Two related findings from the 2026-06-02 drift audit (#552):
- **#560 (Android)** PlayerController.setQueue() unconditionally
called controller.play() at the end, with no way for
ResumeController to opt out. Cold-boot resume therefore restored
the persisted queue AND auto-started playback, which surprised
users who had paused mid-track before backgrounding the app.
Add `autoplay: Boolean = true` parameter; ResumeController
passes false. Every existing setQueue call site continues to
autoplay (the default is unchanged).
- **#562 (Android)** ResumeController.restore() ran from
MinstrelApplication.onCreate alongside PlayerController's own
init {} block that asynchronously binds the MediaController to
MinstrelPlayerService. On fast devices with slow IPC the
restore could land before mediaController was non-null;
PlayerController.setQueue early-returns on null mediaController,
so the restored queue was silently dropped — the user would
open the app to an empty player after explicitly using "resume
previous queue". Add `awaitReady()` suspend that completes when
the MediaController binding lands; ResumeController awaits it
before calling setQueue.
The two fixes ship together because the autoplay opt-out only
matters once the await fix guarantees the queue actually reaches
the player.
Two findings from the 2026-06-02 drift audit (Scribe parent #552):
- **#577 (Android)** RequestsViewModel.cancel() called refresh() on
BOTH Synced and Queued outcomes. Synced is fine (re-fetch the
canonical list); Queued is offline by definition — the optimistic
removal at the top of cancel() is already correct, and refresh()
on Queued either (a) gets the not-yet-delivered cancelled row
back from the server and snaps it into the list (confusing), or
(b) fails with a transport error and flips the screen to
UiState.Error so the user thinks the cancel failed even though
it's queued. Gate refresh() on outcome == Synced; the mutation
replayer reconciles when connectivity returns.
- **#570 (Android)** LikesRepository.refreshIds() pulled the
server's likes list and INSERTed it into cached_likes — but
never DELETEd local rows the server no longer surfaces. A
cross-device unlike (user likes on web, then unlikes on web)
left the entry visible on Android's Liked tab indefinitely with
no way to clear short of wiping app data. Add
CachedLikeDao.clearForUser + a @Transaction replaceAllForUser
that atomically wipes-then-inserts the user's set; refreshIds()
uses replaceAllForUser so the local cache is exactly what the
server reports. The LOCAL_USER_ID hardcode is its own drift
(#576) and stays for now — fixing it needs threading
AuthStore.userId through the repo.
Five findings + one cancelled duplicate from the 2026-06-02 drift
audit (Scribe parent task #552):
- **#561 (Android)** PlayerController.playbackErrorEventsChannel was
Channel.CONFLATED. The PlaybackErrorReporter coroutine reads it in
a debounce loop that buffers events to coalesce into "Skipped N
unplayable tracks" — but CONFLATED silently dropped every emission
except the latest each time the reader wasn't actively pulling.
A network blip that failed 5 tracks back-to-back surfaced only the
last failure to the snackbar AND only POSTed one playback_errors
row to the admin inbox. Switch to BUFFERED (default capacity 64,
well above any plausible burst rate). Coalescing path now reaches
N > 1 and the admin inbox sees every failure.
- **#563 (server)** systemPlaylistSources rotation whitelist in
playevents/writer.go had drifted behind the migrations. It listed
only for_you + discover; migrations 0021 + 0028 added 6 more
variants (deep_cuts, rediscover, new_for_you, on_this_day,
first_listens, songs_like_artist) that ship as refreshable system
mixes. Plays from those surfaces never advanced the per-user
rotation, so "unplayed first" ordering staled — the same tracks
kept resurfacing. Add all 6 to the map; comment now points at the
migration's CHECK list as the canonical source so future variants
notice the requirement. #573 was the duplicate auditor hit for
the same drift; cancelled in Scribe.
- **#564 (Android)** Android emitted source = "playlist:<variant>"
for system-mix plays from Home and PlaylistDetail, but the
server's rotation matcher keys on the BARE variant string (web
sends the bare form — PlaylistCard.svelte:83). Misalignment meant
system-mix plays from Android never advanced rotation; switching
from web to Android effectively reset the perceived "unplayed
next" ordering. Fix HomeScreen.kt:291 to send bare variant and
PlaylistDetailScreen.kt's play() to prefer systemVariant over the
playlist:<id> tag when the playlist is a refreshable system mix.
User playlists keep playlist:<id> (intentional — rotation only
applies to system mixes anyway).
- **#568 + #569 (Android)** AuthCookieInterceptor was unconditionally
attaching the Minstrel session cookie to every outgoing request
AND wiping the session on any 401. The shared OkHttpClient is also
used by Coil for external image fetches (artwork.musicbrainz.org,
coverartarchive.org, Lidarr /MediaCover URLs); this leaked the
session cookie to those hosts (privacy posture) AND silently
signed users out of Minstrel if any external image host returned
401. Scope both attach + clear to the placeholder.invalid sentinel
host the same way BaseUrlInterceptor was scoped in aec10ce7. Two
new regression tests cover the external-host pass-through. Existing
tests rewritten to make requests through the placeholder URL so
they exercise the in-scope path explicitly.
All five Scribe tasks updated to in_progress at start, will flip to
done after CI green on this push.
