| Summary: | Memory leak caused by 100x100 glyphs | ||||||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Product: | xorg | Reporter: | Clemens Eisserer <linuxhippy> | ||||||||||||
| Component: | Server/General | Assignee: | Xorg Project Team <xorg-team> | ||||||||||||
| Status: | RESOLVED FIXED | QA Contact: | Xorg Project Team <xorg-team> | ||||||||||||
| Severity: | normal | ||||||||||||||
| Priority: | medium | CC: | peter.hutterer | ||||||||||||
| Version: | unspecified | Keywords: | patch | ||||||||||||
| Hardware: | Other | ||||||||||||||
| OS: | All | ||||||||||||||
| Whiteboard: | |||||||||||||||
| i915 platform: | i915 features: | ||||||||||||||
| Bug Depends on: | |||||||||||||||
| Bug Blocks: | 23613 | ||||||||||||||
| Attachments: |
|
||||||||||||||
Created attachment 28584 [details]
testcase
Created attachment 28621 [details]
second version of testcase
Passed wrong destination-picture, fixed now.
Tested, memleak does not occur with Fedora8+intel-2.1.1.
seems to happen with the vesa driver too, most likely not intel specific. However would be great anyway if you guys could fix it ;) (In reply to comment #3) > seems to happen with the vesa driver too, most likely not intel specific. > However would be great anyway if you guys could fix it ;) Thanks for the detailed report, and for identifying it as a non-driver-specific issue. -Carl Seems to be a regression per comment #2, so should probably be considered for 1.7. Created attachment 29101 [details]
valgrind output
How do you determine the p(In reply to comment #7) > Could this be somehow related to bug 23645? Does it only happen if uploading several glyphs at once? Yes. When I upload one-by-one, using the same code-paths everything seems to work fine. Created attachment 29785 [details] [review] 0001-render-Plug-a-memory-leak-in-AddGlyph.-23286.patch The FreePicture call was missing from AddGlyph, this caused the leak. Just adding the line to make it identical to the DeleteGlyph part doesn't leak memory anymore. This patch moves that part out of both into a static function to share between the two but the net effect is just adding the FreePicture call to AddGlyph. I wonder if AllocateGlyph (on the bail code path) is affected by this too. Looking at the code, I'm not 100% sure. Ping? can I have a tested-by or ACK for the above patch please? Pushed to master as f772014c435f56db56520ca13ffa39431684f122, no complaints from xorg-devel about the patch. Extra verification would be appreciated. commit f772014c435f56db56520ca13ffa39431684f122 Author: Peter Hutterer <peter.hutterer@who-t.net> Date: Wed Sep 23 10:53:51 2009 +1000 render: Plug a memory leak in AddGlyph. (#23286) Closing as fixed, thanks for reporting. |
Use of freedesktop.org services, including Bugzilla, is subject to our Code of Conduct. How we collect and use information is described in our Privacy Policy.
Created attachment 28583 [details] xorg log The following sequence of Xrender functions triggers a memory leak when called with 100x100 glyphs: XRenderAddGlyphs XRenderCompositeText32 XRenderFreeGlyphs I only saw the leak with 100x100 glyph images, however not with 96x06 or 104x104. I attached a short testcase, triggering the problem. I am running Fedora rawhide with a xorg-1.7 prerelease, 2.6.31.rc5, and intel-2.8.0.