* perf(epg): render only the timeline programmes near the visible range (J3)
Selecting a live channel rendered every programme block of its schedule
(about 240 for the Xtream mock) while the stream was starting: about 6,000
of J3's 6,199 renderer.domMutationsToPlaying. The ribbon now renders the
blocks, ticks and day dividers within half a viewport of the visible range;
the track keeps the full schedule's width, so positions, the scrollbar and
scroll-to-now are unchanged.
A resize reported before the scroll-to-now must not re-centre the window
(it once jumped to the schedule's start and back); resizes only widen it.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
* test(e2e): sweep the EPG ribbon to see every programme
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
* fix(epg): estimate the ribbon window once per channel or mount
The live estimate followed the centred day and the 30 s now tick while the
ribbon was not yet scrolled: a small scroll across midnight re-centred the
window on the next day's noon and left the visible range empty, and the
host width was re-read (a forced layout) on every tick.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
* test(e2e): keyboard Tab reaches programmes beyond the rendered EPG range
Greptile flagged that windowing the ribbon could strand keyboard users at
the last rendered block. Focusing a block scrolls it into view, which
re-windows before the next key press; the new test walks ten unrendered
programmes past the range with Tab and with Shift+Tab, and fails if focus
ever leaves the ribbon.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
* fix(epg): re-centre the ribbon window in the same pass as a zoom
Review follow-up (Codex): a zoom (button, Ctrl/⌘ wheel, coalesced wheel
burst) or a group expansion changed the scale before the anchored
scrollLeft landed on a later frame, so the window was the previous
centre at the new scale until the next scroll event re-measured it. The
ribbon could flash empty or show the wrong section. The zoom controller
now hands the window the minute its anchored scroll will centre, and a
group expansion the group's centre, together with the new scale
(TimelineWindowController.centreOnMinute).
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
* fix(epg): window the scroll position a zoom can actually reach
Review follow-up (Greptile, Codex): a zoom-out anchored right of centre
at the ribbon's start computed a negative scroll position. The window
centred on it, the browser kept scrollLeft at 0, and with the position
unchanged no scroll event re-windowed, so the visible right-hand part
stayed empty. The centre now uses the position clamped at 0, and once
the frame applies scrollLeft the window re-centres on what the browser
kept, which also covers the clamp at the end of the track.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
* fix(epg): close the programme popover when the window drops its block
Review follow-up (Codex): the ribbon window can remove a focused narrow
block on scroll, and a removed node fires no focusout, so the fixed
tooltip kept showing a programme no longer on screen. The popover is now
a linkedSignal over the rendered items that keeps its state only while
its block (by key) is still rendered.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
* fix(epg): re-measure the ribbon window when the axis origin moves
Review follow-up (Codex): a time offset that carries the first programme
across midnight moves the axis origin and every track position while
scrollLeft and the ribbon stay put, so no scroll event fires and the
window kept its epoch-time centre at a different pixel position. The
window identity now includes the axis start; when only that changes,
the viewport is measured from the ribbon instead of re-estimated.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
---------
Co-authored-by: 4gray <fourgray@proton.me>
Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
* perf(web): keep the NgRx store devtools out of production bundles
app.config.ts imported @ngrx/store-devtools statically and gated it on
AppConfig.production at runtime, so the optimizer kept the module in
main.js for the production, PWA and performance builds. The providers now
come from environments/store-devtools.providers.ts, an empty list in every
build; only the development and electron-e2e configurations swap in the
devtools through fileReplacements. A build-config test keeps it that way.
renderer.initialBytes: 1,608,610 -> 1,596,045 bytes (-12,565) on a local
production build; the baseline is lowered to the measured value.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
* perf(web): cite #1810 as the initial-bytes baseline evidence
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
---------
Co-authored-by: 4gray <fourgray@proton.me>
Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
* fix(workspace): lead header Back to a parent route when the page opened the session
Settings, Discover, actor and in-portal search registered a header Back
that only ran Location.back(). As the first entry of the session (deep
link, reload, restored view) that did nothing in Electron and left the
app in a browser.
WorkspaceBackNavigationService.back(resolveParent) keeps Location.back()
while the previous entry is an in-app one, and while that is unknown
because the Navigation API is missing. Otherwise it opens the page's
parent with replaceUrl, so history Back cannot return to the page:
- Settings: the first workspace view (resolveDashboardPath()).
- Discover: the catalog section it lists (vod for movies, series for TV).
- Actor and search: the portal root, which redirects to its default
section within the same navigation.
The web E2E opens these pages in a fresh tab: a page.goto in the same
tab leaves the previous document behind, often at the parent's URL, so
history Back passed without the fix. Electron covers settings after a
window reload.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
* fix(workspace): lead first-entry Back to the parent without the Navigation API
Review follow-ups (Greptile):
- Without the Navigation API (older Safari and Firefox) back() always
called Location.back(), so a page that opened the session still left
the app. The service now tracks the router's in-app history depth there
(trackRouterHistoryDepth): first navigation 0, push +1, replacement
keeps it, a traversal restores the depth recorded for its entry. Depth
0 opens the parent; an unknown depth (an entry from before a reload)
keeps Location.back().
- Stalker's Discover (movie/tv section), actor and search pages now have
tests that they hand the service the parent under the portal :id.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
* fix(workspace): adopt the router navigation the Back depth tracker missed
Review follow-up (Codex, Greptile): the lazy workspace shell creates the
Back service after the first NavigationStart, so the tracker saw only its
NavigationEnd, left the depth unknown and counted the next push as the
first entry. It now adopts the router's current or last successful
navigation when it starts: a first navigation is depth 0, a later one
leaves the depth unknown (browser history Back), and a late start of the
adopted navigation is not counted again.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
---------
Co-authored-by: 4gray <fourgray@proton.me>
Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
* fix(portals): preserve playlist ownership during detail handoffs
* fix(portals): reload Stalker categories only for a held destination
Review follow-ups (Greptile, Codex): resetCategories() reloaded the
category resource, and the route session calls it on a portal switch
before the destination is resolved and on teardown, so it asked the
portal being left, and a failed destination lookup could let that answer
repopulate the sidebar. resetCategories() now only clears; the session
calls the new reloadCategories() after installing the destination, and
only when a handoff had already put that playlist in the store (the
owner, and so the resource params, did not change).
