The mandatory profile picker checked its re-entry flag before several awaits, so an initialization notification could stack two requireSelection routes; the flag is now claimed before the first await and the late-profiles edge only fires on a real transition. Returning from playback also reran the full metadata loader, replacing the whole detail screen with a spinner and resetting the selected season; playback returns now use the non-loading watch-state refresh.
With the rail collapsed and Libraries expanded, DOWN from the header landed on a mounted but focus-excluded library node and was swallowed, trapping traversal; the focus order now matches the render exclusion and traversal scans past non-focusable candidates.
An action bar rebuilding on count change disposed every node, silently dropping focus and never notifying the TV host; bindings are now reused by stable identity (external focus node, else debugLabel) and the transition is reported. Positional reuse remains the only option for an unlabeled action, but it applies only while the action-list shape is provably unchanged - catalog detail enrichment inserting Request before a focused Trailer used to hand Trailer's focused binding, focus ring and next Select included, to Request. Those actions also carry stable debugLabels so their focus survives enrichment.
The text field consumed back on KeyDown outside the shared handler, letting one press act twice. It now marks the back coordinator like handleBackKeyAction does and keeps the down-only shape solely on TV, where the closing IME session swallows the matching KeyUp. Because onBack can move focus (empty search field -> sidebar), the orphaned KeyUp would otherwise drive app-level back on the new focus chain, so the field arms BackKeyUpSuppressor; the suppressor gained a hardware-level KeyUp observer that clears the armed state in a microtask after the press ends, removing the long-documented pinning hazard where an unconsumed KeyUp left it swallowing the next press.
Queued tracker writes carried no account identity, so rows created under account A replayed against account B after a reconnect; explicit disconnect now purges that service's queue. An in-flight markWatched/reconcile that failed after the purge was re-enqueued anyway (the write scope only tracked profile generation) and replayed into whichever account connected next: Tracker.accountBinding (client identity, rebound synchronously on every disconnect) is captured before each write and re-checked before queueing a retry, and because the check sits in the same synchronous segment as the enqueue, a row that passes it is claimed ahead of the purge in the queue mutex and is still removed by it.
The memoized anime-list snapshot survived writes, so sequential rewatch updates computed from pre-write state; successful writes now evict the entry.
A device-code dialog dismissed by system back never cancelled the poll (blocking new attempts until the deadline), and a user-cancelled sign-in showed the connection-failed snackbar; both now route through one cancel path.
A Jellyfin live retry that failed before open stop-reported the still-current session (recover returns the receiver), killing /Sessions/Playing for the rest of the session; the discard now skips a recovered session identical to the current one. An in-place reload failing before the open boundary left the progress tracker disposed and nulled, so the resumed stream never reported again and the exit flush was skipped; the re-wire now also runs on the rollback path against the restored metadata. Dismissing the Still Watching prompt via back or next also never re-armed the sleep timer; it now counts as the same acknowledgement as Continue.
A cleared skip-marker regex saved as an empty pattern, and RegExp('') matches every chapter title: all synthetic markers became intros or credits, and an end-of-file credits marker could auto-advance the episode. Blank input is now rejected at both validation and save, and the chapter fallback treats already-persisted blank patterns as absent.
Recording indicators never appeared in the EPG and the program sheet kept
offering "Record" for already-subscribed programmes, so rules were created
repeatedly (nine duplicate subscriptions in the linked report). Two defects,
both confirmed against the official web client bundle:
- The sheet's scheduled-state check called
/media/providers/<identifier>/media/subscriptions/mapping/<ratingKey>,
which 404s: PMS mounts that route under the numeric MediaProvider id from
/media/providers. The check failed instantly and unconditionally on every
platform, so the button always read "Record".
- The guide derived its dot solely from cross-matching grab metadata, and
the grab refresh issued right after scheduling raced the server
materializing the grab, wiping the optimistic key before one frame showed.
The grid response already tags subscribed airings with subscriptionID /
grandparentSubscriptionID - the signal the official client renders from.
LiveTvProgram now carries those attributes; the guide checks them first
(grab matching stays as the secondary signal) and keeps local
schedule/cancel actions authoritative until the next grid load. The sheet
resolves Record vs Manage from the tagged rule key without a network call,
falling back to the fixed numeric-id mapping route for untagged airings.
