old GIFs, 22 MB) are removed.
The guides of the map, bookings, costs, files, collab, lists, road trip and
transports now describe the desktop as it is after the redesign: the bars
with their icon buttons and filter popovers, the dialogs with the title in
the head band and status and type as pills, the three booking views and
the booking detail, the compact note cards, the Costs table. Two guides
are new: the booking views (cards, list, timeline, the Trip and Day zoom
and the detail) and the Costs table view. Every changed text is translated
into every locale, and the Plan tab texts call the bookings tab by the
label it shows, Book.
Every picture of the thirteen trip screens is taken again against the new
layout, on one picture day. Scripts follow the new markup; the Dawarich
fixture uploads its recording again when the stored one belongs to another
run, and the spare poll gets its vote from a member so its winner shows
green.
server-launch.mjs skips the server build with E2E_SERVER_PREBUILT, so
several recordings on their own port pairs can share one dist.
vi called the Tracks filter "songs", sv the pending status "commuting", nl
the open-now pill "opening hours", and tr showed "its plan" on the Plan tab
and "being suspended" for pending. The road trip strings added in August
had lost their accents in eight languages (Arrive and Leave, the stay
buttons, the whole day and halfway filters, the avoidance options), uk had
a Latin i in "halfway", hu and tr called No tolls "free of charge", and az
put an en dash between the two ends. sv and vi said "I owe" for I'm owed and
You're owed, sv had "Namnn" as the table's Name, and th said "split
equally" for Split matches total.
The climate hint of the day details and the Unfinished tooltip in Costs
lose their dash in every language.
The PDF preview on the Files tab and the document preview of a Collab note
were about 150 px tall: the object carried flex-1, and in a column without a
fixed height that falls back to the object's own height instead of 74vh.
A tooltip no longer opens when a closing dialog hands the focus back to the
button that opened it. It floated over whatever came next, at the place the
button had before the page moved. Keyboard focus still opens it.
The table view sorts its dates with localeCompare, and two NaN fallbacks and
an unused import go.
The detailed weather asked the forecast API for every day up to 16 days out,
past days included. Once a day is a couple of months back that API answers
nulls instead of an error, and the day details showed 0° while the planner,
which already reads the archive, showed the real figure.
A past day now comes from the archive, with the same daily figures, sunrise,
sunset and hourly series as the climate path, which shares the new helper. A
recent day the archive has no figures for yet goes on to the forecast API as
before, and nulls from either source now answer no_forecast instead of 0°.
Costs gets a second view next to the ledger: the expenses as a table grouped
by category, with name, date, total, persons and days editable in place and
the per person and per day figures worked out beside them. A total with a
payer or in a foreign currency stays locked and opens the expense instead, so
balances and frozen rates are never changed behind the dialog. In the table
view the side column shows a summary by category, day, payer and status. The
switch sits in the bar and the choice is remembered; the ledger stays the
default.
On the Plan tab a place's reservation shows as one Reservation pill whose
colour carries the status and which opens the booking, and the drop line of
a drag that lost its source row now clears on the next mouse move. The
booking detail formats its times in the user's clock, and a hotel without a
check-out runs to the end of the last day in the timeline.
The wiki pages of the trip are rewritten for the new layout, with new
screenshots from the screenshot run (new Plan tab, bookings, transports,
dialogs, costs table). The help center guides of the Plan tab are rewritten
and translated into every locale, and every new string is translated.
The Plan tab was the last trip tab still in the old design. It now reads
like Bookings, Costs, Lists, Files and Collab, without losing a function:
- Day plan: each day is a card with a head band (day tile with weather,
stay and rental pills, a "+" menu for place, stay, transport, transit
and note), rows with "..." menus, white badges and a white route bar.
- Places column: add, search and filters in three rows, a show dropdown,
category and rating filters, and a selection bar at the foot.
- Place inspector, day details and the stay editor on the shared dialog
frame; the place form in the editor look with the name in the head band.
- Clicking a booking or transport anywhere on the plan opens the booking
detail popup first; Edit leads to the editor. The popup is one host
shared with the Bookings and Transports tabs.
- Files and Collab dialogs, notes, polls, links and chat moved to the same
frame; shared DialogShell and dialogParts hold the building blocks.
- The map hover card shows the place photo and follows the theme.
Every new and reworded string is translated into all locales, and a few
older English leftovers (help center, cost split, import hints) with them.
Phone screens keep their own look.
The Bookings and Transports tabs on the desktop take the phone's card
language: a tinted status band, the type chip and framed fields. One bar
holds the search, a filter popover for status, types and travelers, the
view switch and the view options.
Three views over the same filtered entries: cards (four per row at most),
a dense list grouped by day, and a timeline with a trip scale and a day
scale by the hour. A booking opens as a dialog with everything it holds,
the booking code blurred by the existing setting, and "On map".
Work in progress: only en and de are translated, and tests, help guides
and the wiki still follow.
A photo is now decoded only when its header says it is within 40 megapixels,
and bytes TREK cannot shrink, such as a WebP, are sent only up to 3.5 MB. A
scanned PDF page is drawn by its long edge, and pdf.js skips an embedded image
above the same pixel cap instead of decoding it.
The receipt reading is held to bounds in the shared contract, keeps the minus
sign of a discount and leaves such lines out so the Ticket split can save, and
its prompt states the JSON shape, so a server without json_schema still answers
something the parser reads. The model lookup sends the configured key, the
import error names the photo formats, the scanned photo is attached only for
someone who may upload files, and the phone settlement reloads after a receipt
is saved from the trip sheets. Test ids that clashed after the rebase are
renumbered, and the wiki describes the image setting, photo import and receipt
scan.
Whether the AI Parsing model reads images is an admin setting now, and it
lands with the two features that send a photo, so it has something to
control.
The setting is `vision: 'auto' | 'on' | 'off'` on the addon config. On
Automatic a local Ollama is asked what the model can do (`/api/show`), and a
cloud provider means no; Yes and No are applied as set. The old `multimodal`
flag is not read: the admin screens wrote it false on every save, so a
stored false meant nothing. The per-user toggle keeps its on/off meaning and
is read by the extraction now. The lookup is server-side only: capped read,
validated body, never throws, cached per server and model.
GET /api/llm/capabilities answers it for the signed-in caller, and both
photo pickers below ask it.
The booking import takes JPG, PNG and WEBP when the model reads images.
The photo goes as an image to every provider (image_url, an Anthropic image
block, Ollama's native chat with `images`), capped at 1600px, and skips KDE
Itinerary. A scanned PDF, which used to reach the model as empty text, has
its first two pages drawn by pdf-parse and sent the same way. The rest is
the import as it was: background job, review, the file attached to the
booking, the linked cost. On a real qwen3.5:4b a train ticket came back as
bare clock times until the router gave the model every type's field
meanings, and its price as an object inside a string, which read as 4900;
both are handled.
Costs gets Scan receipt: a dialog laid out like the booking import's takes
one photo, read on the import job queue, then reviewed in the existing
expense editor pre-filled with merchant, total, currency, day and lines,
the photo waiting in Receipts & Invoices to be uploaded on save. No table,
no migration, no category guess, no document-type detection.
Brings the v4.3.3 release back into the dev line, so the 4.4.0 release
merges into main without conflicts. All twelve conflicts resolve to the
dev side, which carries the same fixes adapted to dev (migration order
of #2527, dev only code around the costs and PWA fixes). Against the dev
fixes alone, the only change is the version bump to 4.3.3.
A foreign expense without a frozen rate is converted with today's rate.
When the server could not fetch rates itself, it counted the amount 1:1
in the trip currency, although the browser had the rates all along. A
VND bill on an AUD trip was settled as millions of AUD.
Such a row is now left out of the balances, flows, final budgets, trip
totals and per-person summaries. The settlement reports it under
`unconverted` and says which currency its figures are in; the trip
summary lists the ids and the budget prompt says how many are missing.
The browser lends the rates it already holds (`fallback_fx`) when it
saves an expense or a settle-up transfer, and to heal existing rows
through the new POST .../budget/freeze-rates route, which an editor's
Costs view calls for the rows the settlement left out. The server still
prefers its own rate, uses a lent one only while it is quoted against
the trip currency, and never replaces a frozen or explicit rate. The
settlement also takes the browser's display rate (`base_rate`) in place
of a display quote the server cannot fetch; the browser leaves out a
figure the query would refuse. The browser's own fetch gives up after
10 s.
Changing a row's currency while no rate is available no longer keeps
the old currency's rate; the row goes back to unfrozen.
MCP gets the same through the new freeze_budget_rates tool, which only
ever pins the server's own rate.
A foreign expense without a frozen rate is converted with today's rate.
When the server could not fetch rates itself, it counted the amount 1:1
in the trip currency, although the browser had the rates all along. A
VND bill on an AUD trip was settled as millions of AUD.
Such a row is now left out of the balances, flows, final budgets, trip
totals and per-person summaries. The settlement reports it under
`unconverted` and says which currency its figures are in; the trip
summary lists the ids and the budget prompt says how many are missing.
The browser lends the rates it already holds (`fallback_fx`) when it
saves an expense or a settle-up transfer, and to heal existing rows
through the new POST .../budget/freeze-rates route, which an editor's
Costs view calls for the rows the settlement left out. The server still
prefers its own rate, uses a lent one only while it is quoted against
the trip currency, and never replaces a frozen or explicit rate. The
settlement also takes the browser's display rate (`base_rate`) in place
of a display quote the server cannot fetch; the browser leaves out a
figure the query would refuse. The browser's own fetch gives up after
10 s.
Changing a row's currency while no rate is available no longer keeps
the old currency's rate; the row goes back to unfrozen.
MCP gets the same through the new freeze_budget_rates tool, which only
ever pins the server's own rate.
The capture backfill for local Journey uploads asked exifr for
'latitude' and 'longitude'. Those are not tags: exifr derives them from
the GPS tags, and a pick of only the derived names reads no GPS block at
all, so every uploaded photo lost its location and never showed on the
journey map. DateTimeOriginal was revived by exifr as a Date in the
server's time zone and OffsetTimeOriginal was never read, so a photo
taken at 15:52 in France was stored as 20:52 UTC on a server running in
Chicago. Impossible stamps were rolled over into real dates as well
(0000:00:00 became 1899, 30 February became 2 March).
The reader now picks the raw GPS tags and reads the stamps unrevived.
exifCaptureInstant validates the stamp and applies the first well formed
offset, preferring the tag that belongs to the stamp (OffsetTimeOriginal
for DateTimeOriginal, OffsetTimeDigitized for CreateDate). A zone written
onto the stamp itself (' UTC', 'GMT', 'Z', '+02:00') wins over the tags.
Without a usable offset it keeps the old reading in the server zone. A
stamp that is not a real date is dropped, CreateDate stands in when
DateTimeOriginal is unusable, and coordinates are kept only as a valid
pair. A GPS block of zeros, which a receiver without a fix writes, counts
as no location instead of a pin at 0,0.
Photos uploaded before this fix keep their stored time and location for
now; correcting them is left for a later release.
StorageService.withLocalFile now also removes its temp file when the
download itself fails, not only when the consumer throws.
The stamp is text any upload sets, so exifCaptureInstant turns down one
longer than 64 characters before any pattern runs, and the pattern has
no greedy tail in front of an end anchor, which cost quadratic time on a
long run of digits.
The REST uploads (entry, gallery) and the plugin journal.addEntryPhoto call
share this one reader, so both are fixed. The MCP journey tools only
attach provider photos, which never reach it.
The capture backfill for local Journey uploads asked exifr for
'latitude' and 'longitude'. Those are not tags: exifr derives them from
the GPS tags, and a pick of only the derived names reads no GPS block at
all, so every uploaded photo lost its location and never showed on the
journey map. DateTimeOriginal was revived by exifr as a Date in the
server's time zone and OffsetTimeOriginal was never read, so a photo
taken at 15:52 in France was stored as 20:52 UTC on a server running in
Chicago. Impossible stamps were rolled over into real dates as well
(0000:00:00 became 1899, 30 February became 2 March).
The reader now picks the raw GPS tags and reads the stamps unrevived.
exifCaptureInstant validates the stamp and applies the first well formed
offset, preferring the tag that belongs to the stamp (OffsetTimeOriginal
for DateTimeOriginal, OffsetTimeDigitized for CreateDate). A zone written
onto the stamp itself (' UTC', 'GMT', 'Z', '+02:00') wins over the tags.
Without a usable offset it keeps the old reading in the server zone. A
stamp that is not a real date is dropped, CreateDate stands in when
DateTimeOriginal is unusable, and coordinates are kept only as a valid
pair. A GPS block of zeros, which a receiver without a fix writes, counts
as no location instead of a pin at 0,0.
Photos uploaded before this fix keep their stored time and location for
now; correcting them is left for a later release.
StorageService.withLocalFile now also removes its temp file when the
download itself fails, not only when the consumer throws.
The stamp is text any upload sets, so exifCaptureInstant turns down one
longer than 64 characters before any pattern runs, and the pattern has
no greedy tail in front of an end anchor, which cost quadratic time on a
long run of digits.
The REST uploads (entry, gallery) and the plugin journal.addEntryPhoto call
share this one reader, so both are fixed. The MCP journey tools only
attach provider photos, which never reach it.
A euro trip read in dollars books a bill entered in dollars at the rate
of the day, from the euro's quote: 12,345.67 USD at 1.1398 is 10,831.44
EUR. The Costs screen and the settlement then turned those euros back
into dollars with the dollar's own quote, 1 / 0.87732. Frankfurter
rounds the two quotes separately, so they are not exact inverses, and
the bill read $12,346.05 the moment it was saved, a $250.00 one $250.01.
The error grows with the amount, which is why the large rows showed it
first. Paying back what settle-up offered for such a bill left the trip
open with the debt flipped the other way.
Both sides now convert trip money to the display currency with the trip
currency's own quote, the one every entry rate is frozen from. On the
server, settlement() fetches that quote for the REST route and for the
MCP get_settlement_summary tool, which now makes the same call, and a
legacy transfer without a currency reads back through the same factor.