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
---------
Co-authored-by: 4gray <fourgray@proton.me>
Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
* perf(playlist): make the playlist import and shared UI components OnPush
Plan item C6 step 3 for libs/playlist (import/feature and shared/ui): the
twelve Eager components switch to OnPush. Two of them rendered plain fields
written after an await, outside any template event, which only an Eager
check on the next zone tick picked up:
- playlist-item's portal status dot (PWA, after the async portal check)
now reads a signal;
- playlist-info's playlist is backed by a signal behind its existing
getter/setter name, so the EPG source list follows removals and file
picks that land after awaited cleanup and dialogs.
A regression test for each fails on OnPush with the plain field and passes
with the signal. The Stalker import's post-await patchValue needs no change
(see the zoneless checklist). The m3u feature-player components stay Eager
for the playback PR.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
* test(playlist): check the OnPush dialogs without forcing a render
Review follow-ups (Greptile, Codex):
- The portal-status and EPG-row tests forced detectChanges() after their
await, so they passed with plain fields. They now let the fixture render
on its own; with portalStatus back on a plain field the status test fails.
- New: the playlist info dialog enables Save and shows the path after a
native EPG file pick, without a forced render. pristine and valid read
the form's state signals, so the OnPush dialog follows on its own.
- New render spec for the add-playlist dialog with the real URL form: Add
enables after typing and after a patch from outside the child (as an
auto-detect prefill does).
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
---------
Co-authored-by: 4gray <fourgray@proton.me>
Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
* perf(portal): make the portal Eager components OnPush
Plan item C6 step 3 for libs/portal: the nine Eager components in
portal/shared/ui, portal/stalker/feature and portal/xtream/feature switch
to OnPush. Their templates read signals, signal inputs, async pipes and
template-event state; the plain fields they write outside events
(playback request ids, save throttles) are not rendered.
The already-OnPush live channel lists filled their favorites Maps in a
subscription and the Xtream list dropped programme previews after the EPG
mapping dialog, all without marking the view. They now call markForCheck
like the neighbouring handlers do, so a late favorites answer shows its
hearts without waiting for an unrelated check. A regression test for the
Xtream list fails without the call.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
* test(portal): let the favorites handler's markForCheck render the heart
Review follow-up (Greptile): the test forced detectChanges() after the
favorites arrived, so it passed without the handler's markForCheck(). It
now lets the fixture render on its own; removing the call fails it.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
---------
Co-authored-by: 4gray <fourgray@proton.me>
Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
* fix(workspace): keep the macOS header clear of the lights when zoomed out
App zoom scales CSS pixels but not the native traffic lights. At zoom
-3/-4 the header column starts near 29 window pixels, so Back and the
playlist switcher slid under the lights, and the 27px header band let
the lights overlap the context panel below.
A shell-level TrafficLightsClearanceDirective now publishes the lights'
clearance in CSS pixels (84 x 48 window pixels, from the page zoom
factor) on macOS. The header band grows to the vertical clearance (the
rail starts its first link at the same band, replacing the rail's own
zoom listener), and the header's leading padding grows to the
horizontal clearance less the rail column. Both equal the default
layout at 100 %; Windows/Linux and the phone layout are unchanged. The
native position is shared with the main process as
MACOS_TRAFFIC_LIGHTS_POSITION.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
* test(workspace): pass the header clearance poll labels as options
Equivalent to the string form, which Playwright 1.62 also accepts, but
explicit in every version (Greptile review).
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
* test(workspace): check the header switcher before history exists
Since the header's history fallback, a list reached by navigation
leads with Back. The macOS check now takes the switcher on the first
page, which has nothing to go back to, and Back on a detail page.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
---------
Co-authored-by: 4gray <fourgray@proton.me>
Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
* test(performance): add J4 search journey
Measures typing a six-character query into the header search box on
/workspace/search until the global search results settle, on a profile
with the M3U fixture and the mock's existing 12,000-item `large` Xtream
catalog. Counters: bridge calls and SQL statements per search (with a
per-keystroke breakdown), serial IPC depth, DOM mutations, change-detection
ticks, layout shift and long tasks; wall-clock last keystroke to settled
and first keystroke to first result.
Runs in the existing journeys target and the warn-only CI job, whose
summary now prints the per-keystroke table. Moves J3's picsum artwork
blocker into a shared helper.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
* test(performance): J4 settles only on the final term's query
The probe could settle on cards of an earlier term that stay visible while
the final term debounces. The journey now stamps every dbGlobalSearch trace
event in the main process, and the record requires the last query between
the sentinels to be for the final term and to have completed before the
end sentinel. Iterations with a keydown gap over 250 ms (below the 350 ms
debounce) are rejected; gaps and the final query are kept as evidence.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
* test(performance): anchor J4's SQL count to the journey sentinels
renderer.sqlStatementsPerSearch was the difference of test-side samples
taken before the first key and after the end sentinel had been read, so
database work in either gap could be counted. The main process now reads
main.sqlStatements when the start and end sentinels arrive, and the counter
is their difference; sqlStatementsAfterSettled starts at the end sentinel.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
---------
Co-authored-by: 4gray <fourgray@proton.me>
Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
* perf(web): add an opt-in zoneless change-detection build flag
Plan item C6 step 4. app.config.ts takes its change-detection providers
from environments/change-detection.providers.ts, which keeps
provideZoneChangeDetection({ eventCoalescing: true }) for every existing
build. The new electron-performance-zoneless and electron-e2e-zoneless
web configurations are their base configuration plus one fileReplacements
swap to provideZonelessChangeDetection(), so the journeys and the Electron
E2E suite can run zoneless while nothing ships it. zone.js stays in the
polyfills until the flip.
