[Bug]: Tail rolling files test is flaky

[DABH]

🔎 Search Terms

zippedArchive

The problem

[Edit: see bottom of thread; original test may be fixed now but another test may be flaky...]

See #2337 (comment)

This may indicate that the zippedArchive functionality itself is flaky; further investigation is required here. Hoping that the PR authors there or someone else from the community can step in and help investigate and resolve... 🙏

What version of Winston presents the issue?

master

What version of Node are you using?

16, 18

[DABH]

@Pulkit0729 @pixtron, wondering if there's any chance you might be able to look into this? Huge quality of life issue for maintaining the repository, unfortunately (I'm manually re-running lots of tests on PRs to get "good runs" for this flaky test). My biggest concern is whether just the test is flaky, or whether the underlying functionality itself doesn't work sometimes...

[Pulkit0729]

Ok, i will update you on this @DABH

[pixtron]

By reading the test, to me it looks like a race condition in the test itself.

When logged is emitted we are not sure, if the content has already been flushed to disk or not. If the content has already been flushed, a rotate is save and testarchive.log.gz might already exist (compression started or even already finished). Not sure how to fix this test tho.

[Pulkit0729]

@DABH Added a Pr with the fix, it was a silly mistake from my side

[DABH]

It appears that even after merging @Pulkit0729's PR, testtailrollingfiles is still flaky... so going to leave this issue open / update the details since I think the flake might be related to #2337 still (I don't recall ever seeing that test be flaky before #2337 was merged...)