Open Bug 1449576 Opened 3 years ago Updated 3 years ago

Missing crash stack from intermittent timeout in bug 1441580

Categories

(Testing :: Reftest, defect)

Version 3
defect
Not set
normal

Tracking

(Not tracked)

People

(Reporter: ahal, Unassigned)

References

Details

There's an intermittent reftest startup hang in bug 1441580. We should be killing the process and producing a minidump but for some reason that isn't happening.
Here's an example log:
https://treeherder.mozilla.org/logviewer.html#?repo=mozilla-inbound&job_id=170601870&lineNumber=1635

The 'crashinject: exit 0' line tells me that we're at least hitting killAndGetStack and it returns 0:
https://dxr.mozilla.org/mozilla-central/source/layout/tools/reftest/runreftest.py#663

So possibly we just need to call check_for_crashes. Maybe this was a regression from run-by-manifest.
Nvm, checking for crashes should work properly.

The real reason appears to be that 'crashinject' isn't killing the browser. Notice that after the crashinject status comes back there's a 15 minute timeout followed by a mozprocess timeout error. If crashinject had worked, then mozprocess would return right away.
Wonder if this is also a manifestation of bug 1378743. Maybe on Linux/OSX the stack still gets reported but on Windows it doesn't?
See Also: → 1378743
(In reply to Andrew Halberstadt [:ahal] from comment #0)
> There's an intermittent reftest startup hang in bug 1441580. We should be
> killing the process and producing a minidump but for some reason that isn't
> happening.

Please note that the hang is no longer visible since the generic worker has been upgraded on Windows. So it might be even harder to reproduce it now.
You need to log in before you can comment on or make changes to this bug.