In an effort to make upgrading the constantly changing config files easier, Shift defaulted them and merged your true customizations - where ENV variables may not be used.
- Relocate translation files from resources/lang to lang/ via git mv
- Update PHP references to use the lang_path() helper
- Update crowdin.yml source/translation paths
- Update phpstan.neon translationDirectories
The media storage migration commands read/parsed the .env file directly to
check and flip PF_ENABLE_CLOUD. In containerized deploys there is no .env on
disk (config is injected via env vars), so updateEnvFile() threw
'file_get_contents(.env): Failed to open stream' and the scheduled command
exited 1.
- Check the live setting via config_cache('pixelfed.cloud_storage') like the
rest of the app, instead of parsing .env.
- Make the .env write best-effort in ManagesMediaStorageEnv: skip gracefully
when the file is missing or read-only, and still apply the runtime + DB
config-cache updates (the load-bearing changes on a hot server).
- Apply the same fix to the sibling unstable:MediaMoveStorageCloudToLocal.
- Add a regression test covering the no-.env container scenario.
The Cache::remember('profile:following:'.$pid, ...) block that plucks
following_id and appends the caller's own id was copy-pasted across four
call sites, with inconsistent TTLs (1440 minutes vs 1209600 seconds).
Add FollowerService::getFollowingIds($pid), which owns the cache key that
add()/remove() already invalidate, and use it from InternalApiController,
PublicApiController, ApiV1Controller and HashtagUnfollowPipeline. Removes
the now-unused Follower/Cache imports left behind.
Adds a test covering the followed-ids-plus-self result and the
follows-nobody case.
The status- and account-deletion jobs loaded whole collections with
->get() and then looped, running a per-row Notification lookup inside
each iteration.
- StatusDelete / RemoteStatusDelete: resolve associated DirectMessage and
MediaTag ids, fetch their notifications in a single whereIn query,
clear each via cursor (NotificationService::del must run per row for
cache/redis cleanup), then bulk delete the DMs and media tags.
- DeleteAccountPipeline / DeleteRemoteProfilePipeline: stream Story and
Collection deletions with cursor() instead of loading every row into
memory. Per-row file unlink and item deletes are preserved.
Adds StatusDeleteCleanupTest covering DM + notification cleanup, media
tag + notification cleanup, and the no-associations case.
CollectionController::index and StoryComposeController::createPoll had no
route mapping and simply returned $request->all(). Both are unreachable
debug leftovers; remove them. The live poll route maps to
ComposeController::createPoll, which is unaffected.
Two query patterns were copy-pasted across several call sites:
- The 'users who blocked me, plus myself' list used to filter profile
search (UserFilter::whereFilterableId($pid)->pluck('user_id')->push($pid))
appeared in ComposeController (x2) and DirectMessageController. Extracted
to UserFilterService::searchExcludedProfileIds(). Note this is the
inverse of blocks() (who I blocked), so it is a distinct method.
- CustomEmoji duplicate detection (groupBy('shortcode')->havingRaw(
'count(*) > 1')) appeared three times in AdminController. Extracted to a
CustomEmoji::duplicateShortcodes() query scope.
Adds tests for both. No behaviour change.
SeasonalController::getData computed the average posts/likes per profile
by grouping in SQL, then pulling every grouped row into a collection and
calling ->pluck('count')->avg() in PHP. This loaded one row per profile
into memory just to average.
Wrap the grouped per-profile counts in a subquery and let the database
compute AVG(count), returning a single value. Also drops the invalid
SELECT * with GROUP BY (ONLY_FULL_GROUP_BY) by selecting count(*) only.
Adds a test verifying the average-of-per-profile-counts and its
exclusions (remote, wrong type, out-of-range date), plus the empty case.
RemoteAuthService::getVerifyCredentials, getFollowing and getToken made
outbound HTTP requests to a user-controlled remote instance during the
Mastodon login flow with no timeout, no retry and no exception handling.
A slow or hostile instance could hang the request or surface an uncaught
exception.
Wrap all three in timeout(20)->retry(3, 750) with try/catch that returns
false on failure, matching the existing pattern in isDomainCompatible().
Callers already treat a falsy return as a failure; add the missing guard
at the one verify_credentials call site that accessed the result array
without checking it first.
Adds RemoteAuthServiceTest covering connection failure, server error and
success paths.
DirectMessageController@read fetched every matching DirectMessage and
saved each one individually in a loop, issuing one UPDATE per row. On an
active thread this is N queries.
Pluck the matching ids and perform a single bulk update, preserving the
existing response (the list of affected message ids) and updated_at
behaviour.
Adds regression tests covering the marked-read ids, the status_id lower
bound, and sender isolation.
AvatarController@store and BaseApiController@avatarUpdate wrapped the
upload flow in an empty catch(\Exception) block and returned a success
response even when the upload or save failed.
Log the exception and return a real error response (500 JSON for the
API endpoint, a redirect with validation errors for the web endpoint).
Adds regression tests covering the failure path, the success path, and
non-image rejection.
Shlee
Shift
Your Name
dependabot[bot]
dansup