On the client, useExchangeRates takes the trip currency as an anchor
and converts through its cached quote, falling back to the display
currency's while it is missing, so it works offline from cache and
needs no change to the settlement response. The final budget builds
each row from the unrounded trip amount, because a euro cent is worth
more than a dollar cent and 10,831.44 EUR came back as 12,345.68; the
balances and settle-up still net whole trip cents.
A same-day bill in the display currency now reads exactly as typed on
the row, the payer chip, the day header, the totals, Outstanding and
the final budget, on both shells and on the shared trip page, and
paying the amount settle-up offers leaves both balances at zero. When
the rate really moved, the row keeps the value the totals count, with
what was entered and the euros it was booked at underneath, "$801.76 ->
685,26 EUR". One helper builds that line for the desktop list, the
phone list, payments, the final budget breakdown and the shared page;
it prints nothing for an amount of 0 or a bill that reads as typed.
The expense editors share one module for their preview. The split
summary puts the share the row lends or borrows next to the entered
one, "$400.88 -> $390.53 each", and an expense saved without a currency
still opens in the trip currency. Both CSV exports add the amount in
the trip currency when it differs from the display currency.
Kept from before: the desktop totals recompute when rates arrive, a
booking's linked costs are labelled in the trip currency (every linked
row, the link picker and the phone's linked costs), a legacy
transfer is read in the display currency, the shared page reads the
frozen rate in whole cents, a booking import keeps the currency it was
quoted in at a rate frozen before the write (REST and plugin RPC), and
the MCP trip summary, the budget-overview prompt and the per-person
summary add every row up in trip cents at its frozen rate. The
packing-list prompt reads the trip title directly instead of building
the whole summary, which could wait on a rates fetch.
The logo at the top of every TREK mail was an SVG inlined into the HTML as a
data: URI. Gmail strips data: URIs, Outlook blocks them and neither renders
SVG, so notification and password-reset mails showed a broken image with
"TREK" as its alt text. The image data itself was fine; it only rendered in
clients that accept both data: URIs and SVG, such as Apple Mail.
The logo is now a 96x96 PNG, rasterized from client/public/icons/icon.svg
with the corner radius baked in, and MailerService attaches it to both HTML
sends as an inline part. The header img points at it by cid:, so nodemailer
puts the HTML and the image into multipart/related. That is the form Gmail,
Outlook and Apple Mail all display, and it does not depend on the instance
being reachable from the recipient's mail provider. The plain-text admin
test mail carries no HTML and stays without an attachment.
place_regions caches the country and region Atlas resolved for a place,
and a row that was already there was never re-derived. Correcting a
place's location in the trip left the old row in charge, so Atlas, the
dashboard stats and the journey stats kept counting the country the place
had just left, with the trip's dates on it.
A trigger on places now drops the cached row whenever lat, lng or address
actually change, so every writer is covered: the place editor, the MCP
update tools, the plugin RPC and the import enrichment and address
backfill. An edit that writes the same location back keeps the row.
The background lookup that fills the cache also only writes while the
place still sits where it was resolved from. A place moved while its
lookup was running would otherwise get the old answer written back.
Places corrected before this fix would have kept their old country until
they were moved again, because the trigger only sees new edits.
PlaceRegionsRepairJob goes over place_regions once, on the first start
after the update, and remembers that in app_settings. It re-derives each
row from the bundled borders only, never Nominatim, and replaces a row
whose country or region no longer matches where its place is now. A row
the bundle cannot answer for is left alone unless both the coordinates
and the address put the place in another country; then it is dropped and
the next Atlas load looks the place up again. A place without
coordinates keeps no row. Its writes are guarded the same way as the
background lookup, so a place edited during the pass is not overwritten.
The in-memory region cache is now keyed by the country the address names
as well as the coordinates, so correcting only the address of a border
point no longer returns the old answer until a restart.
On dev the trigger takes schema version 244, the slot it has on main, so a
4.3.3 install upgrading to 4.4.0 does not skip it. Web Push moves up to
245 and is idempotent, so a 4.3.3 install gets its table there. An
instance that already ran Web Push at 244 never runs 244 again, so 246
creates the trigger once more; everywhere else that is a no-op.
The phone timeline only put a travel-time connector between two places.
The drive from the arrival airport to the first stop after a flight is
part of the day's route and drawn on the map, and the desktop day plan
shows it under the flight, but on the phone the two rows sat on top of
each other. The drive from the last stop to the departure airport was
dropped the same way, and so was every drive past a booking saved
without a location, like a taxi to the airport or a concert between two
stops.
buildPlanRows now walks the rows the way the route calculation walks
the day: a located booking ends the drive before it at its departure
point and starts the next one at its arrival point, and each leg goes
under the row it leaves from. The route rides straight past a booking
it has no location for, so the drive through it goes under that
booking, as on the desktop, and keeps the mode menu of the stop it
left. Only transport bookings count as located, as in the route
calculation. Two bookings back to back still get no road between them.
A booking whose station is the stop itself, as the transit planner
saves a journey between two of the day's places, leaves a leg of no
length, and that one gets no connector.
On a day that opens with a flight, where the traveller also put the
hotel on the day as a stop of their own, the phone timeline started the
day with a hotel row and the drive from that stop to the next place,
above the flight, and then showed the same drive again in its own spot.
hotelLegsForDay guessed the bookends from coordinates: the first
segment leaving the hotel's spot was the morning leg. The route had
correctly drawn no morning leg on that day, but the stop on the
hotel's spot starts a leg there too, so that hop was taken for it.
The route calculation now tags the two legs it draws from and to the
accommodation, and the timeline reads that tag instead. The hotel end
of a tagged leg still has to be the day's own stay, because the
calculation keeps the previous day's legs until the new day has
routed. The stop to stop connectors leave tagged legs alone, so one
leg can no longer show up in both places.
Since drag-to-day (#891) a place pin swallows mousedown so the map holds
still while a drag starts. Leaflet only clears its "the map was panned"
flag on a mousedown it sees, so after one drag of the map it took every
later click on a pin for the end of that pan and dropped it. Pins stayed
dead until something else on the map was pressed, which is why zooming
out and clicking a cluster brought them back. Trip planner and
collections map were both affected; the GL renderers never were.
Pressing a draggable pin now clears that flag itself by switching the
map's pan handler off and on again, which is what the map would have
done had it seen the press. The click keeps Leaflet's normal route, so
it still closes an open popup on the map and still reaches listeners on
the page, such as an open context menu.
The new test file drives a real Leaflet map, since the mocked
react-leaflet has no pan state to go stale.
The public share page turned every failed GET /api/shared/:token into
"Link expired or invalid". A 5xx during a restart, a proxy 502, a rate
limit, a timeout or a viewer who is briefly offline all told the viewer
the link was dead and sent them back to the owner for a new one, while
the same link loaded fine on the next try.
The endpoint answers 404 with its JSON { error } body for an unknown,
expired or revoked token, so that is now the only answer that shows the
expired screen. A 404 without that body comes from a proxy in front of
TREK (Traefik answers "404 page not found" while the container is
stopped or still starting) and counts as a failed load, as does a 200
that is not the share payload, such as an auth wall's login page, which
used to crash the page. Any failed load shows "This trip could not be
loaded" with a retry button that refetches in place and stays busy
while it runs. The load also ignores a response that lands after the
page moved on to another token.
Both screens use the theme tokens instead of hex literals. New keys are
translated in all 27 locales.
After an update, a tab or a device whose service worker still serves the
previous build asks for chunks the update removed. The server answered those
requests with index.html and a 200, so the import failed with a MIME error,
and everything caching in between was free to keep the HTML as the chunk. A
precache install that met a missing file stored it, and Workbox reuses hashed
entries by URL, so the device showed "A new version is available" on every
load, whatever the reload or the cleared browser cache, and carried the broken
entry into every later release that keeps the file's name.
The SPA fallback now answers a GET for a build file that is not on disk with
a plain 404 (no-store) instead of index.html. A precache install refuses it,
so the worker whose cache is complete stays in charge until the new build is.
index.html is sent relative to the public dir, which also stops an install
living under a dot directory from answering every deep link with a 404.
On the client, the reload after a dead chunk (the automatic one and the
button on the update screen) no longer goes straight back to the worker that
served the broken shell. When that same worker is still in charge and the
server answers, its precache and registration are dropped first, so the
reload reaches the build the server runs now. Offline, and in offline mode,
the shell is kept. Every recovery reload, and the reload of the version
handover, asks for the page under a one-off query parameter, so a proxy or
CDN that keeps index.html against its headers cannot hand back the stale
shell either.
A chunk the entry imports statically fails before any of that code runs.
shell-guard.js, a small classic script loaded ahead of the entry, watches for
exactly that failure and runs the same recovery, once per session.
A stylesheet, font or icon that holds a page raises no error at all, so once
the app has started and the browser is idle, it looks through the precache
for entries that hold a page under a build file's name, fetches each from
the server past every cache and puts it back in place. A repaired stylesheet
is loaded again, and so are the app's stylesheets when a font was repaired,
since the browser does not try a face that failed again in the same
document. Only the headers are read, a precache found sound is not looked at
again until its entries change, nothing is fetched while nothing is broken,
and nothing happens offline, in offline mode or without storage. An answer
that is not the file (an auth wall, its login page, a server error) leaves
the entry for the next launch.
A page the recovery loaded from the network already runs the version the
server reports, so the version handover now records it instead of reloading
it a second time once the new worker takes over. A page loaded past an older
worker that is still in charge, as a hard reload does, still asks for the new
worker and records the version only once that has taken over, so the next
tab or launch cannot get the old shell without a handover.
Dropping the worker ends a Web Push subscription with it, so on such a device
the settings card shows push as off until it is switched on again. A missing
sw-push.js now fails the worker install with a 404 rather than index.html; the
comments that described the old answer say so.
Deleting a trip failed with SQLITE_IOERR_GETTEMPPATH when the container
runs with a read-only root filesystem and no tmpfs on /tmp. The cascade
behind DELETE FROM trips keeps a statement journal, SQLite moves it to a
temp file once it passes 64 KiB, and it looks for a directory for that
file in SQLITE_TMPDIR, TMPDIR, /var/tmp, /usr/tmp, /tmp and the working
directory. None of them is writable there. Installs on the shipped
compose file have a tmpfs on /tmp and never hit it.
Every connection the server opens now comes from openDatabase() in
src/db/connection.ts, which sets temp_store = MEMORY: the main database
(also when it is reopened after a restore), each plugin's own database
and the uploaded backup checked before a restore. reset-admin.js and the
key rotation script cannot import src/ in the image, so they set the
pragma themselves. A test pins that nothing else in src/ constructs a
handle, and one deletes a full trip on the real schema and watches for
temp files.
Since 4.3.0 the guard resolves a name to every address it has, and a
single blocked address refused the whole name. LAN DNS servers commonly
publish a host's fe80:: address next to its IPv4, so a self-hosted IdP on
such a name failed every login with "Requests to link-local /
cloud-metadata addresses are not allowed". Immich, Synology and the other
strict-guard integrations on the same kind of name failed the same way.
Both guards now leave a blocked address out of the list the socket is
pinned to and connect over the rest, and a name that resolves to nothing
else is still refused. A private address among the remaining ones still
needs ALLOW_INTERNAL_NETWORK under the strict guard, and
ALLOW_LINK_LOCAL_IPS works as before.
Leaving addresses out is only safe if the blocked set is complete, so the
strict guard now blocks everything the admin lane does. The AWS IMDSv6
(fd00:ec2::/32) and Alibaba (100.100.100.200, 100.100.100.100) metadata
addresses sit in ULA and CGNAT space, which ALLOW_INTERNAL_NETWORK opened
to any URL a user can type. Both guards also read an IPv4-mapped or
IPv4-compatible address as the IPv4 it carries: Node prints an AAAA
record of ::a9fe:a9fe as ::169.254.169.254, which passed both guards, and
::ffff:100.100.100.200 passed the admin lane. Neither guard connects to a
link-local or metadata address now, and the strict guard never connects
to loopback either.
A transport added from the day-plan sidebar could not be deleted: the edit
window (TransportModal) had no delete action, and the only transport delete
affordances lived on the Transports-tab cards, which that flow never reaches.
Add an onDelete prop and a Delete action in the edit footer, guarded by the
shared ConfirmDialog. Wire it from TripPlannerPage, gated on reservation_edit
(the permission the server enforces for reservation deletion).
A transport added from the day-plan sidebar could not be deleted: the edit
window (TransportModal) had no delete action, and the only transport delete
affordances lived on the Transports-tab cards, which that flow never reaches.
Add an onDelete prop and a Delete action in the edit footer, guarded by the
shared ConfirmDialog. Wire it from TripPlannerPage, gated on reservation_edit
(the permission the server enforces for reservation deletion).
A euro trip read in dollars books a bill entered in dollars at the rate
of the day, from the euro's quote: 12,345.67 USD at 1.1398 is 10,831.44
EUR. The Costs screen and the settlement then turned those euros back
into dollars with the dollar's own quote, 1 / 0.87732. Frankfurter
rounds the two quotes separately, so they are not exact inverses, and
the bill read $12,346.05 the moment it was saved, a $250.00 one $250.01.
The error grows with the amount, which is why the large rows showed it
first. Paying back what settle-up offered for such a bill left the trip
open with the debt flipped the other way.
Both sides now convert trip money to the display currency with the trip
currency's own quote, the one every entry rate is frozen from. On the
server, settlement() fetches that quote for the REST route and for the
MCP get_settlement_summary tool, which now makes the same call, and a
legacy transfer without a currency reads back through the same factor.
On the client, useExchangeRates takes the trip currency as an anchor
and converts through its cached quote, falling back to the display
currency's while it is missing, so it works offline from cache and
needs no change to the settlement response. The final budget builds
each row from the unrounded trip amount, because a euro cent is worth
more than a dollar cent and 10,831.44 EUR came back as 12,345.68; the
balances and settle-up still net whole trip cents.
