Test expectations for gpu tests do not match if acceleration is switched off for small canvas. We should disable switching acceleration when testing so that we can have a consistent set of test expectations. This does not alter the test coverage because we already run canvas test for both hardware and software. The tests got broken in: http://trac.webkit.org/changeset/91599
Created attachment 101785 [details] proposed patch
Comment on attachment 101785 [details] proposed patch View in context: https://bugs.webkit.org/attachment.cgi?id=101785&action=review At first glance, I think this is reasonable. I'm curious what others think. > Source/WebCore/page/Settings.h:364 > + // Hardware acceleration is useful for large canvii where it can avoid the nit: i think it's "canvases"
Created attachment 101786 [details] proposed patch The last patch had missing files. Verified that all canvas-related tests pass with this patch.
ping! GPU bots are red - this needs to be committed ASAP. BTW there are also some unrelated regressions in GPU tests related to rotated layers and scrollbar rendering.
I'm not very familiar with how these settings are managed but it looks reasonable.
Comment on attachment 101786 [details] proposed patch Looks good. r=me
Comment on attachment 101786 [details] proposed patch Clearing flags on attachment: 101786 Committed r91638: <http://trac.webkit.org/changeset/91638>
All reviewed patches have been landed. Closing bug.