The native backend followup test compared and timed the native taste
trainer against trainAutoplayTasteDartInBackground, which 643123661
removed as unused, so the file no longer analyzed. The test now runs the
public Dart trainer on a background isolate itself.
Flutter 3.50 deprecates its built-in Material and Cupertino libraries in
favour of the material_ui and cupertino_ui packages. The imports move
with the official fix (dart fix, migrate_design_widgets); the app's old
two-library theme bridge collapses into one Theme type.
- MaterialExpressiveScope keeps the complete host theme nearest its
child. The reduced copy made for the expressive widgets now shares the
app's Theme type and would hide component themes and extensions.
- LegacyThemeBrightness hands the brightness to the glass package, which
still reads Flutter's built-in theme.
- appLocalizationsDelegates lists the material_ui localization delegates;
the generated list still names the built-in ones.
- Tests: findByTooltip and pageBack replace the flutter_test helpers that
only know the built-in widgets, and several tests that had gone stale
against earlier app changes are repaired.
An integration test still passed the removed artworkForeground argument,
a notifier exposed a getter the Riverpod lint reports on some runs, and
one if statement had no block.
The test still called setRepoRegistryUrl and getRepoRegistryUrl, which
the multi-repository store removed, so the integration tests no longer
compiled. It now sets the list and resolves a URL, which needs no
network.
Every deck now multiplies its rate by playbackSpeed and applies
playbackPitch at once, also mid-song and while paused, and reports
what it can do through PlaybackDeck.tempoSupport.
- audioplayers: effective rate, and a per-player setPitch on the forks'
channel, sent only when it differs from the default and again for
each source so a refusal reports speed only for that source. Direct
output ignores both; Windows and Linux report speed only.
- Old Android engine: setPitch in the vendored audioplayers_android,
applied through MediaPlayer playback parameters with the speed.
- media3: setPlaybackParameters (speed and pitch) replaces setRate; the
AudioTrack buffer is sized for 2.5x.
- libmpv: speed and audio-pitch-correction, and a semitone shift in the
equalizer's af chain: a lavfi rubberband filter where the build has
one, otherwise resampling with mpv's scaletempo.
- Cast reports none; switching decks report their active deck.
Route file hashing, AI lyric reply parsing, cache statistics and Autoplay taste training through native workers. Preserve model persistence and isolated compatibility fallbacks, with native parity and full-route performance coverage.
Add isolated library, download, playback, lyric, and log fixtures with durable-state and idle-work assertions. Native macOS suites pass; measurements exclude occluded-window samples and do not claim Android or iOS runtime performance.
Carry unfinished row displacement into the next stagger and active-row transition. Regressions reproduce a 33.60 px jump on the old code. Keep the intended upward animation in both Material and Mornye.
Accept HTTP 303 redirects to canonical directory paths for GET and HEAD requests. Distinguish unreadable directory listings from local connection-save failures and add diagnostics without URLs or credentials.
Validation: 46 host tests passed, 2 SMB environment tests skipped; 3 Android integration scenarios passed; focused analyzer clean.
A queue write with 64 or more new songs encodes them on another isolate.
If that isolate could not start, every such write failed, and so did each
retry. The writer now encodes those songs on this isolate instead.
A dismissal reset the stored queue revision as soon as it was requested,
so a position save queued before it rewrote the whole row just before the
row was archived. The revision now resets when the dismissal runs, and that
position save stays a scalar update.
The Android background playback integration test's session store takes
the new savePlaybackSession signature, which it no longer matched.
Cancel scans when their source is removed or disabled, and retain offline library indexes when a NAS connection is deleted. Stop waiting on cancelled directory listings and consume late results without publishing them.
Validation: 95 related tests pass, with 2 SMB fixture tests skipped. WebDAV HTTP listing and streaming pass on Android 17 through a non-loopback address. Targeted analyzer and formatting checks pass.
FFmpegService only forwarded each call to FFmpegProbe, FFmpegMetadata,
FFmpegConversion, FFmpegDecryption or FFmpegReplayGain with identical
defaults, so callers now use those modules and the test-only markers move
to the owning statics.
Production removes wishlist, loved, playlist and playlist-track entries through
the batch paths. Device benchmarks now use the production one-item calls as
their per-track baseline.
Compact chrome took five backdrop reads on every scroll frame: the bar
batch plus two stacked reads per side capsule. While folding, each capsule
now has a batch of its own (three reads); at rest the capsules join the
player's batch (one read). A GlobalKey keeps their lenses mounted.
