The following layout test is flaky on Mac10.10
fast/repaint/resize-iframe-text.html
Probable cause:
When it fails, it looks like the resize hasn't finished. This isn't the same as the problems yesterday in issue 635892 .
Could this indicate that the code really isn't working on Mac? Not sure, setting P2 as a compromise.
FYI I'm reverting a CL of mine that affected layout timing in a subtle way: https://codereview.chromium.org/2229303002/
It seems to turn Linux version of the test to flake or always fail. I'm not sure what is the right expectation here so marked it as flake for Linux too.
Thanks, I tried applying your revert patch and fast/repaint/resize-iframe-text.html still crashes on Mac. However possible your patch could explain some of the other flakes I am looking into in http://crbug.com/637245 .
Just reproduced the failure on Linux (remote desktop in Chrome). It seems that window.resize() is restricted by the screen size, so sometimes we don't get the desired window size.
I also encountered this before for the paint/invalidation/window-resize-*.html tests, and fixed that by using smaller window sizes. Will try the same for this test.
The #14 crash should be unrelated. It seems a occasional random issue that might happen in every layout test.
Tried to fix the flake by reducing the window size. However, I couldn't make the test fail with the original code change reverted.
Assigning back to wkorman@ to determine if we could modify the test to make it work and stable, or just remove the test.
This test just became interesting after we landed https://codereview.chromium.org/2702883002/. I encountered crash when I added a visual rect update check in PaintInvalidator.cpp and I need to fix it in my CL.
Comment 1 by foolip@chromium.org
, Aug 10 2016