- watchPlaybackOutcome now tracks a real deadline so slow property reads eat the timeout budget instead of extending it, and a zero probe interval still terminates
- stream strip proxy aborts upstream GETs that go silent for 15s instead of hanging mpv on that segment forever
- anime browser runtime drops a stream proxy that finishes starting after its bridge already died, instead of leaking a listener pointed at nothing
- Download a stream's subtitle tracks to a temp dir and load them into mpv as files instead of streaming from the source URL, so Subsync can use them as the alass reference (mirrors the Jellyfin subtitle preload)
- Detect format from content first, falling back to the URL extension
- A track that fails to download falls back to its URL so the episode still plays
- Remove the temp dir when the next episode loads or the runtime disposes
- Track kill errors separately from exit/spawn so a signal failure (e.g. EPERM) surfaces its own message instead of a bare "did not exit after SIGKILL"
- Attach stop() failures as the cause on the readiness-timeout error so a surviving child's kill failure isn't dropped
- Skip stop() entirely when spawn itself failed, and stop polling once the ready deadline has passed
- Add stopTimeoutMs test hook and a test covering a kill that fails without the child exiting
- Lay source picker label + input inline instead of stacked
- Let detail cover keep natural aspect ratio instead of forced 2:3 crop
- Widen cue rail gutter to match new cue number column width
A hidden overlay window stops producing frames once it is explicitly
hidden and then reloaded (the Yomitan content-script reload right after
startup), so ready-to-show never fires and the content-ready gate keeps
the window hidden forever behind the Overlay loading spinner. Mark
content ready from did-finish-load after a 1.5s grace period so the
overlay always becomes showable; ready-to-show still wins when it fires.
The reported 'APK crash' turned out not to exist: repeated full scans of
all installed extensions load cleanly. The real incident was the app's
quit sequence stopping the bridge (its /stop endpoint) while requests
were in flight - and it exposed that a bridge dying for any reason left
the anime browser failing every call with a raw 'fetch failed' until
the whole app was restarted.
SidecarHandle now reports the child's exit (including to late
subscribers), and the runtime detects a death it did not order: it
drops the dead handle, closes the strip proxy, tells the UI, and the
next request transparently starts a fresh bridge. requireBridge became
an async bridge() that starts on demand, which also covers first use.
Verified live: search -> SIGKILL the JVM -> search recovers.
ffmpeg opens every HLS segment with 'Range: bytes=0-'. The bridge
answers some of those 206 (depending on its cache state), and the proxy
only rewrites full 200 bodies, so the PNG disguise passed through
untouched and whether an episode played was a coin flip: first attempts
died with "no audio or video data played", the same URL played fine
once the bridge had the segments cached.
Reproduced end to end against a private bridge + mpv driven with the
app's exact commands; with Range dropped every segment is stripped and
playback runs.
Right after an episode resolves, the bridge (or the host behind it) can
error on the very first segment fetches and be fine a moment later; mpv
sweeps the playlist of failing segments in ~2s and gives up with "no
audio or video data played" while a manual reload of the same URL plays.
Retry a failed GET once after 400ms and log upstream error statuses so
the next failure names the real cause in the app log.
Some hosts prepend a fake image header (a real 1x1 PNG) to every HLS
segment; ffmpeg probes the segment as a picture and mpv drops back to
idle with no window while the browser claims the episode is playing.
- Route bridge-served m3u8 streams through a local strip proxy that
scans each segment for the first genuine MPEG-TS packet run and drops
the junk in front of it; playlists get absolute origins rewritten so
segment requests come back through the proxy. Non-TS bodies pass
through untouched.
- Only report ok from playEpisode once mpv configures a video output;
an end-file with reason error surfaces mpv's own message (e.g. "no
audio or video data played") in the browser status bar instead of
"Playing". New end-file event plumbed through the mpv IPC client.
- Add PINNED_BUNDLE_SHA256 entry for linux-x64-bundle.zip so Linux no longer refuses to start with "No pinned checksum"
- Fetch the bridge release by its pinned tag (releases/tags/<tag>) instead of listing newest releases, so an upstream publish can't swap in an unverified asset
- Update docs and changelog to describe the pin/tag behavior
- Add a language chip row above the Available list; pick one or more languages to narrow the repo index, or "All" to clear
- Show language display names (e.g. "Japanese") instead of raw tags, and an "N of M" count on the Available heading
- Add language-filter.ts with collectLanguages/toggleLanguage/pruneSelection/filterByLanguage/languageLabel, plus tests
The capabilities probe used a fixed 5s timeout, so a readiness budget
shorter than that could be overrun by one stalled request. Pass the
remaining deadline down instead.
stop() also resolved after the SIGKILL wait even when the child had not
exited, letting a restart race a process still holding the port. Throw
in that case, and keep a failed shutdown from masking the readiness
error at startup.
