RESOLVED INVALID 112875
REGRESSION(r146417-r146431): Substantially more leaks detected on Mountain Lion 
https://bugs.webkit.org/show_bug.cgi?id=112875
Summary REGRESSION(r146417-r146431): Substantially more leaks detected on Mountain Lion
Ryosuke Niwa
Reported 2013-03-20 22:03:45 PDT
http://build.webkit.org/builders/Apple%20MountainLion%20%28Leaks%29 Mar 20 20:20 146431 failure #3715 Failed 9145 total leaks found for a total of 3,025,392 bytes! 203 unique leaks found! 5 failures 11 new passes 23 flakes Mar 20 19:42 146430 failure #3714 Failed exiting early after 20 crashes and 0 timeouts. 3456 tests run. 20 failures 3 new passes 6 unit tests failed or timed out Mar 20 19:15 146426 failure #3713 Failed exiting early after 20 crashes and 0 timeouts. 3459 tests run. 20 failures 3 new passes 6 unit tests failed or timed out Mar 20 18:21 146420 failure #3712 Failed exiting early after 20 crashes and 0 timeouts. 3459 tests run. 20 failures 3 new passes 6 unit tests failed or timed out Mar 20 16:56 146412 failure #3711 Failed 10476 total leaks found for a total of 581,968 bytes! 163 unique leaks found! 11 new passes 28 flakes Mar 20 16:08 146405 failure #3710 Failed 10708 total leaks found for a total of 525,040 bytes! 166 unique leaks found! 14 new passes 32 flakes Mar 20 15:20 146394 failure #3709 Failed 11591 total leaks found for a total of 573,616 bytes! 194 unique leaks found! 14 new passes 29 flakes Mar 20 13:53 146383 failure #3708 Failed 10475 total leaks found for a total of 566,480 bytes! 185 unique leaks found! 14 new passes 32 flakes http://build.webkit.org/builders/Apple%20Lion%20%28Leaks%29 Mar 20 19:52 146430 failure #4230 Failed exiting early after 20 crashes and 0 timeouts. 3626 tests run. 20 failures 3 new passes 6 unit tests failed or timed out Mar 20 19:18 146426 failure #4229 Failed exiting early after 20 crashes and 0 timeouts. 3637 tests run. 20 failures 3 new passes 6 unit tests failed or timed out Mar 20 17:11 146416 failure #4228 Failed exiting early after 1 crashes and 19 timeouts. 21605 tests run. 5005 total leaks found for a total of 542,400 bytes! 196 unique leaks found! 41 failures 12 new passes
Attachments
Add attachment proposed patch, testcase, etc.
Ryosuke Niwa
Comment 1 2013-03-20 22:05:36 PDT
Hm… it seems leaks are coming from __CFURLCache :( I'm not sure what to do. http://build.webkit.org/LeaksViewer/?url=%2Fresults%2FApple%20MountainLion%20%28Leaks%29%2Fr146431%20%283715%29%2F
Stephanie Lewis
Comment 2 2013-03-20 22:10:51 PDT
we usually try to find once specific test case causes the leak and then we file a bug on CFNetwork.
Stephanie Lewis
Comment 3 2013-03-20 22:12:02 PDT
Once a bug is filed we can add a regexp for the leak to the excluded leaks file so it doesn't interfere with the current results.
Ryosuke Niwa
Comment 4 2013-03-20 23:33:37 PDT
Blame list: http://trac.webkit.org/log/?verbose=on&rev=146431&stop_rev=146412 In this range, only the following changesets contain relevant code changes other than http://trac.webkit.org/changeset/146419, which got later rolled out to fix builds: http://trac.webkit.org/changeset/146417 http://trac.webkit.org/changeset/146427 http://trac.webkit.org/changeset/146430 And it appears that 146417 and 146427 are trivial changes too. That makes me think that http://trac.webkit.org/changeset/146430 might be the culprit.
Kentaro Hara
Comment 5 2013-03-21 00:42:09 PDT
(In reply to comment #4) > And it appears that 146417 and 146427 are trivial changes too. That makes me think that http://trac.webkit.org/changeset/146430 might be the culprit. hmm, I don't think r146430 is the culprit. r146430 just changed a rule of integer conversion. michael: thoughts?
Michael Pruett
Comment 6 2013-03-21 06:01:10 PDT
(In reply to comment #5) > hmm, I don't think r146430 is the culprit. r146430 just changed a rule of integer conversion. > > michael: thoughts? I would be rather surprised if r146430 were the source of any leaks since it did only change IDL integer conversion. Aside from the implementation of EnforceRange, currently used only in IndexedDB and internal testing, the only other change of note in this revision concerns the conversion of ECMAScript numbers to IDL long long and unsigned long long types. Specifically +/-infinity are now converted to 0, and values greater in magnitude than 2^64 are converted to integers modulo 2^64. Previously such out-of-range values were mapped to the nearest representable integer.
Ryosuke Niwa
Comment 7 2013-03-21 16:51:40 PDT
It seems like this was a temporary fluke :(
Note You need to log in before you can comment on or make changes to this bug.