269 Commits

Include renames

Author	SHA1	Message	Date
4gray	069b8b3cc9	feat(playback): advanced subtitle support in shared player controls (#1471)	2026-08-23 13:57:46 +02:00
3 people	5b2eb515d1	feat(downloads): track live-TV recordings in the download manager (#1452) ... * feat(downloads): track live-TV recordings in the download manager Embedded MPV recordings were written to disk and forgotten: no list, no reveal/play, no missing-file handling, and the channel/EPG context was lost the moment the recording stopped. Recordings now live beside downloads: - New `recordings` table (no unique index, no playlist FK — recordings survive source deletion; playlist name stored via playlistDisplayLabel). - EmbeddedMpvRecordingTracker persists the lifecycle: start/stop hooks plus a session-snapshot observer for implicit stops (stream-replacement auto-stop, frame-copy helper crash, session error/close); startup repair turns rows a hard kill left behind into playable `interrupted` partials. - Channel/EPG metadata is captured at recording START in all four live hosts (M3U, Xtream, Stalker ITV, unified live tab); a clean stop triggers renderer-side enrichment with every program overlapping the recorded window, keyed by target path — covering recordings that span a program boundary. Provider EPG never reaches SQLite, so post-hoc lookup is impossible by design. - Own RECORDINGS_* IPC surface + RECORDINGS_UPDATE_EVENT ping and a separate supportsRecordings capability gate (the supportsDownloads allowlist is all-or-nothing and stays untouched). Reveal/play shell IPCs are gated on the recordings table, so the renderer-supplied recording directory stays a write-location preference, not a shell-access grant. - Manager UI: `recording` filter chip, "Recording now" queue section (REC pulse, elapsed, live file size — no percentage, the length is unknown), 16:9 channel-logo Recordings library, Needs attention with Remove only, focused detail at /workspace/downloads/recording/:recordingId. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> * fix(downloads): close the stop-enrichment race and repair player stubs Greptile spotted a real ordering bug: the stop IPC returns as soon as mpv acknowledges, while the recording row's terminal-state update is still queued in the tracker. The renderer answers that snapshot with stop enrichment, whose handler only accepts a terminal row — so the covered-program metadata could be silently dropped with "Recording not found". - EmbeddedMpvRecordingTracker.whenSettled() exposes the serialized write chain; RECORDINGS_UPDATE_PROGRAMS awaits it before the terminal-row lookup. Regression covered from both sides: the handler must not touch the database until the barrier resolves, and the barrier must imply a committed row. CI also caught spec stubs that had not learned the new player inputs (my local run-many had been an Nx cache hit, so the failures only surfaced in CI): - Teach the `app-web-player-view` and `app-embedded-mpv-player` stubs the `recordingMetadata` input and `recordingStopped` output across the m3u, Xtream, Stalker, unified-live-tab and web-player-view specs. - The races spec now asserts the metadata argument explicitly instead of matching a two-argument call. - Extract the Stalker and unified-live-tab spec stubs into sibling `*.spec-stubs.ts` files (the pattern ui/playback already uses) so both specs stay under the 1200-line test limit without shaving assertions. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> * test(downloads): make the recordings events spec a module The spec deliberately has no static imports — every dependency is swapped through jest.doMock before the harness's dynamic import — which also made it a TS script rather than a module, so its top-level `registeredHandlers` landed in the global scope and collided with the same-named const in stream-probe.spec.ts (TS2451). Local per-project runs compile the specs separately and stayed green; only the Tier A coverage suite builds them into one program, so CI caught it. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> * fix(downloads): address Codex review on recording lifecycle Four findings from the Codex review, all real: - P1: `addon.stopRecording()` only dispatches — native-view uses `mpv_set_property_async`, frame-copy writes a helper command — so finalizing inside the stop hook could stat a file mpv had not flushed and even unlink bytes still being written. The tracker now treats the hook as a request and finalizes on the acknowledged inactive snapshot, with a 10 s bound so a lost acknowledgement cannot strand the row. Only a recording that never went active has its empty reservation removed. Stop enrichment follows through `whenFinalized(targetPath)` (bounded) instead of merely draining the write queue. - Live file size: `file_size_bytes` is written at finalization only, so the manager's 15 s refresh reported nothing while recording. Active rows are now decorated with a current `fs.stat` size. - Manager-initiated Stop bypassed both player stop paths, so recordings spanning program boundaries kept only the start-time program. `EmbeddedMpvPlayerComponent` now owns the active→inactive edge and emits `recordingStopped` for every trigger; the adapter and legacy toggle no longer emit it themselves. - Startup recovery could terminate a row another live instance was still writing under IPTVNATOR_ALLOW_MULTIPLE_INSTANCES. Rows carry `owner_pid` and recovery skips those whose owner process is alive. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> * fix(downloads): derive the enrichment wait from the stop fallback Greptile caught the seam my previous fix left: the enrichment barrier waited 5 s while the tracker's acknowledgement fallback only finalizes at 10 s, so a stop mpv never confirms let the terminal-row lookup expire early and drop the covered programs with no retry — precisely the case the fallback exists for. The wait is now derived from the acknowledgement bound (fallback + 1 s), with a regression test that fails if the two ever drift apart again. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> * fix(downloads): address the second Codex pass on recordings Four more findings, all real: - P1 (macOS native-view): `StopRecording` clears `recordingActive` *before* dispatching the async property set and restores it if the request is rejected, so the first inactive snapshot is optimistic, not an acknowledgement — the tracker could finalize (and stat) a file mpv was still writing, and a rejected stop would leave the row `completed` while recording continued. An inactive snapshot now has to survive a 1.5 s settle window (three poll cycles); a revived recording cancels the pending finalization. - Removing a failed row unlinked its path unconditionally, which takes the file of a newer recording that reused the freed name within the same timestamp second. The cleanup now runs only while no other row claims it. - The All chip and the header's active badge ignored recordings, so a manager holding only recordings read "All 0" and an active recording never showed up in the badge. - Switching channels auto-stops the recording, but by the time the host handled the stop its `activeChannel`/EPG already described the NEW channel, so the old recording was enriched with the wrong schedule (and an unrelated program could be promoted to its title). The stop event now carries the EPG key captured while the recording was active and every host compares it before enriching. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> * fix(downloads): close the persistence race and two recording UX gaps - Greptile P1: the enrichment deadline (fallback + 1 s) still raced the terminal write — if the tracker queue or the UPDATE took longer than the remaining margin, `whenFinalized` returned while the row was still `recording` and the one-shot enrichment was dropped. The deadline now bounds only the wait for mpv; `finalize()` removes the entry synchronously, so once it has started the wait follows the write itself. - Codex: `RECORDINGS_STOP` ignored `owner_pid`. Session ids restart per process, so under IPTVNATOR_ALLOW_MULTIPLE_INSTANCES stopping another instance's row could stop an unrelated local recording. Foreign rows are now refused. - Codex: the In progress chip counted active recordings while its filter deliberately hid them, so clicking it showed "no matches". Active recordings now belong to that filter — a chip whose count disagrees with its page is a lie. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> * refactor(downloads): drop the enrichment barrier instead of tuning it Three review rounds circled the same class: synchronizing mpv's asynchronous stop acknowledgement with a one-shot program enrichment. Each fix moved the deadline (5 s → fallback+1 s → wait-on-the-write) without removing the reason a deadline existed at all — the handler insisted on a *terminal* row. It never needed one. `openSync('wx')` makes the reserved path exclusive while a recording owns it, so the newest row for that path IS the recording that was stopped, and `finalize()` writes only status/end time/size and never `programs_json`. Enrichment and finalization are therefore order-independent: - `RECORDINGS_UPDATE_PROGRAMS` matches the newest row for the path in any status and awaits only the tracker's write queue, which exists solely to guarantee the INSERT committed (a recording stopped milliseconds after it started). - `whenFinalized`, its deadline constant, and the per-entry finalized promise are gone; the tracker keeps only the settle window and fallback that make *finalization* itself correct. No behavior is lost and the whole timing class disappears with the code. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> * fix(downloads): bind recording finalization to its entry and shield live rows from startup repair Two races from the Codex review: - Tracker timers finalized by reusable session id, so a stop followed by an immediate restart on the same session let the old settle timer finalize the NEW row (marked completed while mpv kept writing) and strand the old row in 'recording'. Finalization is now bound to the exact open entry, and replacing a session's entry arms the old entry's settle timer so an unobserved stop still finalizes it. - reconcileStaleRecordings() runs after the renderer is interactive; a recording started during bootstrap has ownerPid === process.pid and was repaired to interrupted/failed mid-write. Recovery now skips rows the tracker reports as actively tracked (activeRowIds()). Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> * fix(downloads): harden recording startup repair against recycled pids and stale renderer lists Second Codex pass on the recovery path: - A live ownerPid alone no longer shields a row: after a crash the OS can recycle the pid for an unrelated process, which would park the row in 'recording' with no instance able to finalize it. Recovery now also checks (best-effort, ps/tasklist) that the process looks like an IPTVnator/Electron instance; an unreadable name stays conservative and keeps the skip. - The renderer loads before the repair pass runs and may already hold the pre-repair list with a stale Stop affordance; recovery now broadcasts one RECORDINGS_UPDATE_EVENT after changing any rows. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> * fix(downloads): defer teardown finalization behind the flush window and bound the live-size stat Third Codex pass: - A synthetic error/closed snapshot from disposeSession() arrives while the frame-copy helper may still be flushing (0.5 s quit grace + 2 s SIGTERM grace before SIGKILL). Finalizing there statted a file mid-write — short captures became terminal 'failed', longer rows persisted a truncated size, and startup recovery could repair neither. The tracker now defers that finalization behind a 2.5 s flush window; the row stays 'recording' (repairable) meanwhile, and an already-acknowledged stop's settle timer keeps its 'completed' verdict instead of being relabelled 'interrupted'. - The active row's live file size used a bare await stat(): one stat hanging on a dead network filesystem wedged every RECORDINGS_GET_LIST. The probe now mirrors the availability probe's contract — in-flight coalescing plus a 1 s deadline degrading to no size. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> * fix(downloads): unmask recycled recording owners, guard the PWA recording route, and unblock file probes Fourth Codex pass: - Recycled-pid discrimination no longer stops at the process-name family check (any Electron app could shield the row): a live holder must also not provably have started after the recording did (ps -o etime= / PowerShell StartTime). A pid frees only when its previous owner dies, so a recycled pid's holder is always younger than the recording; unreadable evidence stays conservative. - /workspace/downloads/recording/:recordingId gets a supportsRecordings capability guard redirecting the PWA to the manager — RecordingsService never becomes authoritative there, so the detail rendered a permanently blank workspace. - Finalization and startup repair stat through a bounded async probe (3 s deadline, ENOENT/ENOTDIR as the only proof of absence) instead of main-thread statSync: a dead network mount no longer freezes the main thread or the tracker queue, repair leaves unjudgeable rows recoverable, and finalization keeps the requested status with an unknown size rather than branding a likely-good file failed. The 0-byte reservation unlink is fire-and-forget for the same reason. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> * fix(downloads): keep inconclusive recording probes out of Needs attention and bound repair batches Fifth Codex pass: - Recording list decoration now uses the bounded availability variant that preserves 'unknown': a timed-out or permission-errored probe is not proof of absence, so a good recording on a slow mount no longer lands in Needs attention with its Play/Reveal hidden. ElectronRecordingItem.fileAvailability widens accordingly; consumers already gate on === 'missing'. - Startup repair probes its whole batch concurrently, so main.ts awaits roughly one 3 s deadline instead of one per stale row. Cross-process ping propagation under IPTVNATOR_ALLOW_MULTIPLE_INSTANCES stays out of scope (debug-only flag, same single-window design as DOWNLOADS_UPDATE_EVENT) — rationale left on the review thread. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> * fix(downloads): fix duration rounding at hour boundaries and bound owner-process probes Sixth Codex pass: - The recording duration formatter rounded minutes after flooring hours, so 59:45 read '60 min' and 1:59:45 read '1 h 60 min'. One shared recordingDurationLabel() now rounds the total minutes before splitting (both the detail page and the library card used a duplicated copy). - Startup repair's synchronous ps/tasklist/PowerShell ownership probes get a 2 s spawn timeout and are memoized per unique pid, so a batch of rows from one crashed instance costs at most one name query and one start-time query, and a hung process query degrades to the conservative fallback instead of blocking the main thread. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> * fix(downloads): return to the manager through history from the recording detail Seventh Codex pass (single finding): with a validated returnUrl the manager is already the previous history entry, so Back now uses Location.back() instead of pushing a third entry that made the browser Back button reopen the detail; router navigation remains the fallback for direct links — matching the offline-detail navigation. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> * fix(downloads): bound removal cleanup and shell gates, date interrupted rows by file mtime Eighth Codex pass: - RECORDINGS_REMOVE no longer awaits an unbounded unlink of a failed row's leftover reservation: cleanup is raced against the 1 s deadline, so a hung network unlink cannot keep the Remove action busy — the row deletion is what matters. - Reveal/Play swap the synchronous lstat gate for the bounded async availability probe: a dead mount no longer blocks the main process, and only PROVEN absence refuses the action — an inconclusive probe lets the shell try and answer honestly. - Startup repair dates an interrupted row's endedAt from the captured file's mtime (mpv's last write) instead of the repair time, so an overnight shutdown no longer inflates a five-minute capture into an hours-long recording; the repair-time fallback remains when mtime is unreadable. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> * fix(downloads): keep recording-start program metadata fresh across EPG boundaries Ninth Codex pass (single finding): the unified live tab's recordingMetadata computed cached its Date.now() verdict — starting a recording after an EPG boundary snapshotted the previous show. It now tracks the existing 30 s progress tick. The Stalker live layout's currentProgram had the same memoization (feeding recording metadata, the EPG panel summary, and external-player metadata); it gains a 30 s clock tick with interval cleanup in ngOnDestroy. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> * fix(downloads): re-select the Xtream current program against the 30 s tick at recording start Tenth Codex pass (single finding): the Xtream live layout's recording snapshot read withEpg().currentEpgItem, a computed whose Date.now() verdict stays cached until epgItems changes — a recording started after an EPG boundary snapshotted the previous show. The selection logic is extracted as the pure findCurrentEpgItem(items, nowMs), the store computed delegates to it unchanged, and recordingMetadata re-selects with the layout's existing 30 s currentTimeMs tick. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> * fix(downloads): scope stop enrichment to the exact recorded list item Eleventh Codex pass (single finding): the stop-enrichment guard compared only the EPG key, which is not unique for M3U items — two list entries sharing a tvgId (or the display-name fallback) could hand the first item's recording the second item's schedule after a switch-triggered auto-stop. RecordingStartMetadata/RecordingStoppedEvent gain an opaque sourceItemKey (unified tab: item.uid; M3U player: channel.id), captured while the recording is active exactly like the EPG key, carried through the player's stop edge, and compared by the hosts before enriching. Xtream/Stalker keys are already playlist+id-scoped and need no extra key. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> * fix(downloads): derive the M3U start-snapshot program from the active channel's schedule Twelfth Codex pass (single finding): the M3U recording snapshot read the NgRx currentEpgProgram, which retains its last value across a channel switch and through EPG gaps (the mirror effect only dispatches when a program exists) — a recording started on a channel with no airing program could persist the previous channel's title, which stop enrichment deliberately never overwrites. The snapshot now derives the program from the active channel's own schedule against the existing 30 s clock, and an EPG gap snapshots no program. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> * fix(downloads): keep finalizing rows in the recovery ledger and guard the repair update Thirteenth Codex pass (single finding): finalize() removes an entry from the open map before its queued terminal update commits, so activeRowIds() briefly omitted a row still persisted as 'recording' — startup recovery overlapping a clean stop could relabel it interrupted, after which the tracker's status-guarded update could not restore 'completed'. Finalizing entries now stay in a dedicated ledger until the update settles, and the repair UPDATE itself is guarded on status='recording' as a second belt against a finalization that commits between recovery's SELECT and its write. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> * fix(downloads): register update listeners before the initial list load Fourteenth Codex pass (single finding): RecordingsService awaited its initial RECORDINGS_GET_LIST before subscribing to the update ping — a recording transition during that request pinged into the void while the response still reflected the pre-transition state, and recording pings are rare enough that nothing self-healed until the 15 s poll (armed only once an active row is visible). The listener now registers first so the load-state coalescing queues the trailing refresh. DownloadsService had the same latent window and gets the same reorder. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> --------- Co-authored-by: Claude Fable 5 <noreply@anthropic.com> Co-authored-by: 4gray <fourgray@proton.me>	2026-08-23 08:05:25 +02:00
4grayandClaude Fable 5	00aa623b83	fix(downloads): reconnect interrupted transfers and resume without validators (#1446) ... * fix(downloads): reconnect interrupted transfers and resume without validators Xtream panels commonly kill each VOD connection after a byte/time burst (~130-260 MB) and send no ETag/Last-Modified. The validator-only resume path then deleted the partial and surfaced a raw "aborted" failure, so every Retry restarted from zero and large files could never finish. - Resume without a validator through overlap verification: the Range request rewinds by 256 KiB and the replayed window must match the partial's tail byte-for-byte before anything is appended; a mismatch truncates the partial and restarts from scratch (download-overlap.ts). - Reconnect automatically on recoverable interruptions and clean short responses (download-reconnect.ts): progress >=64 KiB past the best attempt resets a 3-stall budget; request-phase failures during reconnects are converted into retained interruptions so automation can never delete a partial. - Extract pure response-header helpers into download-resume-validation.ts to keep download-transfer.ts within the file-size guideline. Verified against a real throttling portal: a 1.6 GB and a 3 GB movie completed through 15 and 23 connection resets respectively. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> * fix(downloads): address review findings on reconnect baseline and small partials - Judge reconnect progress against the previous attempt instead of a high-water mark, so a transfer that legitimately restarted from byte zero mid-loop (overlap mismatch, ignored Range) is measured by its rebuilt file; at most two such regressions are tolerated per transfer to keep the loop structurally bounded (Greptile P1). - Floor reported progress at the partial's retained size while appending, so a response that ends inside the overlap window can never move persisted progress backwards (Greptile P1). - Verify partials smaller than the overlap window in full from byte zero and append, instead of rewriting the .part in place — an early-dying reconnect can now only grow the file (Codex P2). Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> * fix(downloads): keep tolerated regressions off the reconnect stall budget A tolerated restart regression consumed a regression credit AND counted as a stalled attempt (its negative delta is below the progress threshold), so a legitimately rebuilt file that grows in sub-64 KiB steps was failed one reconnect early. Regressions are now charged to their own bounded budget only; the stall budget stays reserved for attempts that genuinely fail to grow the file. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> * fix(downloads): gate success and validator promotion on complete overlap verification Round-3 review findings: - Success is rejected while the overlap verifier has not consumed its entire window: a complete 206 that ends inside the window proves the remote entity shrank, so the transfer truncates and restarts from scratch instead of finalizing the old suffix as a completed file; an early-dying stream stays an ordinary retained interruption (Codex P1). - A verify-append attempt promotes the response's ETag/Last-Modified only after the complete overlap matched — an unverified partial is never blessed with a validator the next resume would If-Range-append onto (Codex P1). - A tolerated restart regression resets stalls accumulated against the discarded representation, so the rebuilt file starts with the full stall budget (Greptile P1). Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> * fix(downloads): carry the known total through total-less reconnect responses A resumed response without a usable total (chunked, or an unsatisfiable Content-Range) erased task.totalBytes, so a reset over that response could no longer classify as a retained interruption and generic cleanup deleted the verified partial. The previously learned total is now carried forward for appending attempts; fresh and restarted transfers still drop it, since it described a discarded file. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> * fix(downloads): never fabricate a total for an unverified retained partial A retained failure with an unknown total persisted totalBytes = bytesDownloaded, so after stalled reconnects over a chunked, validator-less response that kept ending inside the overlap window, Retry's completed-partial shortcut saw the .part size equal the fabricated total and finalized the unverified partial without a request. The fallback is now explicit per call site: only a finalization failure after a complete transfer records its byte count as the total; retained interruptions keep an unknown total unknown, forcing Retry to re-verify. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> * fix(downloads): treat an unsatisfiable resume range as a representation change A range-capable server whose entity shrank below the rewound overlap offset answers 416 before any response body exists, which rejected the request into the generic partial-deleting failure path. The 416 is now recognized as a representation change: the partial is truncated and the transfer restarts against the current entity from byte zero. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> * fix(downloads): treat an indeterminate Content-Range total as unknown `Content-Range: bytes 200-299/*` fell through to the Content-Length fallback, deriving a "total" equal to the end of the selected range — a resumed response ending there was declared complete and the truncated partial finalized. An indeterminate total now yields null, letting the previously known total carry forward and classify the short response as a retained truncation instead. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> * fix(downloads): signal restarts explicitly and keep carried totals informational Round-8 review findings, both fixed at the design level instead of another byte-comparison patch: - Restart epochs (Greptile P1): the transfer layer now reports every rewrite-from-zero via task.transferRestarts (overlap mismatch, shrunk entity, 416, ignored Range), and the reconnect loop opens a fresh progress epoch on that signal — clean stall budget, no baseline. Byte inference could not recognize a rebuild landing near the previous attempt's count; the explicit signal can. Two restarts are tolerated per transfer; an unsignalled regression is an ordinary stall. - Authoritative vs informational totals (Codex P1): completion and truncation decisions now use only the response's own total or its advertised indeterminate range end (`bytes X-Y/*` -> Y+1); a carried total is informational, is dropped once the bytes on disk falsify it, and can flag a short transfer but never authorize finalization. A mid-reset 206 retains the partial even with a falsified total — the response proved range capability — persisting the total as unknown. Also raises the resume spec's jest timeout and tightens its polling: the previous 5 s default flaked on starved CI runners. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> * fix(downloads): keep unproven state fully uncommitted across the resume model Round-9 review findings, closing the remaining commit-before-proof gaps: - The response's total now stays uncommitted (task and row) until the complete overlap matched, exactly like the validator: a persisted total equal to the unverified partial's size let the completed-partial shortcut finalize unproven bytes after a pause, crash, or retained failure (Codex P1). - Retained-interruption persistence syncs the live task with the row: a stale falsified total left in memory made the next reconnect's resume-offset guard reject the retained partial into generic cleanup (Codex P1). - An observed 206 is remembered as task.serverAcceptsRanges, so a request-phase failure (no response at all) can retain an unknown-length partial on that evidence instead of deleting it (Codex P2). Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> * fix(downloads): treat a reset after the final ranged byte as completion A 206 that delivers every advertised byte but ends in ECONNRESET instead of a clean close was classified as an interruption with a falsified total; the reconnect then resumed at EOF, collected a 416, and truncated the complete file — an endpoint that always resets after its last byte could never finish. A retainable failure whose on-disk bytes exactly equal the response's own completion evidence (and the proven total, with the overlap verified) now completes the transfer. Strict equality keeps oversized partials on the generic-failure path. Also splits the transfer error classes and retention classification into download-transfer-errors.ts to stay under the max-lines rule. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> * fix(downloads): never let an indeterminate range end authorize completion Reaching Y of `Content-Range: bytes X-Y/*` proves the selected range was delivered, not that the entity ends there — a range-capping server resetting at its cap would have finalized a truncated movie as complete. The reset-after-final-byte completion now requires the response's authoritative total; indeterminate range ends keep flagging short delivery but such resets stay retained interruptions. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> * fix(downloads): retain any nonempty partial and keep indeterminate ranges incomplete Round-12 findings, resolved by removing the last evidence requirements: - Retention no longer needs a total, validator, or range proof: since overlap verification owns resume correctness, the next attempt can safely prove, resume, or restart over ANY retained partial — deleting bytes is the only unrecoverable outcome. This closes the whole family (refused reconnects, chunked responses, falsified totals, and reconstructed retry tasks losing the in-memory range flag) and removes task.serverAcceptsRanges entirely (Codex P2). - A clean EOF at the advertised end of an indeterminate range (`bytes X-Y/*`) stays incomplete, matching the reset path: reaching Y proves the range was delivered, not that the entity ends there, so a range-capping server can no longer finalize a truncated movie; the shrunk-entity restart likewise requires an authoritative total (Codex P1). Responses with no range and no total keep the clean-EOF completion contract of unknown-length HTTP. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> * fix(downloads): honor 416-confirmed completion and keep proven validators Round-13 findings plus a CI limit: - A 416 whose `Content-Range: bytes */N` equals the partial's size confirms the file IS the complete entity (under If-Range a validator mismatch yields 200, so the 416 also confirms identity): finalize it instead of truncating and redownloading forever (Codex P1). - A validator proven by a complete overlap match is now promoted on the error path too, and retained-failure/pause persistence write resume_validator from the task — later attempts resume via If-Range instead of replaying the 256 KiB window, which stalled out servers whose per-connection cap barely exceeds the window (Codex P1). - download-resume.spec.ts crossed the 1200-line test limit: the shared harness moves to download-resume.test-harness.ts and the overlap-family cases into download-overlap-resume.spec.ts; the runtime spec gets the same CI-load timeout headroom as the resume specs. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> * fix(downloads): settle falsified totals on clean exits and probe EOF after zero-growth replays Round-14 review findings plus the CI build break: - download-resume.test-harness.ts was outside the tsconfig test-helpers exclude glob and broke every app typecheck/build; renamed to download-resume.test-helpers.ts (the excluded pattern). - A clean indeterminate delivery that outgrows a stale carried total now settles that total to unknown on the row AND the live task before persisting or throwing — the reconnect's resume-offset guard would otherwise reject the retained partial into generic cleanup (Codex P1). - A verified overlap replay of an indeterminate range that appends nothing arms a one-shot EOF probe: the next attempt requests the byte after the partial so a compliant 416 (bytes */N) can confirm the file is the complete entity, instead of repeating the rewind until the stall budget fails a finished download; a probe answered with more data is retired unappended and rewound verification resumes (Codex P1). Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> * fix(downloads): retain the partial on an inconclusive EOF-probe 416 A probe at the partial's exact end always collects a 416 when the entity ends there, and the confirming Content-Range length is optional — so a length-less 416 is equally consistent with a complete file, and the unconditional restart redownloaded a likely finished movie every cycle. The probe's 416 now restarts only when a stated total BELOW the partial proves the entity shrank; otherwise the partial is retained. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> * fix(downloads): treat any EOF-offset 416 as inconclusive, not just the probe A validator-backed resume at the partial's exact end IS an EOF request: when its 416 arrived without the optional Content-Range length, the probingEof gate saw false and restartFromScratch truncated the complete partial. The inconclusive-416 retention now keys on the request having started at the partial's end (covering the probe and every If-Range resume alike); a restart still requires a stated total below the partial. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> * fix(downloads): contradiction-proof carried totals and identity-gated 416 completion Round-17 review findings: - A carried total now falls the moment the response's advertised range end contradicts it, before any byte lands — waiting for the bytes to reach it left a pause/exit window where an N/N row let the completed-partial shortcut finalize a truncated file (Codex P1). - The 416 completion shortcut now requires identity proof: an exact-EOF request backed by If-Range, or the EOF probe that follows a fully verified overlap replay. A bare length match on a rewound request proves nothing about whose bytes are on disk; a contradictory 416 (stated total says the rewound range was satisfiable) retains instead of restarting (Codex P2). Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> * refactor(downloads): extract 416 classification to stay under max-lines The previous commit pushed download-transfer.ts to 404 effective lines; the 416 decision moves into classifyRangeNotSatisfiable() in download-transfer-errors.ts with identical semantics, and the unused re-export-only imports are dropped. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> * fix(downloads): promote the proven total on the error path and fix the 416 contract doc - The error-path promotion after a complete overlap match now updates task.totalBytes alongside the validator: a pause landing while the partial sits at a stale carried total otherwise persisted an N/N row that Resume's completed-partial shortcut would finalize (Codex P1). - The download-manager contract doc's rewound-416 paragraph now matches classifyRangeNotSatisfiable(): completion requires identity proof at exact EOF, restart requires a proven shrink, everything ambiguous retains (Codex P1 on the doc). Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> * fix(downloads): correct both 416 classification boundaries Round-19 review findings, one in each direction: - A rewound 416 WITHOUT a stated length now retains: unsatisfiability alone never proves the entity shrank relative to the retained bytes, and the canonical contract reserves restart for stated proof (Codex P1). - A stated total EQUAL to a rewound request's first byte now restarts: the entity ending exactly at the rewound offset makes the 416 valid and proves the partial extends past the entity — equality was being misread as a contradiction, stranding the download in retain forever (Codex P1). classifyRangeNotSatisfiable() gains an exhaustive pure table spec covering every branch. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> * fix(downloads): restart on reset-ended responses that completed a shorter entity in-window The clean-EOF path already restarted when a response delivered its complete authoritative total inside the verification window, but a reset arriving right after that final byte took the retention path and stranded the oversized partial in a stall loop. The catch path now mirrors the shrink detection: authoritative total reached inside an unproven overlap restarts from scratch. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> * fix(downloads): request identity encoding for byte-exact transfers Axios's Node adapter transparently decodes gzip/brotli responses, which would put decoded bytes on disk while Content-Length, Content-Range, and every Range offset speak the encoded representation — desyncing resume offsets and overlap verification on origins that compress. Downloads now always send Accept-Encoding: identity. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> * fix(downloads): drop carried totals an indeterminate range can exactly reach `bytes 200-249/*` can deliver the partial exactly TO a carried total of 250; the <= guard kept that total, and a pause or exit anywhere in that window persisted a 250/250 row the completed-partial shortcut would finalize without EOF proof. The guard is now strict: a carried total survives only when the advertised indeterminate range cannot reach it, which also closes the mid-stream pause window. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> * fix(downloads): arm the EOF probe after a reset-ended verified zero-growth replay A verified overlap replay that reset right at the partial's end (zero growth, indeterminate range, no validator) retained without arming the EOF probe, so every retry replayed the same tail until the stall budget expired — the clean-EOF path's probe arming now has its reset-path mirror. Also splits to stay under max-lines: the classifyRangeNotSatisfiable table spec moves to download-transfer-errors.spec.ts and the DB persist helpers to download-transfer-persistence.ts. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> --------- Co-authored-by: Claude Fable 5 <noreply@anthropic.com>	2026-08-16 07:07:25 +02:00
4grayandClaude Fable 5	7fc9380bff	feat(portals): mark a full season as watched in one click (#1447) ... * feat(portals): mark a full season as watched in one click Series detail pages on both Xtream and Stalker portals get a season-level watched toggle next to "Download season": marking writes full-progress rows for the unwatched episodes only (real durations survive), a fully watched season flips the action to unwatch-all. Persistence goes through new batch IPC channels (DB_SAVE/CLEAR_PLAYBACK_POSITIONS_BATCH, one SQLite transaction with onConflictDoUpdate().run(); the PWA data source rewrites its localStorage blob once). Stalker deliberately bypasses the batch IPC and loops the existing position-mutation queue so legacy-row reconciliation still runs and the queue coalesces to a single reload; partial failures surface a dedicated snackbar. Also removes the dead toggleEpisodeWatched store method, splits season-container/serial-details-playback under the max-lines cap (season-watch-toggle.util.ts, SerialDetailsSeasonWatchService), and classifies *.spec-data.ts fixtures under the test max-lines ceiling (baseline shrinks by main.preload.spec-data.ts). Closes #1442 Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> * fix(portals): guard stale season batches and split partial-unwatch feedback Review follow-up (Codex on #1447): - A season batch completing after the user navigated to another series or playlist no longer writes the old series' rows into the freshly reset position state (episode ids can collide across playlists); the Xtream host captures the playlist/series identity before awaiting and skips the rendered-state mutation when it changed. The DB write is unaffected — it carries its own playlistId. - A partially failed "mark season as unwatched" on Stalker now reports a dedicated SEASON_MARKED_UNWATCHED_PARTIAL message instead of the watch-direction "marked" text; translated into all 18 locales. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> * fix(portals): exclude the playing episode from season marking and count partial saves Second review round (Codex on #1447): - The episode currently playing (inline or in an external session, or with a launch in flight) is excluded from a season's mark-watched batch: the player persists its live position every ~15 s and would immediately overwrite the just-written full-progress row. The button count reflects the exclusion and the action disables when nothing is markable. Unmarking still clears such an episode — the recreated in-progress row reflects live playback truthfully. - A Stalker StalkerSeriesPositionPartialSaveError (scoped watched row saved and published, only legacy cleanup failed) now counts as a watched success instead of feeding false total-failure feedback. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> * fix(portals): gate stale season-batch snackbars on the originating page Third review round (Codex on #1447): a batch resolving after the user navigated away no longer shows its contextless success/error snackbar on the newly opened detail page — the same ownership check that guards the state mutation now guards the feedback too. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> * fix(portals): sync catalog progress badges after toggles and gate Stalker feedback Fourth review round (Codex on #1447): - Any Xtream watched toggle (single episode or season batch) now refreshes XtreamStore.loadAllPositions after persisting — the catalog reads series-progress badges from the store, which otherwise loads positions once per playlist, so returning from the detail kept stale badges. Skipped when the playlist changed mid-flight (the store then belongs to the other playlist; its own init reloads positions). - Stalker's season snackbars are gated on the captured playlist/series identity, matching the Xtream ownership guard — a batch draining after navigation no longer reports on the newly opened page. - Stalker season-toggle specs moved to stalker-series-view.season-watch .spec.ts with their own harness; both prior spec files sat at the 1200-line test ceiling. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> * docs: describe the season watched toggle in CLAUDE.md Fifth review round (Codex on #1447): the canonical Seasons entry in the VOD/Series detail section now covers the bulk toggle, its playing-episode exclusion, both persistence paths, catalog badge sync, and the stale-completion contract. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> * fix(portals): let only the latest positions load patch the Xtream store Sixth review round (Codex on #1447): loadAllPositions is now latest-load-wins — a fetch superseded while in flight (playlist switch before getAllPlaybackPositions resolves) no longer patches the singleton store with the previous playlist's position maps, which could leave the new catalog showing the old playlist's progress badges. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> * docs: reflect the spec-data max-lines classification in CLAUDE.md and AGENTS.md Seventh review round (Codex on #1447): both canonical max-lines descriptions now list **/*.spec-data.ts among the test-ceiling globs so future agents neither treat these fixtures as production files nor remove the exemption unknowingly. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> * fix(portals): parse "N min" durations when marking episodes watched Eighth review round (Codex on #1447): Stalker VOD episodes report durations like "45 min", which parseDuration could not read — bulk (and single) mark-watched then persisted 1/1-second rows. The minute format now parses to seconds, matching what the removed legacy store method already handled. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> * fix(portals): parse compound hour durations and cover the toggle end-to-end Ninth review round (Codex on #1447): - parseDuration now reads the compound "1h 30min" form the Xtream fixtures emit (hour group optional, so "45 min" keeps working) — bulk-marked episodes no longer persist a minutes-only duration. - New Playwright coverage exercises the season toggle through the real UI on both portals: Xtream (category → series detail → mark → reload-persistence → unmark) and Stalker (embedded-series flow, mark → unmark with the item's actual episode count). Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> * fix(portals): refresh Stalker catalog progress badges after watched toggles Tenth review round (Codex on #1447): the Stalker mirror of the Xtream catalog sync — StalkerCatalogFacadeService loads its position maps once per playlist and the runtime bridge only pushes external-player updates, so renderer-initiated toggles left grid badges stale. The series view now calls the facade's new ownership-checked refreshPositions after the season batch (including partial successes) and after single toggles; the reload is latest-load-wins like the Xtream store fix. Optional injection keeps collection-detail mounts outside the catalog working. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> * test(portals): cover the season toggle batch IPC end-to-end in Electron Eleventh review round (Codex on #1447): the new Electron E2E marks a season through the real UI, asserts the eight SQLite rows written by DB_SAVE_PLAYBACK_POSITIONS_BATCH directly through the preload bridge, proves persistence with a full app relaunch (renderer and main process die, so state can only come from the database file), and clears again through DB_CLEAR_PLAYBACK_POSITIONS_BATCH back to zero rows. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> * fix(dashboard): keep watched rows out of the series resume target Twelfth review round (Codex on #1447): a watched position row — a natural finish or a manual/bulk "mark watched" marker — is a completion record, not resumable progress. Continue Watching no longer auto-plays such an episode at its end; the handoff stays detail-only and the series page's quick-start picks the first unwatched episode instead. Card progress bars and SxxEyy badges keep their current source. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> * fix(portals): fail closed on refresh reads and gate batch APIs by capability Thirteenth review round (Codex on #1447): - Position-cache refreshes now use a failure-propagating read (getAllPlaybackPositionsOrThrow through the Electron data source): a transient IPC failure rejects instead of masquerading as an empty list, so a populated store/facade cache stays stale-but-populated rather than being wiped. All load/refresh call sites handle the new rejection (init loads may retry on the next activation; post-toggle refreshes log and keep the snackbar flow). - The season-batch bridge methods joined playbackPositionStorageMethods, so a bridge lacking them degrades to the in-memory path wholesale instead of throwing mid-action. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> --------- Co-authored-by: Claude Fable 5 <noreply@anthropic.com>	2026-08-15 21:03:14 +02:00
4grayandClaude Opus 5	e94cc029eb	fix(portals): time out and fast-fail PWA proxy requests to dead hosts (#1424) ... * refactor(portals): hoist the connectivity guard into libs/shared/host-health The breaker was written dependency-free so both processes that talk to portals could share it. Move the part that has no Electron in it — the state machine, the failure classification, the redirect-attribution helpers and the fast-fail error — into `@iptvnator/shared/host-health` (`scope:shared` / `domain:shared-runtime` / `type:util`). What stays in `apps/electron-backend` is the genuinely main-process part: one guard for the whole process, so both portal IPC handlers see each other's evidence, and the console warning that announces it. Every call site is unchanged; the wrapper re-exports the two types they import. The spec splits the same way — the state machine moves with the class, the singleton and its redirect attribution stay with the wrapper. Register the new project in the coverage policy, which every project with a test target must declare a tier for. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> * fix(portals): time out and fast-fail PWA proxy requests to dead hosts The web backend's proxy routes were bare `axios.get()` calls with no `timeout`, so a provider that accepted a connection and then went silent held the request until the OS gave up on the TCP connection — minutes, rather than the 15/30 s budget the Electron handlers use. Add the same per-route timeouts (Xtream 30 s, Stalker 15 s / 30 s for `create_link`, playlist and XMLTV 30 s). Those numbers are safe for large downloads: on axios' default transport `timeout` bounds the time to response headers and then continues as the socket's inactivity timeout, so a multi-megabyte XMLTV file that keeps delivering bytes is never cut off — only a stalled one is. With requests bounded, run `/xtream` and `/stalker` through the shared breaker, injected via `WebBackendAppOptions.hostGuard` so specs drive it with a clock they own. Playlist and XMLTV downloads keep the timeout but no breaker, matching Electron: a download is one request rather than a catalog fan-out, it is usually the direct result of the user asking for it, and it can outlive the half-open trial window. The breaker is checked before the Xtream URL revalidation, which resolves the hostname — a dead host is where DNS is slow too, and a request admitted and then abandoned by the URL policy hands its token back rather than holding the trial slot. `resetHostConnectivityGuard()` no longer no-ops in the PWA: the breaker lives in the backend process, so it travels to a new `POST /connectivity-guard/reset`, which reads only the origin and never logs the credential-bearing URL. `skipConnectionGuard` now survives the PWA transport too, so Stalker endpoint discovery keeps the exemption it has on the desktop instead of tripping the breaker with its own probes. A fast-fail keeps each route's HTTP 200 `{message, status}` envelope. The Stalker path needs one extra step: `forwardStalkerRequest` turns that envelope into `HTTP Error <code>: …` with a numeric `status`, and the renderer reads both as "the endpoint answered" — which would make discovery walk every candidate and fire lazy repair at a host just declared dead. A prior branch keyed on the shared `isHostConnectivityFastFailMessage` rethrows it bare instead. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> * fix(portals): report exempt Stalker probe responses to the PWA breaker Flagged by the author of #1421 as one of the twelve fixes that landed there after this branch cherry-picked the pre-review commit: an exempt discovery probe must still REPORT, it just must not COUNT. The web backend was skipping the report entirely for a probe, which loses the case that matters. A failure carrying an HTTP response proves the endpoint answered, and this route sets no `validateStatus`, so axios rejects every non-2xx with `error.response` attached — a probe answered with 404 or 500 was therefore dropped instead of clearing the record. Two counted failures either side of it then read as consecutive and opened the breaker in the middle of discovery, which is exactly what the exemption exists to prevent. `reportProviderRequestFailure` now takes `countFailures`, matching the Electron reporter, and both routes always report. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> * fix(portals): read the redirect hop from the transport that followed it `failedAfterRedirect` decided whether a failure belonged to a redirect destination by reading `error.config.url`. That is right for the Electron transport, which sets `maxRedirects: 0` and reissues every hop as its own request, so the hop IS the config URL. It is blind on the web backend, which uses axios' default transport: follow-redirects walks the chain inside one request and `config` is built once, so `config.url` stays the URL we asked for. Verified against the installed axios 1.19.0 with a live server that 302s to a dead port: asked for : http://127.0.0.1:63953/player_api.php config.url : http://127.0.0.1:63953/player_api.php request._currentUrl: http://127.0.0.1:1/dead So the comparison was original-vs-original, found no redirect, and charged two dead destinations to the provider that had answered both times with a 302 — then fast-failed it. Read `request._currentUrl` first and fall back to `config.url`, which covers both transports; Electron's native per-hop requests expose no `_currentUrl` and are unaffected. Also check redirect attribution BEFORE suppressing failure counting for an exempt probe. A 3xx from the guarded endpoint is an answer, so a probe that observed one must clear the record; otherwise a timeout, a probe redirected to a dead destination, and another timeout still read as two consecutive failures. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> * fix(portals): stop axios query params reading as a redirect Codex found that the Xtream breaker never opened at all, and it was right. The web backend passes credentials and the action through axios' `params`, so axios sends `…/player_api.php?username=…&action=…` while the baseline handed to `failedAfterRedirect` is the query-less URL the route built. Verified against axios 1.19.0 with a plain ECONNREFUSED and no redirect anywhere in sight: baseline : http://127.0.0.1:1/player_api.php request._currentUrl: http://127.0.0.1:1/player_api.php?username=demo&… The two normalized URLs differ, so every ordinary failure looked like a post-redirect failure, credited the endpoint, and the breaker could never trip. Compare origin and path, not the whole URL. That keeps what the check is for — an endpoint that answered and sent us elsewhere, including the same-origin `/player_api.php` → `/slow/player_api.php` case — and gives up only a redirect that changes nothing but the query, which is then counted as an ordinary failure. Erring towards counting is the safe direction here. The reason 57 tests passed over a dead feature is the real lesson: `StubHttpClient` threw bare `Error`s, so the guard's redirect check saw neither `config.url` nor `request._currentUrl` and quietly did nothing. The stub now shapes its rejections like axios does, including the query axios appends. With that alone, four existing tests fail against the old comparison. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> * fix(portals): count a hostname that stops resolving, and fix a stale docblock Two from review, one behavioural and one documentation. A name that will not resolve is the host failing to answer — the same evidence as the ENOTFOUND the transport would have raised a moment later. But the SSRF validation turns a lookup failure into a 400 "host could not be resolved", and the release path added earlier handed the token back as inconclusive, so the breaker could never open for a host whose DNS died and every request kept paying for the same dead lookup. `ProviderUrlError` now carries the underlying lookup error internally. A refusal that has one is counted; a genuine policy refusal — private address, bad scheme, credentials in the URL — still only releases the half-open slot, because that says nothing about reachability. The field is internal: `providerUrlErrorBody()` strips it at both call sites, so the client sees exactly the body it saw before, which the test asserts. The docblock on `resetHostConnectivityGuard` still said the PWA channel is unknown and the call no-ops. That stopped being true when this branch implemented `CONNECTIVITY_GUARD_RESET` over HTTP, and a stale contract there is how the next caller silently skips the PWA path. (The edit was in an earlier commit and was lost when the branch was rebuilt on master.) Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> * docs(portals): record the transport-specific redirect contract The redirect section still described one transport: hop-by-hop requests, `error.config.url`, whole-URL comparison. Two of those three are now wrong for the web backend, and this document is the canonical contract — leaving it stale is how the attribution bugs fixed in the last two commits get reintroduced. Says what is actually true: which field holds the failed hop on each transport and why the helper reads both, and that the comparison is origin + path because the web backend's credentials ride in axios' `params` and a whole-URL comparison therefore reported a redirect for every ordinary failure. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> * docs(portals): scope the guard summary to both processes The opening line still defined the breaker as an Electron main-process concern, which contradicted the ownership section below it and is the part a reader skims to decide whether the document applies to them. Names both processes, and records that the web backend had the worse version of the problem first — no request timeout at all — since that is why the timeouts and the breaker had to land there together. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> * fix(portals): declare the shared-interfaces dependency of host-health The new library's manifest listed only `tslib`, but its emitted JavaScript does `require('@iptvnator/shared/interfaces')` — the guard builds its fast-fail message with `buildHostConnectivityFastFailMessage`. Anything resolving the built artifact from its own manifest would have failed with MODULE_NOT_FOUND. The manifest was copied from `shared/logging`, which imports nothing across libraries and therefore needs nothing beyond `tslib`. `shared/m3u-utils` is the right precedent: it imports the same library and declares `"@iptvnator/shared/interfaces": "0.0.1"`. Verified against the build output rather than by inspection — the emitted `host-connectivity-guard.js` requires the module, and the generated `dist/libs/shared/host-health/package.json` now declares it. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> * fix(portals): bound the PWA connectivity-guard reset The reset was a bare `fetch` with no timeout, which is the exact failure this change exists to remove, reintroduced one layer up. Every caller awaits the reset BEFORE issuing the request it is clearing the way for — `retryContentInitialization` awaits it first by design — so a backend or reverse proxy that accepts the POST and then goes quiet would leave Retry doing nothing at all, for as long as the socket stayed open. Bound it with an AbortController and a 5 s timer. The abort rejects, `resetHostConnectivityGuard` swallows it as it already does for any other failure, and the caller proceeds to its real request — which is what "best effort" was supposed to mean. The timer is cleared in a `finally`, and it covers the body read as well as the headers. Five seconds because this talks to the user's own backend rather than a provider: it should answer immediately, and a slow one must not hold up the retry that asked for it. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> --------- Co-authored-by: Claude Opus 5 <noreply@anthropic.com>	2026-08-13 22:27:05 +02:00
