RESOLVED FIXED Bug 59965
[chromium] Resizing a 2d canvas to huge dimensions after compositing crashes with accelerated 2d canvas option enabled 
https://bugs.webkit.org/show_bug.cgi?id=59965
Summary [chromium] Resizing a 2d canvas to huge dimensions after compositing crashes ...
James Robinson
Reported 2011-05-02 15:13:13 PDT
[chromium] Resizing a 2d canvas to huge dimensions after compositing crashes with accelerated 2d canvas option enabled
Attachments
Patch (6.53 KB, patch)
2011-05-02 15:18 PDT, James Robinson 		no flags
DetailsFormatted DiffDiff
View All Add attachment proposed patch, testcase, etc.
James Robinson
Comment 1 2011-05-02 15:18:14 PDT
Created attachment 91991 [details] Patch
James Robinson
Comment 2 2011-05-02 15:19:28 PDT
This crash is showing up with moderate frequency in recent Chrome releases. It only shows up with the --enable-accelerated-2d-canvas flag on but can be triggered by doing something as simple as resizing the window with a fullscreen canvas demo opened.
Kenneth Russell
Comment 3 2011-05-02 19:37:54 PDT
Comment on attachment 91991 [details] Patch View in context: https://bugs.webkit.org/attachment.cgi?id=91991&action=review > Source/WebCore/html/canvas/CanvasRenderingContext2D.cpp:207 > +#endif senorblanco should review this. He just added a short-circuit to DrawingBuffer::reset() to do no work if the size hasn't changed, and I have a feeling that making this contentChanged call in all cases may impact performance.
Stephen White
Comment 4 2011-05-03 12:56:53 PDT
Comment on attachment 91991 [details] Patch View in context: https://bugs.webkit.org/attachment.cgi?id=91991&action=review Looks good (unofficially). >> Source/WebCore/html/canvas/CanvasRenderingContext2D.cpp:207 >> +#endif > > senorblanco should review this. He just added a short-circuit to DrawingBuffer::reset() to do no work if the size hasn't changed, and I have a feeling that making this contentChanged call in all cases may impact performance. I think this is correct. Valid size changes are unaffected; this only changes the case when we're going from valid -> invalid size, in which case we won't be rendering Canvas2D on the GPU anyway and we should definitely tell the compositor.
Justin Novosad
Comment 5 2011-05-03 13:20:35 PDT
*** Bug 60056 has been marked as a duplicate of this bug. ***
Kenneth Russell
Comment 6 2011-05-03 14:03:06 PDT
Comment on attachment 91991 [details] Patch OK, sounds good.
James Robinson
Comment 7 2011-05-03 14:21:39 PDT
Comment on attachment 91991 [details] Patch Clearing flags on attachment: 91991 Committed r85661: <http://trac.webkit.org/changeset/85661>
James Robinson
Comment 8 2011-05-03 14:21:42 PDT
All reviewed patches have been landed. Closing bug.
Note You need to log in before you can comment on or make changes to this bug.