The lake cost only added half again within 12 m of the bank, so the cheapest line round a bend ran one grid cell off the water's edge (7 to 19 m on the big lakes), and pulling it straight could draw it over land. A clearance ramp now makes the water nearer than 80 m (40 m along the shore) steadily dearer, which also centres the route in anything narrower than twice that, the shore style rides a band 100 to 350 m out instead of the bank itself, and a shortcut may not come nearer the shore than the path it replaces. Measured on nine lakes: no sample on land (was up to two), direct keeps 51 to 78 m at its closest (was one cell), along the shore the median is 60 to 100 m (was 30 to 40) and under 1 % of it within 25 m (was up to 22 %); lengths change by under 1 %.
Paddle tours are planned on the water by TREK's own router: a water network
per paddle region, built from the same Geofabrik extract as the obstacles and
searched in a worker thread of the server (no sidecar). Rivers, canals and
lakes (direct or along the shore), portages round weirs, locks by policy,
whitewater by boat, gaps marked. Found portages and locks become points of
their own; an unplannable leg says why and offers a way out.
- Shaping points: a click on the line sets a point that steers the route
without being a stop (hikes and paddle tours); recompute a single leg.
- Lake style direct/shore, suggested from the wind forecast.
- Regions load by themselves where a paddler plans (admin setting: automatic,
on request, off; size cap and a daily count), with requests for the admins,
a notification, and the load followed on the leg.
- Portage from the last point, portages follow footways via the walking
router (or stay straight), "put in here" on slipways and put-ins.
- MCP plan_paddle_route with the REST route's rules.
lint:warnings failed with the total and the files holding the most of the rule, which are old ones the baseline allows; the file that tipped the count was never named, so finding it meant linting by hand. When a rule grows, the ratchet now lints the files that differ from the base twice, as they are and as they were there, and lists every one holding more of the rule, line and message. The base is TREK_LINT_BASE (the CI fetches the pushed-over commit on a push and the target branch on a pull request) or origin/dev; without it the report stays as it was.
The list of who a linked trip lets in took MJourneySettingsSheet past the 300-line limit for a component (max-lines-per-function), which turned the client lint on dev red. It is MTripAccessList beside it now; the markup is unchanged.
Under the contributors, the journey settings on desktop and phone get a section of their own, Through trips, with everyone a trip's journal link lets in, their role and the trip they come through. It is read only there, since the trip's settings manage it, and it leaves out whoever is a contributor already. getJourneyFull answers it as trip_access.
The settings button was a bare round icon next to the trip's back pill; it is the same pill now, with the gear and the word Settings. The settings dialog opens centred again: its fixed height already keeps it from jumping between areas, so it does not need to be pinned to the top.
The dialog's nav buttons are named by their area alone (aria-label), so a hint line naming invites no longer matches the invite button. The sharing cases open the Sharing area first, the wiring and More sheet tests drive the settings button and row, the hook test reads the toasts as keys (renderHook loads no locale), and the lazy list and store registry know the journal panel and store.
linkRecipientUserIds and listTripIdsForJourney had no caller and kept src/db/repositories under its 99 % statement floor. TJLREPO-001..007 cover the link rows, the members list, the reach across trips ending with the membership, the hide-skeletons choice of a link member, the role precedence and leaveJournalLink.
The departure port needed a global binding that every suite building TripMembersModule without AppModule lacked. The roster's repository now deletes the leaver's row on the trip's journal link itself (leaveJournalLink), inside the transaction that removes the member: no edge into journey, no DI, no test stub, and the eight suites the stub touched are back as they were.
The schema inventory tests count the two new tables and their NOT NULL columns, the legacy step map lists the migration as post-legacy, and the TJL scenes give their users numbered names so cases sharing one database do not collide.
lint:i18n-keys only reaches a key through a literal t() or a *Key property: the three save toasts were passed to run() as bare keys and the tab label came out of TRIP_TAB_LABEL_KEYS, so four new en keys counted as unused. They are translated where they are named now.
TripJournalLinksRepository spelled 'viewer' | 'editor' by hand while the generated entity exports the same roles, which knip then counted as an unused export. The repository now derives its type and values from the entity.
Declared in trip-members, the port gave journey an edge into the roster, which reaches journey through other domains: the cycle again, the other way round. In the shared kernel neither domain owns it.
TripMembersService called TripJournalLinkService directly, which made trip-members and journey depend on each other (lint:boundaries). The roster now declares TRIP_MEMBER_DEPARTURE and calls memberLeft in the transaction that removes the member; the journey domain binds it globally (TripMemberDepartureModule), the way JourneyAccessModule binds the gateway's JOURNEY_ACCESS. Suites that build TripMembersModule alone import MemberDepartureStubModule.
budget.service.db and trips.service tests are held at their size baselines; the stub now comes from journey-share-repos, which both already import, so neither suite grows.
index.css is held at its line baseline. The two variables the embedded journal overrides (--nav-h on the desktop, --m-safe-top on the phone) now sit as arbitrary properties on the wrapper that needs them.
useTripJournalLink had grown past the 120-line limit for a function. The member list (candidates per journal, roles, who is new) is now useJournalMemberChoice. What loaded is kept with the journal it belongs to, so a new pick shows nothing until its own people are in, without resetting state by hand. The integration test adds its trip members with Promise.all.
The Share button becomes the trip's Settings, a round button beside the back
pill. The dialog is rebuilt around three areas (People, Sharing, Journal) with
a side navigation, a fixed height and 960 px width on wide screens.