Six findings from the 2026-06-02 multi-system drift audit (Scribe
parent task #552):
- **#553 (web)** Tailwind class fix: web admin playback-errors Delete
confirm button was using `bg-action-danger`, an undefined token —
swap to `bg-action-destructive` to match every other destructive
button. Restored the Oxblood signal that distinguishes Delete from
Cancel.
- **#555 (web)** Type the `source` field on `play_started` in the
EventRequest discriminated union. Server's eventRequest accepts it;
web's TS type was missing the slot, so a "drop extra properties"
refactor could silently strip the source tag and break system-
playlist rotation attribution.
- **#556 + #557 (server)** Coverage rollup whitelist was pinned to
('sidecar','embedded','mbcaa','theaudiodb'); migration 0020 added
'deezer' and 'lastfm' as valid cover_art_source values but those
never got wired in, so albums with art from those providers
silently counted as MISSING in the admin Coverage dashboard. The
rollup test was seeding only the pre-0020 sources, masking the
gap in CI. Extend the query to include deezer + lastfm; seed the
test with one row per valid source (regression-guards future
additions).
- **#558 (web)** Auth gate was blocking /forgot-password and
/reset-password/<token> — both are entered without a session by
definition, so the email-link reset flow was bouncing signed-out
users to /login. Add /forgot-password to the public set and a
/reset-password/ prefix matcher. New tests assert both routes
reach their pages without redirect.
- **#571 (server)** Library scanner was indexing only .mp3/.m4a/.flac
/.ogg while the stream handler (media.go) had been extended to
serve .opus, .aac, and .wav. A user with .opus files in their
library never saw them in artist/album listings because the
scanner skipped indexing — silent data loss. Aligned the scanner
to match the media handler.
Scribe statuses updated to in_progress; flipping to done after the
push since these are mechanical and verified directly against the
cited file:lines.
The APK was shipping with a hardcoded versionName="0.1.0-native"
and versionCode=1 — so the About card never reflected the actual
release, and two same-day re-cuts of the per-day mutable tag
v2026.06.02 looked identical to the update-banner comparator.
Operator wants the iteration restored on the APK side (the docker
tag stays plain per the earlier intentional change).
Scheme:
- Per release: versionName = "${tag}.${commit_count}", e.g.
"2026.06.02.142", where commit_count = `git rev-list --count HEAD`.
Monotonic across the project lifetime, deterministic, no manual
counter to maintain.
- versionCode = commit_count. Monotonic, fits in Int forever (we're
not hitting 2.1B commits).
- Local / debug / dev builds fall back to versionName="dev" /
versionCode=1 so the About card reads honestly.
build.gradle.kts:
- defaultConfig reads MINSTREL_VERSION_NAME / MINSTREL_VERSION_CODE
Gradle properties via project.findProperty with the dev fallbacks.
.gitea/workflows/release.yml:
- android-release: checkout with fetch-depth: 0 (the default shallow
clone would return 1 for `git rev-list --count HEAD`); new
Compute release version step exports name + code as step outputs;
assembleRelease passes them via -P; new job-level outputs propagate
them to the downstream image-release job.
- image-release: Stage bundled APK + version sidecar pulls the
computed version_name from needs.android-release.outputs and
writes it into client/minstrel.apk.version, so the server's
/api/client/version reports the exact string baked into the APK.
Without this the sidecar would say "v2026.06.02" while the
installed APK has "2026.06.02.142" — isVersionNewer would call
the bundled APK older and the update banner would thrash.
The existing isVersionNewer comparator already handles the
4-component shape ("2026.06.02.142" > "2026.06.02.141"), so no
client-side logic changes are needed.
The kdoc on PlaybackErrorReportRequest mentioned the existing
/api/plays/* endpoint. Kotlin's lexer treats /* inside a /** ... */
kdoc as a NESTED comment opener, which then swallows the outer
*/ — so the entire PlaybackErrorReportRequest data class
disappeared from the symbol table and the four call sites in
PlaybackErrorsApi.kt / MutationReplayer.kt / PlaybackErrorRepository.kt
all reported "Unresolved reference".
This is the trap recorded in the project's KSP-could-not-be-
resolved memory; mark it again. Fix is mechanical: rewrite the
prose as `/api/plays/...` so no /* sequence appears inside a
block comment.
Three rule trips from the playback-errors + scrubber commits:
PlayerController.startRadio (4 returns → 2): extract the mid-queue
append branch into appendRadioToQueue(). startRadio just does the
guard checks and dispatches; the helper handles the cursor trim +
addMediaItems. Behavior identical.
PlayerController.onPlaybackStateChanged (4 returns → 2): extract
the duration check + zero-duration error emission + skip logic
into handleZeroDurationIfNeeded(). The listener stays compact (one
return for non-READY, one for repeat-evaluation guard); the helper
owns the failure path.
NowPlayingScreen.ScrubberRow (63 lines → ~50): extract the custom
track Box block into a ScrubTrack(fraction, accent) composable.
The Slider's `track` lambda becomes a one-line call. Pixel output
is identical.
Surfaces client-reported playback failures from /api/admin/playback-errors
in a new admin tab. Tabs: Unresolved (default) / Resolved. Each row
shows track + artist + album, error kind badge, who hit it, when,
optional client-supplied detail, and the absolute file path so the
operator can grep the library mount without leaving the page.