A build-config test pins each *-zoneless configuration to its base plus
the swap and refuses the swap anywhere else.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
* fix(epg): schedule the guide's post-render scroll without zone.js
The programme guide jumps to now once the virtual list first renders rows,
and focuses cells after keyboard scrolls, from afterNextRender hooks
registered in CDK and RxJS callbacks. zone.js followed those callbacks
with a tick; under zoneless change detection a render hook schedules no
render, so the guide opened at midnight (epg-guide.e2e.ts on the zoneless
build). The guide now marks itself when it registers one, which is
harmless with zone.js.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
* docs(performance): record the zoneless flag measurements and E2E run
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
* docs(performance): say the zoneless flag ran with the three implemented journeys
Review follow-up (Greptile): J4 search is still planned, so the flag was
validated with J1-J3 and the Electron E2E suite, not all four journeys.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
---------
Co-authored-by: 4gray <fourgray@proton.me>
Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
Plan item C6 step 3 for libs/workspace: the seven Eager components in
workspace/shell/feature, including the workspace shell root the idle audit
found re-rendering on every idle tick. Their templates read signals,
signal inputs, computed values and template-event state only; the one
plain field written outside the template (categoryLockTarget) is not
rendered. They switch to OnPush without other changes.
Co-authored-by: 4gray <fourgray@proton.me>
Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
* perf(web): make the app root and settings components OnPush
Plan item C6 step 3 for apps/web: the fifteen Eager components switch to
OnPush, among them the app root and the update notification panel that
the idle audit found re-rendering on every idle tick. Their template
state is signals from the settings facades, signal inputs and the shared
reactive settings form.
The checklist flagged the backup import, which patches the form from a
detached file input with no template event. A new spec patches only a
value, which changes no form status, and confirms the OnPush general
section still shows the new theme; it guards that path for the zoneless
flag.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
* fix(settings): re-render OnPush sections when the form changes outside them
Review follow-ups (Greptile, Codex):
- The settings sections read form values in their templates (selected
theme and cover size, epgField.value, form().value.player), and the
parent changes the form outside their events: Discard and backup import
patch it, the store hydrates it, the EPG file picker sets a control
after an await. Under OnPush the section kept the old selection or EPG
status. Each section now marks itself on its form's events
(markSectionForCheckOnFormEvents).
- The value-only patch test no longer forces detectChanges(); with the
fixture rendering on its own it fails without the marking, and so does
a new test for a control set outside the EPG section.
- The zoneless guard counts only changeDetection metadata outside
comments, so a comment naming the strategy is not an Eager component.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
* test(settings): guard the unsaved-changes bar after a save off the sections
Review follow-up (Codex): Save marks the form pristine after an async
store write, also on Backup, Reset or search, where no form section is
rendered. The OnPush page re-renders anyway because pristine and valid
read the form's state signals; the new test checks that on the Backup
page without forcing a render (it waits for the scheduled one).
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
---------
Co-authored-by: 4gray <fourgray@proton.me>
Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
* chore(deps): bump mpegts.js from 1.8.1 to 1.8.2
Bumps [mpegts.js](https://github.com/xqq/mpegts.js) from 1.8.1 to 1.8.2.
- [Release notes](https://github.com/xqq/mpegts.js/releases)
- [Commits](https://github.com/xqq/mpegts.js/compare/v1.8.1...v1.8.2)
---
updated-dependencies:
- dependency-name: mpegts.js
dependency-version: 1.8.2
dependency-type: direct:production
update-type: version-update:semver-patch
...
Signed-off-by: dependabot[bot] <support@github.com>
* test(playback): accept the mpegts.js 1.8.2 error contract
1.8.2 keeps the public ErrorTypes and ErrorDetails exports unchanged, so
the version lock moves to 1.8.2 with the same accepted contract.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
---------
Signed-off-by: dependabot[bot] <support@github.com>
Co-authored-by: dependabot[bot] <49699333+dependabot[bot]@users.noreply.github.com>
Co-authored-by: 4gray <fourgray@proton.me>
Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
* perf(playback): make the web players and M3U player OnPush
Plan item C6 step 3 for playback: the eight Eager components in
libs/ui/playback (video.js, ArtPlayer, HTML5/hls/mpegts, audio player,
web player view, VOD details, sidebar, external-player dialog) and the M3U
video player and VOD detail switch to OnPush. The player libraries' events
already reach the UI through the signal-backed controls adapter or
outputs, and the players' DOM belongs to the libraries.
The M3U video player rendered three plain fields written outside template
events: the channel-number overlay, cleared by a 2 s debounce timer, and
the player choice, written from an IndexedDB read and a settings effect.
They are signals now, and a test checks that the overlay leaves the DOM
when the timer fires.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
* test(playback): let the overlay's signal write schedule its own render
Review follow-up (Greptile): the test forced a render with
fixture.detectChanges() after the debounce timer, so it would pass even
if the signal write stopped scheduling an OnPush render. It now runs the
fixture with autoDetectChanges and only advances the fake timers; with
plain fields under OnPush the overlay never renders and the test fails.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
---------
Co-authored-by: 4gray <fourgray@proton.me>
Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
* test(perf): record which nodes move in a journey's layout shift
J2's renderer.layoutShiftScore went from 0.222 to 0.233 with #1814, and
its evidence only held the recent-input / without-recent-input split, so
the moved element could not be named from a summary. The probe now keeps
the first 20 counted shifts (value, recent input, time since the journey
start and the moved nodes, as J1's late shifts do), and J2 writes them to
evidence.layoutShift.shifts.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
* test(perf): record the horizontal move of layout-shift sources
J2's 0.233 shift on the runner moves main.workspace-content, the header
search field and the header actions with deltaY and deltaHeight 0, so
the move is horizontal and the probe could not show it. Sources now also
carry deltaX and deltaWidth.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
* test(perf): count every layout shift next to the capped list
Review follow-up (Greptile): the score counts every shift but the
evidence lists only the first 20, so a reader could not tell that later
shifts were omitted. The probe now keeps shiftCount, and J2 writes it
beside evidence.layoutShift.shifts.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
---------
Co-authored-by: 4gray <fourgray@proton.me>
Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
* ci(perf): enforce the journey counters stable on master
Promote the J1, J2 and J3 counters that were identical in all 55 measured
iterations of the 11 master runs from 2026-10-03 to 2026-10-04 to
journey-baselines.json, and check them in the Performance journeys job
(still warn-only), in one step together with #1828's two validated J1
entries. None of the new ones has Principle 3 evidence, so each carries a
"guard only, not validated" note that the checker prints with a failure.