Grab parsing also accepts the nested airing under "Video", which PMS uses
for every non-scheduled grab status in JSON.
close#2009
Opening Plezy on a TV box resumes the resident process rather than
cold-starting it, so the Libraries grid kept showing the in-memory
content from the previous session -- hours stale -- until the user
switched libraries and back. Nothing on the resume path refreshed tab
content.
MainScreen now latches genuine backgrounding and, on a resume more than
five minutes later, refreshes LibrariesProvider and sweeps the content
tabs through each screen's in-place Refreshable.refresh() -- skipped
while offline, before startup priming, during playback, or with no
connected servers. LibrariesScreen.refresh() now refetches the selected
library's loaded tabs (the toolbar refresh action) instead of
re-selecting the saved library, which never reloaded them because tab
widgets only reload when the library's globalKey changes.
close#2043
Home hub rows showed ~3 large posters per row while the hub's "see all"
page packed 5 small ones on the same 360dp phone. Hub detail was the
only surface on the padding-aware target-count formula, which resolves
ceil(lerp(5, 2, f)) columns regardless of screen width (and 9+ columns
on TV). Drop the flag so hub detail shares the fixed-extent formula
with every other grid, and delete the now-unused
getMaxCrossAxisExtentWithPadding and its usePaddingAware plumbing.
close#2039
_hasRenderedFirstFrame had five writer files with hand-rolled
snapshot/restore pairs in the reload and channel-zap rollbacks, and the
paired _hasFirstFrame ValueNotifier was reset independently along the
open path. Move both flags into FirstFrameGate with explicit verbs —
markReady, resetUiForOpen, resetRenderedForAttempt, reset,
forceUiReadyOnFailure, snapshot/restore — so the UI-vs-reporting
asymmetry and the transactional rollbacks are enforced by one type.
The uiReady notifier keeps its identity across attempts; the video
surface, controls, and buffering overlay hold it by reference.
Unit-tests the gate's verb semantics.
Google Play rejected 2.16.0 (versionCode 142) for an Android Automotive
navigation dead end: on a car, "Sign in with Plex" resolves to the
in-app QR wait, which offered only Retry — and the AAOS system bar has
no back button, so a touch user could not leave the screen without
killing the app. Add a Cancel action to the QR and browser-polling
states that aborts the attempt and returns to the initial sign-in
actions on every platform.
Verified on an API 34 Automotive emulator: Cancel returns to the
landing with the Jellyfin/Emby paths reachable.
Sleeping the device mid-video with "Ask for profile on app open" enabled
left the app wedged on an unresponsive Choose Profile screen (Nvidia
Shield). The picker is pushed on the root navigator, but the player's
focus self-heal only consulted its own nested profile-session route, so
it yanked D-pad focus back behind the picker whenever it lost it.
Route currency now walks every enclosing navigator (isRouteChainCurrent)
before the player reclaims focus, primes loading-phase navigation focus,
or claims the surface on window focus. The resume-time prompt is also
skipped entirely while a video player is active: waking mid-stream
resumes the stream instead of stacking the picker over the session.
close#2034
Plex Discover's hub, search, and related endpoints ignore includeGuids,
so items opened from those rows carried no tmdb id and the Request
action never appeared — only watchlist items (whose endpoint does
return Guids) could be requested. The gate was frozen in initState from
the row item; it now reads the detail-enriched item, whose metadata
fetch does bring the tmdb id.
close#1959
Searching a name shared by a movie, a show, and audio drowns the wanted
item in episode and track rows. The search screen now offers kind chips
(All, Movies, TV Shows, ...) above the results; chips derive from the
pre-rank candidate pool, and selecting one re-ranks that kind's
candidates with the full display budget, so a kind crowded out of the
trimmed All view still surfaces everything the servers returned for it.
Chip switches never refetch, and the filter falls back to All when a
refined query no longer returns its kind.
close#2001
Connected MDBList accounts were omitted from Explore, so their watchlists could not be selected or displayed. Register the profile-scoped MDBList client as a catalog source with watchlist mutations and search.
close#1888
There was no way to tell from the detail screen that a movie or episode
has multiple versions (theatrical/extended cuts, 1080p/4K encodes); the
only entry point was the hidden Play Version item in the overflow menu.