Addresses CodeRabbit review feedback on the Anime Browser:
- reject repository package/apk names that are not plain identifiers, and
verify the install target resolves inside the extensions directory
- count mpv's %n% option escape in UTF-8 bytes, and escape backslashes in
header values so a trailing one cannot eat the list separator
- key the bridge extension-id cache by APK content, so an in-place upgrade
re-uploads instead of running the previous build
- bound every bridge, release-listing, and download request with a timeout
- enforce the APK size limit while streaming rather than after buffering
- read APK bytes on demand instead of holding a base64 copy per extension
for the lifetime of the browser
- serialize preference mutations and write the file atomically
- handle the sidecar spawn error event, and wait for the child to exit in
stop() before returning
- report a failed Anime Browser bootstrap instead of showing the starting
banner forever
- keep the preferences panel's save confirmation and in-flight multi-select
edits by re-rendering only on a structural schema change
- Add `subminer anime` / `--anime` and a tray entry to open a browser that searches installed Aniyomi extension sources, shows cover art and episodes, and plays into mpv with overlay/mining attached
- Add an Extensions tab to add repos and install/update/remove sources, and per-source settings for sources needing config
- Support searching all sources at once with streaming, per-source results and status
- Prefer Japanese audio/subtitle tracks from the source and keep the primary subtitle slot reserved for Japanese
- Fix window/tray/Dock handling so the browser and mpv can be switched between without quitting the app or losing the Dock icon
- Add anime.repos, anime.extensionsDir, anime.preferredQuality config keys (no bundled repos or discovery)
The query test suite opened each temp database with SQLite's defaults
(rollback journal, synchronous=FULL) while the runtime opens them via
applyPragmas (WAL, synchronous=NORMAL). Every statement therefore ran as
its own fsync-ing transaction.
That is unnoticeable on a fast local disk but dominates on CI: the file
took 40.6s there versus 0.5s locally, and the 106-insert case
'getVocabularyStats pages past hidden rows' crossed the 5s per-test
timeout and failed the quality gate.
Open test databases through a helper that applies the same pragmas as
production, so the tests exercise the runtime's actual SQLite
configuration instead of a slower one.
- Add TsukiHime and AniSkip to feature table; note xz dependency
- Clarify Anki+AnkiConnect is required to mine, not to run
- Rework Windows setup with winget/scoop commands and PATH steps
- Note ffmpeg has no path-override setting, unlike mpv
- Fix screenshot alt text and installation.md anchor link
- Generate config.example.jsonc with a Linux socket path instead of Windows, so the example stays reproducible across platforms
- Document TsukiHime config (`tsukihime.*`), maturity-based known-word highlighting keys, and other config surfaces missing from the reference
- Correct secondary-subtitle auto-load defaults and Anki field-matching (case-insensitive) claims
- Expand Windows installation guide with winget/Scoop package tables and manual PATH setup
- Add docs-site/links.test.ts to catch broken internal links and heading-anchor mismatches
- Fix stale cross-references in jimaku-integration.md and demos.md
- Add formatRofiPrompt to trim trailing whitespace and append a single space, keeping the prompt from running into the input field
- Apply it across all rofi -p usages in picker.ts and history-command.ts
- Add tests for formatRofiPrompt edge cases (empty/whitespace-only prompts)
- Extract buildHistoryEntryActions to build the series action menu, now including a previous-episode option alongside replay/next/browse/quit
- Update docs and changelog entry to describe the new option
- Add tests covering previous/replay/next ordering and omission when last watched file is missing
- AnkiConnect proxy now detects EADDRINUSE, logs a warning, and surfaces an overlay notification instead of crashing video startup
- Background AppImage bootstrap runs Chromium with in-process-gpu so no GPU child survives app.exit() and outlives the FUSE mount, fixing the remaining DrKonqi "Service Crash" case
- Add changelog fragment for the proxy fix; update the AppImage quit fragment with the GPU child root cause
- Merge the five separate sync change fragments (launcher command, UI window, close/quit fixes, active-session double-count, monthly rollup timezone) into a single changes/sync.md
- Fold kana-pos-content-classification.md into annotation-noise-rule-table.md
- Drop the now-stale Unreleased sections from CHANGELOG.md and docs-site/changelog.md
- Tweak tsukihime-english-subtitles.md wording
- Autoplay-ready gate now ignores untokenized subtitle payloads while tokenization warmup is pending, so playback doesn't resume before the first tokenized cue
- Wire isTokenizationReady dep through main.ts and update gate/wiring tests accordingly
- Drop stale release/release-notes.md, add changelog entry under changes/
Background AppImage launches executed directly from a FUSE mount owned by a
short-lived helper's AppImage runtime. On quit, the runtime unmounted the
squashfs while utility children (network service) were still mid-shutdown,
SIGBUSing them and triggering KDE Service Crash notifications on every video
close. Detach a POSIX-sh supervisor instead: mount via --appimage-mount, run
AppRun from the mount, and release the holder only once no process still
executes from the mount. Falls back to the old direct launch on any failure
or with SUBMINER_NO_APPIMAGE_MOUNT_KEEPALIVE=1.
sudacode