A same-day bill in the display currency now reads exactly as typed on
the row, the payer chip, the day header, the totals, Outstanding and
the final budget, on both shells and on the shared trip page, and
paying the amount settle-up offers leaves both balances at zero. When
the rate really moved, the row keeps the value the totals count, with
what was entered and the euros it was booked at underneath, "$801.76 ->
685,26 EUR". One helper builds that line for the desktop list, the
phone list, payments, the final budget breakdown and the shared page;
it prints nothing for an amount of 0 or a bill that reads as typed.
The expense editors share one module for their preview. The split
summary puts the share the row lends or borrows next to the entered
one, "$400.88 -> $390.53 each", and an expense saved without a currency
still opens in the trip currency. Both CSV exports add the amount in
the trip currency when it differs from the display currency.
Kept from before: the desktop totals recompute when rates arrive, a
booking's linked cost is labelled in the trip currency, a legacy
transfer is read in the display currency, the shared page reads the
frozen rate in whole cents, a booking import keeps the currency it was
quoted in at a rate frozen before the write (REST and plugin RPC), and
the MCP trip summary, the budget-overview prompt and the per-person
summary add every row up in trip cents at its frozen rate. The
packing-list prompt reads the trip title directly instead of building
the whole summary, which could wait on a rates fetch.
The logo at the top of every TREK mail was an SVG inlined into the HTML as a
data: URI. Gmail strips data: URIs, Outlook blocks them and neither renders
SVG, so notification and password-reset mails showed a broken image with
"TREK" as its alt text. The image data itself was fine; it only rendered in
clients that accept both data: URIs and SVG, such as Apple Mail.
The logo is now a 96x96 PNG, rasterized from client/public/icons/icon.svg
with the corner radius baked in, and MailerService attaches it to both HTML
sends as an inline part. The header img points at it by cid:, so nodemailer
puts the HTML and the image into multipart/related. That is the form Gmail,
Outlook and Apple Mail all display, and it does not depend on the instance
being reachable from the recipient's mail provider. The plain-text admin
test mail carries no HTML and stays without an attachment.
place_regions caches the country and region Atlas resolved for a place,
and a row that was already there was never re-derived. Correcting a
place's location in the trip left the old row in charge, so Atlas, the
dashboard stats and the journey stats kept counting the country the place
had just left, with the trip's dates on it.
A trigger on places now drops the cached row whenever lat, lng or address
actually change, so every writer is covered: the place editor, the MCP
update tools, the plugin RPC and the import enrichment and address
backfill. An edit that writes the same location back keeps the row.
The background lookup that fills the cache also only writes while the
place still sits where it was resolved from. A place moved while its
lookup was running would otherwise get the old answer written back.
Places corrected before this fix would have kept their old country until
they were moved again, because the trigger only sees new edits.
PlaceRegionsRepairJob goes over place_regions once, on the first start
after the update, and remembers that in app_settings. It re-derives each
row from the bundled borders only, never Nominatim, and replaces a row
whose country or region no longer matches where its place is now. A row
the bundle cannot answer for is left alone unless both the coordinates
and the address put the place in another country; then it is dropped and
the next Atlas load looks the place up again. A place without
coordinates keeps no row. Its writes are guarded the same way as the
background lookup, so a place edited during the pass is not overwritten.
The in-memory region cache is now keyed by the country the address names
as well as the coordinates, so correcting only the address of a border
point no longer returns the old answer until a restart.
The phone timeline only put a travel-time connector between two places.
The drive from the arrival airport to the first stop after a flight is
part of the day's route and drawn on the map, and the desktop day plan
shows it under the flight, but on the phone the two rows sat on top of
each other. The drive from the last stop to the departure airport was
dropped the same way, and so was every drive past a booking saved
without a location, like a taxi to the airport or a concert between two
stops.
buildPlanRows now walks the rows the way the route calculation walks
the day: a located booking ends the drive before it at its departure
point and starts the next one at its arrival point, and each leg goes
under the row it leaves from. The route rides straight past a booking
it has no location for, so the drive through it goes under that
booking, as on the desktop, and keeps the mode menu of the stop it
left. Only transport bookings count as located, as in the route
calculation. Two bookings back to back still get no road between them.
A booking whose station is the stop itself, as the transit planner
saves a journey between two of the day's places, leaves a leg of no
length, and that one gets no connector.
On a day that opens with a flight, where the traveller also put the
hotel on the day as a stop of their own, the phone timeline started the
day with a hotel row and the drive from that stop to the next place,
above the flight, and then showed the same drive again in its own spot.
hotelLegsForDay guessed the bookends from coordinates: the first
segment leaving the hotel's spot was the morning leg. The route had
correctly drawn no morning leg on that day, but the stop on the
hotel's spot starts a leg there too, so that hop was taken for it.
The route calculation now tags the two legs it draws from and to the
accommodation, and the timeline reads that tag instead. The hotel end
of a tagged leg still has to be the day's own stay, because the
calculation keeps the previous day's legs until the new day has
routed. The stop to stop connectors leave tagged legs alone, so one
leg can no longer show up in both places.
Since drag-to-day (#891) a place pin swallows mousedown so the map holds
still while a drag starts. Leaflet only clears its "the map was panned"
flag on a mousedown it sees, so after one drag of the map it took every
later click on a pin for the end of that pan and dropped it. Pins stayed
dead until something else on the map was pressed, which is why zooming
out and clicking a cluster brought them back. Trip planner and
collections map were both affected; the GL renderers never were.
Pressing a draggable pin now clears that flag itself by switching the
map's pan handler off and on again, which is what the map would have
done had it seen the press. The click keeps Leaflet's normal route, so
it still closes an open popup on the map and still reaches listeners on
the page, such as an open context menu.
The new test file drives a real Leaflet map, since the mocked
react-leaflet has no pan state to go stale.
The public share page turned every failed GET /api/shared/:token into
"Link expired or invalid". A 5xx during a restart, a proxy 502, a rate
limit, a timeout or a viewer who is briefly offline all told the viewer
the link was dead and sent them back to the owner for a new one, while
the same link loaded fine on the next try.
The endpoint answers 404 with its JSON { error } body for an unknown,
expired or revoked token, so that is now the only answer that shows the
expired screen. A 404 without that body comes from a proxy in front of
TREK (Traefik answers "404 page not found" while the container is
stopped or still starting) and counts as a failed load, as does a 200
that is not the share payload, such as an auth wall's login page, which
used to crash the page. Any failed load shows "This trip could not be
loaded" with a retry button that refetches in place and stays busy
while it runs. The load also ignores a response that lands after the
page moved on to another token.
Both screens use the theme tokens instead of hex literals. New keys are
translated in all 23 locales.
After an update, a tab or a device whose service worker still serves the
previous build asks for chunks the update removed. The server answered those
requests with index.html and a 200, so the import failed with a MIME error,
and everything caching in between was free to keep the HTML as the chunk. A
precache install that met a missing file stored it, and Workbox reuses hashed
entries by URL, so the device showed "A new version is available" on every
load, whatever the reload or the cleared browser cache, and carried the broken
entry into every later release that keeps the file's name.
The SPA fallback now answers a GET for a build file that is not on disk with
a plain 404 (no-store) instead of index.html. A precache install refuses it,
so the worker whose cache is complete stays in charge until the new build is.
index.html is sent relative to the public dir, which also stops an install
living under a dot directory from answering every deep link with a 404.
On the client, the reload after a dead chunk (the automatic one and the
button on the update screen) no longer goes straight back to the worker that
served the broken shell. When that same worker is still in charge and the
server answers, its precache and registration are dropped first, so the
reload reaches the build the server runs now. Offline, and in offline mode,
the shell is kept. Every recovery reload, and the reload of the version
handover, asks for the page under a one-off query parameter, so a proxy or
CDN that keeps index.html against its headers cannot hand back the stale
shell either.
A chunk the entry imports statically fails before any of that code runs.
shell-guard.js, a small classic script loaded ahead of the entry, watches for
exactly that failure and runs the same recovery, once per session.
A stylesheet, font or icon that holds a page raises no error at all, so once
the app has started and the browser is idle, it looks through the precache
for entries that hold a page under a build file's name, fetches each from
the server past every cache and puts it back in place. A repaired stylesheet
is loaded again, and so are the app's stylesheets when a font was repaired,
since the browser does not try a face that failed again in the same
document. Only the headers are read, a precache found sound is not looked at
again until its entries change, nothing is fetched while nothing is broken,
and nothing happens offline, in offline mode or without storage. An answer
that is not the file (an auth wall, its login page, a server error) leaves
the entry for the next launch.
A page the recovery loaded from the network already runs the version the
server reports, so the version handover now records it instead of reloading
it a second time once the new worker takes over. A page loaded past an older
worker that is still in charge, as a hard reload does, still asks for the new
worker and records the version only once that has taken over, so the next
tab or launch cannot get the old shell without a handover.
Deleting a trip failed with SQLITE_IOERR_GETTEMPPATH when the container
runs with a read-only root filesystem and no tmpfs on /tmp. The cascade
behind DELETE FROM trips keeps a statement journal, SQLite moves it to a
temp file once it passes 64 KiB, and it looks for a directory for that
file in SQLITE_TMPDIR, TMPDIR, /var/tmp, /usr/tmp, /tmp and the working
directory. None of them is writable there. Installs on the shipped
compose file have a tmpfs on /tmp and never hit it.
Every connection the server opens now comes from openDatabase() in
src/db/connection.ts, which sets temp_store = MEMORY: the main database
(also when it is reopened after a restore), each plugin's own database
and the uploaded backup checked before a restore. reset-admin.js and the
key rotation script cannot import src/ in the image, so they set the
pragma themselves. A test pins that nothing else in src/ constructs a
handle, and one deletes a full trip on the real schema and watches for
temp files.
Since 4.3.0 the guard resolves a name to every address it has, and a
single blocked address refused the whole name. LAN DNS servers commonly
publish a host's fe80:: address next to its IPv4, so a self-hosted IdP on
such a name failed every login with "Requests to link-local /
cloud-metadata addresses are not allowed". Immich, Synology and the other
strict-guard integrations on the same kind of name failed the same way.
Both guards now leave a blocked address out of the list the socket is
pinned to and connect over the rest, and a name that resolves to nothing
else is still refused. A private address among the remaining ones still
needs ALLOW_INTERNAL_NETWORK under the strict guard, and
ALLOW_LINK_LOCAL_IPS works as before.
Leaving addresses out is only safe if the blocked set is complete, so the
strict guard now blocks everything the admin lane does. The AWS IMDSv6
(fd00:ec2::/32) and Alibaba (100.100.100.200, 100.100.100.100) metadata
addresses sit in ULA and CGNAT space, which ALLOW_INTERNAL_NETWORK opened
to any URL a user can type. Both guards also read an IPv4-mapped or
IPv4-compatible address as the IPv4 it carries: Node prints an AAAA
record of ::a9fe:a9fe as ::169.254.169.254, which passed both guards, and
::ffff:100.100.100.200 passed the admin lane. Neither guard connects to a
link-local or metadata address now, and the strict guard never connects
to loopback either.
Discussion #2084: a cost bought later (the seat upgrade for a flight) could
not be tied to the booking it belongs to, and a booking held one expense.
- The Costs block of the booking, transport and place forms lists every
linked expense with edit, unlink (the expense stays in Costs) and delete,
and offers the expenses that belong nowhere yet under Link existing
expense. The phone sheets get the same block in their own design, and
their Files row lists the attached files and links an existing one.
- PUT /budget/:id, update_budget_item (MCP) and costs.update (plugin RPC)
take reservation_id / place_id; create and update refuse an id from
another trip on all three paths.
- The booking's mirrored price is the sum of its expenses when they share
one currency (none counts as the trip's), else the first one, with that
currency, and follows every change to a linked expense. The booking
form's price field leaves a booking with several expenses alone.
- Deleting a booking deletes all of its linked expenses and broadcasts each
one, also from the MCP delete tools.
- Cards show the price with its currency.
- Booking and transport dialogs: Booking Type and Travelers as dropdowns at
the top, the URL beside Files (transports gain the URL field). The row is
one shared component; linking and unlinking files go through the store.
Help center pictures and texts for the booking, transport and expense
guides are redone in all 27 languages, five Costs labels that were still
English in 16 languages are translated, and the wiki follows.
Rows get the packing list's filled checkbox, priority, due date and list
as quiet badges, and the assignee as an avatar with the name as its
tooltip. The tasks sit in a card whose head names the filter and holds
the sort, Priority or Due date, which moved out of the sidebar; a second
click on the active one goes back to the manual order. The progress card
matches the packing list's.
The detail pane is a card with a box in its head that ticks the task
off, and Delete and Save changes stay in sight: the To-Do view now fills
the space under the Lists bar instead of a fixed height that pushed the
pane's foot off screen. The detail pane and the new-task form share one
set of fields, and a new list is named in a dialog, from the sidebar
and from the form; the dialog moved to shared/NameDialog for both lists.
The To-Do help pictures are re-rendered, the texts that described the
old sidebar and form follow in all 27 languages, and the wiki page is
updated. The phone screens are unchanged.
The Costs tab opens with the same bar as Transports, Bookings, Lists and
Files: the tab's name, the trip's dates and days in one pill, the
travelers, and Settle up and Add expense on the right.
The desktop packing list takes the phone's look without losing anything
it shows. Category cards get a tinted head band, rows get hairlines and
filled checkboxes, owner, quantity and weight read as small badges, and
the bin and a menu with Move to List, Sharing, Rename and Delete replace
the row of icons. What a row does not use stays dimmed until hover or
focus, so nothing leaves a gap. A click on the name renames, ticked
items too. Add item opens a composer shaped like the next row, Add list
moved into the Lists bar, Add list and Save as template ask for the name
in a dialog, and Remove checked sits beside the progress bar. The Bags
panel becomes a card, and the item menu, the bag picker, the list menu
and the bag member picker share one popover look and close on a click
beside them.