The Play button now becomes a Material 3 Expressive split button when
the item carries more than one version and its server is reachable: the
main segment keeps plain Play (which already resumes the remembered
version), and a narrower chevron segment runs the existing Play Version
flow (version picker, quality picker when the backend can transcode,
preference save). The flow itself is extracted from the context menu
into a shared promptAndPlayVersion helper, and FocusableActionBar gains
per-action spacingBefore so the joined pair can sit tighter than the
rest of the row. Transcode-only quality picking stays in the overflow
menu so the chevron keeps signaling a real version choice.
close#1881
Changing playback speed, shader preset, aspect ratio, or a sync offset in
the player always rewrote the global default, so a show watched at 1.5x
forced 1.5x onto movies, an Anime4K pick followed the user into
live-action libraries, and a file-specific subtitle offset leaked onto the
next title.
Each of those settings now has a scope under Settings > Video Playback >
Remember Player Changes: don't save, everywhere (default, pre-existing
behavior), per library, or per show/movie. Player writes route through the
configured scope; playback start and in-place item changes resolve the
value for the current item, falling back to the global default. Items
without the needed identity (live TV placeholders, Jellyfin/Emby downloads
without a stamped library) keep the global behavior. Resolution consults
only the configured scope, so entries saved under a previous scope become
inert instead of shadowing the new one.
close#1984close#1322close#1616
On Fire TV, invoking Alexa and dismissing its overlay popped the profile
picker over a still-playing session: MainScreen treated every resumed
lifecycle event as an app open and re-applied "ask for profile on open".
The Alexa overlay only produces inactive -> resumed, which the handler
could not tell apart from a real return to the foreground.
Track the deepest lifecycle state since the last resume in a small gate
(ProfileSelectionResumeGate): only a genuine backgrounding (hidden,
paused, or detached) arms the prompt for the next resume. This also
keeps iOS working, where returning from the background arrives as
hidden -> inactive -> resumed.
close#1990
Square-card (music) grids had zero delegate spacing - the only
separation was each card's internal 3px padding, and square shapes are
excluded from the full-card spacing path, so playlist and album tiles
sat nearly edge to edge. Square grids now get an 8px cross/main gutter
(automotive keeps its larger spacing); poster, list, and full-card
layouts are unchanged.
Artist pages showed one flat album grid, unlike Plex's own clients.
Plex listing rows never carry Format/Subformat tags (even with
resolveTags=1), so fetchArtistDiscography follows the unchanged album
listing with one batched /library/metadata/{ids} request - whose rows
do include the tags - and classifies each album individually
(EP/Single, then live, then compilation, case-insensitively). A failed
tag fetch degrades to the flat list, and single-album artists skip the
lookup entirely. Jellyfin/Emby carry no album taxonomy on the wire and
return a single albums group.
The artist screen renders one titled section per non-empty group
(single-group artists keep the flat grid), with the grid focus index
space offset per section so D-pad traversal crosses section
boundaries.
The guide replaced itself with a bare spinner on every window change,
unmounting the day-picker anchor and the guide focus node - after one
day or time change the pickers silently no-oped and the remote went
dead. Picking a day also applied nothing by itself: the window only
moved after the second (time-slot) menu, so backing out of it read as
'the day picker does not work'.
The grid now stays mounted during in-session reloads behind a light
loading overlay, a day pick re-anchors the window immediately (the slot
menu is an optional refinement), D-pad SELECT arms the key-up
suppressor before opening the menu, and the day chip uses the existing
'Tomorrow' translation.
Plex Live TV lost DVB subtitles when 2.10 moved the live path to HLS
(#1983): the transcode starts with subtitles=none, which drops a
tuner's bitmap subtitle streams from the output entirely - unlike
in-band CEA captions, they are separate elementary streams.
The tune response now surfaces the part's embedded bitmap subtitle
streams as selectable tracks in the player's subtitle menu. Picking one
points the part's server-side selection at that stream and rebuilds the
live transcode with subtitles=burn, preserving the time-shift position.
Off stays the default so tuner captions are never auto-burned (#1590),
and channels without bitmap streams keep the native track list. The
selection survives stream recovery by re-mapping onto the re-tuned
session's streams and resets on channel zap. Jellyfin live sessions
record subtitle selection as intentionally unsupported: their one
negotiated URL has no rebuild to deliver it through.
