Bug 9385 – LEAK: Ran `run-webkit-tests --leaks --singly` on svn r14761

• Component: New Bugs
• Status: RESOLVED FIXED

[Henk Poley]

Found a total of 5460 leaks. 5 crashers. 24MB of output, I will attach those or upload an archive to my webserver.

I think I'd like some assistance in sifting through the `leaks` output. And I would also like comments if I'm doing something unneedingly stupid in running the leaks-test with '--singly'.

[Henk Poley]

Created attachment 8794 [details]
Output of `run-webkit-tests --leaks --singly` on SVN revision 14761

[Henk Poley]

7z can be decompressed with 7zX on OS X:
http://sixtyfive.xmghosting.com/

And with 7-zip on windows:
http://www.7-zip.org/

(The zip file would have been 4.5MB, so I thought I'd give this a try)

[Mark Rowe (bdash)]

Created attachment 8816 [details]
bzip2 version of the 7zip file

I've just extracted the 7z file and repacked it as a bz2 file as this is supported out of the box on Mac OS X.

[Gustaaf Groenendaal (MysteryQuest)]

Confirmed. The Nightly has too many leaks at this moment. Made this a P1 priority, since this is getting really bad. I don't know what the cause is, but even with only opening http://nightly.webkit.org/start/ gets me memory leaks in some builds or in some occasions.

[David Kilzer (:ddkilzer)]

(In reply to comment #4)
> Confirmed. The Nightly has too many leaks at this moment. Made this a P1
> priority, since this is getting really bad. I don't know what the cause is, but
> even with only opening http://nightly.webkit.org/start/ gets me memory leaks in
> some builds or in some occasions.

Please try a more recent nightly.  Most of the leaks have been eliminated recently.

[Mark Rowe (bdash)]

The majority of these leaks appear to be identical to the following:

Leak: 0x26a067e0  size=16	string 'LRFCLRFC'
	Call stack: [thread ffffffff]: | 0x2 | start | _start | main | dumpRenderTree | -[NSRunLoop runMode:beforeDate:] | CFRunLoopRunInMode | CFRunLoopRunSpecific | malloc_zone_malloc 

This appears to be outside of WebKit.  Perhaps it's a candidate for the leak exclude list?

As these results are so out of date and contain so many instances of the false-positive mentioned above, I am dropping the priority of this back to P3.  Up to date results with the leak above excluded would be very useful.

[Henk Poley]

I also suspected those to be false positives, but then they could have been canaries (stack boundary markers). I'm not that intimate with Webkit.

I'll see if I can run the tests again with a newer checkout. btw, back then they also told me that "Most of the leaks have been eliminated recently.". So I was somewhat suprised to find that many leaks..

[Eric Seidel (no email)]

This bug is no longer useful being so old.