Help center pictures for Costs and Lists are re-rendered, the Lists
guide texts follow the new controls in all 27 languages, and the wiki
pages are updated. The phone screens are unchanged.
The packing list bar gets an export menu next to the import button, both
now icon only. The menu prints the open view as a TREK styled page (trip,
dates, packed count, one card per category, bags), which the browser can
also save as a PDF, and downloads it as a Markdown checklist or as a CSV
in the exact format the import reads, bags included, so a list works as a
personal template without admin rights.
The import now also reads Markdown checklists: headings become
categories, "- [x]" marks an item packed, a "3x" prefix sets the
quantity and a "(200 g)" suffix the weight. The phone action sheet
offers the same exports through the shared usePackingExport hook.
The help center gets an export guide with new pictures, the import guide
and the wiki page describe the Markdown format, in all 27 languages.
Addresses discussions #1420 and #875.
A searchProvider plugin can now implement an optional suggest() beside
search(). Its places then show up in the place search's typed-ahead list,
after the core suggestions and marked with the plugin's name, on desktop
and phone. An index that cannot take a request per keystroke leaves it
out and keeps answering the full search only.
The child reports which functions each hook object carries at load, so
GET /api/plugin-search/suggest only asks the plugins that implement
suggest: from the second character, three rows at most across all of
them, 800 ms to answer. The feed marks plugins holding
hook:search-provider and the client only asks the route while one is
installed. A plugin row carries its whole place, so picking it skips the
details lookup, which knows no plugin id. The phone's search results now
name their source as the desktop ones do, and the desktop dropdown
scrolls instead of running past the dialog.
Addresses discussion #2221.
A new manifest capability poiCategories (up to 4, icon from an allow-list,
strict #rrggbb color) and the hook poiCategoryProvider.getPois behind the
hook:poi-category-provider permission. GET /api/plugin-pois in the plugin
contributions module answers them in the /api/maps/pois shape, with
search_plugin_pois as the MCP twin. plugin: ids no longer count as Google
place ids. SDK 1.8.0 carries the types, validation and docs.
On the map, one hook merges the built-in categories with the ones active
plugins declare, and the pill shows the plugin chips after a divider on
desktop and phone, scrolling sideways instead of shrinking a chip below
its touch target. A plugin chip searches through a repo that refuses
offline and parses the answer strictly. Its pins, on Leaflet and GL, use
the category's own icon and color, both checked again before they reach
any markup, and the hover card lists the plugin's details as escaped
text. A plugin POI can be added as a place like any other. The admin
panels on desktop and phone show a chip for the new permission and the
categories a plugin would add, from one shared module.
Addresses discussion #1781.
The photo picker's Select all took only the pages already loaded, 50 at a
time, and replaced whatever had been picked on another tab. It now loads
the rest of a day, trip or date-range search before selecting (cancelled
by closing the picker or switching tab), adds to earlier picks, and
Deselect all only clears what is in view. Add stays disabled until that
loading is done. The whole-library tab keeps selecting what is loaded and
shows the count as "N+".
A date-filtered Immich search stopped at 1,000 photos without a word,
because its page budget was absolute; it is now relative to the page
asked for, with an absolute ceiling of 270 raw pages sized so the
picker's deepest page still gets a full answer. Adds go out in batches of
500 ids, under the JSON body limit, and each batch is sent oldest first,
so photos land in an entry in the order they were taken. When a later
batch fails, the photos already saved are counted and the journey
reloads while the error still shows, through one hook for desktop and
phone.
After a provider add, the capture-time lookup runs at most four at a time
per user, shared by all of that user's batches. When it fills a row, a
journey:photos:updated event reloads open journeys once per batch, from
the REST routes and the MCP add tool alike. Device uploads, which arrive
one file per request, get the same lookup without the event.
Addresses discussion #1587.
A fourth built-in channel next to email, webhook and ntfy. Browsers
register per device, the server encrypts per RFC 8291 and signs per RFC
8292 with node:crypto and sends only to known push services through the
SSRF guard. The event matrix decides per category what arrives by push;
admins switch the channel on under Notifications.
VAPID keys come from the environment or are generated once, only while
neither stored row exists, and kept encrypted. A pair that does not
decrypt or match is never overwritten and never swapped for another one:
push stays off with an error in the log and every device stays
registered until the key is fixed. The subject is VAPID_SUBJECT, then an
https APP_URL, then the project URL.
Only chat messages collapse into one notification per trip, and a
replacement still alerts. A 403 removes a device only after three
failures in a row; 404 and 410 remove it at once. Password change, reset
and an admin setting a password drop the user's devices with the other
credentials. The request log redacts subscriptions and endpoints, boot
errors print *** for every secret variable, and logout forgets the
device with a plain fetch so an expired session cannot redirect midway.
Addresses discussion #894.
Settings > General > Language & region gets "Week starts on" (Monday,
Sunday or Saturday), and Admin > User Defaults a default for everyone who
has not picked one. Monday stays the built-in value, so nothing changes
for anyone who leaves it alone.
Both calendar implementations follow it: CustomDatePicker, used by trip
dates, bookings, costs, to-dos and the admin school holidays on desktop
and phone, and the journey date picker, whose weekday headers were
hardcoded English and now speak the app language. The grid math and the
header row live in one helper (utils/calendarWeek.ts). The day names on
the options come from Intl, so only the label and a hint needed
translating. Vacay keeps its own per-plan week start.
The value is validated like the other regional settings: VALID_VALUES for
admin defaults, the display-settings MCP tool, and the managed key list.
Addresses discussion #2029.
Translates the keys dev gained since the branch was cut, the help center
above all, and drops the six keys dev no longer has. Registers th
everywhere a language is listed: notification emails, the rate-limit
message, the placeholder check and the wiki. The two tests that counted
the languages by hand now derive the count from SUPPORTED_LANGUAGES.
Translates the keys dev gained since the branch was cut, the help center
above all, and drops the six keys dev no longer has. Registers sk
everywhere a language is listed: notification emails, the rate-limit
message, the placeholder check and the wiki. The two tests that counted
the languages by hand now derive the count from SUPPORTED_LANGUAGES.
Translates the keys dev gained since the branch was cut, the help center
above all, and drops the six keys dev no longer has. Registers et
everywhere a language is listed: notification emails, the rate-limit
message, the placeholder check and the wiki. The two tests that counted
the languages by hand now derive the count from SUPPORTED_LANGUAGES.
Translates the keys dev gained since the branch was cut, the help center
above all, and drops the six keys dev no longer has. Registers az
everywhere a language is listed: notification emails, the rate-limit
message, the placeholder check and the wiki. The two tests that counted
the languages by hand now derive the count from SUPPORTED_LANGUAGES.
The server already put PLACES_API_KEY ahead of the key saved in the admin
panel, but nothing said so: the variable was in no template and the admin
field kept showing a stored value that no search used.
GET /me/settings now answers env_keys, the variable behind each key the
environment sets (Google, Unsplash, Amap), and null for that key's value.
The admin field turns read-only and names the variable, Test still probes
the key a search resolves to, and a save leaves those names out so the
stored fallback survives. The variable is documented in .env.example,
docker-compose.yml, the Helm chart (as a secret) and the wiki.
Addresses discussion #1881.
On a slow line an OIDC-only login page drew the email and password form while it
asked the server which sign-in it offers, then the SSO button once it knew, and only
then left for the identity provider. Both looked like something to fill in or click.
Until the config answers, the page now shows a short wait instead of the form, and
on an OIDC-only instance it becomes a screen that only says where it is going,
"Taking you to Keycloak…", with nothing to fill in. A repeat visit announces that
redirect at once from the cached config. Should the redirect not have left the
page after eight seconds, a quiet link continues by hand.
The redirect still never runs on a cached config, right after a sign-out or after a
failed sign-in at the provider. That last case now loads the config as well, so an
OIDC-only instance shows its own screen with the error rather than a password form
nobody can use.
"Keep the documents in step with a Nextcloud folder" walked through Nextcloud
alone: its app password, its base folder, the question where the trip should live
in Nextcloud. The dialog offers five stores. The guide now reads for all of them:
how each files things (by tag, in a folder, in a space), what each signs in with
(an API token for Paperless-ngx, an API key and the organisation for Papra, username
and app password for Nextcloud, username and app token for OpenCloud, username,
password and a two-factor code for Synology Drive), that the base folder belongs to
Nextcloud and Synology Drive, and that the trip lives in a tag, folder or space. The
pictures keep Nextcloud as the example, and the text says so. All 23 languages use
the labels of their own interface.
"Add, reorder and rename days" still described the reorder popup as it was before
#2485: Move up, Move down and one Add day at the end, and its picture showed that
popup. The guide has five steps now. The reorder step names the bin on each day and
the two ways of adding one, the next date or Without date, and a new step shows the
question before a day goes, with what goes with it, answered with Cancel. The first
step says that shortening the dates lists the days that go, and the tip says moving
days can be undone and deleting one cannot. All 23 languages use the labels their
own interface shows.
The picture script looked for the Add day button that #2485 removed, so it could not
run any more. It frames the popup itself now, asks about the second day so the
question has places and a booking to list, and clicks the first day a second time
when the steps before left it selected and the first click folded its panel away.
The pictures picked up five faults that the run could not see, because a
passing step only proves the click landed.
The ring vanished on the lightbox: the accent is nearly black and so is
the overlay. It now measures the surface behind it and puts a white
hairline outside itself and the number only where the contrast is too
low, so every other picture keeps the single line it had.
The number fell outside the picture whenever the target was wider than
the frame, which is every full-width row: the frame is known before the
ring is drawn, so the number moves inside far enough to clear the edge.
The satellite imagery was a grey rectangle although every tile came back
200. Leaflet fades a tile in over 200 ms and takes the progress from
`+new Date()`, and the browser's clock is pinned to the picture day, so
the opacity it writes on every frame is always 0. A stylesheet on
Leaflet's own `leaflet-tile-loaded` outranks that inline style. The
satellite guide also waits for decoded tiles now instead of for a tile
element to exist.
The uploaded files looked like nothing anyone would upload. A PDF is
drawn with an issuer band, a reference, a table of fields and a footer,
and the two photographs are gone: a drawn photograph reads as clip art
at any size, so the trip carries a map screenshot with a pin and a
booking confirmation in a browser window, which are graphics in reality
too.
In the panel, a step picture was covered into a 16:10 box and lost its
sides; contained everywhere now, the hero and the result in 16:9.
Beside that: the place guide gained the drag-and-drop route to a day as
its fourth step, in all 23 languages, and `e2e/help/README.md` now holds
the whole visual contract, the traps that cost this round (the pinned
clock, prettier against the client sources, a killed run holding its
ports) and how to find out which pictures a change forces to be retaken.
The whole set, 914 files, made in one pass on one machine against one
calendar: 41 screens, 225 guides, 764 steps. What changed beyond the new
guides: the pictures carry the run day's dates now, the Files tab's
preview shows a real PDF page, the Place details card has its hours, its
ring and its phone number, the trip map and the Atlas carry the Dawarich
control the new bullets describe, and the Files toolbar has Document
sync beside Trash.
Ninety-eight keys in the other 22 languages: the nine guides of this
round, the reworked read-place and import-booking-file, and the two
screen bullets. Every UI label in them is the app's own string in that
language, and where a label is still English in a locale, the text keeps
it English so the reader finds it on screen.
The first round could not photograph a few screens because the service
behind them was not there: Dawarich, AirTrail, a document store, the
booking extractor. They are there now, so the guides are too.
- map-dawarich-trail, dawarich-countries and dawarich-wishes: the run
uploads a recording of the seeded trip, and of the year before it, into
a Dawarich of its own (dawarich-track.ts) and reads it back.
- airtrail-import: four flights are put into an AirTrail account, two of
them connecting in Tokyo, so the picker offers the layover.
- import-booking-file and import-transport-file: the confirmations carry
their facts as schema.org data the way portals send them, the extractor
reads them, and the guides go on to the filled-in form and the card.
- files-sync and document-providers: a Nextcloud folder with two
documents in it, bound to the trip, and the switches in the admin panel.
- read-place: three seeded places carry their OpenStreetMap id, so the
card shows the week's hours, the phone number and the Open or Closed
ring. The rating stays Google's, and the texts say so now.
- map-compass on the MapLibre renderer, which the guide switches to and
back; files-preview opens a PDF, which needs the full browser rather
than the headless shell (channel: chromium on the help-media project).
Each of those takes an address and a key from e2e/help/media.env, which
is not committed; media.env.example and the README say what each one
needs and why the address has to be the machine's own on the network.
The seed's dates are offsets from the day the pictures are taken
(e2e/dates.ts): the trip ends today, so the boarding pass, What's Next
and the forecast have something to show whenever the run happens, and
nobody bumps a date in three files every few weeks. run.mjs fixes the day
once per run so the projects and the workers agree across midnight.
The Links section still described the panel as it was before #2414: pin
or delete, and add again to correct a typo. It names Edit link now, says
the form takes only http and https addresses and what the chip shows, and
it is a panel beside Notes rather than a tab, except on the phone.
shared.other labels the same fallback group on the public share page that
the app itself calls Sonstiges and Otros, so de and es said one thing in
the trip and another beside it. admin.tabs.addons was Russian in uk.
reservations.fromPlan, packing.suggestions, packing.suggestionsTitle,
packing.allSuggested, packing.emptyHint and the packing.suggestions.items
array are referenced by no component; the suggestions block left the
packing list in #14 and the keys stayed behind in all 23 languages. The
VORSCHLAEGE constant beside them was exported and imported nowhere.
With the one structured entry gone, a catalogue value is a string and
nothing else, so TranslationValue says so and the placeholder spec no
longer needs to skip what is not one.