A performance-tools test keeps the job's --only list equal to the journey
entries. Number formatting uses three decimals, the precision of the
layout-shift scores.
J2 renderer.layoutShiftScore is 0.233, not the window's 0.222: every
master run from #1814 (page Back buttons in the header) on reads 0.233.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
* ci(perf): check the journey counters whenever a summary was written
Review follow-up (Greptile): the check ran only after the composite
action succeeded, so a failed job-summary report after a written
summary.json skipped every baseline. It now runs unless the job was
cancelled, as long as the action produced a summary path.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
---------
Co-authored-by: 4gray <fourgray@proton.me>
Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
* fix(electron): show the main window once its document has loaded; enforce J1 IPC and mutation counters
Re-lands #1788, which merged into #1782's branch after #1782 had already
reached master, so none of it is on master.
The hidden main window was shown on ready-to-show only. On Linux under
X11, when the startup scripts run before the window's first frame, the
next frame comes about a second later: nothing is on screen and the
splash's requestAnimationFrame waits, so J1's first card came ~940 ms
after load instead of ~480 ms in most runner launches (18 bridge calls /
1,031-1,033 DOM mutations instead of 15 / 558).
The window is now shown at ready-to-show or the main frame's
did-finish-load, whichever comes first, with the splash colour as its
background so showing before the first paint does not flash. The journey
gate keeps the app's did-finish-load listeners away from its about:blank
detour, as it already does for ready-to-show.
Three dispatched runs on this branch (37192092882, 37192097790,
37192103151) read 15 calls and 558 mutations in all 18 iterations,
stable: true. Both become baselines (slack 0), and the Performance
journeys job checks them with check-journey-ratchet.mjs --only.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
* chore(perf): record the evidence PR of the J1 runtime baselines
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
---------
Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
Co-authored-by: 4gray <fourgray@proton.me>
* test(performance): skip every test-only file suffix in the zoneless guard
#1813 added serial-details.test-stubs.ts, whose stub components set
ChangeDetectionStrategy.Eager. The zoneless checklist guard listed only
some test-only suffixes, counted the stub file as production code and has
failed the performance-harness job on master since. It now skips every
`.spec` / `.test` file with or without a suffix, test-setup.ts and
test-stubs/ directories.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
* test(performance): accept multi-segment test-only suffixes in the zoneless guard
Review follow-up (Greptile): `(-\w+)?` allowed one suffix segment, so a
file such as `rail.test-data-stubs.ts` would be scanned as production.
The suffix now repeats, and a classifier test pins which names are
skipped and which ship.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
---------
Co-authored-by: 4gray <fourgray@proton.me>
Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
* docs(portals): document forced MPV/VLC launch and pending-start contracts
The rules #1792 settled for forced external launches and playback-start
bookkeeping lived only in code comments and specs. Record them as
contracts in the owning documents:
- embedded-inline-playback.md: the shared detail-host rules (one external
player per title, unconfirmed teardown cancels, ownership rechecked
after every await, owner-scoped pending state).
- xtream-portal-compatibility.md: the series launch chain, page-token
duplicate guard and queued choice.
- vod-multi-source.md: the movie menu launch and reset follow the copy
the primary button acts on; pending resets are a list of targets.
- stalker-portal.md: the per-series launch queue, the batch-held choice
and the movie launch/reset pending start.
Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
* docs(portals): correct launch-contract claims against the code
- A launching session is published before the launch IPC resolves; what
it lacks until then is an exact closer.
- The series watched/reset batch and the Xtream movie launch gate are
page-wide, not owner-scoped.
- Only the Stalker movie hosts retire a pending start, and that does not
clear the repeat guard of a launch still in flight.
- Name only the specs that exercise the Xtream series rules.
Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
* docs(portals): tighten closer timing and page-wide gate wording
A launching session may already have its closer before the launch IPC
resolves, the Xtream movie launch gate does not cover the inline
player's diagnostic fallback, and the hero-state spec covers only the
external-player and reset rows.
Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
* docs(portals): fix stale session-timing comment and search guard wording
- serial-details-external-launch.ts: Electron publishes a `launching`
session as soon as the launch IPC arrives, not only after the launch
settles. Comment only; no behaviour change.
- stalker-portal.md: the search host's selection check does not include
the content type; a switch to a series supersedes the launch through
the playback owner key instead.
Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
---------
Co-authored-by: 4gray <fourgray@proton.me>
Co-authored-by: Claude Fable 5.1 <noreply@anthropic.com>
* refactor(xtream): split the series details page into focused services
serial-details.component.ts sat at the 400-line max-lines limit and its
playback service and spec were close behind. Move cohesive concerns out
without changing behavior:
- route params, the provider-only flag and the (re)load of the addressed
series go to SerialDetailsRouteService; the component still registers
the effect, so effect order is unchanged
- season descriptions, posters and the TMDB season enrichment go to
SerialDetailsSeasonsService
- actor, Similar and Discover navigation go to injectXtreamDetailNavigation,
shared with the movie page
- the watched toggles and the episode playback payload leave
SerialDetailsPlaybackService for SerialDetailsWatchToggles and
buildSerialEpisodePlayback
- the spec's TestBed moves to a harness and the watched-toggle cases to
serial-details.season-watch.spec.ts
Counted lines: component 400 -> 318, playback service 388 -> 342,
component spec 1196 -> 674.
Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
* refactor(xtream): move VOD route selection and position state out of the page
vod-details-route.component.ts grew from 618 to 741 counted lines and
vod-details-playback.service.ts reached 395. Extract without changing
behavior:
- the route-derived read model (selected movie, category, catalog item,
fallback view, multi-source identity, session and content keys) goes to
VodDetailsSelectionService; the component keeps the same member names
- trailer state and the Similar-rail click move into the hero presenter,
the cancel-download prompt and the progress-ring geometry into the
downloads service, navigation into injectXtreamDetailNavigation
- stored positions (last seen, route row, guarded load) become
VodDetailsPositionState and the external-launch bookkeeping becomes
VodExternalLaunchClaim
- drop the unused MatTooltip import (NG8113) and eight unused imports
Counted lines: route component 741 -> 492, playback service 395 -> 344.
Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
* refactor(stalker): extract series-view positions and watched toggles
stalker-series-view.component.ts grew from 1601 to 1813 counted lines and
its spec reached 1198 of 1200. Move code out verbatim:
- saved positions, their reconcile and the persist/clear writes go to the
component-provided StalkerSeriesPositionsService; the ordering of reads
and writes goes to StalkerSeriesPositionQueue
- episode, season and series watched toggles go to
StalkerSeriesWatchToggleService, the batch core to
runStalkerWatchToggleBatch
- the lazy VOD season loads go to StalkerVodSeasonEpisodeLoader
Every effect stays registered in the component constructor in its original
order; template bindings and public member names are unchanged.
The spec setup moves to a shared harness and the spillover-prefetch and
TMDB season cases to their own specs; the 184 cases of the directory are
unchanged.
Counted lines: component 1813 -> 1136, component spec 1198 -> 670.
Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
* refactor(stalker): extract catalog-detail position and search paging
- stalker-catalog-detail.component.ts (397): the stored VOD position and
its runtime updates become StalkerCatalogVodPosition, the Play/Resume
start becomes startStalkerCatalogVodPlayback
- stalker-search.component.ts (776, baselined): the paging resource, the
accumulated results and the parental-lock bookkeeping become
StalkerSearchPagingController, with pure helpers in
stalker-search-results.util.ts. The spec reaches the moved members
through component.paging; its cases and assertions are unchanged.
Counted lines: catalog detail 397 -> 317, search 776 -> 482.
Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
* refactor(ui): give season-container and vod-details real line headroom
- season-container.component.ts (396): season auto-selection and the
seasonSelected emission move to createSeasonAutoSelectState, called at
the same place in the constructor so effect order is unchanged; the
episode subline becomes buildEpisodeSubline
- vod-details.component.ts (393): the cross-portal Similar loader, the
provider-only download state and the actor/Similar route helpers move to
sibling modules
Inputs, outputs, selectors and public members are unchanged.
Counted lines: season container 396 -> 341, vod details 393 -> 337.
Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
* test(web-e2e): move Stalker portal helpers into fixtures
stalker.e2e.ts was at 1196 of 1200 counted lines. Move the mock endpoints,
scenario MACs and page helpers to stalker-portal.fixture.ts and the
embedded-series steps three tests repeated to
stalker-embedded-series.fixture.ts. Every test stays in stalker.e2e.ts in
the same order, with the same serial mode and OWNED_MACS reset.
Counted lines: 1196 -> 981.
Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
* docs: follow the detail-page extractions
Name the Stalker watched-toggle and position services that now own the
batch and reconcile code, point AUTH_REJECTED_MAC and the scenario MACs at
stalker-portal.fixture.ts, and drop two stale statements about components
sitting at the max-lines cap.
Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
* docs(xtream): note why the seasons service is injected last
Its TMDB enrichment effect keeps the position it had as the first effect
of the component constructor only while it is created after the other
services' effects.
Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
---------
Co-authored-by: 4gray <fourgray@proton.me>
Co-authored-by: Claude Fable 5.1 <noreply@anthropic.com>
UI-24 from the UI consistency audit.
- No-artwork slides paint their gradient in CSS from the slide hue: a light
tint in the light theme, unchanged near-black in the dark one. The dark
gradient under the light page-coloured scrim read as a grey slab.
- The side scrim holds 88% of the page colour up to the slide's right edge
(inset + min(560px, 55%)), so the end of a full slide no longer sits on
about 45%.
- Narrow layout (container <= 720px): a full-bleed 90% scrim behind the text
block, a scrim-coloured text shadow, and an entrance without a fade so
that scrim never flashes the art on a rotation.
- --hero-body is 85% of the heading colour (was 72%).
- Light --app-rating-color #a16207 -> #7a4a00: measured 3.36:1 on the chip
over artwork, now 5.10:1. The details pages share the chip and token.
- Buttons cap at the slide width and end long labels in an ellipsis.
- The page gets one visually hidden h1 ("Dashboard"); slide titles are h2.
- One live region outside the re-created slide announces slide changes;
progress bars are named and VOD ones read "N% watched"; dots are 24px.
dashboard-hero-legibility.e2e.ts replaces every image with a checkerboard
and measures each piece of slide text from the screen in both themes, wide
and narrow, for backdrop, poster, no-artwork and live slides. On master the
worst cases were 2.35:1 (body text) and 2.65:1 (pills).
Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
* docs(performance): inventory the zoneless change-detection migration
Plan item C6 step 2. docs/architecture/zoneless-migration.md lists the 66
production files (67 components) that still set
ChangeDetectionStrategy.Eager, the ten places where a template-read plain
field is written outside an Angular event, the NgZone and
ChangeDetectorRef calls to remove at the flip, and the IPC, player,
observer, timer and dialog paths checked as signal-safe.
On Angular 22 an unset changeDetection already means OnPush, so only the
explicit Eager components re-render on every tick.
zoneless-migration.spec.ts in the performance harness compares the
checklist with the code: a new Eager component, or a converted one left
unticked, fails it.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
* perf(ui): make the libs/ui Eager components OnPush
Plan item C6 step 3 for libs/ui/components and libs/ui/epg: eleven
components (twelve with the EPG trust dialog) set
ChangeDetectionStrategy.Eager and were checked on every tick, among them
the always-mounted EPG progress panel the idle audit found re-rendering
on every idle tick. Their template state is already signals, signal
inputs, immutable dialog data or fields written from template events, so
they switch to OnPush without other changes.
The epg-item-description spec mutated dialog data after creation and
marked only the fixture's host view; it now marks the component's own
view, which OnPush requires. The libs/ui playback and remote-control
components stay Eager for their own PRs.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
---------
Co-authored-by: 4gray <fourgray@proton.me>
Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
* fix(playback): keep the saved Embedded MPV player when the mpv check is inconclusive
On Linux native-view the support check runs `mpv --version` by bare name
and waits for the login shell PATH first. Since #1784 that lookup is
asynchronous with a 10 s budget; when it ran out, the check ran on the
inherited PATH and answered a plain `supported: false`. The settings store
took that as a verdict and persisted the default player over a saved
Embedded MPV selection. The main process probed again once the shell
answered, but nothing restored the setting.