4gray	61fca6f016	fix(dashboard): reuse the detail view's TMDB identity for activity rows (#1423) ... An Xtream activity row is built from its `content` row, and the catalog endpoints that create those rows carry only a title and a poster. So the dashboard hero and the recommendations rail rebuilt their TMDB query from the display title alone, while the detail view had searched with the original title, the release date and often a TMDB id. Without a year `pickConfidentMatch` requires a globally unique exact title, which common titles never satisfy — "Inside Out" matches several films and resolves to nothing, every time. Three `content` columns close that gap next to the existing `backdrop_url`: `tmdb_id`, `release_year`, `original_title`. The detail views back-fill them from what is on screen, the activity SELECTs project them, and `buildDashboardTmdbAttempts` reads them back. Stalker keeps stating the same facts through its stored entry, and rows with neither keep the title-only fallback. Measured against a real profile before building: of 58 distinct Xtream movie/series activity rows, 16 (28%) produce a year-less key — the cohort where a miss is guaranteed rather than likely. Contracts worth preserving: - Per-column, never overwrite. Enrichment supplies the pieces at different times, so a row-level guard would let the first arrival block every later one forever. - `release_year` is the year the PROVIDER stated. The TMDB merge fills the date field when the provider left it empty, so it marks its own substitution with `tmdb_supplied_release_date` and the extractor skips those — making contamination structurally impossible rather than avoided. - The id is stored unvetted: every consumer re-gates it through `assessProviderId`, which re-decides per lookup where a write-time verdict would be permanent. - No media-type column — for Xtream the catalog files movies and series apart, so `content.type` already is the media type. Worker requests now await `getDatabase()` before dispatching. The renderer loads before `initDatabase()` and the worker opens the database file without running migrations, so a query issued during startup on an upgraded install could otherwise hit a schema whose new columns do not exist yet. Not covered: the PWA, whose catalog cache is rebuilt from the API on every load, so a stored id would never outlive the detail view that resolved it.	2026-08-13 18:51:26 +02:00
4gray	0cba49f3e2	fix(dashboard): keep every catalog row per title key when matching (#1425)	2026-08-13 07:31:59 +02:00
4gray	e3f72f7dce	perf(portals): fast-fail requests to portal hosts that stopped answering (#1421)	2026-08-13 07:31:23 +02:00
4grayandClaude Fable 5	5f4d219d3a	refactor(portals): compact credential-free logging for failed Stalker requests (#1418) ... * refactor(portals): log failed Stalker requests as compact credential-free summaries Failed STALKER_REQUEST handlers dumped the entire axios error object (config, request internals, agent state, stack) into the main-process console — ~100 lines per dead-portal request. Extract the existing compact Xtream error formatter into a shared portal-request-error util (action, host, pathname, code, status, message — query string never included) and use it from both handlers. The redundant pre-throw "[StalkerEvents] HTTP Error" line is dropped; HTTP >=400 already reaches the catch block and is now logged once, compactly. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> * fix(portals): never retain an unparseable request URL in error logs The URL guard in stalker.events.ts rejects credentialed portal URLs before the request URL is built, so a malformed row value such as "http://user:secret@" reaches the error formatter as-is. Its fallback copied that raw string into `pathname`, and redactSensitiveData only sanitizes userinfo of URLs it can PARSE — an unparseable one passes through verbatim, password included. Withhold the URL entirely in that branch. Verified: the added regression test fails on the old fallback and passes with the marker. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> --------- Co-authored-by: Claude Fable 5 <noreply@anthropic.com>	2026-08-12 17:54:23 +02:00
4grayandClaude Fable 5	36c2867d36	feat(xtream): recognize more language tags in VOD multi-source (#1417) ... * feat(xtream): recognize more language tags in VOD multi-source The sources popover's language filter and copy chips now read prefixes with Unicode pipe lookalikes, brackets and spaced dashes, Cyrillic tags and MULTI. When a stream title carries no tag, the language falls back to what the stream's visible categories unambiguously state ("EN | Netflix") — discovery aggregates category names per (playlist, stream) in SQL, and category prefixes must pass a known-language gate because everyday category words like new/top/hot are real ISO 639-3 codes. Both signals stay parsed guesses: browse filter and chips only, never ranking, failover or dub-warning inputs. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> * fix(xtream): address Codex review on multi-source language detection Gate the new bracket and dash title forms through isKnownLanguageTag: those positions carry quality/rip tags ([HD], [CAM], NEW -) whose fabricated "language" would outrank and mask a real category-derived one. The legacy pipe form stays permissive. Overlay a late-arriving route category onto the existing route row in the same-key refresh path — cold/direct routes load categories after discovery, and the category is outside the movie key on purpose. The mid-flight case is redelivered by the bind() effect re-running on the controller's sources signal; that tracked read is now documented as load-bearing and pinned by a session spec. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> * fix(xtream): pair brackets and strip the new tag forms when matching Greptile: the bracket prefix chose its opening and closing delimiter independently, so a malformed "[EN)" was read as a language tag. Codex: recognizing a prefix is only half the job — normalizeTitleKeys has to strip the same tag, or the tagged copy never matches the bare one and multi-source cannot offer the film at all. Its leading-tag rule now shares the pipe-lookalike set and, on the pipe branch only, takes the same Latin+Cyrillic any-case alphabet with no required trailing space. Dash and colon keep their uppercase-Latin spaced form: those are ordinary punctuation, and loosening them would amputate "ОНО: Часть 2" the way a case-insensitive rule amputates "It: Chapter Two". Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> * fix(xtream): keep normalization uppercase-only, measured on real catalogs The previous commit widened the pipe branch of normalizeTitleKeys to any case and to Cyrillic, on the theory that nothing but a tag precedes a pipe. Checked against 1.27M real catalog titles that theory is wrong in two ways at once: "Akira | 1988" and "Coco | 2017" put the film's name before the pipe and the year after it, and Russian catalogs write "Момо | Momo" — localized title, then original. The widening corrupted 349 keys and rescued none, so it is reverted. What survives is what the data supports: the pipe-lookalike set (0 changed keys, and correct for panels that use them) and dropping the required space after a pipe (35 changed keys, genuine welded tags like "EN|Dark Shadows" and "|FR|VO|Le dernier empereur"). A leading-tag guard that refused to strip when no letter remained is also dropped: it fixes "AKA | 2023" but breaks "IT - 65", so telling those apart needs a tag vocabulary and belongs in its own change. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> * docs(xtream): cite the measured evidence for the category language gate The gate's rationale named hypothetical category shapes. On a real catalog the four it actually turns away are VOD (5,245 movies), KIDS (1,010), SHOW and WWE — without it the language select offers "VOD" and "KIDS" as languages. Comments, doc and one spec case only. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> * docs(xtream): restore the docblock currentSourceRow lost to an insertion routeCategoryLanguage was added between currentSourceRow's docblock and its signature, so the paragraph describing "the row standing for the source the route is already playing" ended up introducing a function that returns a language string. Moved below; no behavior change. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> * fix(xtream): stop grouping the scan tier, it can drop a matching source content is unique per (category, type, stream), so one stream sitting in several categories is several rows and nothing forces their titles to agree. The GROUP BY added for group_concat let SQLite keep an arbitrary row's title, and the normalized confirmation then rejected the whole stream on a title a sibling row would have matched — the source vanished. The FTS tier can afford that grouping because its window makes it necessary; the scan tier takes no window at all, so it now returns a row per category and their names are merged per stream in TypeScript, which also keeps the rejected sibling's category in the language derivation. Found by Codex. Latent rather than active on the catalog I measured (0 streams currently carry differing titles across categories), but the schema permits it. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> * docs(xtream): record why category names stay scoped to matched rows Codex flagged that the FTS predicate runs before the aggregate, so a sibling row under a localized title contributes no category. True, and deliberate: the field is a guess feeding a chip and a browse filter, and completing it costs measured latency — 0.74s to 2.0s for a correlated subquery on a 3.9GB catalog, 19.7s for a second bounded lookup — to correct a cosmetic guess in a shape that occurs 0 times in 2.7M rows. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> --------- Co-authored-by: Claude Fable 5 <noreply@anthropic.com>	2026-08-12 07:59:07 +02:00
4grayandClaude Fable 5	dded17010d	fix(playback): keep the display awake while built-in players play video (#1405) ... * fix(playback): keep the display awake while built-in players play video Closes #1095. The renderer tracks every playing <video> through document-level capture listeners (element-level release listeners catch the detached-element pause on component teardown) and, while any video is playing and the document is visible, holds a display-sleep lock: a main-process powerSaveBlocker over IPC in Electron — reliable on Linux where Chromium's own video wake lock depends on DE D-Bus inhibitors — and the Screen Wake Lock API in the PWA. The vote is auto-cleared when the renderer reloads or dies. Radio's <audio> deliberately never blocks display sleep; embedded MPV and external MPV/VLC already manage their own inhibition. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> * fix(playback): withdraw the keep-awake vote when the renderer crashes A crash emits render-process-gone while the WebContents object stays alive, so the destroyed listener alone missed it: without a follow-up reload the display stayed pinned awake. Review finding by Codex. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> * fix(playback): keep the display lock for picture-in-picture playback Minimizing the window hides the document but leaves the PiP surface on screen, so the visibility gate was releasing the lock mid-watch. A tracked playing video that owns document.pictureInPictureElement now counts as visible playback, and PiP enter/leave events resynchronize the gate. Review finding by Codex. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> * fix(playback): re-evaluate the wake lock after a rejection masked a state change In the PWA path a hidden-visible round-trip (or pause/resume) while wakeLock.request() was pending got swallowed by the in-flight guard; if that request then rejected, only the flag was cleared and a continuously playing visible video sat without a wake lock until the next unrelated event. State changes arriving mid-flight now queue one re-evaluation on rejection; permanent denials still don't loop because nothing queues a retry without a fresh interleaved change. Review finding by Codex. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> * docs(playback): mirror the display-sleep contract into AGENTS.md AGENTS.md carries its own playback sections (radio, shared controls, PiP), so the keep-awake contract belongs there too. Review finding by Codex. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> --------- Co-authored-by: Claude Fable 5 <noreply@anthropic.com>	2026-08-12 00:23:03 +02:00
4grayandClaude Fable 5	5b211faf73	feat(remote-control): cover live collections, honest volume, status resets (#1399) ... * feat(remote-control): cover live collections, honest volume, status resets Remote control previously worked only on the three routed live layouts (M3U player, Xtream live, Stalker ITV); playing live TV from favorites, recently viewed, or the global collections left the mobile remote inert. - Wire channel up/down, number select, and status publishing into the unified live tab, covering per-portal and global favorites/recent for M3U, Xtream, and Stalker; navigation follows the search-filtered, sorted list exactly as rendered (shared deriveVisibleFavoriteChannels) - Treat non-live status updates as snapshots in the main process so stale now-playing fields are cleared instead of merged forever - Publish a reset snapshot from every integration on destroy, so leaving a live view clears the remote instead of freezing it - Report M3U supportsVolume only for built-in inline playback and no-op volume commands while MPV/VLC/Embedded MPV owns the audio - Publish live status for Stalker radio (same layout, same handlers) and fix its channel-number lookup for non-numeric radio ids Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01NmMT33wgK52QL6JAz468eH * fix(remote-control): review-loop hardening for status honesty - Make the non-live status update an authoritative reset in the main process: only portal survives, supportsVolume is forced false, stray now-playing fields from callers are dropped (Copilot review) - Stop Stalker radio status from leaking an unrelated TV channel's EPG: the ITV-keyed bulk cache survives itv->radio navigation and Ministra ids collide across the two lists, so EPG fields publish for itv only - Publish the reset snapshot when the M3U active channel clears in place (e.g. quitting external MPV), not only on route destroy - Consider a live external session in the M3U volume gate: a diagnostic-recovery MPV/VLC launch owns the audio even while a web player is configured; republish capability on session start/end - Share one REMOTE_CONTROL_RESET_STATUS constant across all four integrations instead of four hand-copied literals Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01NmMT33wgK52QL6JAz468eH * fix(remote-control): external session outranks DASH in M3U volume gate The managed clear-DASH MPV/VLC fallback (Shaka browser-support preflight failure) leaves activeChannelIsDash() true while the external session owns the audio, so the DASH shortcut bypassed the session check and kept advertising remote volume support. The live-session check now precedes the DASH branch; radio stays first because its inline audio element is always mounted and remains audible. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01NmMT33wgK52QL6JAz468eH --------- Co-authored-by: Claude <noreply@anthropic.com>	2026-08-10 08:18:20 +02:00
4grayandClaude Fable 5	9aeb83e515	fix(m3u): forward playlist-level custom headers to external players (#1397) ... * fix(m3u): forward playlist-level custom headers to external players The custom User-Agent/Referer/Origin stored on an M3U playlist only reached the built-in web players (via the Electron webRequest override). MPV/VLC and the embedded MPV player make their own HTTP requests and received only the per-channel #EXTVLCOPT values, so a playlist-wide custom User-Agent was silently dropped for UA-locked providers (#1221). External launch payloads now resolve each header independently: the channel-level #EXTVLCOPT value wins, the playlist-level value is the fallback, blank values count as absent — matching the semantics the unified favorites/recent stream resolver already had. Covers the auto-launch and catch-up effects in m3u-state, the manual MPV/VLC fallback and the embedded MPV payload in VideoPlayerComponent. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_011id2tdJtkJYRYX8dwYYKwL * fix(playback): send Origin as a real VLC header and cover header IPC in E2E Review follow-ups: VLC only used the Origin value as an :http-referrer fallback while MPV already sent it via --http-header-fields; both VLC paths (fresh spawn and RC enqueue) now emit the same buildHttpHeaderFields list, so a real `Origin: ...` header reaches the provider, deduplicated against an explicit headers-map Origin. The legacy origin-as-Referer fallback stays. The dash-clearkey Electron E2E now asserts the new IPC contract (blank channel-level headers arrive as undefined, not empty strings) and gains a scenario that sets a playlist-level User-Agent through the source editor and verifies the captured MPV fallback launch carries it across the renderer/main IPC boundary. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_011id2tdJtkJYRYX8dwYYKwL --------- Co-authored-by: Claude <noreply@anthropic.com>	2026-08-09 23:34:03 +02:00
4gray	ae375e0e8f	fix(settings): protect unsaved edits on window close, quit, and reload (#1394)	2026-08-09 18:45:48 +02:00
4gray	d73acd6bfc	fix(playback): clarify external player launch feedback (#1388)	2026-08-09 13:33:07 +02:00
4gray	d2a83164ec	feat(stalker): protocol-correct auth lifecycle (#1354)	2026-08-03 23:06:47 +02:00
4grayandClaude Opus 5	e197409b10	fix(stalker): only mint a temporary link when the row asks for one (#1364) ... * fix(stalker): only mint a temporary link when the row asks for one `create_link` ran on every Stalker playback. The reference client — the portal's own `player.js`, mirrored by Kodi's pvr.stalker — mints a link only when the catalog row sets `use_http_tmp_link` or `use_load_balancing`; otherwise it plays the static `cmd` that `get_all_channels` / `get_ordered_list` already returned. Neither flag was read anywhere in the codebase, so every channel paid a round trip and gained a failure point the reference client does not have. One helper now owns the decision (`resolveStalkerStaticPlaybackUrl`), used by `fetchStalkerPlaybackLink()` for ITV/VOD/radio, by the download path, and by `StreamResolverService` for Favorites/Recently Viewed. Its guards are deliberately wider than the flags alone and can only route a row back onto the `create_link` path: no row to read flags from, a relative or query-only command (the VOD `has_files` rewrite), a non-HTTP scheme, or a loopback host. An episode always mints, since `series` selects it server-side. Radio joins the same decision, so a station the portal proxies now gets its link instead of playing a URL the portal never meant to serve. Temporary links live ~5 s, so the audit that came with this: favorites and recently-viewed persist the `cmd`, playback positions store ids, and the main-process context map stores headers keyed by origin+path — none replay a resolved URL. Downloads are the documented exception, and honouring the flags shrinks even that, since an unflagged movie now yields a permanent URL that survives retry. `forced_storage` and `play_token` stay unwired, with the reasoning recorded in the docs rather than left ambiguous. The mock's ITV/radio rows now carry both flags, and the new `static-channel-cmd` scenario (MAC 00:1A:79:00:00:0A) serves unflagged rows with a playable command so the e2e can assert that NO `create_link` request reaches the portal — verified to fail when the change is reverted, with a companion test proving the recorder sees a link when one is due. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> * fix(stalker): keep temporary-link flags across VOD normalization Codex P1 on #1364, and it is real. `buildStalkerSelectedVodItem()` narrows a raw portal row to an explicit whitelist, and the two flags were not on it. It feeds both `selectedItem()` — which the VOD playback path reads as `linkFlags` — and, through `createStalkerVodItem`, the download payload. So a flagged VOD row with an absolute HTTP `cmd` arrived looking unflagged and took the static path, playing the portal's non-final URL instead of minting a link. The direction of the failure is what makes it a P1: a dropped flag reads as "no temporary link needed", so the whitelist fails OPEN. Both flags now sit on `StalkerVodSource` / `StalkerSelectedVodItem` and on the whitelist, with the consequence spelled out at the normalizer so the next edit does not quietly undo it, and specs pinning all three normalizers plus a store-level test that a flagged VOD still mints. Also two things from re-reading my own diff: - The radio path called `resolveStalkerStaticPlaybackUrl` and then handed the same row to `fetchStalkerPlaybackLink`, which runs that exact check again. Two copies of one decision is the divergence this PR exists to remove, so the outer call and its now-unreachable guard are gone. - `portal-catalog-facade.ts` spells the flag shape out instead of importing `StalkerLinkFlagSource`; it now says why (`type:util`/`domain:portal-shared` may not depend on `type:data-access`/`domain:stalker`), so the obvious "reuse the type" cleanup does not get made and break the boundary lint. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> * fix(stalker): authenticate before serving a static collection stream Second Codex P1 on #1364, and a regression this PR introduced. `create_link` was also the request that warmed the portal session. Tokens live in memory only (`StalkerSessionService.tokenCache` is a plain Map), and the collection header builder reads the raw `getCachedToken()`. So a cold start from global Favorites or Recently Viewed — the portal never opened this session — took the static path, found no token, and handed a same-host gated stream headers with no `Authorization`: a 403 on exactly the streams the header contract exists for. The same raw accessor cannot tell a token negotiated for a pre-edit identity from a current one. `StreamResolverService` now calls `ensureToken()` before building a static playback. It is the right primitive: handshake + `get_profile` with no link minted, identity fingerprint validated, concurrent callers deduped, and an immediate null for simple portals — and calling it keeps this change out of `stalker-session.service.ts`, which PR 6 (#1354) is splitting. Best-effort by design: a static URL may point at a CDN that needs no credentials, so a failed handshake degrades to the token-less header set instead of costing the user their playback. Both halves are pinned by tests, and removing the call makes the cold-start test fail. The portal routes need no equivalent and do not get one: an item cannot be selected before its catalog has loaded, and every catalog load authenticates. That reasoning is now written down rather than assumed. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> * fix(stalker): warm the session at the choke point; keep downloads authenticated Two more Codex findings on #1364, and the first one shows my previous commit message reasoned too broadly. P1 — I claimed the portal routes are "structurally warm" because an item cannot be selected before its catalog loads. That is true of the routed portal views, but not of the global collection detail, which calls `setCurrentPlaylist()` and `setSelectedItem()` straight from a persisted row with no catalog load in between and then goes through the STORE playback path. A VOD opened from Favorites on a cold start therefore still played a same-host gated stream with no Bearer token. Rather than extend the per-route argument, the warm-up moved to the one place every static return passes through: `fetchStalkerPlaybackLink()` now calls the session before short-circuiting, covering ITV, VOD, radio and downloads at once. `StreamResolverService` keeps its own call — its static branch does not go through that function — but both now share a single primitive, `ensureStalkerSession()` in `stalker-request.utils.ts`, so the two routes cannot drift on when a session is required. Still best-effort, still outside `stalker-session.service.ts` (PR 6 territory). P2 — downloads cannot use that escape hatch at all: the main-process stored header allowlist is User-Agent/Origin/Referer only, no Cookie or Authorization, so a static same-host URL 401s where a minted one worked. `startStalkerVodDownload` now classifies the candidate with the shared `isStalkerStreamCredentialSafe()` and withholds the row — forcing `create_link` — for anything portal-owned. A CDN-hosted movie keeps the permanent URL that survives retry; a portal-hosted one keeps the minted URL that carries its own token. Both fixes mutation-checked: each reverted change fails exactly one test. Docs corrected, including the overreaching "structurally warm" claim. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> * docs(stalker): record the cached-token revalidation trade-off Codex flagged that the static path no longer self-heals a retired token, since `ensureToken` returns a same-identity cache entry without a network call — whereas `create_link` used to refresh it through `makeAuthenticatedRequest`'s auth-failure retry. The mechanism it posits does not exist on stock Stalker: per the 4.9.35 reference, handshake tokens have no TTL, and not sending the watchdog does not invalidate auth (it only clears the admin panel's "online" flag). The real residual vector is another device calling `get_profile` on the same MAC, which is common enough on shared subscriptions to be worth naming. Revalidating on every static playback would cost exactly the round trip this change removes, so it is deliberately not done. Recorded as a known trade-off with its mitigation (a running watchdog still self-heals within a ping cycle) and handed to PR 6, where a refresh on an OBSERVED playback authorization failure belongs. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> * docs(stalker): tighten the token-revalidation trade-off wording Greptile review feedback: the watchdog mitigation was the most important part of that paragraph and sat behind the caveat. It now follows the MAC-sharing vector directly, and the paragraph ends by naming what is actually left uncovered — a same-host static stream played while no watchdog is up — so a future reader can size the residual without re-deriving it. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> * fix(stalker): prefer the live playlist row over a stale favorite snapshot Codex P1 on #1364, and mine. `resolveStalker` reads its portal coordinates as `item.stalkerPortalUrl ?? playlist?.portalUrl` — item first. The create_link branch quietly corrected for that afterwards by re-reading `applyOverride(playlist).portalUrl`, so the row won wherever it existed, which is what the comment right above it already promised: "when the row exists it wins over the item's snapshot of the portal URL (a repaired endpoint must beat a stale favorite)". The static branch I added returns before that correction, so it shipped the stale snapshot. Consequences after a playlist edit: a same-host static URL matching the OLD host gets the newly negotiated token and identity headers sent to the previous portal, and a MAC-only edit pairs the new token with the old MAC cookie — precisely the pairing `stalkerIdentityFingerprint` exists to prevent. Both branches now derive the coordinates once, row-first with the repair override applied, and fall back to the item's snapshot only for a playlist that no longer exists — which is the role `buildStalkerPlayback` already documents for it. Mutation-checked: restoring item-first precedence fails the new test alone. Also documents a local-only e2e hazard found while re-running the suite: `mode: 'serial'` orders tests within one project, but chromium/firefox/webkit run the file concurrently against the same mock server, so one project's beforeEach reset can drop a session another is mid-test on — which is what a lone auth-spec failure that passes on rerun actually is. CI never sees it; the Web E2E job runs --project=chromium alone, and that command is clean (22/22). Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> * fix(stalker): warm the session against the repaired portal configuration Found while auditing my own static branch against the create_link path rather than waiting for the next review round. `executeStalkerRequest` applies the lazy-repair override on its first line, so the create_link path always talks to the configuration a completed repair proved good. The session warm-up I added did not: it handed `ensureToken` the caller's pre-repair row, so a portal whose endpoint or mode had been repaired would handshake against the configuration the repair had already rejected — stranding the session precisely on the portals repair exists to rescue. The override now happens inside `ensureStalkerSession`, mirroring `executeStalkerRequest`'s first line, so every caller inherits the rule instead of each having to remember it. Mutation-checked: dropping the override fails the new test alone. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> * fix(stalker): fall back to create_link when a portal-owned static url has no session Codex P1 on #1364. `create_link` was also the request that could FAIL, and a failure is what triggers the lazy portal repair. A playlist still misclassified as token-free, or pointing at an unrepaired endpoint, used to self-heal on that failure and then play; the static path issues no request, so nothing fires and the stream just 401s. Its suggested remedy — routing a skipped warm-up through `repairPortal()` — cannot be taken literally: a skipped warm-up is the NORMAL case for the many legitimately token-free reseller panels, and probing each of them on every playback would cost far more than the round trip this PR removes. What is decidable without a request is whether we are about to serve a stream we already know will fail. `ensureStalkerSession` now reports whether the session can serve credentialed playback — true for a portal needing no token and for one holding a usable token, false for a full portal left without one — and both static call sites act on it: - foreign-host URL: served regardless, it never needed the session; - portal-owned URL with a usable session: served, as before; - portal-owned URL with no usable session: falls back to `create_link`, which mints a URL carrying its own token AND re-enters the only path that can observe a failure and repair. That covers the unrepaired-endpoint half exactly. The misclassified-as-simple half stays open by construction — no request means no evidence, and "simple portal" is indistinguishable from "misclassified" without one. It belongs with the other reactive-repair work already handed to PR 6: refresh and repair on an OBSERVED playback authorization failure. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> * fix(stalker): require flag evidence before trusting a row as unflagged Two Codex findings on #1364. P1 — legacy persisted snapshots. Favorites and Recently Viewed rows saved before this change went through `buildStalkerSelectedVodItem`'s whitelist, which dropped both flags, and `buildStalkerFavoritePayload` spreads that whitelisted object. So a legacy row is flagless because WE stripped it, not because the portal said no — and the helper was reading it as "explicitly unflagged". With an absolute HTTP `cmd` from a load-balanced portal that meant playing a non-final URL. There is no migration or provenance marker for those rows. A stock portal returns both flags on every row, so their PRESENCE is itself the provenance signal, and it is the only one available without a refetch. `resolveStalkerStaticPlaybackUrl` now requires at least one flag key to be present; absence reads as "no evidence" and routes back to `create_link`, which is the pre-PR behaviour. This costs the optimization on panels that omit the flags entirely — the honest price for not being able to tell them apart from our own stripped rows. Radio is the one documented exception. It has always played a directly usable command without `create_link`, so a flagless radio row keeps that rather than newly minting — a portal whose radio `create_link` never worked would otherwise lose playback it has today. ITV and VOD have no such history and stay conservative. P2 — loopback range. IPv4 reserves all of `127.0.0.0/8`, so `127.0.0.2` was being handed to the player as a real address. Classified by range now, with a test that `127.0.0.1.cdn.example` is still treated as the ordinary hostname it is. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> * fix(stalker): classify every portal-local IPv6 placeholder Codex P2 on #1364, same class as the 127.0.0.0/8 one. `http://[::]/ch/1234_` and the IPv4-mapped loopback forms slipped past the exact-name set and would have been handed to the player as real addresses. Checked how `URL` actually normalizes these rather than guessing at the spelling a portal might use: brackets are kept, `[0:0:0:0:0:0:0:1]` collapses to `[::1]`, and an IPv4-mapped address is rewritten to hex — `[::ffff:127.0.0.1]` arrives as `[::ffff:7f00:1]`. The guard now strips the brackets, matches `::1` and `::`, and decodes the mapped form by its high byte, so the whole of the mapped 127.0.0.0/8 range is covered along with the mapped unspecified address. The dotted tail is still accepted for any engine that leaves it alone. Routable hosts are unaffected, pinned by tests for `[2001:db8::1]` and `[::ffff:203.0.113.7]`. Mutation-checked: dropping `::` and the mapped-IPv4 decode fails five tests and nothing else. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> * fix(stalker): normalize hostname and scheme spelling before the static verdict Two Codex P2s on #1364, both about trusting how a portal spells things. `http://localhost./ch/1234_` — a trailing dot is the DNS root and resolves identically, but `URL` keeps it for names while dropping it for IP literals (`127.0.0.1.` arrives bare, `localhost.` does not). The exact-name check read that as a remote host and would have pointed the player at its own loopback. Stripped before classifying. `HTTP://cdn.example/a.ts` — RFC 3986 makes the scheme case-insensitive. The case-sensitive tests failed SAFE, minting a link instead, but that defeats the contract for a portal that spells it this way, and one whose `create_link` cannot resolve an already-playable row would break. There were five such tests, and only one was on the new static path: the other three live in `resolveStalkerPlaybackUrl`, the create_link RESPONSE resolver, where `ffrt3 HTTP://…` failed to split its solution prefix and a query-only reply was appended to the portal base instead of to the command. That is pre-existing, but it is the same bug in the same shared normalizer, and fixing only the half this PR introduced would leave exactly the divergence this PR keeps removing. All five now go through one `hasHttpScheme()`. The response resolver had only indirect coverage, so it gains a direct spec alongside the static-path tests. Mutation-checked: reverting the dot strip and the case-insensitive scheme fails ten tests and nothing else. Also carries a docblock fix noticed on a read-through: the guard list still pointed at `PORTAL_LOCAL_HOSTNAMES` after the logic moved into `isPortalLocalHostname`, which now covers considerably more than that set. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> * fix(stalker): normalize DNS root dots in the shared credential classifier Codex P2 on #1364, extending the `localhost.` fix into `isStalkerStreamCredentialSafe()`. It compared hostnames literally, so a portal on `portal.example` serving `https://portal.example./movie.mkv` classified its own stream as third-party. Wider than the download guard it was reported against: this predicate is the single rule BOTH the renderer playback-header builder and the Electron main-process fallback use to decide whether a stream may carry the mac cookie and Bearer token. A portal-owned stream spelled with the root dot was getting the credential-free profile and would 401 — pre-existing, and exactly the "only VLC works" class this contract exists to prevent. My PR added two new dependencies on the same predicate (the download static guard and the portal-owned fallback), which is how it surfaced. Both sides are normalized, so it stays symmetric, and it can only widen toward "same host" — never toward handing credentials to a different one. A test pins that `evil.portal.example.` is still rejected. Also carries the authority guard found by probing the same class myself rather than waiting for it to be reported: `http:///ch/1` has no authority and `URL` quietly reinterprets the first path segment as the host, so a malformed command reached the player as a nonsense address instead of going to the portal. `isPlayableHttpUrl()` now requires a non-empty authority. The other exotic spellings I probed were already covered — `URL` canonicalizes `127.1`, `2130706433` and `0x7f000001` to `127.0.0.1`, uppercases and expanded IPv6 normalize too. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> * perf(stalker): classify the static url before authenticating Codex P2 on #1364. Both static call sites awaited the session warm-up and only then asked whether the stream needed portal credentials at all — so a movie or channel on a foreign CDN paid for a handshake whose result was immediately discarded. That is not free: non-`create_link` requests carry a 15 s timeout (`stalker.events.ts`), so a portal that is slow or offline stalled playback of a stream the CDN would have served instantly. Cold Favorites/Recently Viewed starts are exactly where this bites, since that is where the session is not warm already. Classification now runs first. Foreign host returns immediately, portal-owned still warms and still falls back to `create_link` without a usable session. Behaviour is otherwise unchanged; only the order and the wasted wait are gone. Two tests moved with it: the foreign-host case now asserts the portal is not contacted at all rather than merely not asked for a link, and the repaired-endpoint case had been written against a foreign-host command, which under the new ordering correctly never reaches the handshake it was meant to be testing — it uses a portal-owned command now. Mutation-checked: restoring warm-before-classify fails the foreign-host test alone. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> * test(stalker): repoint two handshake tests at the path they claim to cover Self-audit, prompted by the previous round: the reorder exposed one test that was asserting through a path it no longer reached, so I checked the rest of that class rather than assume it was the only one. Two more had the same defect, both mine. `still returns the static url when the handshake fails` (both specs) mocked `ensureToken` to reject, but used a FOREIGN-host command. Now that classification runs before authentication, that command returns before the handshake is ever attempted — the rejection was never exercised and the test passed on the early return instead of the mechanism in its name. Worse, the foreign case is already covered by the test added alongside the reorder, so these were asserting nothing new. Both now use a portal-owned command, which is what actually reaches the handshake, and assert what a throw really produces: `ensureStalkerSession` swallows it, the verdict is false, and the row falls back to `create_link` rather than being served as a known 401. Each asserts `ensureToken` was in fact called, so neither can silently drift back into testing an early return. Docs corrected with them: the "best-effort degrades to the token-less header set" wording described behaviour the reorder removed. A foreign-host URL is now returned before any handshake, and a failed one routes to `create_link`. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> * test(stalker): make the simple-portal skip test prove portal mode Fourth test found passing through the wrong exit, from auditing all ten in the block rather than waiting to trip over another one. `skips the handshake for a simple portal` used a foreign-host command, so the classification step returned before the warm-up was reached. `ensureToken` was indeed not called — but because the host was foreign, not because the portal was simple, and the assertion could not tell those apart. The command is now portal-owned, so the skip can only come from the mode, and the test also pins the returned URL and that no request was made. Mutation-checked properly this time: removing the simple-portal early return from `ensureStalkerSession` now fails this test. Under the old command it would not have. Also records the pattern where the next person will meet it. The decision chain has several exits — no flag evidence, unresolvable command, `series` set, foreign host, unusable session — and more than one can satisfy the same assertion, so a foreign-host command silently stands in for "simple portal" or "handshake failed". Mutation testing does not catch that class: it proves a test is coupled to its target, not that it reached the mechanism it names. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> * fix(stalker): key the radio fallback on flag evidence, not snapshot presence Codex P2 on #1364, and a divergence I introduced myself. `withStalkerPlayer`'s radio branch checks `hasStalkerLinkFlagEvidence(item)` before synthesizing the zero flags. `StreamResolverService` used `??`, which only falls back when the snapshot is absent entirely. A radio Favorite or Recent row persisted before the flags were carried HAS a snapshot — the old whitelist just stripped the flags out of it — so the `??` selected that flagless object, the helper found no evidence, and the collection route began minting for exactly the rows that used to play directly. That breaks portals whose radio `create_link` is unsupported, which is the case the radio exception exists for. The two paths now apply the identical rule. The divergence came from fixing them in different rounds and is precisely the class this PR keeps closing, so the comment on each side now points at the other. The existing radio test carries no `stalkerItem` at all, so it exercises the missing-snapshot arm and stayed green throughout — the same "passes through a different exit" pattern documented in the section above. The new test supplies a present-but-flagless snapshot. Mutation-checked: restoring the presence check fails it alone. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> * fix(stalker): treat every reserved localhost name as portal-local Codex P2 on #1364, the fourth in this class. RFC 6761 §6.3 reserves `localhost` AND every name ending in `.localhost` for the loopback interface, and resolvers honour it — so `http://stream.localhost/ch/1234_` reached the player's own machine instead of being sent to the portal to resolve. Closed the class rather than adding one more name: the suffix is matched, and `localhost.localdomain` goes in with it as the conventional `/etc/hosts` alias for 127.0.0.1 on most Linux systems. Together with the earlier rounds the predicate now covers `localhost` and `*.localhost`, `localhost.localdomain`, `127.0.0.0/8`, `0.0.0.0`, `::1`, `::`, the IPv4-mapped forms `URL` rewrites to hex, and a terminal DNS root dot on any of them. Only the suffix is reserved, so the guard must not over-match: tests pin that `localhost.cdn.example` and `notlocalhost` remain ordinary routable names and keep playing statically. Mutation-checked: dropping the suffix rule and the localdomain alias fails four tests and nothing else. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> --------- Co-authored-by: Claude Opus 5 <noreply@anthropic.com>	2026-08-03 18:36:56 +02:00
4gray	96facd6f49	feat(downloads): queue season episode downloads (#1357) ... * docs(downloads): specify season queueing * docs(downloads): plan season queue implementation * feat(downloads): define episode queue identity * fix(downloads): align episode identity contract * feat(downloads): coordinate season queue submissions * fix(downloads): keep queue coordination provider neutral * fix(downloads): reconcile legacy episode identities * fix(downloads): fail closed on invalid stored coordinates * refactor(downloads): adapt Xtream episode requests * fix(downloads): use canonical Stalker episode ids * test(downloads): cover Stalker adapter reactivity * feat(downloads): add selected season queue action * refactor(downloads): extract season download presenter * feat(downloads): localize season queue feedback * test(downloads): cover series batch queue flow * test(downloads): harden series queue fixtures * docs(downloads): describe season queueing * docs(downloads): clarify season queue IPC contract * fix(downloads): isolate season header build warnings * fix(downloads): label season view toggles * fix(downloads): preserve Xtream episode headers * fix(downloads): fail closed on stale episode state * fix(downloads): align renderer queue safeguards * fix(downloads): block ambiguous episode actions * fix(downloads): accept nullable legacy coordinates * fix(downloads): preserve scoped episode ownership * fix(downloads): probe restored files asynchronously * fix(downloads): bound restored file probes * fix(downloads): release timed out file probes * fix(downloads): bound file probe callers * fix(downloads): refresh stable season skips * fix(downloads): fail closed before provider prep * fix(downloads): preserve retained partial ownership * fix(downloads): reconcile partial cleanup completion * fix(downloads): await authoritative list refresh * fix(downloads): coalesce list refreshes * fix(downloads): preserve specials season identity * fix(stalker): preserve specials season mapping * fix(downloads): distinguish missing Xtream seasons	2026-08-03 08:53:44 +02:00
4grayandClaude Fable 5	b92503feae	feat(stalker): endpoint probing + behavior-based portal mode with lazy repair (#1344) ... * feat(stalker): endpoint probing + behavior-based portal mode with lazy repair Replace the URL-shape guess behind isFullStalkerPortal with real endpoint discovery: at import, probe portal.php -> server/load.php -> stalker_portal/server/load.php (the pasted .php endpoint first) and classify the portal by observed behavior — a token-less itv/get_genres answering data proves a token-free panel, the middleware's plain-text auth failure proves the endpoint enforces the token, confirmed by the real handshake + get_profile. The proven endpoint and mode are persisted. The three diverging portal-mode predicates (import, session service, legacy migration) collapse into one shared helper in @iptvnator/shared/interfaces; executeStalkerRequest becomes the single request choke point (search and the collection stream resolver fold in), and the production-dead makeStalkerRequest copy is removed. Existing misclassified playlists repair themselves lazily: only after a request actually fails with the plain-text auth bodies, HTTP 404, or a terminal handshake error, at most once per playlist per session, and only a configuration discovery proved to answer is persisted — via a minimal portalUrl/isFullStalkerPortal patch, so favorites, recents and playback positions survive. Working reseller panels are never probed or rewritten; there is deliberately no eager one-shot migration, because tolerant portal.php panels cannot be told apart from misclassified canonical portals without probing. The Electron handler now embeds the HTTP status code in the error message (ipcRenderer.invoke strips custom properties from rejections), and probe requests carry silent:true so expected 404s do not toast error snackbars. The stalker mock gains a portal.php-less /ministra host so e2e can prove the 404 fallthrough end to end. Fixes #850, #686, #755. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> * fix(stalker): sync watchdog, PWA proxy errors and cmd resolution with lazy repair Review round 1 (Greptile P1, Codex P1/P2): - A successful repair now re-syncs the ACTIVE watchdog playlist via the new StalkerSessionService.refreshActiveWatchdogPlaylist(): a simple-to-full repair starts the required keepalive mid-session, full-to-simple stops it, and an endpoint change repoints the pings instead of leaving them on the activation-time snapshot. - PwaService.forwardStalkerRequest surfaces the web-backend proxy's normalized { message, status } no-payload envelope as an HTTP error carrying the status, so endpoint discovery and the lazy repair can classify upstream 404s in the PWA too (previously payload unwrapping returned undefined and dead endpoints were unrepairable there). Probe requests pass silent:true and skip the error snackbar. - fetchStalkerPlaybackLink and the collection StreamResolverService re-apply the repair override AFTER the request, so a relative create_link reply resolves against the endpoint that actually answered (the resolver keeps the /stalker_portal path segment as base, so this matters beyond origin). Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> * fix(stalker): parse candidate URLs and tie repair overrides to their source config Review round 2 (Codex P2 x2): - Endpoint candidates are now derived from the parsed origin + pathname: a pasted URL carrying a query or fragment (host/c?key=value) no longer gets /portal.php bolted onto the query, which made every probe hit /c and persisted the non-API URL. - A repair override is tied to the failing configuration it replaced. Playlists carrying anything else (the user edited the portal URL or mode through the playlist dialog) drop the override and re-arm the once-per-session probe latch, so edited metadata is used verbatim and may repair again if it fails. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> * fix(stalker): auth-gated probes, normalized offline fallback, mock docs sync Review round 3 (Codex P1 x2, P2): - A probe answered with HTTP 401/403 now classifies the endpoint as auth-required and attempts the real handshake instead of skipping the candidate: non-standard middlewares answer 401 where the stock server sends HTTP 200 + plain text, and such portals authenticated fine before discovery existed. - The unreachable-host import fallback normalizes the pasted URL (origin + pathname) before the legacy /c -> portal.php rewrite, so a query or fragment can no longer make it persist the browser page URL - a 200 HTML answer from /c is not a repair trigger, which would have left the playlist empty for good. - The stalker mock-server README and architecture doc now describe behavior-based discovery and the /ministra host instead of the retired URL-shape rule and its "known inconsistency" note. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> * fix(stalker): recognize JSON auth failures and guard repairs against mid-probe edits Review round 4 (Codex P1 + P2): - isStalkerAuthFailureResponse() recognizes the JSON envelope some panels answer instead of the plain-text body ({js:{error:"Authorization failed"}} / {js:{msg:...}}). Probe classification treats it as auth-required instead of token-free data, and the lazy-repair trigger fires on it at runtime — previously such a portal was persisted simple with no repair path at all. - A repair is committed only after re-reading the persisted row and verifying it still carries the configuration that failed: a user who edits the portal URL (or deletes the playlist) during the multi-second probe now wins over the in-flight repair result for the old URL. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> * fix(stalker): probe past endpoint 5xx, sibling fallbacks, identity-aware repair guard Review round 5 (Codex P2 x3): - A probe that fails with a RESOLVABLE HTTP status keeps discovery going: a broken /portal.php handler answering 500 must not hide a healthy sibling endpoint. Only status-less failures (true network level) stop the loop. The Electron handler now gives real HTTP 5xx responses the same parseable "HTTP Error <code>" message shape as 4xx, so the renderer can tell them apart from ECONNREFUSED/timeouts after ipcRenderer strips the object shape. - Standard fallback candidates for a nonstandard pasted endpoint (.../cp/api.php) derive from its DIRECTORY, so recovery probes hit /cp/portal.php instead of /cp/api.php/portal.php. - The repair's row re-verification also compares the MAC and all Stalker identity fields: a probe authenticated as the old identity must not install its token/watchdog or persist onto a row whose credentials were edited mid-probe. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> * fix(stalker): reactivation-safe watchdog, wider JSON auth phrases, per-config probe latch Review round 6 (Greptile 4/5 concern + Codex P1/P2): - setCurrentPlaylist applies the repair override before feeding the watchdog and store state: re-activating the portal route with the stale NgRx meta no longer stops or repoints the repaired keepalive back to the broken configuration. - The structured js.error/js.msg fields accept the full phrase set the session service recognizes (Invalid token, Auth failed, bare unauthorized/authorization) — panels answering those envelopes were still classified token-free. Plain-text body matching stays narrow on purpose (HTML false positives). - The once-per-session probe latch is keyed by the SOURCE configuration fingerprint (endpoint, mode, MAC, identity) instead of the playlist id: a repair discarded because of a mid-probe edit no longer blocks the edited configuration from repairing, while stale snapshots of an already-probed configuration still cannot loop the probe. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> * fix(stalker): identity-aware override invalidation and timeout-tolerant probing Review round 7 (Greptile P1 + Codex P2): - The repair override records the identity fingerprint the probe authenticated as. Editing the MAC or any Stalker identity field afterwards drops the override, the per-config probe latch AND the cached token, so requests and watchdog pings never pair the edited identity with a session negotiated for the previous one. - A status-less probe failure that is a TIMEOUT (renderer budget, axios request timeout, ETIMEDOUT) continues to the next candidate — one hanging handler must not hide healthy siblings; connection-level failures (refused, unresolvable host) still stop discovery, so dead hosts keep failing fast. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> * fix(stalker): watchdog pings authenticate as the persisted row Review round 8 (Greptile 4/5 concern): The watchdog held its activation-time playlist snapshot for the whole session, so portal metadata edited (or repaired) mid-session kept the keepalive authenticating as the previous identity/endpoint — its pings could keep the old session alive and repopulate the playlist-scoped token cache with a token for the pre-edit identity. Each ping now resolves the playlist from the persisted row first (the single source of truth), falling back to the snapshot only when the store cannot be read, and refreshes the snapshot on every successful read. Any edit — identity, endpoint or mode — reaches the keepalive within one ping cycle; a row now marked simple (or deleted) stops the watchdog. The in-flight guard is claimed before the row read so overlapping pings cannot double-fire. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> * fix(stalker): identity-tagged tokens, watchdog override overlay, retire-on-failure Review round 9 (Greptile 4/5 concern + Codex P2): - The session token cache is tagged with the identity fingerprint (MAC + all Stalker identity fields) the session was negotiated for; ensureToken re-authenticates instead of handing an edited identity the previous token. The fingerprint helper is shared (stalker-identity.utils) with the repair layer's override/latch checks. - Watchdog pings overlay the repair layer's in-session override on the resolved row (registered decorator, no import cycle): a simple-to-full repair whose persistence is pending or failed no longer reads the stale row and stops the freshly started keepalive. - makeAuthenticatedRequest retires a failed token even on the no-retry path (watchdog pings), so a dead session is never handed to the next caller. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> * fix(stalker): pending authentications are identity-scoped Review round 10 (Greptile 4/5 concern): pendingAuth entries carry the identity fingerprint they authenticate as. A request for an edited identity no longer adopts an in-flight result negotiated for the previous identity: it waits the old authentication out (a competing handshake would strand it with a dead token on strict portals) and then negotiates its own session. This was the last id-only-keyed session structure — override, probe latch, token cache, watchdog snapshot and pending auth are now all identity-aware. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> * fix(stalker): atomic repair persistence, full probe history, normalized offline classify Review round 11 (Codex P2 x3 + P1 docs): - The repair's row verification and patch now run ATOMICALLY inside the per-playlist write queue via the new PlaylistsService.transformPlaylistMeta(): a user edit that is queued but not yet committed wins over the repair — the transform sees the edited row and aborts instead of overwriting it. Write failures after a successful verification keep the session-only override, read failures discard the repair. - The per-playlist probe latch keeps EVERY attempted source fingerprint, so alternating edits (A -> B -> A) cannot evict a fingerprint and let stale snapshots re-run discovery. - The unreachable-host import fallback classifies the normalized origin+pathname, so a query merely mentioning /server/load.php cannot make a panel URL look canonical and abort the offline import. - docs/architecture/stalker-portal.md documents the actual probe sequencing: any resolvable HTTP status (incl. 5xx) and timeouts continue, 401/403 classify as auth-required, only connection-level failures abort. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> * fix(stalker): collision-proof session fingerprints Review round 12 (Greptile P1): identity values are unrestricted strings, so the delimiter-joined fingerprint could alias distinct identity tuples (serial "a|b" + empty device vs serial "a" + device "b") and bypass the identity invalidation. Both the identity fingerprint and the repair source fingerprint are JSON-encoded now; regression test pins the exact aliasing pair. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> * fix(stalker): preserve URL authority in normalization; document per-config latch Review round 13 (Codex P1 docs + P2): - normalizeStalkerPortalInputUrl mutates the parsed URL (clear query/ fragment, trim pathname) instead of rebuilding from origin, and the candidate builder swaps only the path — file: URLs (origin "null") no longer make the builder throw, and basic-auth credentials are not silently dropped before probing. - The canonical docs and the repair service JSDoc now describe the actual loop guard: at most one probe per SOURCE CONFIGURATION (endpoint, mode, MAC, identity) per playlist per session, not once per playlist. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> * fix(stalker): HTTP 401/403 failures trigger the lazy repair Review round 14 (Codex P1): discovery classifies 401/403 endpoints as auth-required, but the repair trigger accepted only 404 — a legacy playlist misclassified token-free against an HTTP-auth-gated middleware could never reach discovery and stayed unusable. 401/403 now qualify; endpoint-specific 5xx still do not. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> * fix(stalker): re-enter repair for edited configurations after a pending probe Review round 15 (Codex P2): a request carrying an edited configuration that raced an in-flight probe only awaited it and inherited its outcome — the edited fingerprint stayed unattempted and the first request failed without triggering its own discovery. repairPortal now re-enters after awaiting the pending probe, so the per-config latch decides: already attempted -> reapply, never attempted -> own probe. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> * fix(stalker): probe history remembers outcomes so restored configs repair again Review round 16 (Greptile P1): the per-config latch kept A's fingerprint after an edit to B dropped A's override, so restoring A left it latched with nothing to reapply — broken until restart. The history now stores each probe's OUTCOME (override or null): a restored configuration reinstalls its remembered repair without a second discovery, and the anti-ping-pong property (A<->B alternation never re-runs discovery from stale snapshots) is preserved. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> * fix(playlist): serialize deletion behind the per-playlist write queue Review round 17 (Codex P2): deletePlaylist bypassed serializePlaylistWrite, so a queued mutation (e.g. the Stalker portal repair's conditional transform) finishing after an unserialized delete could upsert the row back and resurrect the playlist. Deletion now runs through the same queue: queued writes commit first, the delete lands last, and a transform enqueued after the delete reads a missing row and aborts. Regression test pins the write-then-delete ordering. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> * fix(stalker): reinstalled repairs re-sync the watchdog like fresh ones Review round 18 (Greptile P1): the restored-configuration branch reinstalled the remembered override without the watchdog refresh the fresh-repair path performs — if the intermediate edit stopped the keepalive, the restored full-portal session recovered requests but never its pings. The reinstall now calls refreshActiveWatchdogPlaylist with the override applied, symmetric with a fresh repair. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> * fix(stalker): discarded probes retry once their configuration is restored Review round 19 (Greptile P1): the pre-probe history reservation survived the row-mismatch discard, so restoring the original configuration hit the latch with nothing to reinstall — lazy repair stayed disabled for the session. Probe records are now explicit (override / no-change / discarded): a discarded configuration probes again once one cheap row read confirms the row was RESTORED to it, while stale snapshots of it stay declined without a discovery run. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> * fix(stalker): IPC-safe transport errors, repairable profile path, nested base paths Review round 20 (Codex P2 x4): - The Electron handler throws a real Error for axios failures without a response: Electron serializes rejections via toString(), so a plain object arrived as "[object Object]" and discovery could not tell a timeout (keep probing) from a dead host (stop). - isAuthorizationError parses HTTP 401/403 out of the IPC-wrapped message, so an expired-token 403 retires the token and re-authenticates instead of surfacing as a plain failure. - The account-info full-profile path (which bypasses executeStalkerRequest) routes repair-trigger failures through StalkerPortalRepairService and retries with the repaired playlist, so opening the dialog can fix a stale endpoint. - resolveStalkerPlaybackUrl derives the installation base from the endpoint's API suffix instead of a fixed stalker_portal|c|portal allowlist: relative create_link replies now resolve correctly under arbitrary discovered installations such as /cp/server/load.php. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> * fix(stalker): strict probe data shape, mode-aware profile retry, docs API name Review round 21 (Codex P1 docs + P2 x2): - Probe classification requires the real get_genres shape (array, or a {data: []} envelope without an error) instead of a bare `js` key: a 200 error envelope ({js:{error:"Unknown action"}}, {js:false}) no longer ends discovery on a broken candidate and persists an empty catalog. - After a repair that flips the portal to simple mode, the account-info retry re-enters the mode routing and uses get_main_info instead of handshaking against a token-free panel again. - docs/architecture/stalker-portal.md names transformPlaylistMeta and its atomic source-check invariant (plus the serialized deletion) rather than the race-prone updatePlaylistMeta. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> * fix(stalker): account dialog re-routes after a simple-to-full repair Review round 22 (Codex P2): fetchViaMainInfo runs through executeStalkerRequest, whose lazy repair retries the SAME action, so a repair proving the portal is actually full left the dialog calling get_main_info — canonical installations publish subscription details only through handshake + get_profile, leaving the dialog empty. The routing is now symmetric with the full-to-simple case: an empty main-info result whose repair flipped the mode re-enters the profile flow. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> * fix(stalker): row-gate override reinstall; document mode-based account routing Review round 23 (Codex P2 + P1 docs): - Reinstalling a remembered override now requires the persisted row to actually carry that configuration again. A stale request for A while the row holds an unrelated C no longer resurrects A's override, which would retry against B and repoint the active watchdog away from C. (The edit-back-to-A case stays as documented: there the row IS A.) - docs/architecture/stalker-portal.md and CLAUDE.md describe account-info routing by the observed portal MODE instead of the endpoint shape — a token-enforcing portal.php is a full portal now — and note the mode-change re-routing in both directions. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> * fix(stalker): share the auth-failure predicate; prefer profile over partial main-info Review round 24 (Codex P1 + P2): - isAuthorizationError now reuses isStalkerAuthFailureResponse, so the phrases discovery and the lazy repair already classify as auth failures (Access denied., Unauthorized request., and their JSON envelopes) also retire the session token. Previously a full portal expiring with either phrase kept its dead token: the repair rediscovered the same endpoint/mode, recorded no-change, and every later request stayed broken. - After a simple-to-full repair, even a PARTIAL get_main_info answer no longer wins over the profile flow — expiry and tariff live only behind handshake + get_profile. The partial facts are kept only if the profile path itself publishes nothing. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> * fix(stalker): keep a literal c installation directory in candidate derivation Review round 25 (Codex P2): the /c landing-page rewrite ran after the endpoint file was stripped, so `/tenant/c/portal.php` collapsed to `/tenant` and the sibling probes went one level too high, rejecting a valid portal whose installation directory is literally named `c`. The rewrite now applies only when the pathname itself ends in `/c` (no endpoint file); pasted endpoints strip only the file part. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> * fix(stalker): route rejected post-repair main-info retries to the profile flow Review round 26 (Codex P2): a simple-to-full repair during fetchViaMainInfo makes executeStalkerRequest retry the same action against the repaired full portal, and installations that do not implement get_main_info answer 404 — the rejection escaped before the repaired-mode check, so the dialog failed instead of switching to get_profile. The rejection is captured and reaches the same check; without a mode change it is rethrown unchanged. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> * fix(stalker): full predicate for wrapped denials; record the removed store prop Review round 27 (Codex P2 + P1 docs): - The repair trigger applies the shared auth-failure predicate to the error MESSAGE too, so authentication's wrapped structured denials (Error('Profile error: Access denied.')) reach the repair instead of bypassing it and leaving a healthy sibling endpoint unprobed. - docs/architecture/stalker-store-api-baseline.md records makeStalkerRequest as removed, with the reason it gets no facade alias: it was production-dead and held a fourth private copy of the portal-mode branch that the shared predicate exists to prevent. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> * fix(stalker): complete auth predicate for wrapped error messages Review round 28 (Codex P2): the plain-text BODY matcher deliberately knows only the three middleware phrases, so passing an error message through it let authenticate()'s wrapped denials — Error('Profile error: Invalid token') / 'Auth failed' — bypass both the repair trigger and the session auth predicate. A dedicated isStalkerAuthFailureMessage() applies the wide phrase set to controlled error strings, while arbitrary portal bodies keep the narrow matcher that cannot false-positive on HTML pages. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> * fix(stalker): reject denied profiles during confirmation; document all repair triggers Review round 29 (Codex P2 + P1 docs): - Full-portal confirmation validates the get_profile envelope with the shared structured predicate: a handshake can hand out a token whose profile still answers {js:{error:"Invalid token"}}, and authenticate() inspects only msg/block_msg — discovery would have persisted an unusable endpoint and stopped before the healthy sibling. authenticate() now returns the raw profile response for that check. - The canonical lazy-repair contract lists the complete trigger set: the plain-text bodies AND their JSON envelopes, HTTP 404, HTTP 401/403, and terminal handshake/profile errors. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> --------- Co-authored-by: Claude Fable 5 <noreply@anthropic.com>	2026-08-02 18:01:45 +02:00
4grayandClaude Fable 5	65f81b7110	fix(pwa): bring the Stalker transport to parity with Electron (#1348) ... * fix(pwa): bring the Stalker transport to parity with Electron The self-hosted PWA's /stalker proxy now derives its portal requests from the same shared identity and URL builders as the Electron main process: MAG User-Agent/X-User-Agent, full STB cookie (mac + stb_lang + timezone + serial-derived __cfduid), SN header, JsHttpRequest=1-xml defaulting, and the sn-only-on-get_profile rule. macAddress/token/serialNumber are control params consumed into headers and never echoed into the portal's query string (handshake keeps its candidate token — protocol content). The stalker-mock-server /stalker route mirrors the new contract through the same shared builder. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> * test(stalker): forward the full identity header set in the mock /stalker mirror Greptile review: the synthetic portal request kept only the cookie and Authorization from the generated identity, so mock handlers could never validate the SN/MAG-UA/Accept/Language/Connection headers the real proxy sends. Forward the complete set, lowercased the way Express normalizes incoming headers. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> --------- Co-authored-by: Claude Fable 5 <noreply@anthropic.com>	2026-08-02 14:51:36 +02:00
4grayandClaude Fable 5	fc7f23b229	feat(playback): forward portal Cookie/Authorization to built-in players (#1335) ... * feat(playback): forward portal Cookie/Authorization to built-in players The web players (HTML5/hls.js, Video.js, ArtPlayer, Shaka) could only ever receive User-Agent/Referer/Origin, so any Stalker stream gated on the portal session cookie or Bearer token played exclusively in external MPV/VLC — the long-running "only VLC works" cluster (#849, #910, #732). - request-header-overrides.service: the scoped override now carries Cookie and Authorization, attached only to requests on the exact stream origin, in-memory only, dropped on replace/clear. Unscoped (playlist-level) calls drop credentials fail-closed; control characters in header values are rejected. Chosen over session.cookies.set(): jar cookies attach only to credentialed requests, which would force withCredentials into every engine and break against the Access-Control-Allow-Origin:* IPTV panels send, and jar scoping is port-blind. - WebPlayerViewComponent is now the single owner of the scoped override for every built-in player: it extracts the full header set from the resolved playback, configures the override BEFORE handing the source over (players render only once the source exists), and clears the scoped layer on destroy. HtmlVideoPlayerComponent's own three-header call is removed — it would overwrite the credentialed override. - Stalker VOD, series episodes and radio now build the same portal header set ITV already had (they previously carried no portal headers at all); same-origin playback sends the real User-Agent alongside X-User-Agent. - Stream classification is host-based via one shared predicate (isStalkerStreamCredentialSafe): same-host port changes and scheme upgrades keep the portal profile (the #1158 class), a foreign host or https->http downgrade keeps the credential-free KSPlayer profile. The main-process fallback context uses the same predicate so isStalkerDirectStreamProfile can no longer discard renderer headers. - setUserAgent bridge gains an optional credentials parameter; preload, ipcMain handler and ElectronBridgeApi updated together. - stalker-mock-server: gated-stream scenario (MAC 00:1A:79:00:00:09) whose create_link returns a local /stream/gated/video.mp4 that 403s without the mac cookie + current Bearer token; new Electron e2e proves a built-in player actually plays it (and that the gate refuses bare requests). Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> * fix(playback): apply header override to Stalker radio, redact mock cookie log Address Codex review feedback on #1335: - The radio branch of the Stalker live layout renders the dedicated audio player, never WebPlayerViewComponent, so the resolved portal headers were built but never applied — an auth-gated radio stream still 403'd. The override sync is extracted into ElectronStreamHeadersService (single owner of the scoped override slot, with clear-only-while-owning semantics so a destroyed consumer cannot wipe a newer consumer's override), applied by WebPlayerViewComponent for video players and by the radio branch before the audio element gets its URL. The service feature-detects the bridge method so partial bridges behave like the PWA instead of throwing. - The gated-stream mock no longer logs the raw Cookie header on 403 — presence only, matching the Authorization logging. - The gated scenario now serves an audio fixture for radio create_link and the Electron e2e covers the radio path end-to-end (bare request 403s, built-in audio player advances past the gate). Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> * fix(playback): claim radio header ownership before awaiting the IPC Codex round-2 P2: leaving the radio route while the header IPC was still in flight left the portal cookie/token installed — ngOnDestroy saw a null scope URL (it was recorded only after the await) and could not clear the override. Ownership is now claimed synchronously before awaiting, destroy invalidates the pending playback continuation, and the apply's stillCurrent verdict is honored. Regression test covers destroy-during-pending-IPC. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> * fix(playback): carry portal headers into collection playback Codex round-3 P1: Stalker channels opened from Favorites/Recently Viewed resolved through StreamResolverService.resolveStalker(), which returned no portal headers — the video path handed the header owner an empty set and collection radio bypassed it entirely, so auth-gated streams still 403'd from collections. - resolveStalker() now builds the same profile as the live layout via the shared classifier: portal-owned streams get mac cookie/Bearer token/MAG UA/portal Origin+Referer, foreign hosts keep the credential-free KSPlayer profile (both create_link results and direct radio URLs). - UnifiedLiveTabComponent applies the scoped override for radio before the audio element gets its URL (ownership claimed before awaiting the IPC, round-2 lesson), and clears it on close and destroy. - Regression tests: resolver header profiles for portal-host and foreign streams; unified tab radio apply-then-clear. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> * fix(playback): release the radio override when a new selection mounts no player Codex round-4 P2: after radio installed its credentials, selecting an item that never mounts a player surface (external video playback, failed resolution) left the old Cookie/Authorization installed — no WebPlayerViewComponent, close, or destroy cleanup runs on that path. Both radio hosts (unified collection tab and the Stalker live layout, which has the identical hole) now release the previously owned radio scope at the start of every new selection; the slot-ownership semantics keep this a no-op when another playback already owns the override. Regression test in the live-layout spec pins the failed-selection path. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> * docs(playback): state the exact override release points Codex round-5 P2 flagged that the media 'ended' event does not clear the scoped override while the player stays mounted. That is deliberate, not a gap: a mounted player still owns the session — replay or a seek into an unbuffered range must keep working against a gated stream, and clearing on 'ended' would 403 exactly the streams this PR fixes. The credentials only ever travel to the exact origin that issued them, and every dismount path (channel/source change, player close/destroy, radio close, playerless selection) releases them. The security doc and the release note now say precisely that instead of the ambiguous "cleared when playback ends". Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> * docs(playback): fit the release note back under the 400-character cap Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> --------- Co-authored-by: Claude Fable 5 <noreply@anthropic.com>	2026-08-02 10:51:12 +02:00
4grayandClaude Fable 5	297e9fbef8	fix(stalker): send cmd in the reference MAG wire format (#1334) ... * fix(stalker): send cmd in the reference MAG wire format A real MAG sends cmd unencoded and the portal decodes its query exactly once, so a cmd that already contains percent sequences (%3A tokens, pre-encoded path segments) must pass through untouched. The previous encodeURIComponent transport (2c032cd3c, 0.22) double-encoded such cmds (%3A -> %253A): strict portals and reseller panels that compare cmd literally, and stock create_link handlers matching the decoded value, saw a different string than a real STB sends. The new shared encodeStalkerCmdValue() reproduces the reference wire bytes: % passes through verbatim, characters the WHATWG URL serializer keeps raw in a query stay raw (so the bytes survive the axios/new URL transport unchanged), and everything else is percent-encoded. That preserves the 0.22 injection protection - &, # (and ; for PHP setups with a ; argument separator) inside cmd cannot append or truncate query parameters; they decode back to the original byte server-side. Both transports now share the format: the Electron query builder is extracted to buildStalkerRequestUrl() and the web-backend /stalker proxy appends cmd to the portal URL itself instead of letting axios turn slashes into %2F (the opposite divergence). Also unifies the two divergent response-side cmd normalizers: the cross-portal collection resolver now uses the Stalker store's normalizeStalkerPlaybackCommand/resolveStalkerPlaybackUrl, so playing from Favorites/global collections resolves relative (/media/...) and query-only (?token=...) create_link replies against the portal base instead of handing the player a bare relative path. The mock portal's create_link response gains mock-only cmd_received/ query_keys_received diagnostics; a new Electron e2e pins the contract end-to-end (single decode, injection blocked). Unit corpus tests cover the encoder, the Electron builder, the web-backend proxy, and the resolver. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> * fix(pwa): sanitize portal URL before appending stalker cmd A registered portal URL carrying a fragment would swallow the appended cmd (everything after # is never transmitted), and a trailing bare '?' produced '??cmd='. Drop the hash and pick the separator from the sanitized href before appending. Flagged by Greptile/Codex on #1334. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> * refactor(pwa): make stalker cmd append visibly query-only for CodeQL Rebuild the /stalker request URL through the URL object and concatenate the encoded cmd strictly behind a literal '?', so static analysis can see the tainted value never reaches host or path (js/request-forgery alert on the previous separator ternary). Behavior unchanged; the fragment/bare-'?' regression tests still pin the wire format. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> --------- Co-authored-by: Claude Fable 5 <noreply@anthropic.com>	2026-08-02 07:38:11 +02:00
4gray	aba89d64cf	fix(downloads): resume interrupted Xtream VOD transfers (#1329) ... * fix(downloads): resume interrupted Xtream VOD transfers * fix(downloads): validate partials before resuming * fix(downloads): propagate headers to episode transfers	2026-08-01 22:01:10 +02:00