Rendered in place, its fixed overlay sat inside the trip page's stacking
context, below the navbar's z-[200]: the bar painted over the lightbox's
header, so Open in new tab, Download and the x were unreachable. The two
document previews beside it already portal into the body for this reason.
The tests read the lightbox through baseElement now, since a portal puts
it outside the container Testing Library returns.
- The lightbox header's button group is the right thing to ring, but its centre
falls between the buttons where the picture underneath takes the pointer, so
the hover never landed. The pointer goes on a button, the ring stays on the
group.
- The collection the place is saved into was matched by a name another guide's
list starts with.
- help-media gets four minutes a guide instead of two and a half. A guide is a
whole task and a few of them wait on the network more than once; the
nearby-places one runs two POI searches and opens the place form on top.
The + at the end of a place row belongs to the open day, and closing the day's
details panel closes the day with it, so there was no + to click and the guide
waited for one until the test gave up.
- place-to-collection found the place already in a list, because the Collections
screen's guides run earlier and leave it there. It clears the place first now,
with the same helper its cleanup uses.
- The road trip rail leaves out a day with nothing to drive, which its own
helper says, and the fold step was the last place still pointing at day 1.
- The transport form's location search took whatever the geocoder ranked first,
and for a landmark that is as often a shop down the road. It picks the
suggestion that answers the query.
Every one of these passed when its own file was run alone and failed when all
of them ran against one seeded trip, because another screen's fixture had put
something in the way.
- The Bookings fixture names a booking after a place, so the Assign File dialog
offered two entries opening with the same words.
- The Lists tab has a Shared button, which the loose name of the navbar's Share
also matches, so opening a trip on that tab never settled.
- A traveller's avatar is named by the initial it draws, not by the username;
the title carries the name.
- Departure and Arrival are what a booking gets when it crosses days, so the
guide about them points at the flight, not at the train that starts and ends
on one day.
- The What's Next stop was timed from the wall clock of whoever runs this,
while the browser's clock is pinned to the day the pictures are taken on.
Both now come from the same constant.
- The map was panned with the arrow keys, which Leaflet only reads while its
container has focus, and focusing it from here did not take. It is dragged
now, from a point checked to be free of markers, because a press on a marker
drags the place onto a day instead.
The duplicate-trip guide dropped its copy by position, keeping whatever the
trips endpoint happened to answer first. The copy survived, and three files
later the journey dialog offered two entries starting "Weekend in Lisbon" and
the create-journey guide could not say which one it meant.
The copy is now dropped by name and the delete is checked, and the journey
dialog picks its trip by accessible name rather than by a substring another
trip shares. Each file passed on its own before this; only the whole suite in
one run showed it.
Forty-five values were rewritten after a second pass over the running app found
the old text naming a control that is not there or describing one in the wrong
order. Each locale's existing translation was edited rather than replaced, so
the register and the vocabulary of the block it sits in are kept.
All ten screens of a trip are now in every language TREK speaks: 2134 help keys
in each of the 23 locales. i18n:parity:strict passes for the first time since
the help center landed, which also un-skips every CI job that was waiting
behind it.
Place details, Day details, Map, Transports, Bookings, Lists, Costs, Files,
Collab and Road Trip: 77 guides over 360 steps, each one a task a reader can
follow, with the Playwright script that will photograph it. The fixtures they
need go in beside ensureDaysFixtures, and every screen is a child of the trip,
so the panel folds them under it.
The capture now pins the browser's clock to the day the seeded trip ends. A
picture has to look the same whoever takes it and whenever: a day later the
trip is over, My Trips moves it out of the boarding pass, What's Next has
nothing left to list, and a forecast for a day in the past is not a forecast.
Only `Date` is pinned, so timers keep running.
trip-days is green end to end after its locators were checked against the
running app; its route guides read day 6, whose two stops sit next to each
other, because day 1's are separated by the flight and have no leg of their own.
Place details, Day details, Map, Transports, Bookings, Lists, Costs, Files,
Collab and Road Trip: 77 guides and 360 steps, written against the running app
rather than against the components, so every label a guide names is a string the
reader will actually see. Places, Day details, Map and Files are translated into
the other 22 languages; the rest follow.
Also reworded the Settings overview: it called TREK self-hosted, which the
repository's own wording rule forbids and which nothing was running to catch,
because the test that does needs i18n parity and parity has been red.
Two cases replaced window.location wholesale to spy on reload. The copy is flat,
the real object's fields are getters on its prototype, so the replacement has no
href for a relative request URL to resolve against and the DELETE the leave-trip
case waits for never reached the mock server. It also stayed replaced for the
rest of the file, which took six later cases down with it.
The leave-trip case asserts the request, so it keeps the real location: jsdom's
reload only logs. The transfer case asserts the reload, so it keeps the stub,
and the original is put back after every test.
Nobody saw this: the client coverage job needs i18n parity, which has been red.
'2 sub-screens' stopped identifying one button the moment a second screen had
two of them, and the trip is about to have a dozen. The toggle is looked up
within the row it belongs to instead, which says what the assertion means.
`collab.notes.expand` was never added to the catalogue, and `t()` answers with
the key itself, so the `|| 'Expand'` fallback never ran and the tooltip read
collab.notes.expand. `common.expand` says the same thing and is already
translated everywhere, so the button uses that, and it now carries the label
for assistive tech as well.
What the ? in the navbar opens, what an overview and a guide contain, how the
search and the screen switcher work, where the pictures come from and what
happens on a screen that has no help yet. Linked from the sidebar, from the
features list and from In-App-Help, which documents the other thing and was
being confused with it.
All eight guides of Trip > Places now have their pictures, taken from the real
app, plus the hero of the screen, and the block is translated into the other
22 languages.
Making the pictures showed three texts describing a UI that is not there, and
they are corrected: the file import dialog has no Enrich places via Google
switch (only the list import has one, and only with a Google key), a GPX import
closes with a message while only KML and KMZ leave a summary behind, and both
list providers are always offered. The search and the details column need no
Google key either: they fill themselves from the TREK index, OpenStreetMap,
Wikipedia, Wikivoyage and Wikimedia, and only ratings are Google's.
The script needed the same honesty. ConfirmDialog and the two import dialogs
are portals of their own with none of Modal's classes, so they get their own
locators, and create-place now looks up a place whose entry actually carries a
picture, opening hours and a description, instead of photographing an empty
details column.
Every context file carried its own copy of the same little guide factory and
the same context literal. Thirteen screens of a trip would have made thirteen
copies of it, which is both the duplication a quality gate counts and the
hand-mirrored state the repository asks us not to keep, so the shape moves into
defineScreen and a context file becomes data: ids, counts, docs, cross-links.
Four doc links never resolved. Trip-Planner and Sharing-and-Collaboration are
not wiki pages (Trip-Planner-Overview and Trip-Members-and-Sharing are), the
places import anchor is importing-multiple-places, and the offline anchor was
written in GitHub's slug, which is not the one the in-app reader produces.
Nothing caught them, so the registry test now checks every slug and every
anchor against wiki/, with the heading slug itself lifted out of HelpPage so
the reader and the test cannot drift apart.
A help icon on every screen opens a panel with the screen's overview, its
sub-screens, task guides with a framed picture per step, search across guides
and the wiki, and a switch to any other screen's help. Guides exist for the
dashboard, Vacay, Atlas, Collections, Journey (list, journal, studio), Settings
(all tabs), Admin (all tabs) and the trip frame; the trip's own screens are
in progress. Pictures are produced by Playwright (client/e2e/help), texts in
all 23 languages.
2026-09-24 14:05:42 +02:00
7416 changed files with 712866 additions and 153823 deletions
@@ -20,8 +20,9 @@ Run from the repo root unless noted. **Use `npm run ... --workspace=<pkg>`, not
```bash
npm run dev # build shared, then watch shared + server + client concurrently
npm run build # build shared → server → client (order matters)
npm run lint # lint all three workspaces (server and shared run eslint --fix and rewrite files; client is check-only)
npm run lint # lint all three workspaces, check-only (npm run lint:fix applies the fixes in shared and server)
npm run format # prettier --write all three
npm run lint:knip # knip over client/server/shared (knip.jsonc): unused files, exports, types, deps may only shrink against scripts/ci/knip-baseline.json (CI gate; --update lowers it)
```
Tests:
@@ -32,9 +33,34 @@ npm run test --workspace=server # vitest run (server); also t
npm run test:e2e # server e2e (boots Nest against a temp SQLite)
npm run test:cov # coverage (lcov) for all four packages incl. plugin-sdk
cd client && npm run test# client vitest run
cd client && npm run e2e # Playwright (local only)
cd client && npm run e2e # Playwright (CI runs the public and app projects; the screenshot and help-media ones are local)
cd client && npm run lint:pages # enforce the Page pattern (CI gate)
cd client && npm run theme:lint# flag styling that bypasses appearance tokens (local only)
cd client && npm run lint:rtl # physical left/right sides may only shrink per file (CI gate)
cd client && npm run lint:dup # lines in copied code may only shrink per file, at SonarCloud's thresholds (CI gate)
cd client && npm run lint:pairs # desktop and phone views of a feature import its shared hook; every phone screen is listed (CI gate)
cd client && npm run lint:size # no source file past 1000 lines, longer ones only shrink (CI gate; server has the same)
cd client && npm run lint:strict # strictNullChecks + noImplicitAny errors may only shrink per file; new files start at zero (CI gate)
cd client && npm run lint:warnings # eslint (typed) errors fail, warnings per rule may only shrink, new rules start frozen (CI gate)
cd server && npm run lint:warnings # the same for the server, against server/scripts/eslint-baseline.json (CI gate)
cd server && npm run lint:format # Prettier outside the baseline list (CI gate; shared has the same in lint-prettier.yml)
cd server && npm run lint:strict # strict type errors per file may only shrink, a new file has none (CI gate)
cd client && npm run lint:format # Prettier outside the baseline list (CI gate)
cd client && npm run lint:layers # components never import pages/ or mobile/ (CI gate)
cd client && npm run lint:offline # views may not import src/api/ beyond the baseline (CI gate)
cd client && npm run lint:skips # no .only; skipped tests may only shrink (CI gate)
cd client && npm run lint:i18n-keys # every key the client names exists in en; unused en keys may only shrink (CI gate)
cd server && npm run lint:boundaries # no new import cycles or cross-domain reach-ins (CI gate)
cd server && npm run lint:dialect # SQLite-only SQL spellings may only shrink (CI gate; rules in server/CLAUDE.md)
cd server && npm run probe:pg # dialect helpers + the repository statements it reaches against Postgres (CI job; needs TREK_PG_PROBE_URL)
cd server && npm run lint:tx # no new method writing twice outside one transaction (CI gate)
cd server && npm run lint:test-sql # raw SQL fixtures in tests/ may only shrink; seed through tests/helpers/factories (CI gate)
cd server && npm run lint:test-mocks # vi.mock of config, realtime and db modules in tests/ may only shrink (CI gate)
cd server && npm run lint:test-new-service # hand-built services and repositories in tests/ may only shrink (CI gate)
cd server && npm run lint:query-api # MikroORM QueryBuilder calls in the repositories may only shrink (CI gate)
cd server && npm run lint:mcp-zod # inline zod in the MCP tool files may only shrink; fields derive from @trek/shared (CI gate)
cd server && npm run lint:service-http # { error, status } results and their caller-side branches may only shrink; services throw DomainError (CI gate)
cd server && npm run lint:response-contracts # route handlers without @ResponseContract may only shrink (CI gate)
cd client && npm run theme:lint # styling that bypasses appearance tokens may only shrink per file (CI gate)
npm run typecheck --workspace=server # tsc --noEmit (also client/shared; server also has typecheck:tests)
npm run i18n:parity:strict --workspace=shared # CI gate (exit 1 on drift)
```
Dev ports: the server listens on **3001**; the Vite dev server proxies the API, websocket, uploads, MCP and OAuth paths to it (see `client/vite.config.js`). There is no Node version pin — the CI workflows and the Dockerfile define what is tested. `server/.env.example` is the env-var reference.
Dev ports: the server listens on **3001**; the Vite dev server proxies the API, websocket, uploads, MCP and OAuth paths to it (see `client/vite.config.js`). `.nvmrc` pins the Node major the Docker image runs and every CI job reads; `engines` in each package.json states the range that installs (npm only warns outside it). `server/.env.example` is the env-var reference.
## Server invariants
@@ -62,8 +88,8 @@ Nest owns everything: every domain is a DI module under `server/src/nest/<domain
- **`/mcp` bodies are raw by design** — the parser wrappers exempt `/mcp` so the MCP SDK reads the untouched stream; never `@Body()` a `/mcp` route.
- **Auth is default-deny**: global guards run as `APP_GUARD`s; public routes opt out with `@Public()`, and a boot-time ratchet (`validate-route-guards.ts`) refuses any public route not on its allow-list.
- **Trip-scoped routes** verify trip access (404) and the permission (403) — `@UseGuards(JwtAuthGuard, TripAccessGuard)` + `@RequirePermission('<action>')` per handler — and forward `X-Socket-Id` to the broadcast so the originating client doesn't echo its own change. **Never gate a multipart upload with a guard** (the client sees ECONNRESET instead of 403 — check in the handler).
- **DB** (`server/src/db/`): `better-sqlite3`, WAL, FK on. Migrations are positional and append-only. `NODE_ENV=test` gives each vitest worker an isolated `:memory:` DB.
- **Realtime**: `/ws` is a Nest gateway in `server/src/nest/realtime/` — inject `RealtimeService`; `server/src/websocket.ts` is a legacy re-export stub.
- **DB** (`server/src/db/`): `better-sqlite3`, WAL, FK on. MikroORM's migrations (`db/migrations/`) are the only schema source, **including for tests** — the hand-kept `schema.ts`/`migrations.ts`/`seeds.ts` test builders are gone. Seeding is MikroORM's too (`db/seeders/`), applied at boot from `buildApp()`; the ORM shares the one better-sqlite3 connection (bound in `server/src/db/orm-driver.ts`) rather than opening its own. Queries go through repositories (`server/src/db/repositories/`) and `UnitOfWork` (`server/src/nest/database/unit-of-work.ts`) for transactions — there is no `DatabaseService`. `better-sqlite3` itself is ESLint-restricted to `src/db/` plus a short, named allow-list (`server/eslint.config.mjs`), so a domain reaching for the driver directly fails lint. `NODE_ENV=test` gives each vitest worker an isolated `:memory:` DB.