`EmbeddedMpvSupport` now carries `inconclusive`. The native service sets it
on a missing mpv while its probe has only seen the inherited PATH; the IPC
handler declares that state before probing and registers the re-probe
before the check, so a throwing check cannot leave it stuck. Every other
answer stays final.
Consumers no longer settle on an inconclusive answer: the settings store
keeps the saved player, and the command palette and the settings search
probe again on their next use instead of caching it for the session.
Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
* fix(playback): keep asking for Embedded MPV support while the answer is inconclusive
Keeping the saved player on an inconclusive answer left a mounted player
stuck on it: the session controller asked for support once, in its
constructor, and the session effect never starts while unsupported, so the
player did not recover after the login shell answered. The settings page
held its one answer the same way.
`watchEmbeddedMpvSupport()` asks again every 3 s until the answer is final
or the surface is destroyed. The player controller and the settings page
facade load support through it, so playback starts by itself and the
Embedded MPV option appears without reopening the page.
Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
* fix(settings): follow an inconclusive Embedded MPV answer to a final decision
The settings store checked a saved Embedded MPV selection once. After an
inconclusive answer it kept the selection and never looked again, so when
mpv turned out to be really missing the player stayed on Embedded MPV
instead of falling back to the default one.
The store now follows the answer with `watchEmbeddedMpvSupport()` until it
is final and only then decides. It acts on an answer only while Embedded
MPV is still the saved player, so a player picked meanwhile, also while
the first answer was pending, is never overwritten.
The watch backs off from 3 s to 30 s between rechecks, so a login shell
that never answers does not keep the app polling at the first rate, and it
no longer schedules a recheck after its answer handler stopped it.
Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
* fix(settings): keep the settings search following an inconclusive Embedded MPV answer
The settings page asks the search service for Embedded MPV support once,
when its search facade is created. After an inconclusive answer the service
only probed again on its next call, so with the page left open the
Embedded MPV rows stayed unsearchable after the login shell answered,
while the player option on the same page already updated.
The service now follows the answer with `watchEmbeddedMpvSupport()` until
it is final, which updates the open page and the command palette alike. A
call made while the answer is still inconclusive restarts the watch, so it
asks at once as before.
Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
* fix(settings): follow Embedded MPV support for search only while the settings page is open
The settings search service is provided in the root injector, so the
watch it started on its first use had no owner: with a login shell that
never answers it kept asking every 30 s until the app quit, long after
the settings page or the command palette that needed the answer was
closed. A failed recheck also ended the watch as if it were a final
answer, hiding the Embedded MPV rows for the rest of the session.
The service now separates the two uses. `ensureEmbeddedMpvSupportLoaded()`
is a single request again, for the command palette. The settings page
calls `followEmbeddedMpvSupport()` and ends it when the page is destroyed.
Only a final answer is kept: after an inconclusive one or a failed
request the next use asks again, for the palette's player commands too.
Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
---------
Co-authored-by: Claude Fable 5.1 <noreply@anthropic.com>
A mixin included from another Sass module (`@use 'x'; @include x.m`) now
lands where it is included, as a same-file mixin already did: its
top-level weights meet the including rule's family, and its family
becomes that rule's, in the order Sass writes them out. Include sites
resolve through the existing `@use`/`@forward` scope resolution, the
definition Sass resolves is the one that runs, and an include inside a
mixin body resolves where that mixin runs. The header's "Not traced" list
keeps what stays out (positional arguments, content blocks placed by
another module's mixin, a name two `@import`ed files define).
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Load Roboto 600/700 and DM Sans 700 so Cyrillic and Greek headings render
real semibold and bold faces instead of a synthetic bold, keep
<html lang> in step with the UI language, and move every font weight onto
the 400/500/600/700 scale (JetBrains Mono at 500 or lighter; the dashboard
LIVE badge now uses the interface font at 700).
Add the `styles:font-weights:validate` ratchet guard and its CI step. It
reads stylesheets much as Sass and the browser do (cascade, layers,
mixins, content blocks, `@extend`, `@at-root`, `:is()`/`:where()`,
keyframes) and lists what it deliberately does not trace in its header.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Every push to a pull-request branch starts the CI matrix and both review
bots, and runs from several open pull requests queue behind one another.
Move the fix rounds off GitHub: a branch is reviewed with the Codex and
Greptile CLIs until both are clean, then pushed once.
- AGENTS.md: the rule, linked to the procedure
- agent-workflow.md: commands, loop, stop conditions and exemptions
- agent-context-map.md: route the topic to the workflow document
Co-authored-by: Claude Fable 5.1 <noreply@anthropic.com>
* ci(i18n): fail on new English-identical translations
The drift check only warned about locale values identical to English, so
untranslated strings kept landing. It now fails on any such value that
tools/i18n/identical-en-baseline.json does not record for that locale and
key. The baseline captures today's 2,015 entries: legitimately identical
values (brand and technical names, language autonyms, PIN) and the
existing debt. An entry only covers the English text it recorded, so
copying reworded English into a locale fails too.
Baseline entries that are no longer identical are reported, not fatal.
`pnpm run i18n:baseline:update` rewrites the baseline deliberately; CI
runs `pnpm run i18n:validate` (node tests, then the check) and never
rewrites it. `--fail-on-identical` remains as a strict audit.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
* fix(i18n): keep the baseline intact on failed updates and strict audits
`--update-baseline` now writes nothing while any locale is unreadable or
has missing or extra keys, so an incomplete translation cannot reshape
the baseline. `--fail-on-identical` no longer reads the baseline it
ignores, so a damaged file cannot block a strict audit.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
---------
Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
* fix(import): consistent add-source forms with masked passwords and URL errors
- Mask the Xtream password in add and edit (and the Stalker one in edit)
behind a shared PasswordVisibilityToggleDirective: one translated
"Show password" label, state in aria-pressed, type="button".
