NixReaperandClaude Sonnet 5
0a50e0ffb6
CI / lint (push) Canceled after 0s
CI / typecheck (push) Canceled after 0s
CI / unit-tests (push) Canceled after 0s
CI / integration-tests (push) Canceled after 0s
CI / container-build (push) Canceled after 0s
CI / dependency-audit (push) Canceled after 0s
1. test_ingestion_cache_invalidation.py: the job was created through
the test's db_session fixture, whose commit() only ends a savepoint
that gets rolled back at teardown (see tests/conftest.py) -- it's
never a real commit. _run_ingestion looks the job up through its
own session_scope() (a genuinely separate connection), which could
never see it, so the very first _require_job call raised "job
vanished mid-run" every time. Worse, test_failed_ingestion_does_not_
invalidate_the_cache still passed, coincidentally -- it expected a
RuntimeError and got one, just the wrong one, masking the bug.
Fixed by having _make_job create the job through session_scope()
itself, matching test_ingestion_retry.py's existing pattern.
2. Same class of bug as the video-collision fix in the last commit:
FAKE_MOVIE's fixed imdb_id ("tt0133093"), not just its video, also
collides across two different fake movies sharing it unmodified --
uq_external_id_source is a real unique constraint on
(provider_id, entity_type, external_id). Fixed by also overriding
imdb_id on the second movie in both affected tests.
All caught by running the actual integration suite against Postgres --
these paths were previously only collect-only'd locally (no local DB).
Co-Authored-By: Claude Sonnet 5 <[email protected]>