Don't unconditionally repaint compositing layers when their size changes
Created attachment 178389 [details] Patch
<rdar://problem/12842382>
Comment on attachment 178389 [details] Patch Eek!
http://trac.webkit.org/changeset/137215
The test added in this patch seems to be failing on EFL port: --- /home/buildslave-1/webkit-buildslave/efl-linux-64-debug-wk2/build/layout-test-results/compositing/repaint/resize-repaint-expected.txt +++ /home/buildslave-1/webkit-buildslave/efl-linux-64-debug-wk2/build/layout-test-results/compositing/repaint/resize-repaint-actual.txt @@ -12,9 +12,6 @@ (position 8.00 8.00) (bounds 402.00 207.00) (drawsContent 1) - (repaint rects - (rect 0.00 1.00 402.00 1.00) - ) ) ) )
I didn't think EFL even did compositing. Has that changed? It looks like this port doesn't have the code to track repaint rects on GraphicsLayers. The test should be marked as failing in TestExpectations.
(In reply to comment #6) > I didn't think EFL even did compositing. Has that changed? It looks like this port doesn't have the code to track repaint rects on GraphicsLayers. The test should be marked as failing in TestExpectations. Yes, we have enabled accelerated compositing in WK2 EFL about 1 month ago (using coordinated graphics). Thank you for the indication about tracking repaint rects. I have already marked the test as failing in TestExpectations until we implement the missing functionality.
The fix for bug 104626 undid most of this optimization.
Created attachment 178971 [details] Patch
http://trac.webkit.org/changeset/137526