Skip to content

flaky test-regress-GH-814_2 #34527

Description

@Trott
  • Version: 15.0.0-pre (current master branch)
  • Platform: fedora-last-latest-x64
  • Subsystem: test, fs, libuv

What steps will reproduce the bug?

Run test/pummel/test-regress-GH-814_2.js on fedora-last-latest-x64 in CI.

How often does it reproduce? Is there a required condition?

Not every time. Running https://ci.nodejs.org/job/node-stress-single-test/151/ right now to see how often.

What is the expected behavior?

The test should pass.

What do you see instead?

The test crashes with -9, which I believe means libuv is complaining about a bad file descriptor.

21:31:32 not ok 2805 pummel/test-regress-GH-814_2
21:31:32   ---
21:31:32   duration_ms: 11.27
21:31:32   severity: crashed
21:31:32   exitcode: -9
21:31:32   stack: |-
21:31:32     /home/iojs/node-tmp/.tmp.2804/GH-814_test.txt
21:31:32     
21:31:32   ...

Additional information

Something similar to this was going on with readstream last year and I think @addaleax fixed it. Maybe fs has the same/similar issue, perhaps when explicit fs.close() is not used?

Activity

  1. added
    fsIssues and PRs related to file-system APIs and the fs module.
    libuvIssues and PRs related to the libuv dependency or the uv binding.
    flaky-testIssues and PRs involving tests that fail intermittently in CI.
    on Jul 27, 2020
  2. Trott commented on Jul 27, 2020

    @Trott
    MemberAuthor

    @nodejs/testing @nodejs/fs @nodejs/libuv

  3. richardlau commented on Jul 27, 2020

    @richardlau
    Member

    Looks like a copy-paste error with the RUN_TESTS parameter for that stress run. I've corrected it and started a rebuild: https://ci.nodejs.org/job/node-stress-single-test/152/

  4. addaleax commented on Jul 27, 2020

    @addaleax
    Member

    The test crashes with -9, which I believe means libuv is complaining about a bad file descriptor.

    Signal 9 is SIGKILL. Looking at the test, this is most likely due to the process running out of memory.

  5. Trott commented on Jul 27, 2020

    @Trott
    MemberAuthor

    The test crashes with -9, which I believe means libuv is complaining about a bad file descriptor.

    Signal 9 is SIGKILL. Looking at the test, this is most likely due to the process running out of memory.

    Yes, I'm more-than-mildly embarrassed that my brain didn't realize that a crash wouldn't be a uv error code but the signal code instead.

    So if it's probably running out of memory, then I guess the thing to do might be to add a free memory check before running like we do for some other tests.

  6. Trott commented on Jul 27, 2020

    @Trott
    MemberAuthor

    So if it's probably running out of memory, then I guess the thing to do might be to add a free memory check before running like we do for some other tests.

    But not if we don't/can't replicate the problem, I guess....

  7. addaleax commented on Jul 27, 2020

    @addaleax
    Member

    PR to replace the test(s) with a more modern one: #34530

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Metadata

Metadata

Assignees

No one assigned

    Labels

    flaky-testIssues and PRs involving tests that fail intermittently in CI.fsIssues and PRs related to file-system APIs and the fs module.libuvIssues and PRs related to the libuv dependency or the uv binding.

    Type

    No type

    Projects

    No projects

      Milestone

      No milestone

      Relationships

      None yet

      Development

      No branches or pull requests

      Issue actions