- **Realtime**: `/ws` is a Nest gateway in `server/src/nest/realtime/` — inject `RealtimeService`. Tests observe broadcasts through a `FakeRealtimeService` (`server/tests/helpers/fake-realtime.ts`), never by mocking the transport module.
- **MCP**: tools are declared in `<domain>.mcp.ts` files beside each module, on the `server/src/nest-mcp/` decorator layer. A validation or permission change on a REST route must land in the parallel MCP tool in the same change. See `MCP.md`.
- **Addons** are admin-toggleable feature modules keyed by `ADDON_IDS` in `server/src/addons.ts`.
- **Plugins** (`server/src/nest/plugins/`): one forked child process per plugin, permission-gated RPC (a handler is registered only if the plugin holds the unlocking permission), install-time manifest/signature/egress/version-range gates. The protocol tables are **generated** into `plugin-sdk/` and `shared/` by `gen:plugin-facts`; `check:plugin-facts` is the CI gate. Author tooling is `plugin-sdk/`; the `trek-plugin-dev` skill covers authoring.
@@ -75,14 +101,16 @@ Nest owns everything: every domain is a DI module under `server/src/nest/<domain
Offline-first, with a layered data flow. A Page never owns state directly:
- **Page pattern** (enforced by `lint:pages`, spec in `client/src/pages/PATTERN.md`): a `*Page.tsx` is a **wiring container** composing a co-located `use<Page>()` hook; it must not call React state/effect/memo hooks itself.
- **Data flow**: `store (Zustand, client/src/store/) → repo (client/src/repo/) → api (client/src/api/) | Dexie (client/src/db/offlineDb.ts)`. Writes go through `sync/mutationQueue.ts`: optimistic Dexie write, then a replay with an `X-Idempotency-Key` header on reconnect. `store/slices/remoteEventHandler.ts` applies incoming WebSocket events. Components never call the API directly.
- **Styling**: semantic Tailwind tokens (`bg-surface*`, `text-content*`, `border-edge*`, `bg-accent*`) or the underlying `var(--token)` variables — never color literals, arbitrary-value color classes or numeric inline `fontSize`, so user-chosen scheme/transparency/text-size keep working. `theme:lint`flags bypasses; suppress intentional exceptions (map/PDF/brand) with a `theme-lint-disable` line comment.
- **Data flow**: `store (Zustand, client/src/store/) → repo (client/src/repo/) → api (client/src/api/) | Dexie (client/src/db/offlineDb.ts)`. Writes go through `client/src/sync/mutationQueue.ts`: optimistic Dexie write, then a replay with an `X-Idempotency-Key` header on reconnect. `client/src/store/slices/remoteEventHandler.ts` applies incoming WebSocket events. Components never call the API directly.
- **Styling**: semantic Tailwind tokens (`bg-surface*`, `text-content*`, `border-edge*`, `bg-accent*`) or the underlying `var(--token)` variables — never color literals, arbitrary-value color classes or numeric inline `fontSize`, so user-chosen scheme/transparency/text-size keep working. `theme:lint`holds the bypasses per file to a baseline that only shrinks (CI gate); mark intentional exceptions (map/PDF/brand) with a `theme-lint-disable` line comment, which is counted too.
- **Reading direction**: Arabic runs the app right to left, so spacing, insets and alignment that follow the text use the logical forms (`ms-`/`me-`/`ps-`/`pe-`/`start-`/`end-`/`text-start`/`border-s`/`rounded-s`, `marginInlineStart`, `insetInlineEnd`, `textAlign: 'start'`). `lint:rtl` counts the physical ones per file against `client/scripts/rtl-baseline.json`, which may only shrink; geometry that is physical on purpose in new code (maps, measured positions, time axes) carries an `rtl-lint-disable` comment.
- **PWA**: `vite-plugin-pwa` + Workbox cache tiles, API and uploads; `prebuild` generates icons.
## Shared contracts & i18n
- A route is "done" only once its contract lives in `shared/` and both sides import the inferred types. Edit the Zod schema, rebuild shared, then server (validation + DTO types) and client (typed requests) pick it up.
- Locale files live in `shared/src/i18n/<locale>/`, one file per domain; `en/` is canonical. When you add or change a key, **add a real translation to every locale** — parity fails CI on missing keys, and an English placeholder is not acceptable in a non-`en` locale.
- Count-bearing strings are plural groups: the key holds the general form and `key.one`, `.few` … the forms the language's `Intl.PluralRules` selects. Pass `{ count }` as a number and never choose between two keys with `count === 1`; details in `shared/CLAUDE.md`.
## Project direction
@@ -90,7 +118,7 @@ These principles come out of a full-repo audit and shape all new code:
- **Protect the crown jewels.** The client's offline-first data core (repos + mutation queue + Dexie), the out-of-process plugin sandbox, the Zod contract layer, and the auth primitives are production-grade. Extend them; never route around them or rewrite them in anger.
- **No "modern shell over legacy core."** The debt pattern is a modern frame delegating to the old approach underneath (on the client: feature components calling the raw API past the offline core). New code goes all-in on the target architecture; when touching a legacy seam, migrate it rather than adding another wrapper.
- **Single source of truth over manual synchronization.** Never add a hand-mirrored copy of state, schema, or contract (server↔Dexie↔Zustand, schema↔SQL, the plugin contract). Derive from one source, or guard the copy with a parity test that cannot silently skip.
- **Single source of truth over manual synchronization.** Never add a hand-mirrored copy of state, schema, or contract (server↔Dexie↔Zustand, schema↔SQL, the plugin contract). Derive from one source, or guard the copy with a parity test that cannot silently skip — the server entities are the precedent: `npm run gen:entities`/`check:entities --workspace=server` (`server/scripts/generate-entities.ts`) regenerate them from the live migrated schema instead of hand-editing, and a parity test (PARITY-009) pins entity-vs-DB drift (delete rules, index names) at zero skips. The Kysely table types the repositories query through are generated the same way (`gen:db-types`/`check:db-types`, `server/scripts/generate-db-types.ts`).
- **DI over global mutable module state.** New server code injects its dependencies; new client code avoids `window`-global buses and mutable module state.
- **Use the runtime's native idioms.** Prefer React 19 features (`useOptimistic`, Actions, `use()`/Suspense) and Nest subsystems (providers, pipes, guards, `@nestjs/schedule`) over hand-rolled equivalents.
- **Fail closed, gates stay on.** Security switches default to safe; misconfiguration must refuse, not degrade. Never lower a quality gate to land a change — no new `any`, no new `eslint-disable`, no downgrading rules to `warn`, no lowered coverage thresholds.
@@ -106,7 +134,14 @@ These principles come out of a full-repo audit and shape all new code:
## Quality gates in CI
`.github/workflows/test.yml` runs, per package: typecheck (server also `typecheck:tests`), `lint:check`, `lint:pages`, `check:plugin-facts`, `i18n:parity:strict`, the S3 contract suite, coverage for all four packages, then a SonarCloud scan. `lint-prettier.yml` additionally runs `lint` + `format:check` on `shared/`. `plugin-sdk` has no lint script and is outside the eslint gates. **Not in CI**: `theme:lint`, Playwright, `check:gl-split` — run them locally. There are no git hooks; nothing runs before a commit except you. Other workflows build Docker images, scan them, and publish `trek-plugin-sdk` on `plugin-sdk-v*` tags.
`.github/workflows/test.yml` runs, per package: typecheck (server also `typecheck:tests`, client also the strict-mode ratchet `lint:strict`), lint (client and server through `lint:warnings`, which also holds the warning count per rule), `lint:pages`, `lint:rtl`, `lint:size` (client and server), the server's `lint:strict`, `lint:boundaries`, `lint:tx`, `lint:dialect`, `lint:test-sql`, `lint:test-mocks`, `lint:test-new-service`, `lint:query-api`, `lint:mcp-zod`, `lint:service-http` and `lint:response-contracts`, the client's and server's `lint:format`, `lint:layers`, `lint:offline`, `lint:dup`, `lint:pairs`, `lint:skips`, `lint:i18n-keys` and `theme:lint`, shared's `contracts:open`, the repo-wide `lint:knip` (in server-quality), `check:plugin-facts`, `i18n:parity:strict` (with the untranslated-strings ratchet), the S3 contract suite, the Postgres probe (`postgres-probe`: every dialect helper on SQLite and Postgres, and the statements every repository method reaches with sample arguments against a Postgres service; the SQL errors per method and the list of methods it cannot measure are held to a baseline that only shrinks and fails while it is still unmeasured; see `server/CLAUDE.md`), the browser end-to-end flows (Chromium and WebKit), a client build with `check:gl-split`, coverage for all four packages, then a SonarCloud scan. It also builds the Docker image and boots it until `/api/health/ready` answers (`docker-image`), and lints and renders the Helm chart against the Kubernetes schemas and runs `docker compose config` on every compose file (`deploy-manifests`). `lint-prettier.yml` additionally runs `lint` + `lint:format` on `shared/`, both check-only. `plugin-sdk` has no lint script and is outside the eslint gates. **Not in CI**: the screenshot and help-media Playwright projects; run them locally. There are no git hooks; nothing runs before a commit except you. Other workflows build Docker images, scan them, and publish `trek-plugin-sdk` on `plugin-sdk-v*` tags.
- **`ci-ok` is the required check.** The branch ruleset requires `ci-ok` (plus `lint` and `check-target` from their own workflows), and `ci-ok` needs every job above except the scan. It fails if any of them failed or was cancelled and passes if they passed or were skipped. A new gate job counts only once it is listed in `ci-ok`'s `needs:`.
- **Guides name paths that exist**: the `changes` job runs `scripts/ci/doc-paths.mjs`, which resolves every backtick path in the CLAUDE.md files, `README.md`, `server/src/nest/README.md` and the wiki against the tracked files; the missing ones are listed in `scripts/ci/doc-paths-baseline.json`, which only shrinks (`--update` after a fix).
- **What runs is decided by `changes`, not by `on.paths`**: `scripts/ci/changed-areas.mjs` maps the changed files to the `code`, `image` and `deploy` areas (tests: `npm run test:ci-scripts`). A path filter on the workflow would leave `ci-ok` unreported on a docs-only PR and block it. Add a new top-level file the build reads to the right area there.
- **Runners and Node**: jobs run on `ubuntu-24.04`, not `ubuntu-latest`, and every `setup-node` reads `.nvmrc`. Third-party actions (anything outside `actions/*`) are pinned by commit SHA with the tag in a comment; Dependabot keeps both current.
- **Security scans are advisory.** `security.yml` runs `npm audit` (what the image installs, plus `plugin-sdk`) on every PR and push into dev and main, and Docker Scout on the built image for main. Neither is required: a newly published advisory would turn every open PR red at once. Fixes are held as floors in the root `package.json``overrides`; npm ignores an `overrides` block in a workspace `package.json`, so never put one there.
- **Releases** are dispatched by hand and refuse a commit whose `ci-ok` is not green. `docker-dev.yml` (any branch, usually dev) is the canary: it builds the dispatched commit as the next minor or major `-pre.N` and moves only `:latest-pre` and `:<major>-pre`. `docker.yml` (main only) bumps the version, commits and tags it, builds both platforms from exactly that commit and moves `:latest`, `:<major>`, `:<major>.<minor>` and `:<version>`. Its `bump=auto` only finalizes a prerelease in flight; without one it stops and asks for patch, minor or major. Ship a release that migrates or changes a lot as a prerelease first, and as a minor, so installs pinned to `:<major>.<minor>` can wait it out.
**SonarCloud** (project `liketrek_TREK`, the built-in Sonar way gate; the run takes 20–25 min, so get it right before pushing) measures **new code only** on a PR; any failing condition blocks it:
@@ -117,7 +152,7 @@ These principles come out of a full-repo audit and shape all new code:
| Security / reliability / maintainability rating | A |
| Security hotspots reviewed | 100% |
- **Duplication is the one that bites.** Desktop and phone shells (`client/src/components/<X>` ↔ `client/src/mobile/screens/.../M<X>`) are deliberate mirrors that are already heavily duplicated. Every line added identically to both counts against the 3% budget, so a feature touching both shells fails on its own. Put the logic in ONE hook/module (precedents: `useInstanceSettings`, `useRangeBypass` under `client/src/components/Admin/`) and leave only markup in each shell. Locale files are excluded from duplication, so i18n additions are free.
- **Duplication is the one that bites.** Desktop and phone shells (`client/src/components/<X>` ↔ `client/src/mobile/screens/.../M<X>`) are deliberate mirrors of each other's markup; their logic lives in one shared hook per feature. Every line added identically to both counts against the 3% budget, so a feature touching both shells fails on its own. Put the logic in ONE hook/module (precedents: `useInstanceSettings`, `useRangeBypass` under `client/src/components/Admin/`) and leave only markup in each shell.`lint:dup` finds the same copies in `client/src` before the push and holds every file to `client/scripts/dup-baseline.json`; `lint:pairs` holds each pair in `client/scripts/feature-pairs.json` to its shared hook and makes every new phone screen either join a pair or say why it has no twin. Locale files are excluded from duplication, so i18n additions are free.
- **Coverage is per new line across all four packages.** A new file without tests drags the PR under 80%. Files in `sonar.coverage.exclusions` don't count either way.
- **Ratings are about new issues.** One new bug drops reliability; one new vulnerability drops security. Don't introduce hotspot patterns (`Math.random` in anything security-adjacent, `child_process` through PATH, backtracking regexes, secret-looking strings). Path-wide suppressions live only in `sonar-project.properties` with a written justification; one-off findings are resolved on SonarCloud, never with an inline `// NOSONAR`.
