Closed
Bug 1093328
Opened 10 years ago
Closed 10 years ago
mozcrash may flood test log if many minidumps found
Categories
(Testing :: Mozbase, defect)
Tracking
(Not tracked)
RESOLVED
FIXED
mozilla36
People
(Reporter: gbrown, Assigned: gbrown)
References
Details
Attachments
(1 file)
1.29 KB,
patch
|
ted
:
review+
|
Details | Diff | Splinter Review |
In bug 1091916, an Android x86 crash was introduced for certain xpcshell tests. When the crash occurs, many (over 1000) minidump files are created. See https://tbpl.mozilla.org/php/getParsedLog.php?id=51563406&tree=Mozilla-Inbound&full=1. When this happens, mozcrash dutifully calls minidump_stackwalk for each one, flooding the test log with crash report after crash report. Eventually, we hit: Output exceeded 52428800 bytes, remaining output has been truncated (output was 52458724 bytes) command timed out: 14400 seconds elapsed running ['/tools/buildbot/bin/python', 'scripts/scripts/android_emulator_unittest.py', '--cfg', 'android/androidx86.py', '--test-suite', 'robocop-1', '--test-suite', 'robocop-2', '--test-suite', 'robocop-3', '--test-suite', 'xpcshell', '--blob-upload-branch', 'mozilla-inbound', '--download-symbols', 'ondemand'], attempting to kill process killed by signal 9 It is not clear to me if the time-out is caused by all the time spent dumping crash reports, log processing, or something else. And I don't understand why we are generating so many minidump files. But I think we can handle this situation better: If more than N minidump files are found by mozcrash, let's just process and report the first N. N = 10, maybe?
Comment 1•10 years ago
|
||
That makes sense. This is a pretty broken situation to get into.
Assignee | ||
Comment 2•10 years ago
|
||
Is 10 a good number? Could/should we limit this in breadpad also, or in addition? https://treeherder.mozilla.org/ui/#/jobs?repo=try&revision=7e4fe39a7ac7
Attachment #8516705 -
Flags: review?(ted)
Comment 3•10 years ago
|
||
Comment on attachment 8516705 [details] [diff] [review] do not report more than 10 crash reports at a time Review of attachment 8516705 [details] [diff] [review]: ----------------------------------------------------------------- Seems reasonable for an arbitrary limit. Might make that a .warn or something more serious.
Attachment #8516705 -
Flags: review?(ted) → review+
Assignee | ||
Comment 4•10 years ago
|
||
Changed to .warning - thanks. https://hg.mozilla.org/integration/mozilla-inbound/rev/e80e4fa3a93d
https://hg.mozilla.org/mozilla-central/rev/e80e4fa3a93d
Status: NEW → 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
•