Closed
Bug 1060527
Opened 10 years ago
Closed 10 years ago
We don't check for crashes if we get an exception in ProcessHandler
Categories
(Testing :: Mochitest, defect)
Testing
Mochitest
Tracking
(Not tracked)
RESOLVED
FIXED
mozilla36
People
(Reporter: jgriffin, Assigned: ahal)
References
()
Details
Attachments
(1 file)
1.45 KB,
patch
|
jgriffin
:
review+
|
Details | Diff | Splinter Review |
When a job times out, we normally deliberately crash the process so we can get a stack trace from it. See https://tbpl.mozilla.org/php/getParsedLog.php?id=46820280&tree=Mozilla-Central&full=1 for an example of this working normally on Windows (search for 'crashinject').
On some jobs, though, it looks our deliberate crash is causing mozprocess to throw an error, and when this happen we don't invoke mozcrash. See bug 900112 and https://tbpl.mozilla.org/php/getParsedLog.php?id=46961924&tree=Fx-Team (again, search for 'crashinject').
Reporter | ||
Comment 1•10 years ago
|
||
Ahal, can you take a look? Let me know if it's not an easy fix.
Blocks: 900112
Flags: needinfo?(ahalberstadt)
Updated•10 years ago
|
Assignee | ||
Comment 2•10 years ago
|
||
I don't understand this IO Completion Port stuff very well, nor do I have a windows environment handy to test on. Just based on the logs and source code though, I think this patch might fix the issue. I have no idea why this is only happening some of the time though and I also don't know if this will have undesired side effects.
Here's a try run:
https://tbpl.mozilla.org/?tree=Try&rev=a17103a474a5
Flags: needinfo?(ahalberstadt)
Reporter | ||
Comment 3•10 years ago
|
||
Comment on attachment 8482748 [details] [diff] [review]
mozprocess_signal_completion
Review of attachment 8482748 [details] [diff] [review]:
-----------------------------------------------------------------
This looks reasonable to me as well. I think there's a race condition somewhere with our handling of Windows processes, such that killing a process (as with crashinject) can sometimes trigger this code path. This won't fix the race condition, but will hopefully render it harmless.
Attachment #8482748 -
Flags: review+
Assignee | ||
Comment 4•10 years ago
|
||
So sorry, this completely slipped past me. Thank you harth for your bugzilla todo app that found this.
Here's a try run:
https://treeherder.mozilla.org/ui/#/jobs?repo=try&revision=043cba306d20
Looked good, so landed:
https://hg.mozilla.org/integration/mozilla-inbound/rev/70a6139f5517
Assignee: nobody → ahalberstadt
Status: NEW → ASSIGNED
Comment 5•10 years ago
|
||
Status: ASSIGNED → RESOLVED
Closed: 10 years ago
Resolution: --- → FIXED
Target Milestone: --- → mozilla36
You need to log in
before you can comment on or make changes to this bug.
Description
•