MemoryLogOutput extended LogOutput but was never wired as a Logger output (storage happens in the printer); DeletionNotifier/WatchStateNotifier's forServer/forItem filtered streams and WatchStateEvent.mediaType had no production consumers (tests now filter .stream directly); context.hiddenLibraries/profileSettings and toPlexUrl had zero call sites; MonoTokens.splashFactory was never read; OptimizedMediaImage's enableTranscoding/cacheKey chains, FocusableMediaCard.width/height/forceGridMode, TvBrowseRail's constant-zero gap functions, and media_image_helper's scaleFactor were never varied by any caller. The sha1 LRU stays (rebuild-hot artwork URLs); TvRailTrailing.none and the mono-theme copyWith stay live (lerp delegates to copyWith). Also carries the MusicPlayContext.id argument drops in the music screens and remaining sweep test updates.
plex_client.selectStreams' allParts=false branch, download_manager's both-branches-true conditional and unscoped getAllPinnedMetadata tail, playback_source_resolver.preferOffline, and update_service's unreachable catches had no reachable path; PlaybackSession.result/streamHeaders, PlaybackContext.clientScopeId, FileInfoStreams.audioStream, PrefsRepairOutcome.settingsReset/sessionsSalvaged, and JellyfinLiveSessionTracker.playSessionId were write-only; PlexConnection.directUrl, PlexServer.isOnline plus its parse-and-persist presence field, SafStorageService.createDirectory, FullscreenStateManager.stopMonitoring/dispose, the play_queue_launcher re-export shim, track-selection's constant params, and PlexApiCache.unpinForOffline/isPinnedRatingKey (matching the earlier Jellyfin removal) were dead API.
classifyPosition's return enum was discarded by its only production caller (rearm is the side effect that matters); the latch's tolerance/retry knobs were never tuned; _loadAdjacentEpisodes' result was unread by all three callers; countsAsApplied carried a subsumed disjunct and _syncMediaControlsAvailability a stale-generation guard with no suspension point before it; _SettingsView.audioSync/subtitleSync could never be current (both routes early-return into the sync bar); PerformanceStats.dvRpuOutputTooSmall/dvPlaybackReason and ContentStrip.chaptersLoaded were written but never read.
getServerBoundPlexClient/getClientForLibrary existed only for their own unit tests; MetadataEditAdapter.backend and Plex's unreachable prefKey fallback and music-kind branches had no production path (edit is double-gated on supportsKind); MainScreenFocusScope's focus aspect, isSidebarFocused, and focusContent were write-only; resolveActivePlexIdentity.preferredAccount was never passed; ProfileRegistry.get was unused outside tests — its removal exposed a stale test contract, now pinned correctly: list() deliberately never serves plex_home rows.
The screen's player stream subscriptions were enumerated by hand in four places (re-wire, rollback, dispose, plus a parallel nulling list) kept in sync only by prose — _cancelPlayerStreamSubscriptions is now the single cancel-and-null authority, living beside the fields. PlayerChromeFocusTarget was a one-value enum with request/consume plumbing, now a bool with identical post-frame consumption timing. The desktop tap and controls-overlay tap duplicated the click-toggles-playback/double-click-fullscreen contract around shared state, now one helper. SyncOffsetControl shipped two layouts of which production only ever rendered compact — the full layout existed solely for tests, which now cover the shipped layout. TrackControlsState's never-non-null onLoadSeekTimes/onSyncOffsetChanged sheet-compat callbacks are gone.
persistAndBindConnection's addToManager/visibleServerId parameters and post-commit manager phase had no production caller (the Jellyfin flow defers runtime pickup to the profile binder) — the function is void now. A failed Plex account add threw a StateError that PlexPinAuthFlow re-rendered as a raw 'Bad state:' string beside the already-rendered inline error; the flow's normal return already stops the spinner, so the inline error is the single surface.
The pause/resume/ticker state machine (three gates, lifecycle observer, TickerMode subscription, timer sync) was triplicated across the What's On, Recordings, and Guide tabs — now LiveTvRefreshMixin, with the guide's drift catch-up layered on its hooks and the lifecycle test asserting the same contract against the shared implementation. Six hand-rolled multi-server loops share forEachLiveTvServer (the channel/favorites loaders deliberately iterate per DVR entry and say so). Recordings' Completer-based load drain became the same whenComplete coalescing shape the screen's channel loader already uses.
TvosMenuPolicyPublisher batched an idempotent fire-and-forget publish behind a transaction-depth counter that could never exceed one across its three non-nesting call sites — the call sites now publish directly and the pure predicate keeps its test coverage. AppleTvRemoteTouchService exposed 14 injectable knobs of which only six are exercised anywhere (verified per knob); the other eight are private constants or internal construction now.