- **Check it yourself.** The `sonarqube` MCP is connected: `list_pull_requests` → `get_project_quality_gate_status` shows each condition's value; `search_duplicated_files` + `get_duplications` show what tripped duplication; `search_sonar_issues_in_projects` lists new issues. Test files are excluded from analysis — when duplication fires, the source is the problem.
@@ -43,7 +43,7 @@ A self-hosted, real-time collaborative travel planner — with maps, budgets, pa
<div align="center">
<img src="https://github.com/mauriceboe/trek-media/releases/download/readme-assets/TREK1.webp" alt="TREK, a tour of the app" width="100%" />
<img src="https://github.com/mauriceboe/trek-media/releases/download/readme-assets/TREK2.webp" alt="TREK, a tour of the app" width="100%" />
</div>
@@ -168,9 +168,9 @@ Most of what follows is an addon an admin switches on or off. Lists, Costs, Docu
- **Admin panel**: users and invites, the permission matrix, packing templates, categories, addons, plugins, API keys, MCP tokens and OAuth sessions, backups, storage, audit log, and GitHub releases
- **Backups**: manual or scheduled hourly, daily, weekly or monthly, with retention in days. The zip carries the at-rest key, so a restore can decrypt its own secrets
- **Storage**: pluggable storage backends per content category — keep everything on local disk, or add S3-compatible backends and replicate any category to them, configured entirely from the admin panel
- **Notifications**: a per-user matrix of events against in-app, email (SMTP), webhook and ntfy, plus any channel a plugin registers
- **Notifications**: a per-user matrix of events against in-app, email (SMTP), webhook, ntfy and Web Push (on every device you turn it on for, even with TREK closed), plus any channel a plugin registers
- **Appearance**: light, dark or follow the OS, seven colour schemes plus a custom accent, transparency, compact density, reduce motion, and text size per tier
- **23 languages**: en, de, es, fr, it, nl, hu, ru, zh, zh-TW, pl, cs, ar (right to left), br, id, tr, ja, ko, uk, gr, sv, vi, ca
- **27 languages**: en, de, es, fr, it, nl, hu, ru, zh, zh-TW, pl, cs, ar (right to left), br, id, tr, ja, ko, uk, gr, sv, vi, ca, az, et, sk, th
- **In-app help**: the wiki ships inside the image and is served from disk at `/help`, so the docs match the version you are running
</td>
@@ -223,7 +223,7 @@ Open `http://localhost:3000`. On first boot TREK seeds an admin account — if y
@@ -53,4 +53,5 @@ See `values.yaml` for more options.
-`FORCE_HTTPS` is optional. Set `env.FORCE_HTTPS: "true"` only when ingress (or another proxy) terminates TLS. It enables HTTPS redirects, HSTS, CSP `upgrade-insecure-requests`, and forces the session cookie `secure` flag. Requires `TRUST_PROXY` to be set.
- Set `env.TRUST_PROXY: "1"` (or the number of proxy hops) when running behind ingress or a load balancer. Required for `FORCE_HTTPS` to detect the forwarded protocol correctly. In production it defaults to `1` automatically.
-`COOKIE_SECURE` is auto-derived (on when `NODE_ENV=production` or `FORCE_HTTPS=true`). Set `env.COOKIE_SECURE: "false"` only during local testing without TLS. **Not recommended for production.**
- Web Push needs no chart settings: TREK generates its key pair on first start and keeps it in the database on the data PVC. To bring your own pair, set `env.VAPID_PUBLIC_KEY` and `secretEnv.VAPID_PRIVATE_KEY` together (and optionally `env.VAPID_SUBJECT`); changing the pair later means every device has to turn push on again.
- Set `env.OIDC_DISCOVERY_URL` to override the auto-constructed OIDC discovery endpoint. Required for providers (e.g. Authentik) that expose it at a non-standard path.
@@ -8,15 +8,28 @@ Scope: the **`@trek/client`** workspace (React 19 + Vite + Zustand + Tailwind PW
npm run dev # Vite dev server; proxies the API/ws/uploads/MCP/OAuth paths → http://localhost:3001
npm run build # prebuild generates PWA icons, then vite build
npm run typecheck # tsc --noEmit (CI)
npm run lint# eslint . (lint:check in CI)
npm run lint:strict # tsc over tsconfig.strict.json (strictNullChecks + noImplicitAny): errors per file may only shrink against scripts/strict-baseline.json, a new file starts at zero (CI gate; --list shows them, --update lowers it)
npm run lint # eslint . (CI runs lint:warnings instead, which fails on the same errors)
npm run lint:warnings # eslint (typed), failing on any error and on any warning count per rule above scripts/eslint-baseline.json and on any entry above its count (app code, tests and eslint-disable'd messages counted apart; a `/* eslint <rule>: ... */` config comment fails outright; --update lowers it). Frozen as warnings: floating/misused promises, `.catch(() => {})` (trek/no-swallowed-catch), react-hooks v7 recommended, complexity/max-depth/max-params/max-lines-per-function (120 per .ts function, 300 per .tsx), `use*Store()` without a selector, `window.__*` and window.dispatchEvent buses (trek/no-window-globals), no-console, no-non-null-assertion; the local rules live in scripts/lib/eslint-rules.mjs
npm run lint:pages # enforce the Page pattern (CI gate)
npm run lint:rtl # physical left/right sides and rtl-lint-disable comments may only shrink per file (CI gate; --list shows them, --update lowers both baselines)
npm run lint:size # no source file or stylesheet past 1000 lines, no test (Playwright specs and fixtures under e2e/ included) past 2000, counted wrapped at 120 columns; longer ones may only shrink (CI gate; --update lowers the baseline)
npm run lint:format # Prettier: every .ts/.tsx/.mjs/.css file under src/, tests/, e2e/ and scripts/ outside scripts/format-baseline.json must be formatted, the list only shrinks (CI gate; npx prettier --write <file>, then --update)
npm run lint:layers # imports go downwards: components never import pages/ or mobile/, nothing under the views imports a view; per-file counts may only shrink (CI gate; --list, --update)
npm run lint:offline # no view file (components/mobile/pages/hooks) imports src/api/ beyond scripts/offline-baseline.json, which only shrinks (CI gate; --list, --update)
npm run lint:pairs # each desktop/phone pair in scripts/feature-pairs.json imports its shared hooks, and every file under src/mobile/screens is listed there, in a pair or under mobileOnly with a reason (CI gate; --list prints the pairs)
npm run lint:dup # copied code: the lines of a file inside a block that repeats elsewhere (100 tokens over 10 lines, SonarCloud's thresholds; .ts and .tsx read as one language) may only shrink against scripts/dup-baseline.json (CI gate; --list shows each block with its other side, --update lowers it)
npm run lint:skips # no .only anywhere; skipped/todo/fixme tests per file may only shrink against scripts/skip-baseline.json (CI gate; skipIf/runIf and Playwright's test.skip(condition, 'why') are fine, but skipIf(true) and runIf(false) count as skips)
npm run lint:i18n-keys # every translation key src/ names exists in shared en, and en keys nothing in src/, server/src or plugin-sdk/src reaches may only shrink per en file against scripts/i18n-unused-baseline.json (CI gate; --unused lists them, --update lowers the baseline)
npm run test# vitest run (tests/** + co-located src/**/*.test.{ts,tsx}); also test:unit / test:integration / test:coverage
npm run e2e # Playwright (local only; e2e:report opens the last report)
npm run e2e # Playwright (CI runs --project=public --project=app; e2e:report opens the last report)
npm run shots # Playwright screenshot project (shots:promote to accept)
npm run theme:lint # theme conformance audit (theme:lint:strict exits 1; local only)
npm run check:gl-split # after build: fails if one chunk bundles both mapbox-gl and maplibre-gl (local only)
npm run theme:lint # colour literals, palette classes, raw text sizes, z-index literals above z-50 and dark_mode reads may only shrink per file, theme-lint-disable markers too (CI gate; --list shows them, --update lowers both baselines)
npm run check:gl-split # after build: fails if one chunk bundles both mapbox-gl and maplibre-gl (CI gate)
```
Every baseline ratchet above also fails on an entry that allows more than the tree holds: a count above what its file or rule has now, or a file that is gone. Run the check with `--update` in the change that made it smaller, so a fixed file cannot grow back to its old entry and a new file at a deleted path does not inherit one.
Single test: `npx vitest run src/store/slices/budgetSlice.test.ts`, or `npx vitest run -t "optimistically adds the place"`.
The dev server needs the API on **:3001** — run `npm run dev` at the repo root to start both.
@@ -32,15 +45,16 @@ The layering is **component → feature hook → store/slice → `repo/` → `ap
- **`src/store/`** — Zustand. `tripStore.ts` is composed from slices in `store/slices/`; `slices/remoteEventHandler.ts` applies inbound WebSocket events to local state.
- **`src/repo/`** — per-entity repositories mediating between the network and the offline cache: online, call REST and upsert into Dexie; offline, read Dexie. Always go through a repo for trip data.
- **`src/api/client.ts`** — the single Axios instance, typed from `@trek/shared`. New API surface goes in a per-domain `api/<domain>.ts`, not the god module.
- **`src/db/offlineDb.ts`** — the Dexie schema. Database and table names are the on-device contract — renaming one orphans every user's cache.
- **`src/sync/`** —`mutationQueue.ts`: offline writes do an optimistic Dexie write (temporary **negative id** for creates), then `flush()` replays REST on reconnect with the mutation's UUID as the **`X-Idempotency-Key`** header (matching the server interceptor) and reconciles the row. The rest of the directory drives reconnection sync, the auth gate that stops background loops before logout swaps the DB, offline preferences, and map pre-download (raster and vector prefetchers).
- **`src/db/offlineDb.ts`** — the Dexie schema. Database and table names are the on-device contract — renaming one orphans every user's cache. Versions are append-only: a new `this.version(N)` adds its stores to `HISTORY` and a case to `UPGRADE_CASES` in `tests/unit/db/offlineDb.upgrade.test.ts`, which opens the previous version with seeded rows in fake-indexeddb and upgrades it to the current one; a guard there fails on a declared version without both. Never edit a shipped version's stores or `upgrade()`.
- **`src/sync/`**: in`mutationQueue.ts`, offline writes do an optimistic Dexie write (temporary **negative id** for creates), then `flush()` replays REST on reconnect with the mutation's UUID as the **`X-Idempotency-Key`** header (matching the server interceptor) and reconciles the row. Writes to one entity keep their order: a write parked as `failed` or `conflict` holds back the later writes to the same `resource` + entity id until the user retries, resolves or discards it, and only that action deletes a parked row (trip eviction never does). Every row `enqueue()` writes carries `schemaVersion` (`MUTATION_SCHEMA_VERSION` in `offlineDb.ts`) and `buildVersion`; bump the constant when a row's fields or the way the replay reads them change incompatibly. A flush replays its own format and older ones (an unstamped row is format 1) and skips a row from a newer format without sending, marking or dropping it, holding back the later writes to its entity, so a tab still on an older bundle never replays a shape it does not know. The guard exists only from the release that introduced `schemaVersion` on; a bundle older than that replays every row, so the first bump of `MUTATION_SCHEMA_VERSION` must not ship while a rollback target older than the guard is still in play. A skipped row stays `pending` with no retry or discard path, so on a rolled-back install it holds its entity until the user clears the cache. Repos whose writes carry no version token (visits, day clears, tours, driving settings) ask `mutationQueue.mustQueue()` before an online write and queue it while an older write to the entity is still queued, so a retried write cannot land over a newer one made here online; places and packing items send `X-Base-Updated-At`, so the server refuses the stale replay with a 409 instead. `flush()` takes the Web Lock `trek-mutation-flush:<db name>` so only one tab replays at a time; a tab that finds it taken waits for it, and a `flush()` called while this tab's flush runs or waits for the lock gets that same flush back (with one more pass asked for), so `flush()` never resolves while a tab is still replaying and a re-seed chained onto it cannot lay older server rows over those writes (the in-tab flush alone where the API is missing). The rest of the directory drives reconnection sync, the auth gate that stops background loops before logout swaps the DB, offline preferences, and map pre-download (raster and vector prefetchers).
## Rules for new code
The offline core is flagship work surrounded by a periphery that ignores it. New code extends the core's discipline, not the periphery's shortcuts:
- **One data path, no layer skips.** Never import `api/client` (or raw fetch/axios) from a component or modal — especially never *write* from one. A long tail of existing files bypasses the layering and no lint ratchet holds the line yet; that is debt, not precedent.
- **One data path, no layer skips.** Never import `api/client` (or raw fetch/axios) from a component or modal, and especially never *write* from one. `lint:offline` enforces it: a file under `components/`, `mobile/`, `pages/` or `hooks/` may not import anything from `src/api/` except types. The long tail that already does is listed in `scripts/offline-baseline.json`; the list only shrinks (a file that stops leaves it via `--update`) and is debt, not precedent. A domain that is online-only on purpose (admin, auth) still puts its calls in a store or repo, not in the view.
- **Offline-first is the product promise.** New domains get a repo with read-through and offline writes (temp id → optimistic Dexie write → idempotent queue entry). If a domain is deliberately online-only, say so explicitly.
- **Imports go downwards** (`lint:layers`, rules in `scripts/lib/layers.mjs`). A component never imports `pages/` or `mobile/`, a shared hook no view, and stores, repos, sync, api, db and the helpers no view at all; a file directly in `src/` is the layer of its name, so `src/types.ts` follows the `types/` rules. A model or helper both sides need moves down (`utils/`, `hooks/`, `types/`); the counts in `scripts/layers-baseline.json` are debt, not precedent.
- **No god components or god hooks.** The page pattern is a floor, not a ceiling. Decompose hooks by concern, split renders into memoizable sections, and don't pass huge prop bags with inline-arrow callbacks.
