This test on the IE9 test center reveals a bug in WebKit: http://samples.msdn.microsoft.com/ietestcenter/html5/canvas/canvas-text-font-002.htm The problem is in CSSPrimitiveValue::computeLengthDouble() which assumes the passed-in style has a usable font().
Not a regression from Safari 5.0.2, but crashing on an ietestcenter test is not good.
I think this can be fixed by calling newStyle->font().update() after setting the font description on the newStyle.
Created attachment 68848 [details] Proposed patch Patch using mitz's suggestion.
Comment on attachment 68848 [details] Proposed patch Clearing flags on attachment: 68848 Committed r68343: <http://trac.webkit.org/changeset/68343>
All reviewed patches have been landed. Closing bug.
http://trac.webkit.org/changeset/68343 might have broken GTK Linux 32-bit Release
Reopening since this broke GTK+.
CC'ing GTK+ people for a look..
Created attachment 68898 [details] Proposed patch v2 Essentially the same patch, except do the update() even if !canvas()->computedStyle() (I guess if the element isn't attached to the document.) Tested on Qt and GTK+.
Comment on attachment 68898 [details] Proposed patch v2 Rejecting patch 68898 from commit-queue. Failed to run "['./WebKitTools/Scripts/webkit-patch', '--status-host=queues.webkit.org', 'apply-attachment', '--force-clean', '--non-interactive', '--quiet', 68898]" exit_code: 2 Cleaning working directory Updating working directory Failed to run "['WebKitTools/Scripts/update-webkit']" exit_code: 2 Full output: http://queues.webkit.org/results/4033166
Comment on attachment 68898 [details] Proposed patch v2 False rejection. Bug in the queue. One of the cluster nodes seems stuck.
Comment on attachment 68898 [details] Proposed patch v2 Rejecting patch 68898 from commit-queue. Failed to run "['./WebKitTools/Scripts/webkit-patch', '--status-host=queues.webkit.org', 'apply-attachment', '--force-clean', '--non-interactive', '--quiet', 68898]" exit_code: 2 Cleaning working directory Updating working directory Failed to run "['WebKitTools/Scripts/update-webkit']" exit_code: 2 Full output: http://queues.webkit.org/results/4074125
Comment on attachment 68898 [details] Proposed patch v2 One of the cq nodes is misbehaving. It should right itself shortly. I can't access it to manually reboot it at this time.
Comment on attachment 68898 [details] Proposed patch v2 Rejecting patch 68898 from commit-queue. Failed to run "['./WebKitTools/Scripts/webkit-patch', '--status-host=queues.webkit.org', 'apply-attachment', '--force-clean', '--non-interactive', '--quiet', 68898]" exit_code: 2 Cleaning working directory Updating working directory Failed to run "['WebKitTools/Scripts/update-webkit']" exit_code: 2 Full output: http://queues.webkit.org/results/4066158
Sorry. Will wait until I know the CQ node is healthy again.
Comment on attachment 68898 [details] Proposed patch v2 I repaired one of them just now. Git got very upset by some test rebaselinings this morning. It was failing to update.
Comment on attachment 68898 [details] Proposed patch v2 Please leave the string "Reviewed by NOBODY (OOPS!)." in your ChangeLog if you want this to be handled by the commit-queue.
Committed r68513: <http://trac.webkit.org/changeset/68513>
Comment on attachment 68898 [details] Proposed patch v2 > Please leave the string "Reviewed by NOBODY (OOPS!)." in your ChangeLog if you want this to be handled by the commit-queue. D'oh! Thanks Eric.