- Give the Xtream server URL its own mat-error and a neutral hint instead
of the EPG file error; give the M3U URL a mat-error.
- Use "Playlist title" in every add form, "MAC address" casing, a single
ellipsis in "Validating portal…" and one "Add playlist" submit label;
translate the method radiogroup's aria-label.
- Show Stalker refusals inline under the portal URL (role="status", like
the Xtream connection test), translated in the template and cleared by
edits; translate the snackbars for outcomes that close the dialog.
- Translate new strings into all locales; reuse the identical Stalker URL
error translations; fix MAC casing and ellipses; drop unused keys.
- Unit specs per form, edit-dialog spec, new add-source-forms web E2E;
update E2E locators; UI guidelines Forms section; Stalker contract.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
* fix(import): mask the password again when an add form is cleared
Clear erased the password but left the visibility toggle on, so the next
password typed in the Xtream or Stalker form showed in plain text.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
---------
Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
#1773 added `ambiguousTimelineCompletions` and `timeline` to
`JourneyMainIpcCaptureState`, while #1774 merged the J3 playback record
spec with a fixture of the old shape, so the spec no longer type-checks
(TS2739). The harness runs it through tsx without type checking, so CI
stayed green.
The timeline holds one start per counted call, matching `callsByMethod`
and `callsBeforeSentinel`, so the fixture stays a capture a real run
could produce.
Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
* perf(electron): look up the login shell PATH without blocking the main thread
fix-path ran $SHELL -ilc env synchronously right after the first load.
With a typical zsh profile that held the main thread for 1-2 s, while the
database worker's ready message and the renderer's first IPC calls waited,
so the launch journey's first card came that much later. Use shell-path's
async shellPath() with fix-path's fallback, so the resulting PATH is the
same and the main thread stays free.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
* docs(perf): name the PR that made the login shell PATH lookup async
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
* fix(electron): let bare-name player spawns wait for the login shell PATH
With the lookup now asynchronous, an external player launched (or the
Linux embedded MPV support check, which runs and caches a bare
`mpv --version`) within the first seconds could see the inherited PATH.
The OPEN_MPV_PLAYER / OPEN_VLC_PLAYER handlers and every embedded MPV
handler now await waitForLoginShellPath() (settled lookup, at most 10 s,
immediate on Windows), restoring the guarantee the blocking lookup gave.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
* fix(electron): wait for the login shell PATH only for bare-name spawns
External players wait only when they resolve to a bare name (no
configured path, no well-known install found); a path to an executable
starts at once. Embedded MPV waits only on Linux and only for support and
prepare, which run the cached bare-name `mpv --version` check; sessions
and controls never wait.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
* fix(electron): wait for the login shell PATH only before the mpv probe
Embedded MPV support and prepare waited on every Linux call, although
getSupport() returns before the bare-name `mpv --version` probe for the
frame-copy engine, native Wayland, a disabled feature or a cached result.
willProbeLinuxMpvExecutable() now gates the wait on the probe actually
running.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
* fix(electron): stop waiting for a login shell that already timed out once
After the first wait for a hung `$SHELL -ilc env` runs out, later
bare-name player launches and Linux mpv probes proceed at once instead
of each waiting the full limit again.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
* fix(electron): bound login shell PATH waits by the lookup's own budget
Replace the latch on the first expired wait with a deadline set when the
lookup starts (10 s). Every wait ends when the lookup settles or the
deadline passes: a launch retried while the shell is still within its
budget waits for the PATH again, and once the budget is spent no launch
waits, so a hung shell still delays at most the first seconds.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
* fix(electron): re-probe a missing mpv once a late login shell answers
When the PATH lookup runs out of budget, the Linux embedded MPV support
check probes `mpv --version` with the inherited PATH and caches a
missing result for the rest of the session. waitForLoginShellPath() now
reports whether the lookup settled; after a timed-out wait the handler
forgets a cached "missing" once the lookup finishes, so the next support
check probes again with the login shell PATH.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
* fix(electron): share one deadline among waits before the PATH lookup starts
A bare-name launch that waited before the lookup was scheduled started
its own 10 s limit, so while startup was stuck every retry paid the full
delay again. The first early wait now sets the shared deadline; the
lookup still replaces it with its own budget when it starts.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
* fix(electron): re-probe mpv after a late login shell PATH either way
A Linux mpv probe that ran on the inherited PATH can be wrong in both
directions: the login shell PATH may add mpv or drop the directory the
inherited one found it in. forgetLinuxMpvExecutableProbe() now clears a
found result as well as a missing one.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
* fix(electron): skip the login shell PATH wait for Flatpak host launches
In Flatpak, players start through `flatpak-spawn --host`, which resolves
the name with the host's PATH; the sandbox's login shell lookup cannot
change it. The launch handlers now decide from the same launch context
the player uses: no wait in flatpak-host mode, otherwise only for a bare
player name.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
---------
Co-authored-by: 4gray <fourgray@proton.me>
Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
* fix(dashboard): scroll a focused rail card fully into view
Chromium skips its focus scroll when 32px or more of the element already
shows, so Tab onto the last source card of a rail that overflows by less
than a card left it half-hidden under the edge fade. The rail track now
handles focusin and scrolls to the first card-start snap position that
reveals the whole card; a plain "nearest" scroll is not enough because
mandatory snapping can round it back (seen on the live channel rail).
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
* fix(dashboard): keep mouse clicks on partly hidden rail cards
A mouse press focuses the card link on mousedown. Revealing the card at
that moment could slide it from under the pointer when the target snap
position overshoots (the live channel rail moves 316px for a 306px
card), so the click landed elsewhere. The rail now reveals a card only
for keyboard and programmatic focus, using the CDK FocusMonitor origin.
