Skip to content

investigate flaky addons/stringbytes-external-exceed-max/test-stringbytes-external-exceed-max-by-1-hex #17849

Closed

Description

@Trott
  • Version: v10.0.0-pre
  • Platform: alpine35-container-x64/
  • Subsystem: test

https://ci.nodejs.org/job/node-test-commit-linux/15185/nodes=alpine35-container-x64/console

not ok 1910 addons/stringbytes-external-exceed-max/test-stringbytes-external-exceed-max-by-1-hex
  ---
  duration_ms: 7.576
  severity: crashed
  stack: |-
    oh no!
    exit code: CRASHED (Signal: 9)
  ...

Activity

  1. added
    flaky-testIssues and PRs involving tests that fail intermittently in CI.
    on Dec 24, 2017
  2. added
    addonsIssues and PRs related to native addons.
    testIssues and PRs related to Node.js core tests and test infrastructure.
    on Dec 25, 2017
  3. bnoordhuis commented on Dec 25, 2017

    @bnoordhuis
    Member

    Signal 9 = SIGSEGV. If you can obtain a backtrace and a disassembly of the call site + info registers, I can probably figure out what the issue is.

    Educated guess: missing nullptr check somewhere.

  4. apapirovski commented on Jan 3, 2018

    @apapirovski
    Contributor

    Given that this seems to be happening solely on Linux-based containers — is it possible this has to do with memory overcommitting? Signal 9 is SIGKILL, afaik, which would signal that the malloc test succeeds (due to overcommitting) but then the memory is not actually available when it tries to actually use it later on in the test.

    (SIGSEGV is 11, right?)

  5. Trott commented on May 8, 2018

    @Trott
    MemberAuthor

    Might have a totally different cause then the Linux SIGSEGV (which I haven't seen in a while now), but it timed out on macOS today:

    https://ci.nodejs.org/job/node-test-commit-osx/18382/nodes=osx1010/console

    not ok 2007 addons/stringbytes-external-exceed-max/test-stringbytes-external-exceed-max-by-1-hex
      ---
      duration_ms: 120.392
      severity: fail
      exitcode: -15
      stack: |-
        timeout
      ...

    Host was test-requireio-osx1010-x64-1. Tests seem to run a significantly faster on the macstadium hosts. This one typically takes 4 seconds or so on macstadium but can take 10 or 50 seconds on requireio. If we're moving off of requireio, then we can probably ignore this timeout until we see it on macstadium. (Are we moving to exclusively using macstadium, @nodejs/build?)

  6. apapirovski commented on Jun 1, 2018

    @apapirovski
    Contributor

    @nodejs/build Anyone want to investigate the memory overcommit suggestion and disable it on Alpine machines? It makes it impossible to detect available memory which tests like this one rely on.

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

    addonsIssues and PRs related to native addons.flaky-testIssues and PRs involving tests that fail intermittently in CI.testIssues and PRs related to Node.js core tests and test infrastructure.

    Type

    No type

    Projects

    No projects

      Milestone

      No milestone

      Relationships

      None yet

      Development

      No branches or pull requests

      Issue actions