1193738 - Use mozcrash.kill_and_get_minidump in xpcshell harness

Status: RESOLVED DUPLICATE of bug 1316309

(Testing :: XPCShell Harness, defect)

Description

[(not currently active) Ted Mielczarek]

Since bug 890026 finally landed, we should switch the xpcshell harness to use mozcrash.kill_and_get_minidump instead of Automation.killAndGetStackNoScreenshot (which would remove the harness' dependency on automation.py, which is nice).

Comment 1

[Chris Manchester (limited bugmail, email directly)]

Ted mentioned taking this on IRC (if that's no longer true I'm happy to do it).

Comment 2

[(not currently active) Ted Mielczarek]

It's cool, I'll do it.

Comment 3

[(not currently active) Ted Mielczarek]

https://treeherder.mozilla.org/#/jobs?repo=try&revision=d8a6ee6fff3a

Comment 4

[(not currently active) Ted Mielczarek]

Attachment: MozReview Request: bug 1193738 - use mozcrash.kill_and_get_minidump in xpcshell harness. r?chmanchester

bug 1193738 - use mozcrash.kill_and_get_minidump in xpcshell harness. r?chmanchester

Comment 5

[Chris Manchester (limited bugmail, email directly)]

Comment on attachment 8647603 [details]
MozReview Request: bug 1193738 - use mozcrash.kill_and_get_minidump in xpcshell harness. r?chmanchester

https://reviewboard.mozilla.org/r/16019/#review14339

::: testing/xpcshell/runxpcshelltests.py:760
(Diff revision 1)
>              if self.checkForCrashes(self.tempDir, self.symbolsPath, test_name=name):
>                  if self.retry:
>                      self.clean_temp_dirs(path)
>                      return

We get here now when a test times out but is eligible to retry, so we print the dumps and turn the job orange.

All this retry logic is already pretty convoluted, but a thing to do would be to not check for crashes if we're going to retry.

Comment 6

[(not currently active) Ted Mielczarek]

https://treeherder.mozilla.org/#/jobs?repo=try&revision=8cf7f60f999c

Comment 7

[(not currently active) Ted Mielczarek]

This is looking a lot greener. I'll try another push with a forced hang to sanity check that it works on Try, and I'll see if I can update our existing hang test to more fully test this case.

Comment 8

[(not currently active) Ted Mielczarek]

https://treeherder.mozilla.org/#/jobs?repo=try&revision=e1543315a655 is that push with a forced hang. It's *close*, in that we kill the process and get a stack, but we hit an error trying to kill the process again. I'll fix that up and get a new patch up.

Comment 9

[(not currently active) Ted Mielczarek]

ahal has a patch in bug 1316309