Journal links one of your journals from the Journey addon to the trip. Every
trip member is ticked as a reader by default, each can be raised to editor or
left out, and the people a save would let in for the first time are named
before it saves. The link can be changed or removed. With a link the trip gets
a Journal tab on desktop and phone that shows the Journey addon's own view,
live through its sync and trip:journal:changed. Nothing shows with the addon
off.
members.shareTrip and nav.share are gone with the button that used them.
The link table keyed by the trip id made both entities a one-to-one, and the
test typecheck failed with TS2589 in the itinerary factory. It now holds its
own id with trip_id UNIQUE, as trip_invite_tokens does, and its members hang
off link_id.
The factory error was not this table's alone: any further relation on trips,
journeys or users tipped the same check. insertRow and createRow now take a
caller's overrides as their own argument instead of a spread checked against
RequiredEntityData, so a new relation no longer breaks the itinerary and tour
factories.
The suites that hand-build TripMembersService get a journalLinksStub.
A trip can now show one journal from the Journey addon as its own: any
journal its owner picks, and one journal on as many trips as they like.
The trip members chosen for the link get into the journal as viewers or
editors, see it in their own journal list and follow it live, for as long
as the link holds them and they stay on the trip.
That access is never copied into journey_contributors. It is read where
journal access is read: the contributor repository answers with the
stronger of a direct role and a link role, the journal list and the live
update recipients take the link members in, and the photo access check
goes the same way. Every one of those reads starts from linkReach, which
only counts members still on the linking trip, so unlinking, deselecting
someone and their leaving the trip all end exactly what the link gave.
A member who leaves is also taken off the link, so coming back later does
not quietly reopen the journal.
Linking or changing the members takes the journal's owner and the new
journal_link trip permission (trip owner by default); unlinking takes
either. /api/trips/:tripId/journal answers the link as each member may see
it, /candidates says who a journal would be opened to for the first time,
and four MCP tools do the same over the same service. A trip:journal:changed
event tells the trip to look again.
A flag set on dragstart could still be unrendered when a quick drag reached
the next card, and the drop was lost. The category now travels in the drag
under its own type, read from the first dragover on and at the drop.
Categories showed in the order they were created, which a list kept per day
cannot undo once the days swap. A grip in each category's head band moves
the whole category onto another's place on desktop. A category has no
position of its own, it shows where its first item is, so the move reorders
the items as one block through the item reorder the list already has: no
new column, and the order reaches everyone the way a dragged item does. The
other view's categories keep their order.
The time placement of a journey entry moves whole into journey-entry-order,
and a moved entry only gets a sort_order when it actually moves, so a patch
never carries an undefined one. The Synology login comment shrinks to a line.
On the phone the Transports tab never rendered a booking's notes, so what
someone typed into a flight was only to be found by opening its sheet,
while the booking cards next to it showed theirs. Both cards now take the
notes from one part, sized on the caption tier so the text size setting
reaches them.
The public OSRM hosts answer 429 above about one request a second. The trip
overview and the road trip waited that out and asked again; the day route
took the first refusal as final and drew straight lines between the stops
until the next edit, which is what a day with a change of mode or a quick
switch between days ran into. Both the day route and the overview now go
through one retry helper: a rate limit is waited out, as long as the host
asks when it says, and a refusal the host meant still ends at once.
The login went to auth.cgi with account and password JSON-quoted, the way
the FileStation calls on entry.cgi take their values. entry.cgi reads JSON
where it can; auth.cgi takes a value as it comes, so DSM looked for a user
named "trek" with the quotes and refused every account. Both go plain now.
The test verdict sat truncated next to the dialog buttons ("The credentials
were re..."), with nothing to read the rest by. It wraps to two lines and
shows the store's own reason on hover; the phone sheet prints it below.
Closes#2666
The index matches a term anywhere inside a place's category text, so the
bare names behind the pills caught whatever was spelled like them: "park"
brought car parks and an amusement park, "garden" furniture stores and beer
gardens, "bar" barbers, "pub" publishers and public squares, "market"
marketing agencies. In Vienna four of sixty "Nature & parks" pins were
parks. Every pill now names its Overture categories by full path, a place
counts only when its path is that path or lies under it, and a page of
near misses only falls through to Overpass. Terms Overture no longer has
(nightclub, nature_preserve, shopping_center ...) got their current names.
Closes#2663
A new entry always went to the end of its day, so writing up 1 am after
2 am left the night out of order, and changing an entry's time never moved
it. A timed entry now lands before the first entry of its day that starts
later, on create and when its time or day changes; untimed entries still
go last, and the hand-made order of the others stays as it was. A save
that leaves time and day untouched keeps an entry where it was dragged.
Closes#2679
The public page showed places.duration_minutes under every stop, and that is
60 on every place nobody set a road-trip stay for, so a shared plan read "1 h"
everywhere. The stay is now the stop's start to end time, per visit, past
midnight included, and a stop without an end shows none.
Closes#2676
The OpenAI-compatible client always asked for 4096 tokens, and a document
whose bookings needed more came back as "no reservations" with nothing that
pointed at the cap. LLM_MAX_OUTPUT_TOKENS now sets the cap for every
provider; unset keeps 4096 and 8192 so a small local model's context still
fits. A finish_reason of "length" is reported like Anthropic's max_tokens
stop, naming the limit and the variable that raises it.
Closes#2677
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.
Maurice