One server with overlapping libraries (movies in both an HD and a 4K
library, shows duplicated per language) returns search results that
cannot be told apart, forcing users to open each copy to find the
right one (#1970).
Each search row now carries a source line under the summary: backend
icon, server name, and library name, shown whenever the owning server
has more than one library. Plex rows name their section inline, with
sectionKey-only rows back-filled from the loaded libraries; Jellyfin
and Emby send no library field on search hits, so search always runs
the per-library scoped fan-out that previously only served
hidden-library exclusion and stamps every hit with its library. The
Jellyfin detail lookup adds a best-effort /Items/{id}/Ancestors stamp
so full metadata stays attributed, row refreshes merge instead of
dropping the stamp, and the mapper no longer misreads SeriesStudio or
ParentId as library identity.
Episode rows trade one summary line for the source line so the
wide-thumb row keeps its height, and the badge paints slightly below
the line-box center to sit on the text's optical middle.
close#1970
On TV, pressing UP in the library Collections or Playlists grid snapped
the list back to the top and dropped focus on the tab chips, making long
lists impossible to navigate. Default directional focus traversal scrolls
the found card into view via Scrollable.ensureVisible, whose
outer-scrollable pass routes through the NestedScrollView coordinator and
resets the inner grid position to zero on every UP press.
Give the shared paginated card grid explicit per-card d-pad navigation,
matching the browse tab: managed per-index focus nodes, row/column moves
that request focus directly, first-row UP to the tab bar, and
first-column LEFT to the sidebar. Focus changes now scroll only through
FocusableWrapper's delta-based auto-scroll.
close#1977
BaseLibraryTabState documented an abstract loadData() that every tab "must implement", but three of the four tabs override loadItems() entirely and carried never-invoked empty loadData stubs; only the recommended tab exercised the contract, so the class docs described an extension surface that did not exist.
loadItems() is now the one overridable hook, and the shared load transaction (generation tracking, localized error mapping, post-frame onDataLoaded) lives once in the protected runLoadTransaction helper that the recommended tab and the focus tests route through. No tab's load behavior changes.
The library quick picker and the libraries dropdown each hand-rolled their own BackendBadge + server-name label rendering with divergent show-when policies: the picker only labeled libraries whose titles collided, so a uniquely-titled library on a multi-server list showed its server in the dropdown but not in the picker.
Both surfaces now share library_server_label.dart — one label widget, one grouping loop over groupLibrariesByFirstAppearance, and one policy: group headers whenever the visible list spans more than one server, per-row labels only in ungrouped lists. The picker's duplicate-title heuristic is gone; its test now pins the unified policy.
profiles_view.dart rebuilt the exact merged-profile view ActiveProfileProvider already computes — the same four source streams, the same merge/avatar derivation, plus a hand-rolled combineLatest4 — and the profile switch screen was its only consumer while already reading the provider for activeId.
The screen now renders from the provider (new connectionsByProfile/connectionsById/plexHomeByConnectionId getters) and gates loading on provider initialization; visibleProfileConnections moved to profile_merge.dart for profile_detail_screen; profiles_view.dart is deleted. The switch-screen tests initialize the provider up front like boot does, using a timer-less PlexHomeService subclass so start()'s periodic refresh timer cannot trip the widget-test pending-timer invariant; the deleted pipeline's merge assertions were ported to profile_merge_test and active_profile_provider_test.
Three new Linux startup cases, one isolate each (a second VideoPlayerScreen
in one isolate never reaches initialize):
- A custom mpv config naming vo/gpu-context/gpu-api cannot detach the
embedded renderer: the writes are withheld by name while ordinary entries
(sub-scale) still land.
- Unparseable stored subtitle colours (named, 3-digit hex) reach the wire
canonicalized to the defaults, proving the OPT_COLOR sanitization.
- A refused sub-color write no longer aborts initialization: the write is
attempted, contained, and playback continues past the styling block.
On TV, a movie carrying a full set of attributed scores pushed the resolution and audio labels past the right edge of the detail metadata line, which silently clipped them.