4gray	760099358b	feat(downloads): redesign download manager (#1313) ... * docs(downloads): specify manager MVP redesign * docs(downloads): plan manager MVP implementation * docs(downloads): tighten manager validation plan * fix(downloads): keep renderer download state global * fix(downloads): make active count accessible * feat(downloads): derive queue and library view model * test(downloads): close view model coverage gaps * fix(downloads): stabilize malformed view model data * refactor(downloads): isolate library navigation * fix(downloads): report library navigation failures * feat(downloads): add ready-to-watch library * feat(downloads): add active download queue * feat(downloads): finish manager MVP * docs(downloads): clarify detail-first offline behavior * docs(downloads): plan detail navigation follow-up * fix(downloads): open completed movies in details * test(downloads): cover pending series navigation * fix(downloads): honor the global cover size * fix(downloads): prefer local playback in shared details * fix(downloads): preserve external launch priority * fix(downloads): prefer local playback in Xtream details * test(downloads): cover offline detail journey * docs(downloads): document offline detail behavior * docs(downloads): format detail navigation plan * fix(downloads): open Stalker items in provider details * docs(downloads): clarify Stalker navigation fallback * fix(xtream): isolate reused detail identities * fix(xtream): ignore stale VOD positions * fix(downloads): keep offline Xtream playback available * docs(downloads): clarify provider playback availability * docs(downloads): design missing-file recovery * docs(downloads): plan missing-file recovery * feat(downloads): derive completed file availability * feat(downloads): recover missing completed files * feat(downloads): refresh missing local files * feat(downloads): separate missing files from ready media * feat(downloads): surface missing files for recovery * refactor(downloads): simplify ready cards * test(downloads): cover missing-file and series journeys * feat(downloads): finish missing-file recovery * docs(downloads): design offline detail views * docs(downloads): plan offline detail views * feat(downloads): persist offline metadata snapshots * fix(downloads): complete metadata snapshot bridge contract * feat(downloads): manage offline metadata snapshots * fix(downloads): harden metadata snapshot updates * fix(downloads): restrict snapshot artwork * fix(downloads): guard restart artwork URL * fix(downloads): refine artwork URL checks * feat(downloads): expose offline metadata updates * fix(downloads): keep metadata service change focused * fix(downloads): preserve metadata error conventions * feat(downloads): derive offline detail content * fix(downloads): preserve unknown episode coordinates * feat(downloads): add focused offline detail routes * fix(downloads): ignore fragments in shell route state * fix(downloads): normalize fragments before queries * feat(downloads): open ready cards in offline details * fix(downloads): use native disabled card styles * feat(downloads): enrich offline detail metadata * fix(downloads): harden offline metadata resolution * fix(downloads): preserve stalker provider titles * fix(downloads): distinguish stalker metadata seeds * fix(downloads): stabilize offline metadata refresh * fix(downloads): throttle sparse metadata refreshes * fix(downloads): type metadata language settings * feat(downloads): render offline movie and series details * fix(downloads): harden offline detail interactions * fix(downloads): close offline detail edge cases * feat(downloads): hand off to provider-only details * fix(downloads): preserve stalker provider handoff * feat(downloads): capture metadata at download time * fix(downloads): preserve snapshot source semantics * fix(downloads): preserve episode snapshot identity * docs(downloads): document offline details flow * docs(downloads): clarify stalker provider fallback * test(downloads): cover offline detail journeys * test(downloads): stabilize offline detail selectors * style(downloads): format changed files * docs(downloads): clean design spec formatting * fix(downloads): preserve offline library ownership * test(downloads): fix Windows workspace navigation * test(database): preserve Electron tsconfig resolution * perf(downloads): avoid blocking file availability probes	2026-08-01 18:09:31 +02:00
4grayandClaude Fable 5	46c58f4f57	fix(stalker): keep session headers on same-host redirects (#1322) ... * fix(stalker): keep session headers on same-host redirects Since 0.22 requestWithValidatedRedirects stripped Cookie/Authorization whenever a redirect changed the *origin*, so a portal answering with an http->https upgrade or a port move lost the MAC cookie and Bearer token mid-session. Real Stalker/Ministra servers then reply with a plain-text "Authorization failed." body: categories fail to load, create_link never resolves, and no player receives a stream URL (#1158 regression window). Scope credential stripping to the host instead: same-host scheme/port redirects keep headers, basic auth, params, and request bodies; a redirect to a different host still drops all of them, preserving the original hardening intent (no credential leaks to third-party hosts). Also adds the Stalker API compatibility roadmap produced by the 2026-08-01 protocol audit (.plans/, force-added like earlier plans). Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> * fix(electron): strip credentials on same-host https-to-http downgrades Review follow-up (Greptile P1 + Codex on #1322): the host-only check kept Authorization/Cookie/basic auth/params/body when an https request was redirected to http on the same host, replaying a TLS-obtained session in cleartext. Treat that downgrade like a host change: strip credentials and refuse to replay request bodies. Scheme upgrades and port moves on the same host keep headers — the actual #1158 scenarios. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> --------- Co-authored-by: Claude Fable 5 <noreply@anthropic.com>	2026-08-01 17:36:44 +02:00
4grayandClaude Fable 5	9bb1811984	fix(playback): escape commas in mpv header fields on the TS paths (#1323) ... * fix(playback): escape commas in mpv header fields on the TS paths PR #1321 fixed the comma truncation of --http-header-fields inside the native embedded-mpv addon, but the same OPT_STRINGLIST parsing bites three TypeScript call sites that join header fields with ',': - external MPV CLI launch (--http-header-fields=...) - external MPV instance reuse (set_property http-header-fields) - embedded MPV frame-copy loadfile options (opt.http-header-fields); the helper's %len% quoting protects only the option-list level, mpv still stringlist-parses the value afterwards The Stalker MAG user agent contains "(KHTML, like Gecko)", so strict portals received a truncated X-User-Agent and rejected live streams with HTTP 400 (#910). Escape backslashes and commas per field with the same scheme as the native fix, via a shared util. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> * test(playback): cover the reused-instance IPC header escaping Review follow-up (Greptile on #1323): the regression coverage only exercised the CLI spawn path. Capture the JSON IPC traffic of a second, reused mpv launch and assert the escaped MAG user agent survives the set_property http-header-fields transport, so later serialization changes cannot silently reintroduce truncated headers. net is mocked with a passthrough default because the VLC specs bind a real ephemeral port via createServer. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> --------- Co-authored-by: Claude Fable 5 <noreply@anthropic.com>	2026-08-01 16:19:38 +02:00
wretchwatson	814161274b	fix(playback): escape commas in mpv http header fields (#1321)	2026-08-01 15:17:48 +02:00
4gray	2ac0de752f	fix(skills): align repository guidance with implementation (#1315) ... * docs(skills): design implementation synchronization * docs(skills): plan implementation synchronization * fix(release): filter internal notes from public body * docs(release): synchronize release workflow guidance * fix(stalker): normalize catalog series flags * fix(stalker): preserve progress with scoped episode IDs * fix(playback): expose strict position persistence * docs(stalker): record series position compatibility * test(skills): validate repository skill contracts * fix(database): keep SQL trace values private * docs(skills): refresh Nx and SQLite ownership * docs(skills): align provider and UI guidance * docs(skills): tighten validated guidance * docs(release): require exact release pushes * style(electron): remove trailing blank line * fix(ci): classify repository skills coverage	2026-07-31 08:00:59 +02:00
4gray	78df3e7dbb	fix(portals): match Greek titles whichever sigma the provider typed (#1310) ... Greek Σ has two lowercase forms — medial σ and word-final ς — and neither the candidate query nor the confirmation treated them as one letter. The GLOB scan built each character's class from a one-way reach that only arrived at ς when it started from ς, so a request for "ΑΣ" never admitted a stored "Ας". Classes are now built from a fold group — every character sharing an uppercase form — derived by scanning the cased ranges at module load the way ACCENTED_BY_BASE already is. It generalises past sigma on its own: dotless ı folds with i, long ſ with s, historic Cyrillic letterforms with В Д О С Т Ъ Ѣ. Only the 24 groups of 767 that a per-character fold would miss are kept. Admitting the row was only half of it. normalizeTitleKeys then compared "ασ" against "ας" and discarded it, because toLowerCase picks the sigma form by position. Both SQL tiers already folded them together — SQLite's trigram tokenizer does full Unicode folding natively, unlike LOWER() — so the JS confirmation was the only tier that did not, making this a pre-existing gap on the FTS path as well. Normalization now rewrites ς to σ after lowercasing, which is what Unicode case folding does. Guards unchanged: a case mapping that changes length (ß → SS, İ) or a GLOB metacharacter still returns null rather than a partial pattern.	2026-07-29 23:14:59 +02:00
4grayandClaude Opus 5	063662028a	feat(portals): find the same movie in your other playlists (#1286) ... * feat(portals): find the same movie in your other playlists A movie that exists in several imported Xtream playlists now shows a "Sources N" chip on its detail page and in the player. Switching playlist mid-film keeps the timecode, a preferred source can be pinned per movie, and a failed stream offers the alternatives instead of a dead end. The governing rule is that a guess is never presented as a fact. Every metadata value carries where it came from — `api` (the provider said so), `parsed` (inferred from the title) or `probe` (we contacted the stream). Facts render as plain tags, guesses are prefixed `~` in a warning colour, and an unknown value renders no tag at all plus a "check" affordance. Ranking and failover read through `factualOnly()`, so a filename claiming 4K is structurally unable to outrank a source that was actually reached. A probe that could not complete reports "unknown", never "unavailable". Scope is deliberately narrow: Xtream to Xtream, movies only, Electron only. Stalker never reaches the `content` table and M3U is a JSON blob whose search forces live content; both are additive later, since the candidate type already carries all three portal kinds. In the PWA every entry point is gated off and the chip renders nothing. Auto-failover is opt-in and off by default. Each source is tried at most once per session, so it terminates structurally, and the switch is never silent — the toast names the new playlist, offers an undo, and warns that the dub may differ only when both sides state an audio track as fact. Notable details: - Playlist names are routinely the pasted URL, credentials included. They are never rendered raw; a short host-only label is derived instead. - Quality is derived from pixel width, not height: a 2.39:1 1080p master is 1920x800, and bucketing that by height would publish "720p" as a fact. - Switching is a single `inlinePlayback.set()` so the player and engine survive and re-seek; the carried position is read before the 15s persistence throttle so it does not rewind. - Sources from one playlist collapse into a group, since the same film often appears there several times under different stream ids. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> * fix(portals): stop stale source resolutions from committing Addresses three defects Greptile found in the multi-source review. **Concurrent switches committed out of order.** Selecting a second source before the first resolution returned let the slower request overwrite the newer selection and repoint Undo at itself. `switchTo` now takes a sequence number and drops its result if a newer switch already committed. **Stale switches crossed movie sessions.** Navigating to another film while a resolution was in flight let the continuation activate the old film's source inside the new controller — and restart it from that session's zero resume position. The controller is now snapshotted per operation and the movie session is revalidated after every await. `check()` had the same hazard across its two awaits and is guarded the same way. **Short titles skipped discovery entirely.** The trigram tokenizer cannot index tokens under three characters, so "Up", "It" or "Us" produced an empty MATCH expression and the query was discarded before SQLite was consulted — the chip could never appear for those films. Discovery now falls back to a bounded scan when FTS structurally cannot serve the title; the existing two-tier normalized confirmation still rejects loose hits like "Upgrade". Each fix carries a regression test; all three were mutation-checked by removing the guard and confirming exactly those tests fail. The previous test asserting that short titles return nothing encoded the bug and has been replaced. The host spec passed 400 lines, so its fixtures moved to a shared module and the race suite into its own file. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> * fix(portals): make the pin decide playback and keep failover going Second round of Greptile review findings. **A pin had no behavioural effect.** Loading a stored pin only decorated the row: Play still started the route's playlist and failover ranking ignored `isPinned`, so "make this the main source" survived a restart as an icon and nothing else. The primary action now starts from the pinned source when one is set, and the pin outranks everything else in failover ranking. **Failover stopped at the first unresolvable candidate.** An expired account or a failing `get_vod_info` on the top-ranked source ended the attempt, and since production calls `failover()` only once — on the original playback failure — a healthy lower-ranked source was never reached. It now continues through untried candidates. `switchTo` reports why it stopped so the loop can tell "could not resolve, try the next one" from "something newer owns the screen"; without that distinction a superseded switch would have spun forever, because only the former marks the candidate tried. **Identity ignored enrichment.** The key was `playlistId:contentId:title`, so when `get_vod_info` added a TMDB id and release year to an unchanged title the host saw no change, never reloaded, and kept yearless discovery and title-only pin keys — a `tmdb:`-keyed pin could never be found. The key now covers every field that affects matching. **A server refusing HEAD read as unavailable.** Some stream hosts answer 405 or 501 to HEAD yet serve the media over GET. The probe now retries once with the ranged GET the main process already supported, instead of caching a working source as failed and penalising it during failover. Greptile also flagged a missing token check after the resolve await in `switchTo`; that guard landed in 4db3a2fd and sits on the line directly below. Answered on the thread rather than changed. The host service passed 400 lines again, so the pin, probe, switch-notice and current-row concerns moved into focused modules beside it. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> * docs(portals): record the behaviour the review rounds changed The architecture doc and CLAUDE.md described the feature as first written, not as it now behaves: pins were documented as a stored preference without saying they decide playback, failover was described as stopping at the first unresolvable candidate, the probe as HEAD-only, and discovery as pure FTS with no mention that short titles cannot be tokenized at all. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> * fix(portals): invalidate the session while the movie identity is empty The staleness guard added in 4db3a2fd bumped the session only inside `load()`, which leaves a window the guard does not cover: route navigation empties the movie identity first, and `load()` for the replacement runs only once a title is knowable again. A resolution completing in that interval still carried a session number that matched, so it passed the check and started the previous movie's source over the page the user was navigating to. The binding effect now bumps the session as soon as the identity goes null, so anything already in flight is invalidated at the moment the old movie stops being the one on screen rather than when the next one finishes loading. `lastMovieKey` is deliberately left alone: returning to the same movie should not re-run discovery, and the controller's state is still correct — only the in-flight operations needed invalidating. Regression test added and mutation-checked: removing the bump fails exactly that test. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> * fix(portals): stop the source list from losing the real alternatives Six review findings, all in how multi-source decides what to show and what it is playing. Discovery: the current playlist is now excluded in SQL rather than after the fact, so its own duplicate rows can no longer spend the whole row budget before a single alternative is read. The short-title scan matches the token as a word instead of a substring and orders by title length in a wider window, so "Titanic" and "The Italian Job" cannot push the real "It" out of it. Session: metadata enrichment re-runs discovery for the film already on screen. That is a refresh, not a new session — a second identity key (playlistId:contentId) now separates the two, so the source the user switched to keeps playing and stays named, the tried set stays burned, the position survives and a switch in flight still commits. Resume: the multi-source controller no longer records the engine's pre-seek timeupdate at ~0. The playback service's one-shot latch now reports whether the position can be believed, and until it can, the requested start time stands in — so a switch during the initial seek does not restart the film. UI: the in-player sources picker gets the same auto-failover setting and match kind as the detail page's, instead of always rendering the default and dropping the toggle. The caption counts distinct playlists, not stream variants, since the popover groups a portal's copies under that portal. Session mechanics and the pin toggle move into their own modules to keep the host service inside the line budget. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> * fix(portals): keep the playing row when the refined year rejects it Follow-on from keeping the session across a rediscovery. The rerun can legitimately drop the row that is playing: enrichment supplies the release year, and the year gate then rejects a copy the yearless search had admitted — "Dune" 1984 while the user is watching the 2021 film. Off the list is right; it is not the same film. Off the screen is not. It is what is streaming, so it stays as a row and keeps the playing badge, rather than letting the caption name a playlist that is not sending any bytes. Also covers the new session key directly in the identity spec. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> * fix(portals): stop a pin write from landing on the next movie Two findings from the review of the previous round. A pin write is an IPC round-trip, and the user can navigate during it. The continuation then applied one film's answer to another film's controller — and because unpinning returns "nothing pinned", it would clear the pin the new movie had just loaded and its Play action would quietly stop starting from the preferred source. It now commits only while the same film is still on screen, like every other async path here. The short-title scan drops its row limit. FTS keeps its window because it ranks by relevance, so what it keeps is what matters; a scan cannot rank, so a window there silently decides which valid sources the user is allowed to see. It also bought nothing: the GLOB cannot use an index, so SQLite reads every row either way and the limit only truncated the answer. What bounds the scan is its predicate — reaching it means the whole title is one or two characters. The switch-notice type moves to the module that builds it, which also removes a circular type import between the two. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> * fix(portals): keep external playback, the pin and the resume point honest Five findings from the round-5 review. An external player launched for an alternative carries that playlist's ids, so the page disowned its own session: the primary button never became Stop, stopping found nothing to stop, and another click opened a second player. Multi-source now tells playback which source is actually active, and the matcher accepts either that or the route's own stream. Stop also has to beat the pin. The primary action consults the pin first — that is what makes a pin decide where playback starts — but while a session is running the same button reads Stop, and consulting the pin there made the control do the opposite of its label. A pinned source started from the Resume button resolved at zero, because nothing reports a live position until the first timeupdate. The controller is now seeded from the persisted position, one-way: a live value always wins, since the stored one lags it and applying it would rewind. A pin whose write failed was still shown as pinned, promising a preference that reopening the movie would not have. Portal failures in this path logged raw errors. An Xtream error message carries the stream URL, and that URL is built out of the username and password, so they now go through the redacting logger. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> * fix(portals): make every alias of a pin agree, and stop losing rows Four findings from the latest review pass. A pin lookup accepts several aliases of the same movie, but a write only touched the most-trusted one — so after enrichment the title alias still pointed at whatever was pinned before, and a reopen that read it (because TMDB had not landed yet, or its request failed) started the source the user had just replaced. Writes now go to every alias. That alias set was also missing one. Enrichment supplies the year as well as the id, so a pin set before either existed is stored yearless; the candidate list skipped that form entirely and orphaned the row. Discovery could lose whole playlists: one playlist listing a film in dozens of categories produces identically ranked rows that fill the window before another playlist is read. The collapse now happens in SQL, before the limit, rather than in TypeScript afterwards where the missing rows are already gone. And an abandoned source pick finishing late cleared the spinner from the row the user was actually waiting on. Removes `isExhausted()` from the host service — no caller outside its own tests, where the assertion above it already proved the same thing. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> * fix(portals): probe like playback, and stop the pin answering for a remake Five findings from the latest review pass. Writing a pin to every alias — last round's fix for stale aliases — was wrong in the other direction: `title:{base}:` is shared by every remake, so a known-year decision stored there answers for a different film. Pin Dune (2021), open Dune (1984) before its year arrives, and it would start the 2021 source. A write now clears every alias and stores only the canonical key, which retires the stale ones without making any of them ambiguous. The probe checked a bare URL while playback sends the playlist's User-Agent, Referer and Origin. A panel that requires them answers 401/403, so a stream that plays perfectly was reported dead and penalised in failover ranking. The switch toast interpolated the raw playlist name. Users routinely name a playlist after the URL they pasted, so that line could put credentials over the video; the notice now carries the same safe label the rows use. External players have no timeupdate, so their polled position IS the live one. Feeding it through the seed — which stops at the first value — froze the resume point where playback started, and a switch an hour in rewound to the beginning. And auto-failover concluded "nowhere to go" when a stream failed before discovery answered, stranding the user on the error screen. Moves `switchTo` into the session module, which is where the rest of the switch mechanics already live, and splits the route spec along the same rendering/behaviour seam the other suites use. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> * fix(portals): re-check the movie after waiting, and follow the alternative Three findings, two of them regressions from the previous round. Awaiting a pending discovery before failover let the user navigate during that wait: the continuation then ran against whatever controller was current and could answer one film's playback failure by starting another film's alternative. Both waits — failover and pinned Play — now re-check that the same movie still owns the screen. Pinned Play also needed the wait it did not have. Pressing Play while the pin lookup was still out concluded "nothing is pinned" and started the route's own source, making a persisted preference depend on worker latency. And the position bridge still accepted only the route's ids, so an external player running an alternative had every progress update discarded: the resume point stayed where playback began and a switch an hour in rewound the lot. The session matcher and the bridge now share one ownership predicate, since a page that shows Stop for a session whose progress it throws away is the bug in two halves. The test for the external case previously set the position signal directly, which bypassed the very filter that was broken; it now drives the bridge. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> * fix(portals): stop a remake matching, and let a pin survive its own playlist Three findings from the latest review pass. `normalizeTitleKeys` strips bracketed segments as tag noise, so "Dune (1984)" normalizes to exactly "dune" — an EXACT match for the 2021 film, ranked above every fuzzy one, with the year never consulted because that tier skipped the gate. Auto-failover could switch the user to the other film entirely. The year is now read out of brackets too, and a stated disagreement rejects the row on either tier. Playback positions are keyed by (playlist, stream), so watching through a pinned alternative stores progress under ITS ids while the page loads the route copy's row. Starting the pin therefore resumed from a position belonging to a different copy — usually zero. It now loads its own. And a pin can point at another copy of the film inside the playlist being viewed, which discovery excludes wholesale: the pinned row was absent from the list, so nothing showed as pinned and Play ignored the preference. The pin is now read before discovery, which keeps that one row. Moves the pin-shaped decisions into the pin module, where the persistence helpers already live. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> * fix(portals): keep a pinned play, a same-playlist copy and Check honest Five findings from the latest review pass. Reading a pinned source's own position is a database round-trip, and the user can navigate across it — the continuation then handed one film's source id to whichever movie now owned the screen. Guarded, like every other await here. Allowing a pinned copy to live in the current playlist made "is this the route's own source?" a two-part question, and the ownership check still asked only about the playlist: an external session for that copy was disowned, so Stop vanished and its progress was dropped. The yearless title alias is shared by every remake, so clearing every alias before a write could delete a different film's pin. Writes and unpins now touch only keys that name one film — plus the ambiguous row this session actually read, which is the one the user is looking at and the one whose absence would make an unpin come back. Restart left the seeded position in the controller, so a failure before the first timeupdate resolved the next source back at it. And the alternative rows on the playback-error screen had a Check button wired to nothing at all. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> * fix(portals): stop a rediscovery restoring the pin it started with A same-movie rediscovery read the pin, then held that snapshot across its source lookup and applied it afterwards. A pin made while the lookup was out was therefore overwritten by the older value: the row and the primary Play action named a source the database no longer held. The snapshot is now applied as soon as it is read, so a later write simply wins on ordering rather than needing to be detected. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> * fix(portals): write the pin before retiring it, and keep the badge honest Three findings from the latest review pass. Repinning cleared the old rows and then wrote the new one, so a write that failed after the clear left nothing persisted while the row still showed the old pin. The order is reversed: the new key is stored first and the stale ones retired only once it landed. Lookups are most-trusted-first, so a leftover alias never outranks what was just written. Starting a source from the picker, or letting a pin decide the primary Play, never recorded the movie as recently viewed — unlike every other way of playing it. And closing an alternative's player and pressing Play started the route stream while the controller still marked the alternative active, so the picker and caption named a source that was not running. Moves the discovery pass into the session module beside the switch and failover mechanics, and splits the pin spec along the persistence/playback seam, both to stay inside the file-size rule. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> * fix(portals): stop claiming playback, a cached answer and a resolution Three findings, all of them the same rule: never state as fact something the app has not established. The "Playing from …" caption appeared as soon as discovery marked a source active — before Play was pressed, and again after the player was closed. It now requires a player that is actually running. Probe answers were cached by URL alone, but the request now carries the playlist's headers. Two playlists sharing a stream URL could therefore be told the other's answer, marking a source dead without ever asking it. And any width below 900 was labelled 480p, published with `api` provenance: a 640x360 stream stated 480p as a fact, and a 720x576 PAL source likewise. Widths below HD only resolve with the height — 720 is NTSC 480p or PAL 576p — so an unrecognised shape now carries no quality tag at all. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> * fix(portals): match the sub-HD formats, and drop the caption on failure Two follow-ups to the previous round, both the same rule again. The 800-wide band still answered from the width alone, so 800x600 and 800x450 were labelled 480p — published with `api` provenance, so read as a measurement. Sub-HD formats are now matched against known shapes with the same 5% tolerance the height path uses, and anything unrecognised carries no tag at all. And "Playing from ..." survived a playback failure: the inline host stays mounted while the diagnostic is on screen, so the page named a source for a stream it had just reported it could not play. The caption now clears on failure and returns when the engine produces time again. Splits the route playback spec along the "what it does" / "what it claims" seam and lifts the repeated active-source stub into one helper. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> * fix(portals): let the height veto a width match, and hold the failure state Two follow-ups to the previous round, both in code it introduced. A width that matched exactly one sub-HD format ignored the height entirely, so 640x480 came back as 360p — a measurement the numbers contradict. The height now vetoes, but only in the direction that can be wrong: cropping removes lines, so a SHORTER frame is a letterboxed master of that format and the width still names it, while a taller one is a different shape and gets no tag. That keeps the reason width is preferred in the first place. And picking a source off the error screen cleared the failure state before the switch resolved, so an alternative that could not be resolved left the diagnostic on screen while the caption went back to claiming playback. The flag now clears only once a switch actually starts something. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> * fix(portals): release the resume latch when the target cannot be reached Carrying a position into a shorter cut of the same film — two hours into a 90-minute source — leaves the engine unable to ever report that time, so the one-shot latch never released: every position save was suppressed for the rest of the session, and the impossible start time kept being reported to multi-source for the next switch. The latch now also opens when a known duration puts the requested point out of reach, while a reachable one still waits as before. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> * fix(portals): say "Playing" only while something is playing `isActive` means "the source a switch or Play would use". Discovery sets it the moment the page opens and it survives closing the player, so it could not back the two claims the UI made in the present tense: the "Playing from" caption and the source row's Playing badge. Both appeared on a page where nothing had started, and came back after the player was closed. `playbackLive` is now that statement, and both read it. Inline it needs a timeupdate — `inlinePlayback()` is only the REQUEST to play, non-null while the engine is still opening the stream and still non-null after it fails — and external it needs the session past `launching`. A row that is merely selected reads "Current" (new key, filled for all 19 locales). Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> * fix(portals): start a never-watched pinned source from the beginning Positions are keyed by (playlist, stream). When the pin points at a copy the user has never opened, the lookup returns nothing and the controller was left holding the ROUTE copy's position — so Play dropped them 42 minutes into an unstarted film, and the first save wrote that timecode back under the pinned source's key, making it permanent. The spec asserted the old behaviour, so it is flipped rather than extended; a second case covers the host that supplies no lookup at all, where "never watched" was never established and the position must be left alone. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> * fix(portals): describe the copy the primary button will actually play Two gaps found by review. A pin makes the primary button play a copy the page never loaded a position for — positions are keyed by (playlist, stream). The label, timecode and Restart affordance still came from the route copy's row, so the button could read "Resume 42:18" and start an unwatched copy at zero, or read "Play" and jump into the middle of one already watched. `createPrimaryActionPosition` lets the pinned copy's row govern, including when that row is absent: never watched is an answer, not a fallback to someone else's progress. A manual source switch also mounts a DIFFERENT stream in the same host while marking the new source active at once, so the previous stream's timeupdate was still vouching for it — the caption and the badge claimed the new source while it was still opening. That path now clears the latch like Play and Restart do. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> * fix(portals): keep the route's own resume point, and honour a closed pin Two more from review, both variations on "selected is not playing". `vodPlaybackPosition` followed whichever copy last reported — so after an alternative played, Resume and its label described that copy's row while starting the route's stream, jumping it to a timecode nobody reached in it. It now splits: `vodPlaybackPosition` stays the last position seen (the progress bar and the switch handoff want the stream on screen), and `routePlaybackPosition` holds the route copy's own row for everything that acts on the route's stream. `pinnedSourceAwaitingPlay` skipped the pin whenever its row was active, but `isActive` means selected — the pinned row stays selected after its player is closed, so the next Play went to the route copy and ignored the stored preference until the page was reopened. It now takes `playbackLive` too. The host service crossed the 400-line cap on the way, so the four derived alternative counts moved into `vod-multi-source-counts.ts`. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> * fix(portals): keep the primary button honest across navigation and pins Four follow-ups from review, all consequences of splitting the position signals. - Route reuse (the Similar rail) cleared only `vodPlaybackPosition`, so the button kept the previous movie's Resume label — and start point — until the new lookup landed. Both signals and the playback latch now reset together. - The primary button's fall-through past an unresolvable pin reached the service directly, skipping the bookkeeping a route start needs: the controller kept the alternative's timecode and the old stream's timeupdate still vouched for the new one. It now goes through the route's own wrappers, and Resume seeds the controller with the ROUTE copy's position. - `alternativePlaylistCount` counted the playlist being watched whenever it held a second copy, so "also found in 2 other playlists" could mean one. - The pinned copy's stored row went stale the moment the user watched it; its live position now wins while it is the one playing. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> * fix(portals): do not spend a source's failover turn on mere selection `setActiveSource` marked the source tried, but discovery calls it the moment the page opens and a pin or the picker can call it before anything plays. So opening a movie burned the route copy's turn: if a pinned alternative then failed, failover skipped a healthy untouched source — and with only one alternative, reported the options exhausted. Selection and attempt are now separate. `setActiveSource` selects; `markPlaying` also spends the turn, and only the three places that really start playback call it. `runFailover` additionally retires whatever is on screen before picking, so the failing source is spent however it got there — relying on the start paths alone would leave one hole per path, and the cost of missing it is a ping-pong between two sources. One existing spec asserted the old behaviour (a route copy burned by a switch it never played); it now plays first, so it still covers what it meant to — that the tried set survives a rediscovery. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> * feat(portals): carry VOD source pins through playlist backup The new pins table was invisible to backup: exporting a playlist and re-importing it on a new machine silently dropped every "main source" choice, with nothing in the archive to say the choice had ever been made. Pins now ride along under the playlist they point AT — carrying them anywhere else would restore a preference for a portal the archive never contained. `matchKey` names the film rather than the portal, so it survives untouched and only the playlist id is remapped to the imported copy. `sourcePins` is the one optional collection in the Xtream user state: archives written before multi-source existed simply do not have it, so its absence is age rather than damage. Only a wrong type is rejected, and pins without a usable match key or content id are dropped, since writing one would occupy the unique key of a film it does not describe. Adds `DB_LIST_VOD_SOURCE_PINS` through the usual six seams (operation, worker case, event, preload, bridge contract, service). Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> * test(portals): follow the normalized restore state's new collection `normalizeXtreamPendingRestoreState` now always emits `sourcePins`, like every other collection it canonicalizes, so three specs that assert the exact normalized shape had to follow. Adds coverage for the sanitizing itself: a pin without a usable match key or content id is dropped, and a non-string `updatedAt` is discarded rather than carried. Caught by CI, not locally — the earlier full run served `playlist-shared-ui` from the Nx cache, so it reported green on a stale result. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> * refactor(portals): lift the VOD route's orchestration out of the component The details route had grown to 864 lines — the repository's hard maximum is 400, and while the file predates the rule, a baselined exemption is not a budget to spend. Three component-provided services now hold what the component was accumulating: `VodDetailsMultiSourceUiService` (the playback-evidence latch, the caption, the primary button's position, source actions and the failover toast), `VodDetailsSimilarService` (the rail and its cross-portal lookup), and `VodDetailsDownloadsService`. The component keeps its public API, so the template and the existing specs are untouched. 864 -> 566 lines. The downloads move also fixes a latent bug: `downloadVod` and `playFromLocal` read `route.snapshot.params`, which is stale once the router reuses this component for detail-to-detail navigation (the Similar rail) — so a download started from a film reached that way fetched the previous one. They now read the same route-params signal everything else uses, with a regression test. Also from review: pins are applied on the FRESH-import path too. A new playlist has no content when the archive is read, so its user state is parked and replayed after the import — the merge path I wired first never ran there, and every pin was dropped. A failed pin write now propagates instead of being ignored: the backup entry is reported failed, and the parked state is kept so a transient failure can be retried rather than silently losing the preference. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> * fix(portals): read array-shaped codecs, bound short-title scans, honour alias clears Three from review. `info.video`/`info.audio` are declared — and sent by the mock server and many panels — as string arrays, but the resolver only read the ffprobe object shape. Every array response therefore lost the provider's codec, so those source rows showed no codec fact and the "dub may differ" warning could never fire. `readStreamInfo` now accepts both, and states nothing when the provider stated nothing. The FTS-empty fallback scan matched only the FIRST token, which is fine for a one-word short title but not for "I Am": every catalog row containing the word "i" came back for TypeScript to throw away — a full scan of a large catalog on the single database worker, just to open a detail page. Every token must now appear. `writePin` reported success when the canonical write landed but retiring the old alias failed. Lookups read aliases before the canonical key, so reopening the movie before enrichment would start the source the user just replaced, with the icon promising otherwise. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> * fix(portals): write a pin and retire its aliases in one transaction Split across two calls, a half-failure had no honest outcome. Reporting success left a surviving alias to win the next lookup and start the source the user had just replaced; reporting failure — which the previous round changed it to — left the canonical row durable while the UI showed a pin that was no longer the stored one. Review was right both times, which is the tell that the two-step shape was the problem. `setVodSourcePin` now takes the keys to retire and does both inside one `db.transaction()`, with the synchronous `.run()` form the better-sqlite3 driver requires there (issue #1137's lesson). `retireKeys` rides through the worker op, the IPC contract, the preload bridge and the service, so there is one call and one outcome. Also corrects the architecture doc: the scan path is reached whenever no token clears the trigram minimum, not only when the whole title is one or two characters — the claim the previous commit's code change had already falsified. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> * fix(portals): tell a superseded pinned play from an unusable pin Double-clicking Play while a pinned source resolves put both handlers into `playPinnedSource()`. The second supersedes the first, so the first returned `false` — which the route read as "no usable pin" and answered by starting the route source over the playback the second click had just begun. `playPinnedSource` now reports `played` / `superseded` / `unavailable`, and only `unavailable` falls through. This is the same distinction `runFailover` already draws between "keep going" and "stop, something newer owns the screen"; the pinned path simply never had it. The host crossed the 400-line cap again on the way, so the pinned-play errand (wait out an in-flight discovery, re-check the session, start the source) moved into the pin module beside `playPinned`, and the pin-toggle commit went with it. 388 lines. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> * fix(portals): restart honours the pin, and a switch replaces the player Three from review, all in the pinned-playback seam. A pinned copy watched through resolved to its stored seconds, so the button read Play — the label uses the in-progress rule — and then started near the end. Both now go through one `isResumablePosition`, so the label and the start point cannot disagree. Restart sat beside a Resume that honours a foreign pin, but called `playVod` and started the ROUTE copy — silently switching the user's playlist. It now restarts whatever the primary button acts on, falling back to the route source only when there is no usable pin. Switching sources left a running external player alone. With MPV or VLC and instance reuse off the backend spawns a second detached process, so both sources kept playing and Stop owned only the newer one. Also merges master, and puts the five host specs on a shared harness — they each carried the same 31-line TestBed, which is what pushed two of them over the file-size cap as cases were added. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> * fix(portals): find short Unicode titles, and stop two pickers racing Five from review. Greptile's P1: a short non-ASCII title was undiscoverable. SQLite's `LOWER()` and GLOB classes are ASCII-only, so "он" never matched a stored "Он" and the source simply never appeared. ASCII tokens keep the word-boundary GLOB; a non-ASCII token falls back to a substring test against both the folded and the as-typed form, which the normalized confirmation afterwards makes safe. A probe now retries the ranged GET for 400 and 403, not just 405/501 — those are what a WAF returns for an unexpected HEAD on a URL it serves happily over GET, and calling that source dead also ranked it below worse ones. Three races, all the same shape as ones fixed earlier in this branch: - a pinned play awaiting its resume lookup did not notice a source picked across it, and finished last, replacing the user's choice; - two overlapping switches both saw the same external session, both awaited its close, and both launched — two detached players again; - the primary button showed the ROUTE copy's Resume while the pinned copy's row was still loading, so a click started somewhere else entirely. Also puts the races spec on the shared host harness, which is what keeps it inside the file-size rule now that it carries two more cases. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> * fix(portals): close the player we launched, not the one we now own Three follow-ups, two of them to last round's own fixes. The external-session close was defeated in exactly the case it was written for: `switchToSource` marks the DESTINATION active before handing playback over, so by the time the service ran, the process still playing no longer looked like ours and was left running beside its replacement. The service now remembers the ids it launched with, independently of what is active. The ASCII/Unicode branch was decided from the NORMALIZED token, which folds diacritics — "Ça" arrived as "ca", looked like plain ASCII, and took the GLOB path while the stored title still read "Ça". Decided from the raw token now. Backup restore upserted archived pins but never removed the playlist's existing ones, so a present-but-empty collection left stale preferences alive — unlike the playback positions cleared beside it. An absent collection (an older archive) still means "no opinion" and is left alone. Four files crossed the size cap on the way; the split ones now share `title-sources.spec-data.ts` and `playlist-backup.xtream-fixtures.ts`, and the external-session ownership moved to its own module. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> * fix(portals): absent is not empty, and every start claims the generation Four more from review, three of them defects in last round's fixes. The restore normalizer materialized `sourcePins: []` for archives that never had the field, so "absent means no opinion" became "this archive says there are no pins" and a merge cleared the user's. Absent now stays absent. My test for that behaviour had passed for the wrong reason — it stubbed an empty pin list, so the clear was skipped whether or not the guard worked. `startGeneration` was claimed only by the switch path, so a plain Play, Resume or Restart could be overtaken by a switch still awaiting its close. Every start claims it now. Raw and normalized tokens were paired by position, which breaks when normalization drops a whole word: "FR: Ça" normalizes to "ca" and got handed the raw token "FR:", sending it down the ASCII branch it cannot match from. They are paired by normalized form instead. And the ambiguous yearless alias (`title:dune:`) is no longer written or retired beside a precise key — it may hold another remake's pre-enrichment pin. It stays available when it is the only key there is, since refusing to pin at all would be worse. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> * fix(portals): a failed close must not leave the page claiming a dead source When `closeSession()` rejected, `startResolvedPlayback` rejected with it and never launched — while `switchToSource` had already marked the destination active and reported the switch as successful. The page then named a source that nothing was playing. The close failure is logged and the replacement starts anyway. A close that rejects usually means the session was already gone, and a possibly-lingering process is the lesser of the two evils: the alternative is a UI that lies about what is on screen. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> * refactor(portals): split the external-playback handoff out of the service Both the service and its spec crossed the 400-line cap with the close-failure handling, so the handoff — deciding which process is ours, closing it, and surviving a close that rejects — now lives in `vod-details-external-session.ts` with its own spec file. Two tests had to start awaiting: replacing a running external player is a round-trip, and the handoff now yields once even when there is nothing to close, so the new playback is mounted a microtask later than before. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> * style(portals): format the extracted external-session module Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> * fix(portals): fold diacritics in the title index Cross-playlist matching compares normalized titles ("Amélie" -> "amelie") against an index built from the raw title, and the trigram tokenizer does not fold diacritics by default. Every accented title was therefore invisible to the FTS path: two identical `Amélie` entries produced no candidates at all. That is the broadest of the Unicode gaps review found, and it predates the short-title work. The tokenizer is fixed at CREATE time, so existing databases recreate and rebuild the index once behind a migration marker. `remove_diacritics` needs SQLite 3.45+, so support is probed on a temp table first: an older runtime keeps its working index untouched and the migration is not recorded as done, leaving a later version free to upgrade it. Case folding for non-ASCII remains impossible in stock SQLite — "ОН" cannot find "Он" by any available predicate — and is documented as the known limit rather than patched around again. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> * fix(portals): clear a playlist's pins by playlist, not by key list Restoring over a playlist reused the keyed clear, which caps its input at MAX_KEYS_PER_LOOKUP to bound an IN clause. A playlist with more than eight pinned movies therefore kept the surplus while the call still reported success, and the restore then wrote the archive's pins on top — leaving the union of two states, which is neither the one the user asked for. Clearing is now a dedicated delete-by-playlist operation with no key list to truncate, and it refuses a blank playlist id rather than deleting everything. A failure fails the entry instead of being swallowed: `listForPlaylist` returns `[]` on error and `clear` returns `false`, so ignoring the result made a failed read indistinguishable from "there was nothing to clear". Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> * fix(portals): keep a pin readable under every identity of its film Two defects in the pin/position subsystem, both reported in review. A pin was stored under the movie's most-trusted key alone and its other keys retired. But a movie's identity GROWS: the film keyed `tmdb:438631` today was `title:dune:2021` before enrichment, and reopening it cold asks for the poorer key first. The preference was therefore ignored until enrichment landed — and permanently when enrichment is off or never answers. The decision is now written under every key in `write` (never the yearless form, which every remake shares), one upsert per key plus the leftover retirement in the same transaction. `setVodSourcePin` also reports failure for a pin with no usable key instead of claiming a write it never made. The primary button asked whether the pinned copy's position had loaded by testing presence rather than identity, so re-pinning left it wearing the previous copy's timecode until the new lookup returned. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> * docs(portals): record the key-addressing limit a pin write cannot close Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> * fix(portals): fold non-ASCII case in the scan, and read years as tags I was wrong about SQLite twice over, and both errors cost matches. GLOB character classes are NOT ASCII-only. `patternCompare` reads them as UTF-8 code points, so `'Он' GLOB '*[Оо][Нн]*'` is true — only `LOWER()` is ASCII-only. The scan tier now folds the case in JavaScript, where Unicode case mapping is real, and hands SQLite one class per character. A short Cyrillic or Greek title stored in a different case is found instead of being silently absent from the Sources chip. The builder returns `null`, leaving the substring tests as the whole answer, for a token holding a GLOB metacharacter (GLOB has no escape character) or a case mapping that changes length. The FTS tier is untouched and still cannot fold — that needs a stored normalized-title column. The movie's own year came from `extractYear`, which reads a year from anywhere in the title. That is right where a year is a search hint, wrong where it is an identity: `2001: A Space Odyssey` was treated as a 2001 film, so every genuine 1968 copy failed the year gate and the movie had no alternatives at all — and its pin key moved the moment enrichment supplied the real year. `releaseTagYear` accepts only bracketed and trailing forms; the repo's own TRAILING_YEAR_PATTERN already documented this exact hazard. