A search entry often carries no `format`, so `format_of(None)` produced
`Other("")`, that was stashed at search time and handed back as the
download's format, and conversion had no parser to pick:
4073669 could not be converted from to SRT: no parser for ""
`POST /download` takes `sub_format`, and the API converts on its side, so
asking for srt makes the answer srt whatever the uploader posted. The
format guessed from the search entry is dead weight and goes with it.
The download mock matched only `file_id`, so it never exercised what the
real API returns.
Closes#272
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
`login` posted the credentials as HTTP basic auth with no request body.
The API takes them as JSON, answers basic auth with a 401, and the lane
reported that back as "opensubtitles rejected the configured
credentials" — for correct credentials.
Search was unaffected and hid this: it authenticates with the Api-Key
header alone and never logs in, so candidates were found and only the
download failed.
`mount_login` matched on method and path, so the mock answered a token
to any request shape. It now matches the body.
Closes#271
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
The id lane sent `tmdb_movie_id` and `tmdb_series_id`. Neither is a
parameter of this API, and unknown ones are ignored rather than refused,
so every search silently degraded to a `moviehash` lookup: one result for
a release someone had already hashed, none at all for anything else.
Two further requirements the API documents and answers a 301 to when
missed: parameters sorted by name, and language codes lowercase and
sorted.
The integration tests asserted the old names against a mock, which
answers whatever it is asked for. They pinned the bug rather than
catching it.
Closes#270
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
A release whose only subtitle is a PGS or VobSub track was stuck both
ways: the track satisfied English so no English SRT was ever fetched,
and bitmaps can never feed a translator. §15 is amended to separate
satisfying viewing from providing a translation source, and to carve a
fetch made to obtain a source out of the no-upgrade rule.
Timings come from the disc rather than from alass guessing at the audio.
At import, each non-forced image track's packet timestamps are paired
show-to-clear into a cue skeleton and stored; the fetched source is then
aligned against that skeleton, and the translation made from it skips
the post-translation pass, which could only move disc-exact timings off.
Pairing is validated before it is trusted — even packet count, plausible
durations, sane density for the runtime — because PGS allows several
composition segments per subtitle and a slipped pairing is quietly half
a second out. A track that fails validation gets no skeleton and falls
back to aligning against the video, as does any file imported before
this: there is no backfill.
Closes#268
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
`grabs.infohash` is unique and grabs point at their target
polymorphically, so nothing cascaded them off a deleted movie or series.
An orphan holding an infohash then blocked the row a later grab of the
same release needed: the upsert only reclaims a `vanished` row, so the
insert did nothing while the new title was still flipped to
`downloading`. It never imported, and `/api/downloads` attributed the
torrent to a title that no longer existed.
Deletes now clear the grabs, and the upsert also reclaims a row whose
target is gone whatever its state — which heals the databases that
already have orphans. A row belonging to a live title stays off limits,
so the restart case still cannot let one title steal another's torrent.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Logging in before the first call turned a stale or wrong password into a
hard failure on calls that would have succeeded: an instance that bypasses
authentication for arr's address answers everything, and never needed the
credentials at all. Authentication is now established on the first 403.
The health probe was reading a 403 from `app/version` as healthy, on the
reasoning that a live daemon is all the lamp claims. That is exactly the
state arr cannot use, and it left the lamp green while every grab and every
reaper tick failed on rejected credentials. It now lists torrents through
the configured client, so the lamp fails when arr's own calls do.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
qBittorrent's add call answers `Ok.` and nothing else — no hash, no name,
no duplicate signal — so arr derives the v1 infohash from the magnet or the
`.torrent` bytes before the call and looks the torrent up by it. That also
drops Transmission's numeric torrent id: the hash is the only identity now.
The reaper needs "stopped because a share limit was reached", and
qBittorrent's state field cannot tell that apart from a hand-paused torrent.
So the ratio and idle counters are checked against the limits arr set, and a
torrent stopped by a global limit reads as still seeding rather than being
deleted.
The WebUI needs a login, so `ARR_QBITTORRENT_USERNAME` and
`ARR_QBITTORRENT_PASSWORD` join the env-only secrets; leaving both unset is
valid for an instance that whitelists arr's subnet.
Verified against qBittorrent 5 (WebAPI 2.15.1) in a container: it rejects
`setShareLimits` without `shareLimitAction`, which 4.x ignores, so it is
always sent.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Reverts the temporary docker-branch gating used to probe the runner, and
restores the deploy step now that the image build is verified.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Dokploy built from source on the host, so every deploy paid a full cold
cargo build and the layer cache died with the builder. CI now builds and
pushes to the Gitea registry, and the deploy is a pull.
The Gitea push webhook fires before Actions starts, so autoDeploy on the
Dokploy application is off and this job triggers the deploy itself.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
4cec2b6 edited an already-applied migration to drop podnapisi from the
subtitle_settings default. sqlx checksums migrations, so every boot since
has failed with "migration 25 was previously applied but has been
modified" — the daemon exits 1, swarm keeps the old task, and three
commits never reached production.