The line now sheds its least useful parts when it overflows -- surplus rating badges first, then the whole ratings slot, then quality labels -- instead of hard-clipping the tail. It also leads with the year instead of the redundant "Movie"/"TV Show" label, matching the desktop hero chip order, and the Discover spotlight line gets the same fitting. Branded badge icons are pinned to their SVG viewBox aspect so the fit is measured exactly.
close#1893
Physical-keyboard Escape at root Home now exits window fullscreen on
Windows and Linux the way it already did on macOS, and never arms the
press-back-again quit — so Escape aimed at fullscreen can't close the
app. Remotes, gamepad B, and system back keep the double-press exit.
close#1748
An EOF-driven advance does one cold metadata fetch with a single endpoint
failover and no transient retry. When connectivity to the server drops for
the ~20s that fetch needs (issue log: both plex.direct endpoints connect
timed out, then the running stream's own TLS socket died), the reload
rolled back to the finished episode's last frame: black screen, progress
bar parked at the end, no way forward but the transport controls - while
pressing Next by hand seconds later succeeded. The per-item metadata cache
row could not absorb the blip either, because adjacency comes from queue
containers, so the next episode's row is cold at the exact moment the
transition needs it.
Three changes:
- A failed in-place reload now records its classified failure reason, and
an advance that ran with the completion latch set re-presents the Play
Next prompt when that reason is serverUnavailable. With auto-play
enabled the countdown re-fires the advance up to two times before the
prompt goes manual-only; Watch Together sessions and mid-episode Next
presses (whose rolled-back stream is still valid) keep the existing
handling. playNextRetryPresentation owns the decision and is unit-tested.
- Committing adjacency now best-effort prefetches the next episode's full
metadata row through fetchItem, which writes the exact row playback
initialization falls back to on both backends (Plex: same cache key and
full playback query shape; Jellyfin: the /Users/{uid}/Items/{id} row the
playback bundle reads). A warm row turns a blip at the transition into a
normal start.
- JellyfinClient.fetchItem's documented "pure transport error -> cached
row" fallback was dead code: the HTTP layer wraps transport errors into
MediaServerHttpException, which the first catch rethrew unconditionally.
Status-less, non-cancelled failures now take the fallback; answered
requests (401/403/5xx) and cancellations surface unchanged.
Verified with new contract tests (Plex: cold row fails transiently ->
fetchItem primes -> the same failing fetch serves playback from cache;
Jellyfin: primed row survives a transport failure into fetchPlaybackBundle)
plus the full test/screens/video_player and test/services suites and
analyzer parity.
close#1867
Video on Linux went through a Flutter texture: 8-bit sRGB, which cannot carry
HDR at all, and which forced a whole-window Flutter recomposite for every video
frame. This moves it onto a wl_subsurface stacked below the Flutter surface, with
mpv rendering into an EGL window surface on it through the libmpv render API. The
subsurface is desynchronized, so video and UI now present independently.
With the plane in place HDR follows: the surface is described to the compositor
through wp_color_manager_v1 as the source's own curve and gamut - PQ or HLG,
BT.2020 - carrying whatever HDR10 static metadata the stream actually declares.
The description and the buffer it describes land on the same commit, staged and
validated before mpv is switched, so a PQ frame is never presented labelled sRGB.
A five-second watchdog bounds the one wait a compositor could otherwise leave
hanging. A session that cannot host the plane - X11, or a compositor without
wl_subcompositor - fails initialize with VIDEO_PLANE_UNSUPPORTED naming the
reason: the texture path is gone, and refusing by name beats degrading to
something the user cannot see. An SDR output, a missing capability or an 8-bit
config keep the plane and simply leave it undescribed.
The output's colour state is trusted only when it has been earned. Every landed
property step records itself as it lands; a reset or sequence that cannot
finish downgrades its result to unknown and marks the applied-output cache
untrusted until a clean apply earns it back. A plane whose output state cannot
be named is quarantined - hidden, its description withdrawn - and the
quarantine is recorded state: an unrelated visibility change cannot put a
mislabelled plane back on screen, and only a commit that resolves to a nameable
outcome lifts it. A rect collapsing to zero detaches the buffer exactly as
hiding does, a refused setVideoRect drops the Dart-side sent-rect cache so the
next layout pass retries for free, and a refused tone-mapping pick tells the
user instead of dying in a log.
NVIDIA's Wayland EGL (through at least 610.xx) offers no 10-bit unorm window
configs, so the plane takes half-float as the tier between 10-bit unorm and
8-bit, declares the whole surface opaque so the compositor never reads the
alpha those configs carry, and states GL_RGBA16F rather than a 10-bit lie.
Whether the output is in HDR is read from luminance headroom above its own
reference white rather than from the preferred transfer function, which current
KWin no longer answers PQ for; the margin is half a stop, because KWin reports
an undimmed maximum over a software-dimmed SDR white. Validated on an RTX 4090
(driver 610.57.04) under KWin 6.7.4 with locked-exposure photographs.