Per-row actions (RowActionsMenu):
- Resolve (primary) — modal with Fixed / Ignored dropdown for the
"no further action taken" cases.
- Copy — JSON payload to clipboard with track_id / file_path / kind
/ detail / reporter / client_id / occurred_at. Matches the
operator's "logs with a copy-out function" ask.
- Delete file (danger, modal-confirm) — uses the existing
/api/admin/quarantine/{track_id}/delete-file endpoint AND
auto-stamps resolution='deleted' so a single click closes both
the file and the inbox row.
Deferred to a follow-up: Hide (the existing quarantine flow is
per-user-flag, not a true library-hide), and Re-request via Lidarr
(needs album MBID join — not in the current ListAdminPlaybackErrors
projection).
Also: admin tab list grows from five to six; AdminTabs.test.ts
updated.
Operator hit a track that loaded with zero duration; player just sat
on it. Two things needed: skip the dead track immediately, and tell
the server so the admin inbox can surface the bad file.
PlayerController:
- Player.Listener.onPlaybackStateChanged(STATE_READY) now checks
duration. If it's <= 0 or C.TIME_UNSET, fires a PlaybackErrorEvent
with kind="zero_duration" and calls seekToNextMediaItem (or stop
if it was the last item). Per-item evaluation guard keeps repeat
STATE_READY events (post-seek, post-resume) from re-firing.
- onPlayerError now also surfaces a PlaybackErrorEvent with
kind="load_failed" + the Media3 exception message as detail.
- playbackErrorEvents flow changes from Flow<String> (title only) to
Flow<PlaybackErrorEvent> (track_id + kind + title + detail) so
downstream consumers can both surface a snackbar AND POST to the
admin inbox without duplicating event emission.
PlaybackErrorRepository (new):
- Wraps POST /api/playback-errors with the offline-first MutationQueue
fallback per the standing rule for server writes.
- Reuses AuthStore.clientId for the client_id field — same UUID-per-
install identifier the play-events reporter sends, so support can
correlate playback errors with surrounding plays.
PlaybackErrorReporter:
- Consumes the new richer event shape. Fires the server report per
event (no debounce — the admin inbox should capture every report,
not a coalesced summary). Continues to debounce the user-facing
snackbar in the 2s window so a burst doesn't spam toasts.
MutationQueue / MutationReplayer:
- Adds PLAYBACK_ERROR_REPORT kind + PlaybackErrorReportPayload +
enqueuePlaybackErrorReport entry point + replayer dispatch case
hitting the new PlaybackErrorsApi.
Web admin inbox + UI is the next commit.
New client-reported playback-error log. Surfaces zero-duration
tracks (and future load_failed / stalled kinds) into an admin
inbox so the operator can hide / delete / re-request the
offending track.
Schema (migration 0032):
- playback_errors table with CHECK constraints on the kind +
resolution enums (per the standing rule that new enum values
need a migration to add)
- Partial index on unresolved rows for fast inbox lookup
- ON DELETE CASCADE from tracks + users so cleanup is automatic
Endpoints:
- POST /api/playback-errors: any signed-in user reports. Body
validates track existence + kind whitelist; client_id required
so support can correlate reports from the same device.
- GET /api/admin/playback-errors?resolved=false&offset=&limit=:
admin list with join to track/album/artist for table render
without per-row round-trips. Pagination capped at 200/page.
- POST /api/admin/playback-errors/{id}/resolve: admin marks
resolved with a resolution enum string.
Auto-resolve on Hide/Delete/Re-request from the inbox row is
driven from the web client (two sequential calls) — keeps the
existing track-action endpoints unchanged.
Operator: the slider height clamp shifted the surrounding layout
above and below — not what they intended to change. Revert the
Modifier.height(20.dp) and remove the SCRUB_SLIDER_HEIGHT_DP
constant; the Slider goes back to its M3-default 48dp interactive
component height so adjacent rows sit where they did before.
The slim 4dp custom track stays — that's what addresses the
"puffy bar" feel — and the thumb still sits on it as a visible
14dp circle, with 17dp empty vertical space above and below.
That's the M3 standard layout the operator wants restored.
The M3 Slider default track is 16dp tall and the Slider itself
expands to the 48dp interactive-component minimum, so the 14dp
thumb we'd already shrunk to a flat circle was still sitting in
the middle of a fat horizontal pill with lots of empty space
above and below. Operator framing: "puffy, not a tool."
Two changes:
- Custom 4dp rounded track replaces SliderDefaults.Track. The
thumb (14dp) now reads as visibly taller than the bar — the
classic "handle on a string" cue that says "tool, draggable."
Also drops M3's stop-indicator dot which the web scrubber
doesn't have.
- Clamp the Slider's vertical footprint to 20dp via Modifier.
height. 14dp thumb + 3dp clearance each side, vs the default
~17dp empty above and below. Touch area stays usable since the
drag axis is horizontal — pulling left/right anywhere on the
thin bar feels natural, and Slider's gesture detector still
responds to a tap anywhere along its row.
Keeps an Android flavor (slightly thicker than the web's 2-3px
hairline; rounded caps; accent fill) without reading as bulky.