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> * fix(portals): cover a letter spelled two ways in lower case Greek Σ lowercases to σ, but a word-final sigma is written ς and is equally a lowercase of it, so a class built only from the character in hand knew one spelling of two. Each class now also carries the uppercase form's own lowercase, which reaches the other one. One-way on purpose: σ → Σ → σ never arrives at ς. Left so because ς is only correct at the end of a word, which is exactly where the request's last character sits — the pair that occurs in real titles is covered, and closing the other direction needs a fold table. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> * fix(portals): let an exact title keep a number that is part of its name Both match tiers weighed the same year, taken from a trailing four-digit tail or a bracketed tag. On the exact tier that rejects the very copy it was meant to confirm: reaching it means both titles are the SAME string, so the trailing digits belong to both, and comparing them against a release year out of metadata makes "Blade Runner 2049" disagree with its own stated 2017 — the genuine alternative disappears at the moment enrichment lands, which is when the user has most reason to expect it. The exact tier now reads the bracketed form only. Brackets are never part of a name, so "Dune (1984)" is still rejected against 2021. The base tier is unchanged: it has just stripped a trailing year, and that year is the only thing separating "Dune 1984" from "Dune 2021". Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> * fix(database): verify the title index folds, rather than trust the marker `createTables` declares content_title_fts with the plain trigram tokenizer, and the diacritics migration declares it again with folding. Two sources of truth for one tokenizer: if the table ever went missing after the marker was written, `CREATE TABLE IF NOT EXISTS` would restore the unfolded form and the migration would skip it on the marker alone. The upgrade now reads the live table's own DDL from sqlite_master and rebuilds unless it really folds. A degraded index is invisible from the outside — discovery just stops finding "Pokémon" for "pokemon" — so the record has to be checked against the thing it describes. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> * fix(portals): give the route's own row the facts the page already has Two provenance defects found in review. The current-source row is never resolved — nothing needs to fetch a URL for the stream already playing — so it carried no provider metadata at all, while every alternative got its facts from the resolve preceding playback. `audioDiffersFactually` requires a fact on BOTH sides, so the "dub may differ" warning was structurally unreachable on the commonest switch there is: route to alternative. It could only ever fire between two alternatives that had both been resolved. The row now carries what `get_vod_info` already told the page, via a `providerVodMetadataOf` mapper shared with the resolver so the two cannot describe one movie differently. Quality bucketed every width from 900 to 1199 as 576p, so a 960x540 stream — an ordinary 540p encode — was published as "576p" with `api` provenance: a measurement its own pixels contradict, from the one field that is supposed to mean the provider said so. That range holds two standard formats, so it is matched now rather than bucketed, exactly as the sub-HD sizes already were. A width matching no known format yields no tag and a check chip, which is the honest answer. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> * fix(portals): let the height veto a width-derived quality, and refresh route facts Both of these are gaps I saw and chose not to close last round; a reviewer was right that neither survives its own reasoning. The shape check only ran below 1200, so the HD ranges kept publishing wrong-but-confident labels: 1440x1080 is anamorphic 1080 and 1600x900 is 900p, and both were "720p" with `api` provenance — the provenance that means the provider said so. Ranges are fine up there, the standard widths really are far apart, but only once a known height can veto the answer. Same rule the matched formats already used: a shorter frame is a letterboxed master, a taller one is a different shape and gets no tag. And the route row picked up provider facts only when discovery reran. On a sparse panel `get_vod_info` can answer with no year and no TMDB id, so the movie key is unchanged, nothing reruns, and the row keeps stating nothing — leaving `audioDiffersFactually` one-sided and the dub warning unreachable on exactly the switch it exists for. It now takes those facts on without rediscovering, merged onto the existing row so a probe result already sitting there survives. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> * fix(portals): a codec is not a dub, and two waits needed a switch guard Three findings from review. The "dub may differ" warning compared audio CODECS. AAC and AC3 routinely carry the same dub, and two AC3 tracks can carry different ones, so it fired on every identical-language re-encode and stayed silent on the dub changes it exists for — wrong in both directions, which is worse than absent, because a warning people learn to ignore is not a warning. Worse, the previous commit made it reach the common route-to-alternative switch for the first time, so the false claim was about to get louder. It now reads a new `audioLanguage`, taken from the track's language tag and never from the codec. `audio` stays as a display fact. Few panels tag a language, so the warning is usually silent — the same answer the rest of this feature gives when it does not know. `failover()` validated only the session across its wait for a discovery in flight. The session moves when the FILM does, so a source the user picked — or the route stream they restarted — during that wait was then treated as the thing that failed and switched away from. It claims and rechecks a switch generation, as the pinned path already did. And the scan's ASCII branch could not find "Ça" from a folded "ca", while the non-ASCII branch found "Ca" from "Ça" — so whether two playlists could see each other depended on which one was open. Each ASCII letter now carries its accented forms, derived by decomposition rather than tabulated, so it cannot drift from the normalizer. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> * test(portals): pin what the declared audio shape can and cannot say The array shape the mock server and many panels send carries a codec and no language, so the dub warning is silent for every source arriving that way. Asserted rather than assumed, alongside the ffprobe shapes that do carry one — otherwise a later reader sees an unused field and wires the codec back into the warning. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> * fix(portals): a failed pin read must not export as "no pins" Three findings, all in code from this session. Backup called the lenient `listForPlaylist`, which turns a failed read into `[]`. Since `e0ebbeaf` made restore treat `sourcePins` as authoritative — clearing the playlist's pins before applying it — an export whose read failed produced a file that looks complete and wipes every pin on restore. Losing them is bad; losing them through the one feature meant to protect them is worse. Backup now uses a strict listing that throws, so the export fails instead. The diacritic map stopped at U+024F, which is tidy and leaves Vietnamese out: `ố` is U+1ED1, the normalizer folds it to `o`, and the scan filtered those rows out before confirmation. Latin Extended Additional is included now; the filter decides what belongs, so the range only has to be wide. And two panels spelling one language differently (`eng` vs `en`, or `en-US`) raised a dub warning between identical tracks. Tags are canonicalized before comparison — 639-2 collapses to 639-1, both German forms meet at `de`, regions drop, and `und` becomes nothing. Anything that survives longer than three characters is not a language code, so the comparison is declined rather than guessed. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> * fix(portals): stop two backup paths from deleting pins they never read Two data-loss paths, both P1, both mine. A web export wrote `sourcePins: []`. Pins are Electron-only, so out there we cannot read them — which is not the same as knowing there are none, and restore treats the collection as authoritative. A backup made in the browser was therefore an instruction to delete every pin the moment it was imported on the desktop. There are three answers here, not two: pins exist, there are none, and "could not look". The last omits the field, exactly as an archive written before pins existed does. The same rule now covers Electron with the bridge method missing. I had written a test asserting the unreachable-store case resolves to an empty list "because a backup made there is complete". That reasoning was wrong: empty was true of what the runtime could see, never of the playlist. Restore also cleared the playlist's pins and then wrote the archive's one by one. A write failing partway left the previous pins already gone and only a prefix applied — a state belonging to neither, reported as a failure the user could not undo. `DB_REPLACE_VOD_SOURCE_PINS` does the clear and every write in one transaction, so the playlist ends up as the archive describes it or exactly as it was. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> --------- Co-authored-by: Claude Opus 5 <noreply@anthropic.com>	2026-07-29 21:53:42 +02:00
4gray	3c342bc555	test(performance): preserve delayed worker samples (#1305)	2026-07-28 23:25:52 +02:00
4grayandClaude Opus 5	99a85da6b0	feat(packaging): register IPTVnator as the .m3u/.m3u8 handler (#1301) ... * feat(packaging): register IPTVnator as the .m3u/.m3u8 handler Every runtime path for an OS-supplied playlist existed, but no packaging metadata claimed the file types — so the OS never offered IPTVnator as a handler and `open-file` could not fire from Finder. `fileAssociations` declares one entry per extension. Electron Builder derives all three platform registrations from it: macOS `CFBundleDocumentTypes` (the prerequisite for `open-file`), the NSIS registry entries, and, on Linux, the desktop entry's `MimeType` plus `/usr/share/mime/packages/iptvnator.xml` for deb/rpm/pacman. Neither platform needs a dedicated icon — both fall back to the app icon. Declaring `MimeType` under `linux.desktop.entry` would not have worked: Electron Builder assigns the association-derived value *after* spreading that object, so an explicit key there is silently overwritten. The per-association `mimeType` fields produce the same entry through the supported path. Registering the types also exposes a gap in the delivery side. The generated Linux `Exec` ends in `%U`, so file managers hand over a percent-encoded `file://` URI rather than a path, which `createPlaylistOpenRequest` would have resolved into a bogus relative path. It now decodes a `file://` candidate before the extension check. Suppressing the `%U` instead would mean putting an exec code in `linux.executableArgs`, which also passes it to the app as a real argument. Verified on macOS against a signed packaged bundle: Launch Services lists the app as a `public.m3u-playlist` handler, and an LS-initiated open imports the playlist both on a cold launch and against the already-running process. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> * fix(playlist): open every playlist of a multi-file selection `%U` is the plural exec code, so selecting several playlists in a Linux file manager is one launch carrying one argument per file. Both argv paths called `extractPlaylistOpenRequestFromArgv`, which returned at the first match, so everything after the first playlist was silently discarded — a gap this PR itself opened by making the desktop entry reachable in the first place. The extractor is now plural and returns every match in argument order, and the queue gained `enqueueAll` so a selection is pushed under a single flush: a delivery that fails partway leaves the untouched remainder queued in arrival order rather than interleaved. Covered by unit tests over a mixed argv (percent-encoded `file://` URI, a non-playlist argument, a second URI) and by a new Electron E2E that launches with two playlist arguments and asserts both are imported. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> --------- Co-authored-by: Claude Opus 5 <noreply@anthropic.com>	2026-07-28 21:53:43 +02:00
4grayandClaude Opus 5	f80eb4d1b9	fix(playlist): open playlists handed over by the OS (#1299) ... Opening an .m3u/.m3u8 file from the command line or a file association did nothing. The renderer parsed `process.argv` and sent an `OPEN_FILE` IPC event that had no `ipcMain` handler and no preload channel, so `sendIpcEvent` logged it as an unknown type and dropped it. The path now belongs to the main process, which is where the OS actually delivers it: - argv is parsed on first launch (skipping the executable and Chromium switches) and normalized to an absolute path; - macOS gets an `open-file` listener registered before `whenReady`, since Launch Services never puts the path in argv; - the single-instance guard forwards a second launch's argv and working directory instead of discarding them, so opening a playlist against a running app works too. Requests are queued in the main process until the renderer subscribes to the `OPEN_FILE` push and drains the queue, which closes the startup race. The import itself reuses the existing file path, so persistence, playlist-scoped EPG and the navigation to the new playlist behave exactly like a dialog import; a failed open now surfaces a snackbar instead of silence. Co-authored-by: Claude Opus 5 <noreply@anthropic.com>	2026-07-28 20:28:29 +02:00
4gray	a2fafcfc08	test(performance): add end-to-end Xtream benchmark harness (#1300) ... * docs(performance): plan Xtream benchmark * feat(xtream-mock-server): add deterministic 100k fixture * style(xtream-mock-server): apply repository formatting * fix(xtream-mock-server): harden performance fixture data * feat(xtream-mock-server): add performance control plane * docs(performance): correct Xtream capture plan * fix(xtream-mock-server): harden performance controls * fix(xtream-mock-server): harden control lifecycle * feat(performance): add Xtream preload markers * feat(performance): trace Xtream main phases * feat(performance): mark Xtream store publications * feat(performance): trace Xtream database phases * feat(performance): trace Xtream delete cancellation * feat(performance): capture Xtream phase attribution * feat(performance): mark Sources Xtream refresh * test(performance): define Xtream benchmark evidence contracts * test(performance): add Xtream benchmark runner * test(performance): surface failure evidence writes * test(performance): align database read clock * test(performance): preserve capture failure contracts	2026-07-28 08:08:07 +02:00
4gray	bfad82c26c	fix(settings): stop settings silently reverting on restart (#1272) ... Settings live in the renderer's IndexedDB, and two failure modes made them look saved while nothing reached disk. A second app instance sharing the same userData directory cannot take the Chromium storage lock, so its renderer reads defaults and every write is dropped. The app now holds a single-instance lock and focuses the running window instead of starting a rival copy. The lock is requested after the userData override so E2E runs with their own data dir keep independent locks, and after Squirrel event handling. IPTVNATOR_ALLOW_MULTIPLE_INSTANCES=1 opts out for local CDP debugging. updateSettings() patches in-memory state before persisting and the submit path had no rejection handler, so a failed write produced an unhandled rejection and no user-visible feedback. SettingsStore now records which half of the round trip failed, and the settings page surfaces it through a dismissible error snackbar; the dialog stays open on failure so the save can be retried. Two follow-ups from review, both wider than the report: - a second launch now re-creates the main window when the lock owner has none left, so closing the last window on macOS no longer leaves a second launch quitting silently with nothing on screen - App.onMainWindowCreated() re-runs window-owned bindings for every rebuilt window, so the downloads broadcaster stops holding a destroyed window. This also fixes the same bug on the pre-existing dock `activate` path. Closes #1156 Closes #102	2026-07-27 23:14:30 +02:00
4grayandClaude Opus 5	0334296f15	fix(electron-backend): make test suite and lint host-agnostic on Windows checkouts (#1176) ... * fix(electron-backend): make test suite and lint host-agnostic across Windows/Linux checkouts Windows checkouts (core.autocrlf=true) had 13 pre-existing jest failures and 5 Windows-only lint errors in electron-backend while Linux CI was green: - embedded-mpv-native-source.spec: normalize CRLF after readFileSync so multi-line source assertions match on autocrlf checkouts - worker-runtime-paths.spec: build expected candidate paths with path.join instead of hardcoded POSIX strings - external-player-launch-context: join darwin-only paths (.app bundle executables, Homebrew Caskroom) with path.posix.join so simulated darwin platforms resolve correctly on win32 hosts (no-op on macOS) - app.spec: build packaged-navigation fixtures with path.resolve + pathToFileURL; file:///tmp/... is not a valid win32 file URL - lint target: quote the eslint glob. The unquoted ** was expanded by the POSIX shell on Linux (shallow match), so CI linted only a subset of files while Windows passed the literal pattern to ESLint and linted the full tree - the hosts checked different file sets - fix the 5 errors full-tree linting surfaces: prefer-const in epg-worker.service and database.worker-connection, no-unsafe-finally in external-player-session-registry and embedded-mpv-native.service (rewritten as catch-swallow with identical semantics, pinned by new regression tests), intentional no-control-regex in the recording filename sanitizer; drop stale unused disable directives - document the quoted-glob convention in CLAUDE.md Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> * docs: mirror the quoted-lint-glob convention into AGENTS.md AGENTS.md already mirrors the neighbouring max-lines/baseline paragraph from CLAUDE.md, and it requires coding conventions to stay in sync between the two files. The quoted-glob rule landed only in CLAUDE.md, so agents bootstrapping from AGENTS.md could reintroduce a host-dependent lint target. Also corrects the wording in both copies: the shallow expansion happens on macOS as well as Linux — /bin/sh has no globstar on either — and the target still exits 0 with a broken glob, which is why this went unnoticed. Adds the file-count check that catches it. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> --------- Co-authored-by: Claude Fable 5 <noreply@anthropic.com>	2026-07-27 23:05:22 +02:00
4grayandClaude Opus 5	19b592badb	fix(epg): make manual EPG mapping lookups actually fail soft (#1291) ... The mapping handlers documented a fail-soft contract but only honored half of it. Four of them returned the database operation's promise from inside the `try` block without awaiting it, so the rejection escaped the `catch`: the try block exits before the promise settles. A `getDatabase()` failure was caught, but a SQLite error in the operation rejected the `ipcMain.handle` promise, and the renderer call threw instead of receiving `null` / `{success:false}` / `[]`. Adding `await` in handleGetEpgMapping, handleSetEpgMapping, handleDeleteEpgMapping and handleSearchEpgChannels closes the gap. resolveChannelIds and handleGetEpgMappingsBatch already awaited correctly and are unchanged. This is pre-existing — the same shape predates the epg.events.ts split in 636545cb, which preserved semantics faithfully and inherited the bug. Adds epg-mapping.service.spec.ts, the first coverage these handlers have had: table-driven over all six functions against both failure modes, plus the guard clauses and queryByResolvedChannelIds remapping. Verified to fail on the old behavior — reverting the four awaits fails exactly the four operation-rejection cases. Co-authored-by: Claude Opus 5 <noreply@anthropic.com>	2026-07-27 22:15:50 +02:00
4gray	5932e71cb9	fix(electron-backend): process zero-delay database cancellation (#1295)	2026-07-27 21:59:20 +02:00
4grayandClaude Opus 5	d2fd27b535	test(performance): deflake the real-timer event-loop delay spec (#1293) ... * test(performance): deflake the real-timer event-loop delay spec The real-timer capture spec failed under full-suite parallelism because `monitorEventLoopDelay()` records nothing on its first internal timer tick - that tick only seeds the previous timestamp, so the first delay sample lands on the second tick. Condition-based arming therefore needs two event-loop turns inside its fixed 50ms wall-clock budget, and a machine running 10 Jest workers stretches a single turn past 20ms. The capture then degraded to a documented `event-loop-delay-arm-timeout`, which is the intended graceful path, while the spec asserted the happy path of that race and turned an environmental outcome into a red build. Retry the real-runtime capture within a 5s budget instead. The real `node:perf_hooks` runtime and the real 20ms block are kept, since the fake harness returns a hard-coded histogram max and never measures anything. Every attempt still asserts a contract: instrumentation never breaks the wrapped work, and a null delay must carry a documented arm/flush timeout rather than being silently null. Budget exhaustion warns instead of failing, so load can no longer produce a false failure. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> * test(performance): assert the real delay measurement unconditionally Codex flagged that budget exhaustion still passed the test, so a regression that made arming or flushing time out on every attempt would have been reported as a warning rather than a failure - removing the only assertion backed by Node's real histogram and a genuine event-loop block. Drop the tolerant retry loop. The arming deadline is read through the injectable `readMonotonicMs()`, so scaling only that clock leaves the wait bounded by its other limit, the 50-poll ceiling, which is ~25x the two event-loop turns arming actually needs. Everything else stays production code: the real `monitorEventLoopDelay()` histogram, real `setTimeout()` polling, and real epoch/CPU/ELU boundaries. The scaled clock reaches nothing but the wait budgets, since its only other consumer records phase events and this spec records none. The test now always asserts a real measurement and hard-fails otherwise. Verified by mutation: forcing arming to never arm fails it, and dropping the deliberate block fails it at maxMs 3.8ms against the 10ms floor. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> --------- Co-authored-by: Claude Opus 5 <noreply@anthropic.com>	2026-07-27 21:33:08 +02:00
4grayandClaude Opus 5	518964c57f	refactor(electron-backend): split the last four files over the max-lines cap (#1288) ... Follow-up to #1278, which split four of the files the electron-backend lint target had been silently skipping. Four were left over; this splits them, so no file in the project sits above the 400-line cap outside the baseline. None are added to tools/eslint/max-lines-baseline.mjs — the baseline only shrinks. Shared setup moves to *.test-helpers.ts, the suffix tsconfig.app.json already excludes, so the production build never sees jest globals. - remote-control.events.spec.ts 588 -> 188, plus remote-control-http.spec.ts. The mock registry moves to remote-control.test-helpers.ts; each spec keeps its own jest.mock() factories, which resolve the mock-prefixed exports. - downloads.events.spec.ts 530 -> 182, plus downloads-actions.spec.ts. The jest.doMock setup is not hoisted, so the whole harness moves to downloads.test-helpers.ts behind setupDownloadsEventsHarness(). - http-server.spec.ts 448 -> 377, plus resolve-static-file-path.spec.ts. The resolveStaticFilePath cases were already self-contained. - worker-performance-capture.spec.ts 408 -> 149, plus worker-performance-capture.resilience.spec.ts, matching the .concurrency and .histogram siblings. Test bodies are unchanged; the helpers keep the same identifiers in scope so the splits are a move, not a rewrite. Verified with the glob quoted locally (that quoting is #1176's, not part of this change): 245 files, 0 max-lines errors, and the only remaining lint errors are the five #1176 fixes. Both tsconfig.app.json and tsconfig.spec.json typecheck clean. Note: worker-performance-capture.concurrency.spec.ts is flaky on master independently of this change — it asserts a real 20ms event-loop block and failed 4/4 clean-master runs here. Co-authored-by: Claude Opus 5 <noreply@anthropic.com>	2026-07-27 20:21:24 +02:00
4gray	24f0dee6f0	test(performance): add formal M3U import benchmark (#1287) ... * test(performance): add formal M3U import benchmark * test(performance): harden formal capture validity * test(performance): address benchmark review feedback	2026-07-27 10:34:22 +02:00
4grayandClaude Opus 5	26331676f9	fix(epg): await mapping queries so lookups fail soft (#1279) ... The four EPG mapping handlers wrap their DB call in try/catch but return the promise instead of awaiting it. An async function *adopts* a returned promise rather than awaiting it, so the catch block only ever fired when getDatabase() itself threw — a rejection from the underlying query escaped to the IPC caller instead of returning the intended null / {success:false} / [] fallback. Add the missing await to handleGetEpgMapping, handleSetEpgMapping, handleDeleteEpgMapping and handleSearchEpgChannels, matching handleGetEpgMappingsBatch and resolveChannelIds in the same module, which already awaited and so already failed soft for both cases. This restores the contract stated in the module's own doc comment: a mapping lookup must never take down an EPG request. The behavior predates the max-lines split in #1278, which carried it over verbatim from epg.events.ts. The new spec drives the real IPC handlers captured from ipcMain.handle, so it asserts the contract that matters: the caller gets a fallback, not a rejected promise. Reverting the four awaits fails exactly those four handlers and leaves the already-correct batch handler green. Co-authored-by: Claude Opus 5 <noreply@anthropic.com>	2026-07-27 08:43:52 +02:00
4gray	e1c4853a39	Merge pull request #1280 from 4gray/agent/perf-exact-process-memory ... fix(perf): make process memory captures comparison-safe	2026-07-27 02:57:02 +02:00
4gray	636545cbb7	refactor(electron-backend): split four files under the max-lines limit (#1278) ... Split epg.events.ts (514), the embedded MPV frame-copy adapter (428) and two of its specs (547, 539) below the 400-line hard limit, and teach the baseline generator to skip files that already carry a justified file-wide eslint-disable max-lines. The generated baseline list is unchanged: 128 entries before and after. No behavior change.	2026-07-27 02:16:02 +02:00
4gray	2455165043	chore(electron): route the worker heap probe	2026-07-27 02:15:58 +02:00
4gray	d1e268f1dd	chore(electron): add perf-only worker heap probe	2026-07-27 02:15:58 +02:00
4gray	a3a8f6e90c	fix(perf): optimize benchmark worker builds	2026-07-26 23:25:09 +02:00
4gray	df7eee7e52	fix(perf): exclude profiler flush from cancel latency	2026-07-26 23:14:57 +02:00
4gray	da3b657277	fix(perf): make worker profiling request scoped	2026-07-26 22:37:52 +02:00
4gray	0579d253c4	fix(electron): fail closed on unknown performance completions	2026-07-26 22:37:52 +02:00