More attempts to get CI to be less flaky#18652
More attempts to get CI to be less flaky#18652benjyw wants to merge 6 commits intopantsbuild:mainfrom
Conversation
|
Harrumph, this test failure is not at all clear to me: https://github.com/pantsbuild/pants/actions/runs/4585391355/jobs/8097532089?pr=18652 @stuhood could this be a byproduct of #18632? Or something it should have fixed but didn't? |
kaos
left a comment
There was a problem hiding this comment.
Using arbitrary delays with time.sleep() is very unfortunate.
Can we come up with a more predictable way to synchronize the condition we are actually waiting for?
5b4b5f8 to
3b90a21
Compare
f8d6c7b to
b3774de
Compare
b3774de to
f12c805
Compare
| # Ensure the test invocation of Pants doesn't restart itself after it writes the tailored files, | ||
| # and then chokes on the invalid BUILD symbol. | ||
| result = rule_runner.run_goal_rule( | ||
| TailorGoal, args=["--alias-mapping={'fortran_library': 'my_fortran_lib'}", "::"] | ||
| TailorGoal, | ||
| global_args=["--no-watch-filesystem", "--no-pantsd"], | ||
| args=["--tailor-alias-mapping={'fortran_library': 'my_fortran_lib'}", "::"], | ||
| ) |
There was a problem hiding this comment.
This is not an integration test, so no external processes or pantsd are spawned: everything runs in process, and file invalidation happens synchronously during the write_files call. So these arguments have no effect.
|
|
||
| # See comment in `test_pantsd_invalidation_pants_toml_file`. | ||
| time.sleep(15) | ||
| time.sleep(30) |
There was a problem hiding this comment.
I made a recent attempt at bumping these, without luck.
It might be a good idea to assign me a ticket to more deeply investigate these, and then skip them for now.
cf0b9dc to
cc071d3
Compare
Based on tests that still fail in a loaded CI environment, even after #18556