- **Prefer React 19 idioms**: `useOptimistic`, `useActionState`/`useFormStatus`, `use()` + Suspense, `useSyncExternalStore` for external subscriptions (`hooks/useNetworkMode` is the reference).
- **Optimistic writes must reconcile** — rollback + user-visible toast on failure. No `.catch(() => {})`, no bare `catch {}`.
@@ -50,19 +64,25 @@ The offline core is flagship work surrounded by a periphery that ignores it. New
- **Rendering security**: never interpolate user content into HTML strings (map popups included — build via DOM + `textContent`, or `escapeHtml` from `@trek/shared`); untrusted markdown gets `rehype-sanitize`; never `rehype-raw` near it.
- **No `window` event buses or global mutable `window` state.** No `any` at boundaries: WS payloads and map renderer props get real types.
- **Hygiene**: error boundaries around new shells/routes (`components/shared/ErrorBoundary.tsx`); every async `.then(setState)` needs a cancelled flag or `AbortController`; no routine `eslint-disable exhaustive-deps` (a missing dep has already shipped wrong money on screen); no sequential-await N+1 fetch loops; search for an existing utility before writing a duplicate.
- **Theming**: use the semantic Tailwind tokens defined in `tailwind.config.js` (`bg-surface*`, `text-content*`, `border-edge*`, `bg-accent*`, status colors) — no palette classes, hex literals, arbitrary-value color classes, or invented CSS vars. Only `applyAppearance()` in `src/theme/` mutates `<html>` styling (see `src/theme/README.md`). `theme:lint` catches literals and`bg-[#...]`-style classes but not named palette classes — review those by hand.
- **Theming**: use the semantic Tailwind tokens defined in `tailwind.config.js` (`bg-surface*`, `text-content*`, `border-edge*`, `bg-accent*`, status colors) — no palette classes, hex literals, arbitrary-value color classes, or invented CSS vars. Only `applyAppearance()` in `src/theme/` mutates `<html>` styling (see `src/theme/README.md`). `theme:lint` counts literals,`bg-[#...]`-style and palette classes, raw `text-sm`- and `text-[13px]`-style sizes, z-index literals and `dark_mode` reads per file against `scripts/theme-baseline.json`.
## Big-picture pieces
- **Maps** (`src/components/Map/`): two interchangeable renderers — Leaflet and Mapbox/MapLibre GL — chosen at runtime by `MapViewAuto.tsx`. Keep both in sync when changing map features, and keep `mapbox-gl` and `maplibre-gl` in separate chunks (`check:gl-split`). The raster prefetcher fetches tiles `no-cors` so custom tile providers without CORS headers keep working; the vector prefetcher deliberately uses `cors` because it has to read the responses — don't "align" one with the other.
- **i18n** (`src/i18n/TranslationContext.tsx`): `en` is bundled; every other locale is a dynamic `import('@trek/shared/i18n/<locale>')` so Vite emits one chunk per locale. Strings live in `shared/`, never here.
- **Mobile shell** (`src/mobile/`): below the phone breakpoint (`useIsPhone`) `App.tsx` wraps routes in `MobileShell` and the `M*` screens under `mobile/screens/` take over. A UI change to a domain with an `M*` twin usually needs both — and the shared-hook rule below.
- **i18n** (`src/i18n/TranslationContext.tsx`): `en` is bundled; every other locale is a dynamic `import('@trek/shared/i18n/<locale>')` so Vite emits one chunk per locale. Strings live in `shared/`, never here.`t()` returns an unknown key unchanged, so `lint:i18n-keys` (`scripts/i18n-keys.mjs`) resolves every literal key in `t()`, `tHtml()`, `tr()`, `<TransHtml html>`, `translateApiError` and `*Key` props against `shared/src/i18n/en`; a template whose interpolations only choose between string literals is expanded and each branch checked as a key; any other template key is read with each interpolation as one key segment and passes by itself only when it reaches between 1 and `MAX_IMPLICIT_MATCHES` (30) en keys; a wider one, or one with no fixed dotted prefix, needs a `DYNAMIC_ALLOWED` entry keyed by file and template that says what bounds the value, and an entry the check no longer needs fails. Pass counts as numbers, and let a plural group pick the wording instead of switching keys on the count.
- **Mobile shell** (`src/mobile/`): below the phone breakpoint (`useIsPhone`) `App.tsx` wraps routes in `MobileShell` and the `M*` screens under `mobile/screens/` take over. A UI change to a domain with an `M*` twin usually needs both, and the logic change goes into the sharedhook (see "Desktop and phone views" below).
- **Plugins UI** (`src/components/Plugins/`): third-party surfaces render only inside the sandboxed `PluginFrame` iframe (postMessage bridge, live theme-token sync). Host-rendered contribution points (widgets, schedule rows, map layers, planner columns/actions) render server-normalized primitives only — plugin markup never runs inline. Authoring is covered by the `trek-plugin-dev` skill.
- **Managed installs** (`src/managed/index.tsx`): the attachment point for screens that only exist on a centrally administered install. Empty here by design — an operator replaces it at build time. Don't put features there.
## Desktop ↔ mobile shells and the SonarQube duplication gate
## Desktop and phone views
Every desktop component with a phone twin is a deliberate copy that is already heavily duplicated. CI's SonarCloud gate allows **≤ 3% duplicated lines on the PR's new code**, so adding the same lines to both shells fails the PR by itself. Put logic in one shared hook/module (`useInstanceSettings`,`useRangeBypass` under `components/Admin/`are the pattern) and keep only markup in each shell. Full gate rules in the root `CLAUDE.md`.
Most features have a desktop view (`components/`, `pages/`) and a phone view (`mobile/screens/`). They share one hook or model per feature that holds the logic (state, effects, loading, handlers, validation, derived values, store and repo calls); each view keeps only its own markup, layout and dialogs or sheets. `useInstanceSettings` and`useRangeBypass` under `components/Admin/`show the shape. `scripts/feature-pairs.json` lists every pair with its hooks.
- **Changing a feature:** change the shared hook, not one view. A behaviour that really differs between the two views is a parameter of the hook, so both stay visible in one place and its test covers both.
- **A new feature with both views:** write the hook first under `components/<Feature>/` (or `hooks/`/`utils/` when it is not bound to one feature; a page's own logic stays in its `use<Page>` hook and models in `pages/<page>/`, which the phone shell imports too), give it its own test, build both views on it and add the pair to `scripts/feature-pairs.json`.
- **A new phone screen:** list it in `scripts/feature-pairs.json`, as a view of the pair whose hook it runs on or under `mobileOnly` with the reason it has no twin. A file under `mobile/screens/` that is in neither fails `lint:pairs`.
Two CI gates hold this: `lint:pairs` fails when a listed view stops importing its pair's hook or a phone screen is not listed, and `lint:dup` holds the lines inside copied blocks to `scripts/dup-baseline.json`, so a copy that looks alike fails too. SonarCloud allows **≤ 3% duplicated lines on the PR's new code**, so the same lines added to both views fail a PR by themselves. Full gate rules in the root `CLAUDE.md`.
| Step picture frame | `guide.ts``frameFor` | 16:10, at least 960 px wide, 64 px padding around the target, centred on it, never larger than the largest 16:10 frame the viewport holds (1728 × 1080). A target wider than that, a full-width row for instance, is therefore cut at the sides, and that is accepted: the narrow frame is the format of the whole library, and a picture that widens to fit one row is at a different magnification from the one beside it. A target taller than the frame is anchored at its top, so a long section shows its heading |
| Ring around the target | `guide.ts``ring` | 3 px in the app's `--accent`, 12 px radius, 6 px outside the element, the rest dimmed 30 %. A white hairline goes outside the ring and the badge only where the surface measured behind them has too little contrast against the accent, which is the lightbox and the other near-black overlays |
| The step's number | `guide.ts``ring` | a 28 px disc in the accent on the ring's top-left corner, outside it. Where that spot is not inside the frame, which is every target wider than the frame, it moves inside far enough to clear the frame's edge. The number is part of the picture; the panel numbers the step again beside it and puts nothing on top of the picture |
| Raster tiles | `guide.ts``settle` → `showLoadedTiles` | loaded tiles are forced visible before every shot. Leaflet fades a tile in over 200 ms and takes the progress from `+new Date()`, and the clock is pinned, so without this the satellite imagery and every raster basemap are a grey rectangle although each tile came back 200 |
| Drag & drop | `guide.ts``ringDrag` (a step with `dropTo`) | source ringed, destination ringed dashed, a bowed arrow from the one to the other, both undimmed |
| Hover | `captureGuide` | the target is hovered before the shot (tooltips, hover-only buttons show); `hover` overrides where the centre is wrong |
| Result picture | `guide.media.result` | the full frame after the last step, pointer parked in the corner |
| Hero (screen overview) | `captureHero` | the full frame of the screen as the reader finds it |
The app's own appearance is the default theme, English, light mode, the seeded
demo data (`e2e/screenshots/seed.ts`) plus the fixtures in `fixtures.ts`. Do not
change the theme, the language or the zoom for a picture.
## How a screen gets its pictures
1. Register the screen and its guides in `src/help/contexts/<screen>.ts` and
`src/help/registry.ts`; write the English texts as a block in
`shared/src/i18n/en/help.ts` (`// ── Screen: <id>`). Guide ids are global: one id,
one picture directory, so never reuse an id from another screen
(`registry.test.ts` refuses duplicates).
2. Write `e2e/help/<screen>.guide.ts`: one `GuideScript` per guide, exactly one
`StepAction` per registered step, `start` that opens the screen, `cleanup` that
puts the seed back. A step's `target` is the element the text tells the reader
to use; `prepare` gets there, `act` does what the text says.
3. Run `npm run help:media -- <screen>.guide.ts` (own ports 5183/3011, fresh
database, the seed, then the file). Fix until every guide passes; a failing
step is a text that no longer matches the UI.
4. Look at the pictures in `e2e/.tmp/help-media/`, then `node e2e/help/promote.mjs`.
5. Translate the block into every other locale (`i18n:parity:strict` is the gate) and
run `src/help/registry.test.ts`, which checks every promised picture is on disk.
## What the fixtures may look like
The documents and pictures the guides upload are drawn in `fixtures.ts`, because
no real person's paperwork and no photograph may live in this repository. They
still have to look like something a traveller would really put into a trip, or
the picture teaches the reader that the feature is a toy.
- **A document is a document.** `drawPdf` takes an issuer band, a title, a
reference line, a table of labelled fields and a footer, which is the shape of
every confirmation, ticket and voucher. Four lines of Helvetica on white is not
a document; it was the first attempt and it was thrown out.
- **A picture is a graphic, not a painting of a photograph.** A hand-drawn
market or landscape reads as clip art at any size, however it is shaded, and it
was tried three times. What works is the kind of image people actually upload:
a map screenshot with a pin and a search field, a booking confirmation in a
browser window, a printed network map. Those are drawings in reality too, so a
drawing of them is honest.
- **Every date comes from `e2e/dates.ts`**, never from a calendar typed into a
fixture. A confirmation whose check-in is two days before the trip's last day
keeps making sense next month.
- The names of the files are part of the seed: guides find their row by name, so
renaming a fixture means renaming it in the guide too.
## Changing the style
`guide.ts` is shared by every picture, so anything changed there is a property of
the whole library, not of the guide in front of you. Before changing it, work out
which pictures actually come out different and re-render exactly those, one
`--grep` away:
```bash
npm run help:media -- --grep "pictures: (files-trash|files-preview)$|hero: trip-files"
```
The staging directory answers most of those questions without a browser, because
the frame is in the file:
```bash
# which pictures were taken at the frame's maximum, i.e. cut at the sides
python -c "import struct,os;[print(r,f,struct.unpack('>II',open(os.path.join(r,f),'rb').read(24)[16:24])) for r,_,fs in os.walk('e2e/.tmp/help-media') for f in fs if f.endswith('.png')]"
```
A change that alters every picture (the frame, the ring's geometry, the viewport)
is a full run of about four hours on one worker. It is not a thing to do twice,
so look at three or four pictures from a subset run before starting it.
## Rules learned the hard way
- Never run `npx prettier` over `client/` sources: `client/.prettierrc` says
`semi: true` and the sources are written without semicolons, so it rewrites
whole files. Nothing in CI formats the client.
- A killed run leaves its server holding 3011 and 5183, and the next run dies
if(!body.connected)thrownewError(`Dawarich is not reachable from the media backend: ${body.error??''}${body.errorDetail??''} (is ALLOW_INTERNAL_NETWORK=true?)`)
body:'Takkyubin from the hotel front desk, **before 11:00**, and the bags are in Kyoto the next afternoon.\n\n- One bag per person\n- About 2000 JPY a bag',
category:'Transport',
website:'https://www.global-yamato.com/en/hc/',
}
/** An .example host never resolves, so the chip shows its own glyph offline too. */
// Plain text on purpose: a bold run in the question would wrap the paragraph
// in one more element and move the card away from `pollCard`'s ancestor depth.
question:'Which evening do we keep for the kaiseki dinner?',
options:[`${short(-5).split(',')[0]}${ordinal(-5)}, after Fushimi Inari`,`${short(-4).split(',')[0]}${ordinal(-4)}, after Arashiyama`,'Either, decide on the day'],
}
/** The seeded poll nobody has voted on yet. */
constSEEDED_POLL='Ryokan or city hotel in Hakone?'
constCHAT={
mine:'Ryokan check-in is 15:00, so let us aim for the 13:10 train out of Kyoto.',
/** Typed before the picker opens, so the picture shows the emoji being added to it. */
emoji:'Two weeks to go ',
flag:'🎌',
theirs:'Drop bags. I want to be at Senso-ji before the crowds.',
answer:'Agreed, bags at the hotel first, then Asakusa.',
Some files were not shown because too many files have changed in this diff
Show More
Reference in New Issue
Block a user
Blocking a user prevents them from interacting with repositories, such as opening or commenting on pull requests or issues. Learn more about blocking a user.
Maurice
