pytest-asyncio 0.23 is incompatible with pytest 9 (it calls the removed
Package.obj attribute, causing an INTERNALERROR during collection).
Bump to ~= 1.1 and set asyncio_default_fixture_loop_scope to silence the
1.x deprecation warning.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
The .gitattributes file was stored with CRLF endings, which prevents
git from parsing its rules reliably on Linux/macOS. Add a self-referential
eol=lf rule and renormalize the file to LF in the index.
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
Windows developers with core.autocrlf=true (or the default) will have
git check out shell scripts with CRLF endings. When those files are
copied into a Linux Docker container they fail to execute:
exec /entrypoint.sh: no such file or directory
Adding eol=lf attributes for *.sh, entrypoint.sh, and
docker/init_scripts/init ensures the repository always stores and checks
out these files with Unix line endings, regardless of platform or
git config.
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
- Use CollectionSchema instead of ReturnType<typeof collectionsStore.getCollection>
in AddRoms.vue and RemoveRoms.vue (simpler, per gantoine review)
- Wrap bulk INSERT in a savepoint so a concurrent duplicate-key violation
is caught via IntegrityError and ignored rather than aborting the transaction
- Only bump Collection.updated_at in remove_roms_from_collection when rows
were actually deleted (rowcount > 0), matching add_roms_to_collection behavior
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
Starlette's TestClient (httpx-based) does not expose body kwargs on the
delete() convenience method; client.request(\"DELETE\", ..., json=...) is
the correct approach. Also switch datetime.utcnow() to
datetime.now(timezone.utc) to silence Python 3.13 deprecation warnings.
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
- Switch content= to data= for DELETE requests (Starlette TestClient is
requests-based and does not accept content= keyword argument)
- Fix test_bumps_updated_at to record time before the API call and use >=
comparison, avoiding false failures when MariaDB truncates DATETIME to
whole seconds and creation/update land in the same second
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
- Replace client.delete(json=...) with content=json.dumps()+Content-Type header
(Starlette TestClient does not forward json= on DELETE requests)
- Adjust duplicate-name test to expect HTTP 500 matching CollectionAlreadyExistsException
- Add description="" to collections created without it to satisfy Pydantic schema
- Strip tzinfo before comparing updated_at to avoid offset-naive/aware TypeError
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
Replace full rom_ids list replacement with atomic POST/DELETE endpoints
that add or remove individual ROMs from a collection. This prevents
concurrent rapid clicks from overwriting each other (last-write-wins).
Also fix missing session.flush() in add_rom_user() and add collection
endpoint tests.
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
Pylance infers dict[str, Unknown] from the {**old, "highest_award_kind": ...}
spread, which then fails to assign to list[RAUserGameProgression]. Wrap in
typing.cast so the TypedDict type survives the reassignment.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Callers now pass the full platform dict and rom.fs_extension; the service
normalizes the extension (optional leading dot, case-insensitive) before
checking the compressed-archive skip set, so ROMs stored with bare
extensions like "zip" correctly hit the skip path.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
RAHasher was being spawned for every hashable ROM regardless of file
type. When the source file is a zip/7z/tar and the RA platform needs
an on-disk disc image (PSX, PS2, PSP, Saturn, Dreamcast, Sega CD,
3DO, PC-FX, Neo Geo CD, TurboGrafx CD, Atari Jaguar CD, Wii), the
subprocess fails with "Unsupported console for buffer hash: {id}"
after paying full process-spawn overhead per ROM — a serious slowdown
when indexing large zipped collections (e.g. myrient PS2/PSP sets).
calculate_hash now short-circuits those combinations with a debug log
and no subprocess. Raw disc images (.iso, .chd, .cue/.bin) and
archives on cartridge platforms still go through RAHasher as before.
Also centralize COMPRESSED_FILE_EXTENSIONS in utils/filesystem.py so
roms_handler (is_compressed_file / hashing), rahasher (skip logic),
and feeds (PKGi passthrough) share one source of truth. The shared
set adds .rar, which is_compressed_file now recognizes too.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
get_rom now also fetches Named_Snaps, Named_Titles, and Named_Logos
when the matching MetadataMediaType (SCREENSHOT, TITLE_SCREEN, LOGO)
is in SCAN_MEDIA. Box art is still fetched unconditionally — it drives
url_cover and libretro_id. Matching extras are appended to
url_screenshots so the scan_handler artwork priority loop picks them
up without further changes. All enabled listings are fetched
concurrently.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
SGDB's url_cover used to clobber whatever the artwork priority loop
picked, which meant user-set priorities (e.g. libretro > sgdb) and
manually uploaded covers were silently overridden. SGDB now only
replaces the current url_cover when it outranks every other source
that produced one under SCAN_ARTWORK_PRIORITY, and never over a
manual cover preserved on UPDATE/UNMATCHED scans. Default artwork
priority gains sgdb at the top so existing "SGDB wins" behavior is
preserved for default configs.
On the /search/roms endpoint, libretro is now an enrichment source
alongside SGDB instead of a primary match source: it decorates
entries resolved by IGDB/Moby/SS/Flashpoint/Launchbox with
libretro_id and libretro_url_cover, mirroring how SGDB works.
get_matched_roms_by_name is removed from the libretro handler since
nothing else calls it.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Commit 3991e1b6e removed `@with_details` from `get_roms_by_fs_name` but
left the body using the `query` parameter that decorator was supposed
to inject, so every scan hit `'NoneType' object has no attribute
'filter'` and crashed the platform identification task.
Make the function self-contained: build `select(Rom)` directly and
eager-load only `Rom.platform`, the one relationship the scan loop
actually needs (via `rom.platform_slug` / `rom.platform.fs_slug`).
Keeps the prior commit's intent of avoiding the heavy `with_details`
eager-load on every batch.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Collapse the two parallel id lists and their mirrored chunked-update
loops into a `flips: dict[bool, list[int]]` keyed by desired state, and
drop unused rom assignments in the related tests.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Adds the libretro thumbnail repository as a first-class artwork source so
region-correct box art (PAL/Europe, Japan, etc.) can be matched directly
to ROM filenames, addressing rommapp/romm#3239.
Implementation follows the SGDB handler pattern (artwork-only, no game
metadata): MetadataSource enum entry, scan-time fetch wired into the
SCAN_ARTWORK_PRIORITY loop, /search/roms integration, MatchRom dialog
chip + cover selection, and a heartbeat flag.
Matching is exact case-insensitive against the directory listing first
(so a ROM named "(Europe)" lands on the (Europe) artwork), with a
JaroWinkler fuzzy fallback at 0.8 that strips parenthetical tags from
both sides. Listings are cached in Redis with a 24h TTL.
`libretro_id` is persisted on the Rom model as the SHA1 hex of the
matched libretro filename — stable across scans, distinct per region,
indexed for lookup. Migration 0077 adds the column.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
