test: mark test-fs-rmdir-recursive flaky on win

[mhdawson]

Refs: #41201

From recent reliability reports this is now the most
common failure by far in CI runs. Mark the test as
flaky until the issue is resolved.

Signed-off-by: Michael Dawson mdawson@devrus.com

[lpinca]

The issue might be fixed by #41545 so I would wait before landing this.

[mhdawson]

@lpinca thanks, landed #41545. Will see if the problem still recreates before landing this.

[mhdawson]

@bcoe any chance you could have a quick look at parallel/test-fs-rm to see if it has similar issues. It is also flaky but does not fail quite as often.

parallel/test-fs-rm

[bcoe]

@mhdawson I did notice a couple missing awaits in test-fs-rm.js, which were included in this PR:

6b9d2ae#diff-596c3dc1f2c2b0c3dc8f6c087b9e0f1910188ff28eaee22800367acc0e9a4f43L188

I'm not 100% sure they're the cause of flakes. But I could imagine them causing some weirdness depending on timing. It seemed like the step that fails is actually the tmpdir helper attempting to cleanup after itself, perhaps the cleanup step fails on some operating systems because there are other processes attempting file operations at the same time?

Should we keep this open for a little bit and see if the flakes have gone away? How often were we seeing the failures.

[mhdawson]

Should we keep this open for a little bit and see if the flakes have gone away? How often were we seeing the failures.

[mhdawson]

Should we keep this open for a little bit and see if the flakes have gone away? How often were we seeing the failures.

That was my plan. I'll leave open for a week or so to see how the flakes look on new PRs. If things are green (keeping my fingers crossed) then I'll go ahead and close this.

[mhdawson]

@bcoe I'll also say that today things look better in the CI so I'm hopeful. Thanks for your help on this one.

[bcoe]

@mhdawson if I'm reading the daily reports correctly, I'm still seeing quite a few flakes in test-fs-rmdir-recursive.

One thought I had was serializing the tests, so that the promise API and callback API are not being exercised a the same time -- my hunch continues to be parallel operations causing issues in Windows due to contention on files.

In the very least, perhaps we will get a clearer picture of which test is failing to cleanup.

[mhdawson]

There have been a few failures within the last few days, but they might be from PRs that were not rebased? I figure we should wait another week and then take a look at the reliability report again.

[mhdawson]

@bcoe unfortunately from the latest reliability report - nodejs/reliability#185. It still looks like failures are occurring with parallel/test-fs-rmdir-recursive. The latest failure was just yesterday and the PR being tested was only opened yesterday as well.

@lpinca I'm going to propose we land this PR. I'll agree to keep an eye on the CI and if we no longer see the test failing in the next week or so I'll back out the change, otherwise we can do that as part of whatever future fixes/updates there are to make the test more reliable on windows.

[lpinca]

@mhdawson I agree, go ahead.

[nodejs-github-bot]

CI: https://ci.nodejs.org/job/node-test-pull-request/42206/

[mhdawson]

Landed in 7faf763