Who tone-maps is a user choice. The default is the compositor: photographed on a
400-nit HDR output against a PQ chart it keeps 400 -> 1000 nits monotonic and
separated where the player leg flattens them, because the player path drives
mpv's legacy vo_gpu, whose own standalone output scores the same. The gap is the
renderer, not the wiring.
The decision itself - what the source carries, what the output supports, what to
tell mpv and what to tell the compositor - lives in hdr_metadata.h, free of
Wayland and GTK so its luminance validation can be tested without a display
server. Sending an incoherent luminance set is a protocol error that disconnects
the client, so the rules are worth a unit test.
The deb, rpm and pacman packages now declare wayland-client, wayland-egl and EGL:
the plane links them directly and bundle-libs.sh deliberately never bundles them,
since they are coupled to the running compositor and GPU driver.
lib/dev/harness_main.dart is a second entrypoint for measuring this on hardware -
it drives one clip with scripted mpv properties and reports the colour state mpv
actually settled on. Nothing imports it, so it is tree-shaken out of the app.
Verified on a Steam Deck against an external 400-nit HDR display: the compositor
reports PQ / BT.2020, the connector carries HDR_OUTPUT_METADATA, and against mpv
vo=gpu-next on the same frame the shipped build sits 4.90 counts away overall -
closer to the reference HDR player than to its own SDR fallback.
The unused-code gate flags debugSetCrashReporterReady (referenced nowhere)
and the formFactorScaleForTesting/rootShellForTesting wrappers (referenced
only from tests, which the lib-scoped check cannot see). Delete the dead
setter and let the car-scale test build the real rootShell and
FormFactorScale directly.
Vertical D-pad/arrow navigation stepped through the flat channel list,
which is number-sorted across servers. With overlapping channel numbers
from multiple DVRs, focus interleaved source groups and could dead-end
before the last displayed row. Derive the up/down order from the same
grouped rows the guide renders.
close#1843
With "Default to Favorite Channels" enabled and no favorites stored —
or only favorites left over from a since-rebuilt lineup — the favorites
filter reduced the guide to zero channels and GuideTab rendered just
the timeline bar: no rows, no message, no sign a filter was active.
Users read it as Live TV being broken; the Aug 8 report in #887 shows
44 channels and 447 grid programs loading in the log while the
screenshot shows a blank guide and 0 favorite channels.
When the filter removes every loaded channel, the guide tab now shows
an empty state naming the cause with a "Show All Channels" action that
clears the filter. The action stays D-pad reachable: the tab-bar focus
handoff falls through to the action's focus node while the empty state
replaces GuideTab, and activating it hands focus back to the restored
guide content. Favorites that match no loaded channel get the same
treatment as an empty favorites list.
Verified: flutter test test/screens/livetv/, analyzer parity,
clean_translations --check --strict, and slang codegen freshness.
Refs #887.
Android TV returns to the platform keyboard for single-line fields; the
Flutter overlay stays for multiline and explicit call sites. The bugs
that forced the overlay (#1051, #1079) were an engine show/bind ordering
race, now repaired at the app level:
- MainActivity retries a soft-input show the engine dropped while the
FlutterView was not yet served (flutter/flutter#177360), rebinds the
IME key session once at first show, and consumes leaked D-pad keys
while the keyboard is visible (bounded restartInput budget) so focus
cannot wander behind a stuck keyboard.
- The platform text-input hint is activation-based, so gamepad pause and
the pre-IME D-pad intercept track a live session instead of mere field
focus.
- While a session is live with the keyboard away, Back closes it and is
consumed once, Select re-raises the keyboard, and arrows keep
caret-aware edge-escape navigation instead of dead-ending.
ffmpeg's reconnect loop deliberately retries 503 without bound (#1520), so a
server that keeps refusing the stream at open time left a silent black screen:
ExoPlayer fell back to MPV, MPV reconnected forever, and no error ever reached
the screen. A new open-phase watchdog arms on the first 503 seen before any
frame renders and, after 20s without one, synthesizes a server-http-503 error
that shows an actionable dialog. Mid-stream 503s and live TV keep their
existing ride-out paths.
close#1830
The clock renders through the existing formatClockTime helper driven by
MediaQuery.alwaysUse24HourFormatOf, so it follows the OS 12/24-hour
setting instead of introducing an app preference. It re-arms a one-shot
timer onto each wall-clock minute boundary rather than polling, and
resyncs on resume because a suspended process runs no timers.
