potiuk opened a new pull request, #68968: URL: https://github.com/apache/airflow/pull/68968
Both failures came from one **Pendulum2 special-test** run (`Tests (AMD)`, v3-3-test). That job runs in a freshly rebuilt env, so the Python bytecode cache is cold and first imports are slow. The job passes normally — these are timing flakes, not regressions — so this just makes the two tests tolerant of slow/cold runners. **`test_cli_run_time`** asserted the best of **3** `airflow --help` startup samples was under a tight **5s** bound; on the loaded/cold runner the whole-provider-load startup consistently sat just over (min 5.49s). Now takes the min of **5** samples and uses a looser **8s** threshold — still catches gross startup regressions, no longer flakes on small fluctuations. **`TestProviderManager`** builds the full `ProvidersManager`, importing every provider hook. The first such import in the process pays a one-time cold-import cost — notably `tableauserverclient`, which self-instruments via `beartype.claw` on first import (**~50s locally**) — which can exceed the default **60s** per-test execution timeout (it timed out at 60s here). Added a class-level `@pytest.mark.execution_timeout(180)`; the tests are sub-second once imports are cached. Test-only; verified both pass locally (provider-manager test paid the real ~50s cold import and passed under the 180s marker). > Should be backported to `v3-3-test` (where it surfaced). --- ##### Was generative AI tooling used to co-author this PR? - [X] Yes — Claude Code (Opus 4.8) Generated-by: Claude Code (Opus 4.8) following [the guidelines](https://github.com/apache/airflow/blob/main/contributing-docs/05_pull_requests.rst#gen-ai-assisted-contributions) -- This is an automated message from the Apache Git Service. To respond to the message, please log on to GitHub and use the URL above to go to the specific comment. To unsubscribe, e-mail: [email protected] For queries about this service, please contact Infrastructure at: [email protected]