Operator framing: the cover art is the main feature of NowPlaying,
not the background. A vibrant accent on the cover (small bright
logo, sticker, stripe) should pop against the background, not be
matched by it. The previous vibrant → muted → dominant fallback
chain often picked a high-saturation accent that covered only a
sliver of the cover, producing gradients that clashed with the
actual image.
Drop to dominantSwatch only — the majority-by-pixel-count color.
If the palette resolves no dominant swatch (extremely rare;
essentially uniform/empty bitmap) the held color stays on the
previous track's dominant, matching the existing "keep previous
on failure" docstring contract.
Operator: tapping Start Radio while music plays previously
reloaded the current track from position 0 because the radio
seed response includes the seed at index 0 and the handler called
setQueue(tracks, 0) — Flutter's playerActions.startRadio does the
same. They want the current track to keep playing untouched,
upcoming queue cleared, radio results appended after.
PlayerController.startRadio now branches on mediaItemCount:
- Empty queue: existing behavior — setQueue from index 0.
- Active queue: keep currentMediaItem, removeMediaItems from
currentIdx+1 to end, then addMediaItems with the radio list.
When the seed is the currently-playing track (the common
"Start Radio on the song I'm listening to" case), drop the
seed from the appended list so it doesn't immediately repeat
after the current track ends.
queueRefs is updated alongside the controller so the cached
TrackRef list stays consistent. Source tag "radio:<id>" is
preserved for the appended items so play_started attribution
stays correct.
Intentional divergence from Flutter — recorded in the docstring
so future ports notice it.
Pull-to-refresh produced a strong full-screen fade on Library
(both tabs) and the Album / Artist / Playlist detail screens
because their Crossfades were keyed on the entire state value.
A refresh emits a fresh UiState.Success with a NEW data instance
(same kind, different content) so Crossfade animated old grid →
new grid even though both are the same Success branch — the
visible result was a flash that read as "broken/heavy."
HomeScreen already keys on `state::class` (4b9d-ish prior fix);
apply the same pattern to the four screens that still flicker.
Inner content reads the outer `state` directly via `val s = state`
so the branch still has access to the typed value. Row-level diffs
are owned by LazyVerticalGrid / LazyColumn via item keys, so the
visual update is smooth and granular instead of a full fade.
Only Loading ↔ Success ↔ Error ↔ Empty transitions animate now —
the intended use of Crossfade. Same-kind state updates flow
through Compose's normal recomposition.
Operator: tabs in the Library view should feel swipeable, not just
tappable. Replace the selectedTab Int state + when-block content
with HorizontalPager whose state drives the PrimaryScrollableTabRow.
Tap routes through animateScrollToPage so swipe + tap share one
source of truth.
Horizontal pager gestures don't conflict with the LazyVerticalGrid
inside each tab (different axes) or with PullToRefreshScaffold's
vertical pull (different axes). HorizontalPager renders only the
current page by default; adjacent tabs remain composed during the
swipe but not eager-mounted at start.
Operator polled another user and reversed the earlier swap to
Material's LibraryMusic. Restore Lucide.LibraryBig and drop the
material-icons-extended Gradle dependency we added for the
intermediate icon, keeping the icon set Lucide-only.
aec10ce7 added a third early return (the placeholder-host bail)
on top of the existing unparseable-baseUrl elvis return, tripping
detekt's ReturnCount ceiling of 2 on the dev test workflow.
Refactor: keep the placeholder-host early bail (it preserves the
no-op cost for external URLs — no AuthStore read, no URL parse),
fold the unparseable-baseUrl case into a `?:` that falls back to
the original URL. Result is two returns and identical observable
behavior — placeholder hosts get rewritten when baseUrl parses,
fall through unchanged when it doesn't.
Existing unit tests cover all three paths and continue to assert
the same outputs.
Tapping the system media notification previously landed on
whatever shell route MainActivity last rendered (Home / Library /
Search) because MinstrelPlayerService never configured the
session-activity PendingIntent, so Media3 defaulted to the
launcher activity entry point. Operator request: tap should go
straight to the full player.
MinstrelPlayerService.onCreate now builds a PendingIntent
targeting MainActivity with an EXTRA_OPEN_NOW_PLAYING flag and
passes it to MediaSession.Builder.setSessionActivity. The flag
also covers the lock-screen card and the Pixel Watch tile —
both use the same session-activity PendingIntent.
MainActivity reads the extra in onCreate AND onNewIntent (so a
warm app gets the navigation too, not just cold launches), flips
a pendingOpenNowPlaying StateFlow, then strips the extra so a
config-change recreation doesn't re-trigger. The App composable
observes the flag and runs a LaunchedEffect to navigate once the
NavHost is mounted — handles both cold start (BootSplash →
resolved → navigate) and warm start. launchSingleTop avoids
stacking copies if NowPlaying is already on top, and the
onOpenedNowPlaying callback clears the flag post-navigation so
later recompositions don't re-fire.
Divergence from Flutter (intentional): audio_service's default
notification tap behavior just opens the launcher activity at
whatever screen it was on — exactly the behavior the operator
asked to improve.
Lucide has no music-library glyph — Library / LibraryBig /
SquareLibrary all read as a generic books-on-shelf icon without
a label. Operator picked Material's LibraryMusic (the canonical
"books + music note" symbol used by every major music app) as
the recognizable alternative.
