[chromium] Remove viewport clear from release builds
Created attachment 102067 [details] Patch
Comment on attachment 102067 [details] Patch Hmm. Debug builds are so slow that it's unlikely we will do extensive testing with them. But I'll leave this decision up to those working on the compositor.
We do run the compositing tests in debug, at least.
Comment on attachment 102067 [details] Patch Rejecting attachment 102067 [details] from commit-queue. Failed to run "['./Tools/Scripts/webkit-patch', '--status-host=queues.webkit.org', '--bot-id=ec2-cq-01', '--port..." exit_code: 1 Last 500 characters of output: 0e94b75603c487eee0ba355e40fcb6b01c81ffa8 r91810 = 851bbe5a9502994f015ea8746c3ec847fc50b2b4 Done rebuilding .git/svn/refs/remotes/origin/master/.rev_map.268f45cc-cd09-0410-ab3c-d52691b4dbfc First, rewinding head to replay your work on top of it... Fast-forwarded master to refs/remotes/origin/master. Updating chromium port dependencies using gclient... ________ running '/usr/bin/python gyp_webkit' in '/mnt/git/webkit-commit-queue/Source/WebKit/chromium' Updating webkit projects from gyp files... Full output: http://queues.webkit.org/results/9245650
Gah, I misformatted the ChangeLog again. D'oh!
Committed r91812: <http://trac.webkit.org/changeset/91812>
Reverted r91812 for reason: Breaks Committed r91821: <http://trac.webkit.org/changeset/91821>
So as it turns out, this clear was the only thing ensuring that the root layer's alpha channel was fully opaque. We mask out alpha writes when drawing the root layer so on the mac bot the alpha channel was in an undefined state. I think a better approach is to clear the default render surface whenever it is created or resized, instead of every frame. I'll look into this if I get a chance. We are paying a lot of bandwidth for this clear currently.