0033 already strips podnapisi from existing rows, and it runs after 0025
seeds row 1, so a fresh database still ends up with opensubtitles alone.
The edit changed nothing except the checksum.
Untracking a season clears `wanted` on every episode (§4.1) and leaves
the files where they were, so a season grabbed once and then untracked
rendered `0/0 eps` beside a column of green check glyphs. The same
counter could also read `10/1`, since the numerator was drawn from the
wanted set the denominator had already shrunk.
An episode now counts once it is wanted, on disk, or both, and the chip
is hidden rather than shown as `0/0` when a season or series accounts
for nothing. `Series.wanted_episodes` becomes `total_episodes`.
status-channel didn't match the real --status-airing token, and the
score heat scale (issue 111), the available check-glyph exemption,
and the score chip component were undocumented. DESIGN.md untouched.
The want button already only shows for state=="missing", so the
chip repeated the same signal. Other non-available states keep it.
Also fixes a pre-existing doc-markdown lint failure blocking CI.
The row and its disclosed panel were siblings in one wrapping flex line,
which only stacked them by accident of flex-basis: 100%. At phone width the
row is told not to wrap so the title truncates instead of pushing the
controls down — and the panel then had nowhere to go but beside the episode.
The row is now its own strip, .episode-line, with the panel a block under
it. Same pairing the movie file row already uses.
Commentary tracks carry the same language as the feature audio, so a file
row rendered `en` two or three times. The chips answer what a file can be
watched in; repeats say nothing the first said.
The floor is a selection filter (§5.5): it keeps the ranking from picking
mud while better candidates are still on the table. Once a release is
grabbed there is nothing left to choose between, so a hard fail at import
condemns a file already on disk and blacklists the release under §5.7 —
over a number the indexer gave before the grab. Size is not post-download
evidence either; §5.6 gives ffprobe audio, HDR, codec and duration.
It bit packs hardest. Pre-grab measures a pack by its total over the
season length; import measures each file on its own, so one short episode
condemned the whole release and reopened the season. And §6.3 outlasts the
fix: a blacklisted release stays rejected however the operator later sets
allow_below_floor, with no delete path.
Post-download the floor now waives instead, whatever the overrides say.
The file imports and carries a waiver naming the floor it missed.
Live download state, inline on the row that owns the item: DESIGN.md §9.8,
GET /api/downloads over arr's own grabs, chips on the movie, season and
episode rows, and a stale season import failure that now yields to a newer
attempt. Closes#262#263#264#265#266#267.
Superseding an import failure with any non-failed grab let a 'vanished'
one take the headline: the torrent has left Transmission, so the season
would show neither the failure nor any progress while the gap is still
there.
A zero download rate is not a stalled torrent: one between peers reads
zero for a poll or two and finishes fine, and at §9.8's 15s cadence that
flicker would raise the one chip reserved for a download that never
finishes. Transmission already decides this with its own stalled window;
carry isStalled and use it.
Four tests force a filesystem failure by freezing a directory to 0o555.
The CI container runs as root, mode bits do not constrain root, and the
rename those tests expect to fail succeeds — main has been red on
a_failed_rename_leaves_the_row_alone since the move-on-root-change work
landed, with nextest's fail-fast hiding the other three.
The guard probes the filesystem rather than the uid: what the tests
depend on is the refusal, and a container can hold CAP_DAC_OVERRIDE
without being uid 0.
DESIGN.md §15 end to end: provider search and download, embedded-track
extraction, translation behind four feature-gated backends, `alass` sync,
sidecar naming and the `.mt` machine-made segment, per-provider and
per-translator daily budgets, and the subtitle lane in /settings.
One sidecar per language, enforced in the schema (#222), so a forced track
never consumes the language's only slot. The OpenAI-compatible backend's
base URL and model are database rows, not bootstrap config (#220) — that
backend is any endpoint speaking the shape, `llama.cpp` included, and an
endpoint that needs no key is a valid configuration.
41 issues, plus #255 landing the provider reorder control on the shared
icon and control vocabulary.
`just ci` green: 778 tests. Migrations 0024, 0025 and 0027-0029 apply out
of order against a database that already has 0030-0032 — verified against
a database built from main's migration set, which is the shape production
is in.
CI's `rust` job is red on main independently of this merge; see #256.
`visibility: hidden` on the disabled reorder button left a control in the
DOM that rendered as nothing. `disabled` already blocks the click and drops
it out of the tab order, so the hiding was decoration; the shared
`.control:disabled` colour says the same thing on screen, and the row does
not reflow because the hidden button reserved its box anyway.
Exposing it surfaced a second problem: `.control-quiet:hover` sits after
`.control:disabled` at the same specificity, so a disabled quiet control
lit accent-bright under the cursor and invited a click that does nothing.
Guarded with `:not(:disabled)` in the shared rule rather than on this one
control.
`refreshProviderEdges` found the buttons by `button:nth-of-type(1|2)`.
Position in the row is not a contract — a third button would have quietly
moved the edge logic onto the wrong control. They carry `data-move` now.
Closes#255
Miguel Palhas
naps62-yolo