Use the Outlined variant: filled icons would clash with the
neighbouring stroked Lucide icons (House, Search, EllipsisVertical),
but Outlined's stroke style matches Lucide closely enough that
the mix is subtle.
Adds the compose-material-icons-extended dependency (version
pinned by compose-bom). R8 strips unused icons in release builds
so the APK cost is just the ones we actually reference.
LibraryBig (stacked book spines) didn't read as "Library" without
the label — easy to misread as a generic stack/columns icon.
SquareLibrary frames the same books-on-shelf glyph inside a
rounded square, matching the visual weight of the neighbouring
House and Search icons better and reading more clearly as a
distinct tappable destination.
Untouched: the RequestsScreen per-row "album"-kind avatar still
uses LibraryBig (parity with Flutter's lib/requests/requests_screen.dart).
Discover suggestion artist images failed to load on Android while
loading fine in the web client. Root cause: BaseUrlInterceptor
unconditionally rewrote every outgoing request's scheme/host/port
to AuthStore.baseUrl. That's correct for Minstrel-bound requests
built with the http://placeholder.invalid sentinel (Retrofit's
frozen baseUrl, every cover-URL builder, ServerImage's
resolveServerUrl). But Lidarr surfaces artist artwork as absolute
URLs to external hosts (artwork.musicbrainz.org,
coverartarchive.org); rewriting those to the Minstrel host
produced 404s that Coil silently fell back from to the User icon.
Web works because the browser fetches the URL as authored. Coil
on Android shares the OkHttp client (and so the interceptor chain)
with Retrofit, which is why the bug surfaced here only.
Add a PLACEHOLDER_HOST companion constant and short-circuit the
rewrite for non-placeholder hosts. Test coverage:
- placeholder host → rewritten to live baseUrl
- absolute external URL → host/scheme/path preserved
- unparseable baseUrl → falls through (no throw)
AuthCookieInterceptor still attaches the Minstrel session cookie
to external requests; external hosts ignore unrecognized cookies
so that's not breaking anything, but it's worth a follow-up DRY
pass to scope auth attachment the same way.
Operator feedback: landscape just stretches the phone-portrait
Compose layout awkwardly — every screen was sized for one column
of cards, so rotation produces wide rows of unrelated content with
big dead bands top and bottom. Until a dedicated tablet/landscape
layout exists, lock the activity to portrait via screenOrientation.
Revisit when a sw600dp resource set + multi-pane layouts land.
The dismissed-latch added a third early return to onPostScroll —
detekt's ReturnCount ceiling is 2 per the project rule. Fold the
NestedScrollSource.UserInput guard into the existing if/else if/else
chain that branches on the drag direction. Behavior is identical;
the source check just becomes the first arm of the expression
rather than an early bail.
74bae74f routed per-artist system mixes through getPlaylist and
always passed the source attribution as the third arg, falling
back to undefined for true user playlists. Vitest's toHaveBeenCalledWith
is arity-strict — playQueue(refs, 0, undefined) is not the same as
playQueue(refs, 0) — so the PlaylistCard contract test failed on
the user-playlist case.
Split the call: pass three args only when a variant tag exists,
two args for user playlists. Preserves source attribution for
songs_like_artist and keeps user playlists source-less as the
test pins.
Operator reproduced black-screen-on-resume by drag-down dismissing
the full player. Root cause: the NestedScrollConnection accumulator
crossed the dismiss threshold, called navController.popBackStack(),
reset accumulated to 0 — but the user's finger was still down and
the pop transition was still running. The next frame's onPostScroll
re-accumulated and re-fired onDismiss(), popping the screen BENEATH
NowPlaying. When that left the back stack empty the NavHost had no
destination to draw, producing a black window until the process
was killed and the activity was cold-launched.
Add a `dismissed` latch that survives until the connection is
disposed (which only happens when NowPlayingScreen leaves the
composition, i.e. the pop completes). After the latch sets we
consume the remaining drag (return `available`) so the underlying
scrollable doesn't paint over-scroll while the pop transitions.
onPreFling also bails after dismissal so the fling can't restart
the accumulator.
The full player previously sat dead-center on wide viewports — cover
+ controls capped at max-w-md left the right two-thirds of the screen
empty while the queue remained behind a drawer toggle. Split the
layout at lg+:
- Extract QueueList from QueueDrawer (header + count + scrollable
list body + empty state); the drawer now wraps QueueList for its
slide-in chrome, and the now-playing route embeds it directly.
- now-playing: at lg+ render the player as a left section + a 384px
(xl: 448px) aside with QueueList. Below lg the layout is unchanged
single-column and the drawer toggle in the bottom row still opens
it (lg:hidden on the toggle button keeps it out of the way once
the panel is permanent).
- QueueList owns the close X conditionally — drawer passes onClose
and the bind:closeButtonRef for focus management; embedded panel
omits both since it has no dismiss action.
QueueTrackRow's drag-to-reorder, click-to-jump, and current-row
highlight all carry over for free since they're owned by the row
component.
