GNOME Bugzilla – Bug 85003
Purify errors (FMR, UMR)
Last modified: 2004-12-22 21:47:04 UTC
Significant memory leak when built/tested using Rational purify tool. The tag used to build this component is June 1st. We did a Free hand testing and executed 80% of test assertions.
Created attachment 9163 [details] Attaching the output generated by Rational purify tool.
I've fixed a few of the UMRs and FMRs but could use more context. 1) Exactly what versions of things are you using ? - my gtk source does not match your line numbers 2) Can you rerun and include slightly more stack for things ? The leaks espcially are impossible to track without seeing who is calling things.
jody, i suspect they're using gtk HEAD
Ok, the FMR are fixed in CVS. The leak is still present until we get more context.
Jody: so is it fair to mark this down?
Yes. The remaining memory issues are minor. It would be nice to track the leak, but not earth shattering. As soon as we get another log with slightly more context it will be fixed.
Attached the purify output which gives more information about memory leaks. HTH
Created attachment 9379 [details] detailed info about memory leaks
Ahhhh.... not quite what I meant. The original format was fine. We need more stack frames, not a more verbose format.
purify wont provide any stack frames. It provided only the info where a memory leak could have happened by specifying the line number. Tested this component on july 14th tag source code and details are: Memory Leaked : 14814 Bytes Potentially Leaked : 1621 Bytes I will be attaching a latest purify log file.
Created attachment 10150 [details] [review] purify log file.
Sorry, I should be clearer. By 'stack frame' I meant the calling function and line number. The current traces have 6 layers of call stack. I need more to find some of the remaining leaks. Please regenerate the trace with traces that go at least 10 layers. Thanks
Ramesh, can you supply this info
Attaching the log file with more stack frames. I have split the log file. The tag used to build is july 24.
Created attachment 10569 [details] purify log file1 with more stack info
Created attachment 10572 [details] purify log file1 with more stack info
Created attachment 10590 [details] purify log file 2
jody: I have uploaded two log-file, will be uploading logfiles for gnome-file-types-properties and gnome-default-application-properties. Uploading files to bugzilla fails sometimes but the attachment would be successful, so i need to wait until i get mail from bugzilla and then attach the next logfile.
Created attachment 10595 [details] gnome-default-applications-properties logfile
jody: Is this data is sufficient? Or need fresh stack?
SPAM as discussed last night. Search for 'SPAM as discussed last night' to catch these all and delete them. :)
With the help of Rational purify tool I have identified the points of mem leak and have fixed them. Am attaching the patch for it and also the brief up of purify output.
Created attachment 13800 [details] Purify Log for gnome-control-center
Created attachment 13801 [details] [review] Patch as stated above
Jody, pls. let me know if this patch is okay to commit. Thanks.
These look safe enough, but I'd rather they wait for 2.2.1 Please commit once gnome-2.2.0 has been released.
Adding some more patches for leaks I've found today.
Created attachment 14219 [details] [review] capplets common patch
Created attachment 14220 [details] [review] theme-switcher patch
Created attachment 14221 [details] [review] libwindow-settings patch
Created attachment 14222 [details] [review] libsounds patch
Here comes one more patch which fixes some gnome-vfs related leaks
Created attachment 14224 [details] [review] patch for capplets/common
Looks like bits of rajkumar's patch was included in my patch that was commited. I'll grab and test the leftovers and make sure to attribute all the changes to the right person in a new patch :-)
Christophe's stuff was also included in the patch I sent to jrb and got included.
Created attachment 14326 [details] [review] leftover patch not commited yet - sent to jrb
Jrb, I think the last of these patches is the only one that hasn't been commited.
Commited to both branches. Close to leak free now. And just one invalid read left in LargePrint themes.