Closed Bug 888196 Opened 12 years ago Closed 12 years ago

Memory increase in endurance testsruns on Win 8 (since June the 16th)

Categories

(Mozilla QA Graveyard :: Mozmill Tests, defect)

Product:
Mozilla QA Graveyard
Component:
Mozmill Tests
Platform:
x86_64
Windows 8
Type:
defect
Priority:
Not set
Severity:
normal

Tracking

(Not tracked)

Status:
RESOLVED WORKSFORME

People

(Reporter: mihaelav, Unassigned)

References

Details

There is an increase of resident memory in endurance testruns on Windows 8 (both 32 and 64 bit) since June the 16th. From the reports it looks that the memory regression is across all tests. Reports: * Win8 32bit 14June: http://mozmill-ci.blargon7.com/#/endurance/report/f2d3b6136b11cb7b6708636cff35af1e * Win8 32bit 16June: http://mozmill-ci.blargon7.com/#/endurance/report/f2d3b6136b11cb7b6708636cff66b0b7 * Win8 64bit 14June: http://mozmill-ci.blargon7.com/#/endurance/report/f2d3b6136b11cb7b6708636cff35c5a8 * Win8 64bit 16June: http://mozmill-ci.blargon7.com/#/endurance/report/f2d3b6136b11cb7b6708636cff66d20e Charts: http://mozmill-ci.blargon7.com/#/endurance/charts?branch=24.0&platform=Windows%20NT&from=2013-06-09&to=2013-06-28 Pushlog: http://hg.mozilla.org/mozilla-central/pushloghtml?fromchange=b197bed90a98&tochange=36da3cb92193 Note: The regression was not reproducible locally (tested on Win 8 32bit). Reports: http://mozmill-crowd.blargon7.com/#/endurance/reports?branch=24.0&platform=Windows NT&from=2013-06-21&to=2013-06-21
Thanks Mihaela. I've triggered the same builds to run again in our CI to see if the regression can be replicated. If not, I suspect a change in the environment that would prove difficult to determine. If it can be replicated, we can try to narrow down the regression range. I'll comment with the results once they're in.
I was able to replicate this regression: 20130614031100: Average resident memory: 97MB http://mozmill-crowd.blargon7.com/#/endurance/report/a1b02004612785c13cf7c6bf1e0678de 20130616031139: Average resident memory: 117MB http://mozmill-crowd.blargon7.com/#/endurance/report/a1b02004612785c13cf7c6bf1e0d979e Please file a Firefox regression bug and set it to block this one. We should also see if we can narrow down the regression range. You should be able to kick off builds on the Mozmill CI for this, let me know if you need assistance.
Depends on: 888888
That's appears to be a merge from birch, and still has 86 changesets. Perhaps you could use the birch builds and see if you can reduce it further? https://ftp.mozilla.org/pub/mozilla.org/firefox/tinderbox-builds/birch-win32/
This is weird. We've run endurance tests on birch tinderbox builds. And the first one that exhibits the increased memory is a symmetrically (I say this because it contains most changesets from the original birch to m-c merge) merge from m-c to birch: http://hg.mozilla.org/projects/birch/pushloghtml?changeset=f07a9e412086 I fail to understand how development is done on FF on different branches. Why would you have the same changesets that you just merged from birch to m-c merged back into birch?
I'm not sure how the birch builds are run, perhaps they are not the same as inbound. Any ideas Henrik? Otherwise, you may need to bisect and build Firefox to reduce this regression range further.
Flags: needinfo?(hskupin)
Whiteboard: [MemShrink]
Whiteboard: [MemShrink]
Sorry for my last comment. I missed to read the commetns before. So why do we continue here, while we have filed bug 888888 on it already for firefox? We should really do the bisect over there. For questions regarding the merge ask ryan who did the merge.
I think we finally have a winner: http://hg.mozilla.org/integration/mozilla-inbound/pushloghtml?changeset=d1fc221d10a0 Which is bug 879624 This looks consistent with what we saw. Only affects Windows 8, and being related to hardware acceleration might explain increased memory consumption. Vladimir is it expected to have increased memory consumption on Windows 8 with your patch for bug 879624 applied? We are running endurance tests. Here are some results from inbound builds: Prior to your patch: http://mozmill-crowd.blargon7.com/#/endurance/report/5aa1ca5e7015e3740d269dc9477271d8 Your patch applied: http://mozmill-crowd.blargon7.com/#/endurance/report/5aa1ca5e7015e3740d269dc94757c7d6
Flags: needinfo?(vladimir)
This should have probably gone into bug 888888
Updated the relevant bug 888888
Flags: needinfo?(vladimir)
Great work guys, turns out this is an expected increase and we should use this as the new baseline.
Status: NEW → RESOLVED
Closed: 12 years ago
Resolution: --- → FIXED
No fix -> WFM. :)
Resolution: FIXED → WORKSFORME
Product: Mozilla QA → Mozilla QA Graveyard
You need to log in before you can comment on or make changes to this bug.