The player header is shared by the mobile and desktop/TV controls, so one
insertion point covers every form factor: the player is fullscreen
everywhere, so it never has an OS clock to defer to. Home is the
exception and only gets one on TV, where a leanback app hides the system
clock; a phone status bar and a desktop menu bar already show the time.
Pausing an episode on one device, finishing it on another and pressing
Refresh left the first device showing the old "minutes left". Restarting the
app showed the right value. Two independent defects produce that, and either
alone reproduces the report.
The first is the watch-state overlay. Every local watch event lands in
WatchStateStore as a patch, and WatchStateSnapshot.apply overwrites
viewOffsetMs unconditionally; isNewerThan only ever orders one patch against
another, never against the server row underneath. Nothing expires a patch and
nothing clears the map except a profile switch, so the Mac's own paused
position kept winning over every subsequent fetch until the process died.
A patch exists to bridge the gap between a local action and the next server
read of that item, so it should stop applying once that read happens. The
store now records the watermark at which a successful authoritative response
returned each key, and suppresses an acknowledged session patch at or below
it. Only a watermark is stored, never the observed state: WatchStateSnapshot
cannot hold a container's leaf counts, and keeping max() per key makes the
order two concurrent responses complete irrelevant. Suppression is a
read-time predicate, so nothing mutates during build.
The barrier covers the parentChain too. patchForItem picks the newest of the
item's own entry and its ancestors', so retiring only the item's entry would
let an older season mark win and render watched/0 -- worse than either the
stale value or the fresh one. An authoritative read of a child already
reflects any container mark that preceded it, so the child's observation
judges its ancestors as well; a newer container action still wins.
Provenance decides what may be suppressed at all. WatchStateEvent now carries
serverAcknowledged, defaulting to false so an unclassified emit site degrades
to today's behaviour rather than silently becoming retireable. An offline
write is owed to the server and a read must never retire it, so it stays
until a WatchPatchPromotionNotifier promotion says the queue replayed it. That
channel is deliberately not a WatchStateEvent: OfflineWatchSyncService reacts
to watched/unwatched by purging queued progress, so replaying one there would
delete a newer rewatch. Promotion matches an exact WatchPatchId -- session
minted for live crossings, derived from the persisted (profile, row, revision)
for queued ones so it still joins after a restart.
Report acceptance is not delivery: PlaybackReportSession resolves true for a
same-state startup heartbeat it drops, so acknowledgement now keys on
onDelivered. A MediaBrowser Started saves play count and last-played date but
not the position, so it cannot acknowledge an offset. No report-derived
watched crossing is acknowledged on any backend -- Jellyfin hard-codes its
threshold and Plex never loads the server pref that would tell it the real
one -- so only an awaited explicit markWatched settles one.
The second defect is that a failed Refresh reported success. Plex _fetchHubs
and the Jellyfin hub legs both degrade a failure to an empty list, and the
library prefetch discarded its failures, so a server whose every hub request
failed was recorded as succeeded; DiscoverProvider then kept the previous rows,
set loaded and surfaced nothing. Worse, the background Continue Watching
refresh wiped the row outright on zero success.
Hub legs now report what they degraded through a HubFetchDiagnostics sink,
which keeps partial rows alongside the failure and leaves every existing
caller untouched. Failures ride through the aggregation results, a leg that
could not run because discovery failed contributes that failure rather than a
successful no-op, and loaded-server ids became succeeded - failed - cancelled
so one bad leg no longer caches a server as covered and blocks its retry. The
toolbar awaits a DiscoverRefreshOutcome and shows the existing unableToLoad
snackbar on failure while the retained rows stay on screen. Rollback after a
mid-pass exception is version-guarded, refilters against the current hidden
libraries and no longer publishes a system shelf the pass never committed.
Observations are staged with the pass and flushed only once the same disposal,
generation and exception checks that authorise committing those rows have
passed, so a discarded or rolled-back response can never suppress a patch.
Also fixes a live data-loss race the promotion work would have built on:
upsertProgressAction stamped a millisecond timestamp and updated the row in
place, so a rewatch queued during an in-flight replay was deleted by id.
Revisions are now strictly monotonic per row, replay deletes and retry updates
compare against them, and the upsert resets the retry fields because a new
revision is a new logical action.
close#1829
edde746