Twitter-style page titles like "(2) 主页 / X" land a literal "/" in
the download filename. Gopeed used to strip such characters silently;
aria2 is strict and passes them through to the OS, which then reads
"/" as a path separator — the save ends up at
`.../(2) 主页 /X-....mp4` pointing at a non-existent sub-directory
and fails with `ERROR_PATH_NOT_FOUND (errNum=3)`.
Sanitize once, at the task-creation boundary, so the DB row, the
downloader command-line `-o` arg, and the post-download
`CheckFileExists(rec.Name, ...)` probe all agree on the same
filesystem-safe value.
- New exported `core.SanitizeFilename` replaces reserved path /
wildcard characters (`\ / : * ? " < > |`) and ASCII control chars
with `_`, and right-trims dots / spaces (Windows strips those
silently, producing a filename that doesn't match the DB row).
Falls back to "download" if every character was illegal.
- `service/download_task.go` `AddDownloadTask` and `AddDownloadTasks`
run titles through `SanitizeFilename` before the
`FindByName` de-duplication check, so both the dedup lookup and
the persisted row see the cleaned value.
- `core/downloader.go` `buildArgs` still calls `SanitizeFilename`
defensively on `p.Name` — cheap, and guards any future path that
bypasses the service layer.
Applies to every downloader (aria2, yt-dlp, BBDown, N_m3u8DL-RE,
mediago) since the fix is in the shared `name` arg-building branch
used by all Schema entries. Pre-existing broken tasks in the DB will
still fail "file not found" on the UI and need to be deleted +
re-created.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Bilibili returns an error page ("出错了") when requests lack a proper
User-Agent. Add browser User-Agent and Referer headers to both
GetPageTitle (service/helpers.go) and fetchTitle (handler/util.go).
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Move ffmpeg execution from Electron to Go Core so conversion works in
both Electron and web server modes.
Go Core backend:
- Add -ffmpeg-bin CLI flag to pass ffmpeg binary path
- Enhance Conversion DB model with status, outputPath, outputFormat,
quality, progress, and error fields
- Create converter executor (service/converter.go) that spawns ffmpeg,
builds args per format/quality, and parses stderr for progress
- Add StartConversion/StopConversion to service with SSE events for
real-time progress (conversion-start/progress/success/failed/stop)
- Add POST /api/conversions/:id/start and /:id/stop endpoints
Supported formats:
- Video: MP4 (H.264), MKV (H.264), WebM (VP9) with CRF quality
- Audio: MP3, AAC, FLAC, WAV with bitrate quality presets
- Quality presets: high/medium/low mapping to CRF and bitrate values
Frontend (converter-page):
- Add format selector (Video/Audio groups) and quality dropdown
- Show per-item status badge, progress bar during conversion
- Action buttons: Start/Stop/Open Folder/Delete based on status
- "Convert All" batch button for pending items
- SWR auto-refresh during active conversions
Integration:
- Pass -ffmpeg-bin from Electron and core:dev to Go Core
- Remove old broken Electron-only convertToAudio code
- Add startConversion/stopConversion to core-sdk and GoApi
- Add i18n keys for new UI elements (en/zh)
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Converted all Chinese-language comments across 47 source files to
English, covering Go backend (apps/core, apps/player), TypeScript
scripts, Electron main process, UI components, and shared packages.
Only comment text was modified — code, string literals, log messages,
and i18n translation values were left untouched.
Also fix pre-existing lint errors in scripts/release.ts and
scripts/utils.ts (node: protocol prefix, unused imports, bare catch).
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
caorushizi