Repository navigation
Conversation
…arning
_make_backend created Database(':memory:') and nothing closed it: the raw
sqlite3.Connection was finalized by the GC at an arbitrary later test, and
filterwarnings=error turned that PytestUnraisableExceptionWarning into a
failure attributed to whichever test was running (observed twice in mutmut's
sandbox — stats collection died at test_adk_backend with 1713 tests green;
GC timing there differs from the normal suite run, where the test passes).
The autouse fixture closes every created Database; the three sync contract
tests become async (module is anyio-based) so the async fixture can apply.
Co-Authored-By: Claude Code <noreply@anthropic.com>
…ents /simplify follow-up: test_codex_backend._make_backend had the identical leaked-Database pattern (altitude finding) — same registry + autouse close, sync contract test converted (async autouse fixture). Comments trimmed to the mechanism; the investigation story lives in the commit history. Co-Authored-By: Claude Code <noreply@anthropic.com>
axisrow
force-pushed
the
fix/close-adk-test-db
branch
from
October 7, 2026 06:34
00cc842 to
1c812c4
Compare
Owner
Author
|
Закрыт как ошибочный диагноз (проверено ревью): регистрируемые базы в adk/codex тестах никогда не открывают соединение — закрывать было нечего; реальный источник unraisable — незакрытые file-backed базы в test_cli_photo_loader/extended. Честный фикс: #1505 (cli_init_patch закрывает базы по завершении теста, покрывая все 13 файлов с фикстурой). Но-оп-механика в main не попала. |
axisrow
added a commit
that referenced
this pull request
Oct 8, 2026
… net) (#1505) * fix(tests): close cli_init_patch databases at test teardown — real unraisable-source hygiene Post-merge review of #1502 proved its mechanism a no-op: the Databases created in test_adk_backend/test_codex_backend never open a connection (lazy DatabaseConnection, backends never read self._db), so closing them could not prevent the GC-unraisable flakes; the real leak sites are file-backed Databases in test_cli_photo_loader.py (19x asyncio.run(initialize()), zero close()) and test_cli_extended.py. Fix at the shared altitude: cli_init_patch registers the handed-in Database in a per-test registry and an autouse finalizer closes them after the test — after teardown nothing can legally use them (closing inside the context manager broke 12 tests that touch db after the with-block), and a failed close logs a warning instead of failing the suite. Note: 12 test_cli_agent_commands failures on this machine are pre-existing on clean main and are a different family — unclosed local async-generator fakes GC-finalized mid-run (_fake_stream), not sqlite; filed separately. Co-Authored-By: Claude Code <noreply@anthropic.com> * fix(tests): leak net closes db before pop; failed close surfaces as teardown error Review of #1505: pop-before-close made a failed close unrecoverable, and the swallowed log line hid the exact flake class the net targets. Close first, pop after — the next teardown retries, and a close failure is a pytest teardown error, not a silent warning. Drop the dead BLE001 noqa and the logging import. Add the only exerciser of the close branch: a deliberately leaked db is verified closed via a module-scope witness fixture. Co-Authored-By: Claude Code <noreply@anthropic.com> --------- Co-authored-by: axisrow <axisrow@users.noreply.github.com> Co-authored-by: Claude Code <noreply@anthropic.com>
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Проблема
_make_backendвtests/test_adk_backend.pyсоздаётDatabase(':memory:'), и никто его не закрывает: сыройsqlite3.Connectionдоедается GC в случайный момент более позднего теста, аfilterwarnings=errorпревращаетPytestUnraisableExceptionWarningв падение, приписанное тому тесту, который в этот момент работал.Проявилось дважды в песочнице mutmut (там другой тайминг GC): статс-сбор умирал на
test_adk_backendпри 1713 зелёных тестах; в обычном прогоне репозитория тест зелёный — потому баг и был невидим.Фикс
_make_backend-ом базы (реестр_created_dbs).async def(модуль anyio), иначе async-фикстура к ним неприменима.Верификация
pytest tests/test_adk_backend.py— 20 passed, 1 skipped (было 18 passed + 3 errors до конверсии).test_chat_stream_emits_text_then_doneс этим самым unraisable-warning (лог живой верификации PR-кноба mutmut); после фикса — прогон статс-фазы проходит (см. следующий прогон в описании PR fix(tooling): mutmut sandbox copies conftest+scripts; document flat-src incompatibility #1501-цепочки).Refs: boxed/mutmut#593 (кноб strip_src_prefix), PR #1501 (конфиг песочницы mutmut).
🤖 Generated with Claude Code