parallel/test-trace-events-api crash on Windows

[Trott]

• Version: v10.0.0-pre (master)
• Platform: win10 vs2017
• Subsystem: test trace_events

https://ci.nodejs.org/job/node-test-binary-windows/16687/COMPILED_BY=vs2017,RUNNER=win10,RUN_SUBSET=3/console

not ok 439 parallel/test-trace-events-api
  ---
  duration_ms: 0.257
  severity: crashed
  exitcode: -1073741819
  stack: |-
    oh no!
    exit code: CRASHED

Not sure if the exit code reveals anything useful about the probable cause of the crash.

@nodejs/platform-windows @nodejs/trace-events

[addaleax]

Not sure if the exit code reveals anything useful about the probable cause of the crash.

That’s 0xC0000005, which is basically the Windows equivalent of a segfault.

I’m not sure where the oh no! is coming from, btw – that might be a more helpful clue.

[richardlau]

I’m not sure where the oh no! is coming from, btw – that might be a more helpful clue.

Probably not -- it's coming from the test runner:

node/tools/test.py

Line 301 in dadddcb

self.traceback = "oh no!\nexit code: " + PrintCrashed(exit_code)

[bzoz]

I cant reproduce this, does this still happen in the CI?

[Trott]

@bzoz It happened in the last week: #20619

[apapirovski]

This is still happening FWIW. I'm labeling with flaky test because we don't really know what's causing it and maybe it'll get more attention.

[Trott]

@nodejs/testing @nodejs/build

[jasnell]

Yeah, as we've been digging in and expanding the trace events stuff, @ofrobots and I have been spotting a number of issues in the current implementation that lead to aborts/segfaults. I know @ofrobots is looking at a few PRs to get things fixed up. Marking the test as flaky for now is the right thing.

[ofrobots]

I suspect #21335 will fix this.

[jasnell]

Stress test: https://ci.nodejs.org/job/node-stress-single-test/1986/ (forgot to select actual platforms ;-) ...)
https://ci.nodejs.org/job/node-stress-single-test/1987/