User feedback: dislikes both the old M3 20dp thumb and the current
10dp slim variant; wants the bare HTML range thumb the web client
shows. Replace SliderDefaults.Thumb with a plain Box(CircleShape +
accent fill, 14dp). Drops the M3 state-layer halo on press and the
implicit elevation/border so the on-screen result matches a
<input type=\"range\" accent-color> rendering. Slider's 48dp hit
slop is intrinsic to the composable, so tapability is unchanged.
PlaylistCard: per-artist system variants (songs_like_artist) all share
one variant tag but exist as one playlist per seed artist; routing
their play handler through systemShuffle(variant) hit the wrong
playlist (or 404). Detect via seed_artist_id != null and fall through
to getPlaylist(playlist.id) for those; still tag the queue with the
variant so source attribution stays correct.
smoothPosition.svelte.ts: new useSmoothPosition() hook mirrors
Android's rememberSmoothPositionMs. player.position only updates
~4 Hz (HTML audio timeupdate), so the seek thumb stepped visibly;
$effect resets on each canonical tick / seek / track-change and a
rAF loop extrapolates at playback rate between ticks.
Wired into both PlayerBar.svelte (mini + expanded seek rows) and
now-playing/+page.svelte. Seek input handler still reads the raw
range value (not smoothed.value) so user drags stay authoritative.
Operator: prior rail disabled empty buckets, so clicking a letter
like 'Z' did nothing if it wasn't loaded yet. AlphabeticalGrid now
accepts a paginated source and walks pages on click until the
target letter surfaces.
- New props: hasMore, onLoadMore. When hasMore=true, every empty
bucket stays enabled (clicking will chase pages); when hasMore=
false, only populated buckets are clickable.
- jumpTo(bucket): if populated, scrollIntoView. Otherwise loop
onLoadMore + tick() until the bucket appears or no more pages.
Loader2 spinner replaces the letter on the pending button;
cursor:wait + aria-busy. Other rail buttons disable while one is
pending so clicks don't stack.
- Library Artists + Albums pass hasMore + onLoadMore through to the
grid. Existing InfiniteScrollSentinel still handles scroll-driven
loading.
Test: new case asserts rail enables empty buckets when hasMore=true
(the click would trigger onLoadMore).
Operator: prior rail only emitted buttons for letters with items, so
jumping to a letter required scrolling to that section first — not
useful as a navigation tool. Rail now always renders the full set so
the page reads as a stable A-Z reference regardless of which letters
are present.
- New bucketFor() classifies the first character into '#' (digits),
'A'-'Z' (letters), or '&' (everything else). Anchor ids switch
from per-letter to per-bucket: alpha-#, alpha-A, ..., alpha-&.
- ALPHABET + railEntries hardcode the full # / A-Z / & order.
- Buttons for empty buckets render disabled with a low-opacity tint
+ cursor:default so they read as 'no entries here' rather than
broken jumps. aria-label changes too — 'Jump to A' (enabled) vs
'C — no entries' (disabled) so screen readers announce state.
- # / & get verbose labels ('numbers'/'symbols') because the bare
glyph isn't readable.
Tests rewritten — 4 cases: full-rail-with-disabled-buckets,
DOM-order, populated-bucket ids, and a separate fixture confirming
digit-starting items bucket under '#'.
Operator UI review: artists/albums layout had a full-row letter
divider per first-character, so a section like '2' (one artist)
left 9 empty cells before '8' (one artist) started a new row. Lots
of dead space across letters with few entries.
Operator chose option 2: drop the dividers, add a side jump-bar.
- AlphabeticalGrid no longer renders per-letter row-spanning
dividers. Items flow continuously across the grid; first item
of each letter gets id='alpha-<letter>' so the rail can target
it via scrollIntoView.
- New sticky vertical alphabet rail on the right. position:sticky
+ top:50% + translateY keeps it vertically centered in the
viewport while the grid scrolls underneath.
- Each rail entry is a focusable button with aria-label='Jump to
<letter>'. Hover/focus tinted with accent.
- Only renders when there's more than one distinct letter so
small libraries (or filter-narrowed views) don't get an empty
rail.
Test rewritten — letters are buttons on the rail, not dividers in
the grid. Three assertions:
- one jump button per distinct first-letter
- DOM order is items first then rail (was rail-then-items before)
- first item of each letter carries the alpha-<letter> id
Three operator-reported issues:
- Header was flex with the nav inside a flex-1 span — when the
right side (search + user menu) grew wider than the left
(wordmark), the nav's centered position drifted off page-center.
Switched to grid-cols-3 with justify-self-{start,center,end} so
the middle column pins to true window-center regardless of side
widths.
- Library nav link pointed to /library, which 308-redirects via
+page.server.ts. Operator reported it didn't navigate. Linked the
nav button directly to /library/artists so SPA navigation skips
the redirect roundtrip. matchPrefix='/library' keeps isActive
matching every Library tab.
- SearchInput placeholder was 'Search artists, albums, tracks…' —
shortened to 'Search' and added a Lucide Search icon inside the
input on the left. Padding adjusted (pl-7) so the input text
clears the icon.
Operator clarification: the 'compact like the app' request was for
Most Played, not Rediscover.
- CompactTrackCard rewritten as a horizontal Row: cover thumbnail
left, title + artist column right. w-72 (~288 px) matches the
Android CompactTrackTile's 176 dp visual weight. Most Played's
existing chunked-rows-in-shared-scroller layout now reads like
the Android multi-row LazyHorizontalGrid.