Adds an Electron E2E that checks the final layout after snapping: Tab and
focus() leave the last source card fully visible, and a mouse press keeps
the rail still and still opens the source.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
* fix(dashboard): tell pointer focus apart without touching the DOM
FocusMonitor toggles cdk-*-focused classes on the monitored track, so a
mouse press on a source card mutated the DOM before the click. The J2
"open a source" performance journey rejects iterations with DOM activity
between its settle snapshot and the click. Read the input modality from
the CDK InputModalityDetector in a focusin handler instead: it only
listens, so the rail stays untouched until the click.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
* fix(dashboard): reveal script-focused rail cards after a mouse click
The input modality stays "mouse" after any click, so a later focus() on
a partly hidden card left it clipped. The rail now skips the reveal only
for focus caused by a press inside the track: the focus has to arrive
within 100ms of that pointerdown (650ms for touch, whose focus comes
with the tap's compatibility mouse events, as in the CDK FocusMonitor).
Only event timestamps are compared, so the DOM still stays untouched
before the click.
The E2E now clicks elsewhere before the script focus, and unit tests
cover a tap and focus() after an earlier mouse press.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
* fix(dashboard): keep an over-wide focused rail card in view
In a window narrower than a card (or under zoom), a focused card could
never fit, so its own snap offset fell short of the needed scroll and
the rail jumped to the next card's snap point, moving the focused card
offscreen. Such a card is now aligned at its own start instead.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
* test(dashboard): select rail internals through stable test ids
The dashboard contract makes data-test-id hooks the supported Electron
E2E selector surface. The rail now exposes -viewport, -track and
-card-link hooks next to its existing ones, and the focus E2E selects
those (and the rail heading by role) instead of internal class names.
The dashboard doc lists the new hooks and records the focus-reveal
contract.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
---------
Co-authored-by: 4gray <fourgray@proton.me>
Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
* fix(ui): destructive confirmations, verb labels and provider icons
Confirmations: ConfirmDialogData.confirmLabel is required, so no dialog can
fall back to "Yes"/"No"; the dismiss defaults to "Cancel" and
`tone: 'destructive'` styles the confirm with .app-destructive-button. Every
caller names its action ("Remove playlist", "Clear", "Refresh playlist",
"Cancel download" with a "Close" dismiss). The confirm button has the
confirm-dialog-confirm test id and drops its no-op color="primary".
The no-op `warn` color input becomes .app-destructive-button on the EPG
mapping, playlist item, error view, EPG/reset settings, delete-all and source
cleanup buttons, and on the unsaved-changes dialog's Discard.
Provider icons come from SOURCE_TYPE_ICONS in shared/interfaces (Xtream
cloud, Stalker cast, M3U playlist_play / link / description / subject) in the
add dialog, auto-import, empty state, playlist switcher, playlist rows,
dashboard source rail, command palette, Sources filters and both reset
summaries. Stalker no longer borrows the Dashboard icon, and Xtream no longer
shares a glyph with M3U URL playlists.
The playlist error view removed a playlist through the stale
PlaylistActions.removePlaylist: it dropped the playlist from state before the
delete ran, swallowed failures, skipped the source activity guard and showed
no toast. It now uses PlaylistDeleteActionService like every other removal,
commits only a completed delete, toasts and goes home. The unused action and
its effect are removed.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
* fix(ui): one provider icon per playlist row, imperative Korean remove label
A restored Stalker or Xtream playlist can also carry a URL, and the row's
independent checks then showed the M3U URL icon next to the provider icon.
The row now switches on resolvePlaylistSourceIconKey(), the precedence every
other surface uses, so each source shows exactly one icon.
HOME.PLAYLISTS.REMOVE now names the confirm button and the row's delete
tooltip; in Korean it read "the playlist has been removed". It now says
"remove playlist", like every other locale.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
* fix(ui): keep the auto-refresh badge on playlist rows with one provider icon
Showing one provider icon per row moved the auto-refresh badge into the M3U
branches only, so a restored Stalker playlist with a URL and auto-refresh
lost it although the URL is still re-fetched. The row now renders one icon
container: the provider icon from the shared precedence, then the badge for
any row with a URL or a local M3U, exactly the rows that showed it before.
The Xtream portal-status dot, used without source health, keeps that corner.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
* fix(ui): let the playlist row's cancel action render in the error color
The row's action buttons set `color: inherit`, and the selected row does so
again with more specific selectors. Both beat Material's token-driven icon
color, so the .app-destructive-button cancel action kept the row color
(selection blue on the active row). Pin the cancel button to
--mat-sys-error in both row states.
The large-deletion Electron E2E now checks the cancel color in both themes;
without this rule it reads rgb(47, 123, 255) instead of the error red.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
* test(ui): give the dialog service spec the now-required confirm labels
ConfirmDialogData.confirmLabel became required, and the spec still built
confirmations without one. Jest only transpiles, so the suite stayed green,
but the "Typecheck Jest spec programs" CI step rejected it.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
---------
Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
* test(perf): count change-detection ticks in the electron-performance build
J1 renderer.cdTicksToFirstCard, J2 renderer.cdTicksToFirstPage and the
J1 idle baseline renderer.cdTicksIdle30s. Angular's ɵsetProfiler is only
reachable through the dev-mode window.ng global, so the electron-performance
configuration alone swaps environment.ts for environment.performance.ts,
which re-exports the production AppConfig and wraps ApplicationRef._tick.
Production and PWA sources and output are unchanged.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
* test(perf): refuse a J1 idle window that opened late after the settle point
Addresses review: the idle window opens in the settle timer's callback while
the settle point is that timer's deadline, so a late callback left ticks
uncounted between the two.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
---------
Co-authored-by: 4gray <fourgray@proton.me>
Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
* test(perf): measure the serial IPC depth before the J1 first card
Adds renderer.ipcSerialDepthToFirstCard to the launch journey: the length
of the longest chain of bridge calls in which each call started after the
previous one completed, among calls that completed before the first card.
The main IPC capture now records the ordered start/completion timeline;
the depth, its lower bound, the chain and the timeline are per-iteration
evidence, and the CI job summary prints the chain.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
* test(perf): keep the IPC timeline consistent around the J2 start marker
A call that started before the start marker no longer records its
completion in the timeline, and completions of a method with calls in
flight both inside and outside the timeline are attributed outside and
counted, instead of skipping the first marker-method completion.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
---------
Co-authored-by: 4gray <fourgray@proton.me>
Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
