Disco DeDisco ca169be0fb fix: CI Postgres teardown — RobustCompressorTestRunner.teardown_databases now force-closes lingering connections before Django's DROP. CI step test-UTs-n-ITs (python manage.py test apps, full suite incl. channels-tagged tests) was failing post-test even when all 1165 tests passed — psycopg2.errors.ObjectInUse: database "test_python_tdd_test" is being accessed by other users / DETAIL: There is 1 other session using the database. Two-step leak: (1) core.settings.DATABASES['default']['conn_max_age']=600 keeps Postgres connections alive in the per-thread pool for 10 min (prod-perf default); (2) Channels' database_sync_to_async (16 call sites across apps.epic.tests.integrated.test_consumers's CursorMoveConsumerTest + SigHoverConsumerTest) runs in a process-wide asgiref threadpool — each worker thread accumulates its own DB connection that outlives the test + sits idle in the pool when teardown fires. Postgres refuses DROP while ANY session targets the row. Local dev unaffected: --exclude-tag=channels skips the consumer tests + SQLite has no DROP step. **Fix** lives entirely in core/runner.py's teardown override — connections.close_all() covers the main thread's runner connection; iterating old_config + running SELECT pg_terminate_backend(pid) FROM pg_stat_activity WHERE datname = %s AND pid <> pg_backend_pid() against each test DB kicks any worker-thread session still pinning the row. Both safe on a green run (DB about to be dropped anyway) + scoped to vendor == "postgresql" so SQLite local-dev is a clean no-op. Prod CONN_MAX_AGE=600 untouched — fix lives in the test runner, NOT in settings. 19/19 lyric UTs green via the new runner path (smoke verify the override is benign on SQLite); Postgres-side validated next CI run. Trap captured: [[feedback-test-teardown-conn-leak]] — symptom signature Ran NNNN tests / OK / Destroying.../ObjectInUse: ... belongs in CI-fail triage notes so future flakes get diagnosed in seconds instead of test-hunting ...

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>

2026-05-21 14:24:41 -04:00

.woodpecker

CI: route test_game_my_sea*.py to test-FTs-room stage — 49 my-sea FTs DRY-reuse the room-shell hex + sea-cross picker (same Selenium surface as test_game_room_* + test_trinket_*), so they belong w. the heavy room flows instead of bloating test-FTs-non-room. Filename-regex partition stays clean (13 room + 24 non-room = 37 total, no overlap)

2026-05-20 13:20:23 -04:00

infra

pyswiss_url env var added to more places throughout ansible vault architecture; staging now has working astro wheel

2026-04-16 12:04:46 -04:00

pyswiss

natus wheel: Semisquare & Sesquiquadrate; Ott orb pair detection; intensity sort

2026-04-27 20:16:47 -04:00

src

fix: CI Postgres teardown — RobustCompressorTestRunner.teardown_databases now force-closes lingering connections before Django's DROP. CI step test-UTs-n-ITs (python manage.py test apps, full suite incl. channels-tagged tests) was failing post-test even when all 1165 tests passed — psycopg2.errors.ObjectInUse: database "test_python_tdd_test" is being accessed by other users / DETAIL: There is 1 other session using the database. Two-step leak: (1) core.settings.DATABASES['default']['conn_max_age']=600 keeps Postgres connections alive in the per-thread pool for 10 min (prod-perf default); (2) Channels' database_sync_to_async (16 call sites across apps.epic.tests.integrated.test_consumers's CursorMoveConsumerTest + SigHoverConsumerTest) runs in a process-wide asgiref threadpool — each worker thread accumulates its own DB connection that outlives the test + sits idle in the pool when teardown fires. Postgres refuses DROP while ANY session targets the row. Local dev unaffected: --exclude-tag=channels skips the consumer tests + SQLite has no DROP step. **Fix** lives entirely in core/runner.py's teardown override — connections.close_all() covers the main thread's runner connection; iterating old_config + running SELECT pg_terminate_backend(pid) FROM pg_stat_activity WHERE datname = %s AND pid <> pg_backend_pid() against each test DB kicks any worker-thread session still pinning the row. Both safe on a green run (DB about to be dropped anyway) + scoped to vendor == "postgresql" so SQLite local-dev is a clean no-op. Prod CONN_MAX_AGE=600 untouched — fix lives in the test runner, NOT in settings. 19/19 lyric UTs green via the new runner path (smoke verify the override is benign on SQLite); Postgres-side validated next CI run. Trap captured: [[feedback-test-teardown-conn-leak]] — symptom signature Ran NNNN tests / OK / Destroying.../ObjectInUse: ... belongs in CI-fail triage notes so future flakes get diagnosed in seconds instead of test-hunting

2026-05-21 14:24:41 -04:00

.dockerignore

ignore .vscode/ in git & docker contexts; workspace tasks/settings are local-only (Windows paths + per-machine venv layout)

2026-05-04 18:17:54 -04:00

.gitattributes

added .gitattributes for real this time

2026-02-17 17:57:20 -05:00

.gitignore

ignore .vscode/ in git & docker contexts; workspace tasks/settings are local-only (Windows paths + per-machine venv layout)

2026-05-04 18:17:54 -04:00

ansible.cfg

Gitea repo init; Woodpecker pipeline init; much CI/CD structure outlined in ./infra, incl. docker-compose.yaml, cicd-playbook.yaml, new cicd vault, & config files for gitea, woodpecker & ansible

2026-02-10 21:16:00 -05:00

CLAUDE.md

rename natus → sky across the codebase — natal chart abstraction is now sky throughout, since chart inputs aren't birthday-gated

2026-05-04 20:36:15 -04:00

Dockerfile

Dockerfile updated to run uvicorn worker class to support asgi (was still gunicorn & wsgi)

2026-03-18 22:03:10 -04:00

Dockerfile.ci

Dockerfile.ci & debug-ci now build working container for CI test image

2026-02-10 22:42:45 -05:00

requirements.dev.txt

requirements: align dev w. prod — add celery, psycopg2-binary, redis to requirements.dev.txt (missing from the CI dev image's pre-installed set, which means the CI fallback pip install -r requirements.dev.txt step would not satisfy these prod runtime deps if the image lagged); bump requests 2.31.0 → 2.32.5 in requirements.txt to align w. the dev pin (eliminates the version-drift wedge where CI's dev-only install would silently downgrade-or-upgrade vs prod's pin)

2026-05-12 20:10:22 -04:00

requirements.txt

requirements: align dev w. prod — add celery, psycopg2-binary, redis to requirements.dev.txt (missing from the CI dev image's pre-installed set, which means the CI fallback pip install -r requirements.dev.txt step would not satisfy these prod runtime deps if the image lagged); bump requests 2.31.0 → 2.32.5 in requirements.txt to align w. the dev pin (eliminates the version-drift wedge where CI's dev-only install would silently downgrade-or-upgrade vs prod's pin)

2026-05-12 20:10:22 -04:00

requirements.wsl.txt

added wsl reqs

2026-01-13 20:58:25 -05:00

Description

No description provided

155 MiB

Languages

Python 45%

JavaScript 37.6%

HTML 9%

SCSS 8.2%

Jinja 0.1%