Keep the media service foreground while paused, release CPU wake locks on pause, and demote the service on stop. Await native state publication before requesting audio focus, honor focus denial, and apply the initial MediaPlayer wake mode.
Validated with 141 playback tests and an Android 17 screen-off pause, resume, and stop integration test.
Apply the existing source-error classification during preparation so recoverable FFmpeg diagnostics do not reject playable audio. Fatal source errors still fail the load.
Validation: 15 player unit tests and two native libmpv tests on macOS passed, including repeated playback and seek with an invalid ID3 BOM.
Windows and Linux now create a libmpv deck through media_kit; macOS,
Android and iOS keep audioplayers. Without libmpv the runtime falls back
to audioplayers.
- The equalizer becomes one libavfilter graph behind mpv's lavfi filter
(input gain, peaking bands, balance, compressor, -0.5 dBFS limiter)
after a planar float conversion. Changes set the af property again.
Filters a libmpv build lacks are probed once and left out; the input
gain then moves to mpv's volume. Filter failures are reported per player.
- Deck volume is scaled by the in-app output volume on mpv's cubic curve,
position updates run on a timer, DSD sources throw when libmpv cannot
open them, and exclusive output has a hook for a later setting.
- The in-app output volume persists across launches on every desktop OS.
- Previews go through the runtime's preview deck, so desktop previews
follow the in-app volume.
- Linux publishes MPRIS over D-Bus and Windows uses SMTC through
smtc_windows, both driven by the audio handler.
- The Linux CI job installs libmpv-dev.
An Xcode "Embed FFmpeg" phase copies build/ffmpeg/macos-universal (or
the host architecture) into Contents/Resources/ffmpeg and signs the
tools; MORNYE_REQUIRE_FFMPEG=1 turns a missing stage into an error.
Linux installs them into lib/ffmpeg and Windows into ffmpeg\, warning
when nothing is staged. A macOS integration test converts and probes
with the bundled tools.
Introduce PlaybackDeck, injected through PlaybackDependencies.createDeck,
so the runtime and AutoMix no longer depend on audioplayers types. The
audioplayers implementation becomes AudioplayersPlaybackDeck and keeps
Android's native USB path; it now owns the release mode and audio context
through configureOutput.
macOS gets the equalizer: the audioplayers_darwin fork shares its tap
source between iOS and macOS. On Windows and Linux, until their own deck
lands, enabling the equalizer reports it as unavailable.
Desktop uses an in-app output gain (playbackOutputVolume) that every deck
multiplies into its own gain, and the volume slider writes it instead of
the system volume. Music haptics, pause on mute and play on headphones
are hidden and disabled on desktop.
Rename the Dart package, Android and iOS identifiers, Rust crates, native libraries, and Discord plugin. Preserve legacy data readers and compatibility aliases.
Update localized branding, documentation, artwork, build tooling, and release artifacts. Route Telegram release notifications to @mornyeupdates and link @mornyecommunity.
Reduce repeated artwork readbacks, lyric layout and subscriptions, and animation rebuilds. Virtualize Library sections and Settings search, bound artwork decoding, batch queue restoration, and honor reduced motion.
Add shared player benchmarks with 60/90/120 Hz budgets and regression coverage for lifecycle, scrolling, and accessibility. Verified with 337 host tests, six Android profile suites, scoped analysis, formatting, and diff checks.
Add integration_test/README.md: how to run one suite, why not to run the folder, performance_driver.dart usage, emulator and fixture requirements, every --dart-define and environment variable, and an index of all 54 files.
Derived from every file in integration_test/, the aggregate entry points, performance_driver.dart, performance_probe.dart and scripts/smb2_fixture_server.py.
Replace 16 private SettingsNotifier fakes that only returned a fixed AppSettings with test/support/fixed_settings.dart. Mornye owner WIP tests and bespoke fakes are untouched.
Drop the gomod renovate rule, the stray em dash in crowdin.yml and the migration wording in check_rust_backend.sh. Rename the SMB fixture servers so test_*.py discovery no longer implies unit tests, and describe supported_locales.dart as hand-maintained since its generator never existed.
Provider and decryption extensions describe the downloaded container, not
its audio codec. A decrypted M4A/MP4 can still carry FLAC, so a requested
FLAC output now reaches the codec probe instead of being preserved early.
Lossy streams keep their native container and still receive metadata.