581775 - Buildbot fails to kill off timed out xpcshell on win7

Status: RESOLVED FIXED

(Release Engineering :: General, defect, P2)

Description

[Nick Thomas [:nthomas] (UTC+12)]

Attachment: Exception on slave

This is spun out from 581709, after we fixed the hang in test_punycodeURIs.js and hit this issue in ipc tests.

See the attachment for the exception generated when buildbot tries to kill off an xpcshell that has hung in test_simple_wrap.js or test_head_wrap.js (possibly bug 580525). It could be that the twisted patch needs adjusting to work on win7. 

Note that we deployed buildbot 0.8.0 to the win7 boxes on 2010-07-22.

Comment 1

[Nick Thomas [:nthomas] (UTC+12)]

I also found a couple of xp machines that are failing to move the tp pageset aside:
http://talos-master02.build.mozilla.org:8012/builders/Rev3%20WINNT%205.1%20mozilla-central%20talos%20tp4/builds/6
http://talos-master02.build.mozilla.org:8012/builders/Rev3%20WINNT%205.1%20mozilla-central%20talos%20dromaeo/builds/4

Comment 2

[Chris AtLee [:catlee]]

Looks like copying _dumpwin32proc.py from the win7-64 machines will fix this.  It at least gets rid of the ProcessExitedAlready error.

Comment 3

[Nick Thomas [:nthomas] (UTC+12)]

I also noticed that we hit this doing the reboot step elsewhere, eg snow leopard
 http://pastebin.mozilla.org/766148

Comment 4

[Nick Thomas [:nthomas] (UTC+12)]

FWIW, the exception doesn't say what process it failed to kill off but there's still an xpcshell.exe, plugin-container.exe, and two python processes running after the fact. That's for when we have issue from bug 580525.

Comment 5

[Chris AtLee [:catlee]]

Fixed by copying _dumbwin32proc.py from the win7-64 slaves to the win7 slaves.