- Rediscover steps back from the compact w-32/w-28 widths to w-40/
w-36 (the pre-PR-#66 sizes) — the size hierarchy still works
with the regular AlbumCard treatment.
Operator UI review on the merged build:
- Hero row felt 'odd' — operator chose remove-entirely on the
AskUserQuestion options. The Playlists row already leads with
For-You and Recently Added leads with the newest album, so the
hero duplicated both anchors. Pulled HomeHeroCard.svelte + all
page-level wiring (systemShuffle/getPlaylist imports + playForYou/
playLatestAlbum handlers) and reverted the within() test scoping
added in 682d7a5e.
- Rediscover tiles step down to w-32 (albums) / w-28 (artists),
matching the Most Played CompactTrackCard visual weight on web.
Reinforces the page hierarchy: fresh content gets real estate,
throwbacks recede.
The new hero card on Home renders the For-You playlist twice — once
at the top in the featured row and once in the Playlists carousel.
The two existing assertions that did screen.getByText('For You')
now match both and fail. Wrap each in within(playlistsSection) so
the test targets the carousel render specifically.
#6 — Adds a 2-up grid above the Playlists carousel:
- Today's pick: For-You playlist card with a Play button that
invokes the system-shuffle endpoint (same path as PlaylistCard
uses for the play overlay).
- Just added: most recently added album card with a Play button
that fetches the album detail and queues the tracks.
Both cards share HomeHeroCard.svelte — cover left, eyebrow + title
+ subtitle + Play right, on a from-accent/15 → surface gradient so
the eye lands here before scrolling into the carousels. Each card
links to its detail page; the Play button is event-stopped so it
plays in-place. md:grid-cols-2 stacks on narrow viewports.
Only renders when at least one of the two pieces of data is
available so the page degrades cleanly on fresh installs.
#9 — new lib/media/dominantColor.ts: load cover image, downsample to
1x1 canvas, read pixel. Approximates the dominant tone via the
browser's bilinear mean — close enough for an ambient accent without
the 5KB ColorThief dependency. Same-origin cover URLs so no CORS
dance. Result cached by URL so revisits are free.
PlayerBar samples the current track's cover and pipes the resulting
rgb into a 2px accent strip above both the compact and desktop
variants. Transparent until the first resolve; 300ms transition on
colour change so track-skips fade rather than snap.
#10 — slight bump to Home tile widths so a typical viewport shows
roughly 5-6 across instead of cramming 8-9: Playlists w-56, all
remaining AlbumCard rows w-48 (Recently Added + both Rediscover
album scrollers). Replaces the wave-2 sizes that were still showing
7-8 across on wider screens.
Second wave of Home visual polish (UI tasks 80/82/84):
- Shell main background gets a 1200x600 radial gradient at the top
using color-mix on fs-iron so the page reads with subtle depth
instead of as a flat slab.
- PlaylistCard moves the title onto the cover with a bottom-to-top
gradient overlay. The server-generated 2x2 collage becomes mood
texture; the Fraunces playlist name becomes the dominant element.
Track count and refresh label stay below the art. Mirror albums
shadow-sm/shadow-lg/ring-accent hover treatment on the art-wrap.
- Tile-size hierarchy: Playlists w-52, Recently added w-44,
Rediscover w-40. Visual weight tapers as you scroll down the
page so the freshest content reads as most important.
First wave of Home visual polish (UI tasks 78/79/81/85):
- CardActionCluster like/queue/menu cluster is now opacity-0 and fades
in on group-hover or focus-within. The group class moves from the
inner anchor to the outer card wrapper so the cluster (positioned
outside the anchor) participates in the same hover scope. PlaylistCard
refresh kebab gets the same treatment.
- AlbumCard drops the year line. Title plus artist is enough on Home
tiles; the album detail page still shows the year.
- HorizontalScrollRow h2 picks up a flex-baseline layout with a trailing
12-wide accent rule (after:bg-accent/60), so section headers read as
chapter breaks instead of identical plain text.
- AlbumCard art-wrap is shadow-sm by default and lifts to shadow-lg
plus ring-accent/40 on hover. Pairs with the existing
group-hover:scale-1.03 for a clean lift-on-mouseover feel.
Previously Recently Added chunked into rows of 25 and rendered each
chunk as its own LazyRow, so chunks scrolled independently. Same
pattern Most Played uses (one LazyHorizontalGrid where all rows
scroll as a single panel) now applies to Recently Added.
- New RECENTLY_ADDED_GRID_ROWS = 2 + RECENTLY_ADDED_GRID_HEIGHT_DP =
440 (matches a 200dp tile × 2 rows + the 8dp inter-row gap).
- New RecentlyAddedGrid composable owns the section header + the
LazyHorizontalGrid. Column-major re-flattening matches Web's
row-major reading order on screen (top row first, then bottom).
- recentlyAddedSection collapses from itemsIndexed-over-chunks to
a single item { RecentlyAddedGrid(...) } in the outer LazyColumn.
- AlbumsRow stays untouched (still used by Rediscover, etc.).
The other multi-section helpers (Rediscover, Last Played, Playlists)
are single-row by design and don't need this treatment.
Three closely-related player polish changes:
1. MiniPlayer (#74) — the bottom bar's progress is a 4dp Box-based
fill, not a Slider. No thumb, no drag handle. Tapping the bar
still expands to NowPlaying where scrubbing lives.
2. NowPlaying scrubber (#75) — keep the M3 Slider (still
interactive for seek) but shrink the thumb from the 20dp default
to 10dp via SliderDefaults.Thumb's thumbSize slot. Slider's own
48dp hit slop is unchanged so tappability stays. Spacers above
and below ScrubberRow drop from 8dp to 4dp.
3. Smooth playhead (#76) — new SmoothPosition.kt with
rememberSmoothPositionMs(). PlayerController.uiState polls the
underlying ExoPlayer position roughly every 500ms; reading it
directly steps the scrubber by half-seconds, which reads as
jumpy. The helper resets to the canonical positionMs on each
tick (and on seek/track-change/duration-change) and runs a
withFrameMillis loop while isPlaying to advance the displayed
value at 1ms/ms between ticks. Pause/end/no-duration short-
circuit the loop. Both MiniPlayer's fill bar and the NowPlaying
scrubber consume the smoothed value.
User-visible: Home flickered continuously after first sign-in until
all sections settled. The top-level Crossfade keyed on the state
INSTANCE — and because each section's flow emission produces a new
UiState.Success(data), Crossfade ran its 300ms fade animation on
every per-section hydration tick. Six sections cascading in over
~1s read as continuous flicker.
Fix: key the Crossfade on state::class. Loading -> Success -> Empty
-> Error class transitions still animate; Success -> Success(with
more sections) recompositions just update the LazyColumn normally
through Compose's standard diff path.
The previous detectVerticalDragGestures modifier on the Scaffold
never fired because the body Column applies verticalScroll, which
wins the touch-slop competition for every vertical drag delta
before the outer detector sees it. User-visible symptom: swiping
down on the NowPlaying screen did nothing.
Replace with a NestedScrollConnection. verticalScroll offers its
over-scroll deltas to the nearest ancestor connection via the
standard nested-scroll protocol; when the column is at the top of
its scroll range, downward drag deltas surface in onPostScroll
unconsumed (available.y > 0). Accumulate those toward a 200 px
threshold and pop the back stack.
- Upward over-scroll or any consumed delta resets the accumulator
so a partial drag-down followed by drag-up doesn't latch.
- onPreFling resets on lift-off so a lazy swipe doesn't dismiss
later after the user has released.
- Slider thumb retains its own pointerInput and is not affected.
- Source filter (UserInput) ignores nested-scroll-driven
animations (e.g. flings from inner scrollables).
CacheSettings.prefetchWindow has shipped at default=5 since M8 but
the prefetcher itself was deferred to a follow-up. Without it,
forward skips re-fetch from the network every time and gapless
transitions stall.
New AudioPrefetcher singleton subscribes to PlayerController.uiState
+ AuthStore.cacheSettings, walks queue[currentIndex+1 .. +window],
and runs each missing track through Media3's CacheWriter against
the same SimpleCache the player reads from (PlayerFactory.simpleCache).
DataSpec uses setKey(trackId) to match PlayerController.toMediaItem's
setCustomCacheKey(id) — without this the player would miss the
cached bytes on read-through.
Reconcile is idempotent: CacheWriter is a no-op when bytes are
already resident, so distinctUntilChanged on (queue, index, window)
gates re-runs and the per-item check is cheap. Window slides cancel
in-flight jobs for tracks that have dropped out (skip-prev, queue
rebuild) so a stale prefetch doesn't keep the network busy.
Eager-constructed via the construct-the-singleton trick in
MinstrelApplication, alongside CacheIndexer and CoverPrefetcher.
Mirrors flutter_client/lib/cache/prefetcher.dart's user-visible
behavior (queue-walk + per-track pin + idempotent reconcile)
implemented with the native Media3 primitives (CacheWriter +
SimpleCache) instead of Flutter's AudioCacheManager.pin.
Previous design: android.yml's release job and release.yml ran in
parallel on every tag push. release.yml polled the gitea release-
download URL for up to 15 min waiting for the APK to appear. In
practice the polling step was completing in 11s — either following a
gitea redirect to a 200 page and writing HTML into the bundled APK
file, or returning empty content silently. Either way the resulting
image shipped without a working APK and the web Settings page's
in-app-update section never rendered.
Fix the race architecturally:
- Move the signed-release-build + Attach-APK steps out of android.yml
and into a new android-release job in release.yml.
- release.yml's image-release job declares needs: [android-release],
so on tag pushes the image cannot start building until the APK is
guaranteed-attached.
- Pass the APK between jobs via actions/upload-artifact@v3 +
download-artifact@v3 (the v2 backend isn't supported on Gitea).
This removes the polling loop entirely — the image-release job just
downloads the artifact, renames to minstrel.apk, writes the
.version sidecar, and continues to docker buildx.
- For main pushes android-release is skipped via its if: condition.
image-release uses so it
still runs (the skipped predecessor doesn't poison the chain) and
the download/stage steps gate themselves on the tag context. Main
images ship without an APK by design, same as before.
android.yml is now testing-only: lint + detekt + unit tests on every
push, debug APK artifact on main. Independent of release CI as
requested.
bvandeusen