Bug 142258

Summary: fast/css/object-fit/object-fit-canvas.html has tiny test failure
Product: WebKit Reporter: Simon Fraser (smfr) <simon.fraser>
Component: CSSAssignee: Nobody <webkit-unassigned>
Status: NEW ---    
Severity: Normal CC: ap, dino, simon.fraser, thorton, zalan
Priority: P2    
Version: 528+ (Nightly build)   
Hardware: Unspecified   
OS: Unspecified   
See Also: https://bugs.webkit.org/show_bug.cgi?id=142630
Bug Depends on: 149828    
Bug Blocks:    
Attachments:
Description Flags
a diff from my machine none

Description Simon Fraser (smfr) 2015-03-03 21:27:26 PST 
fast/css/object-fit/object-fit-canvas.html keeps failing on EWS with a 0.01% image diff.
Comment 1 Simon Fraser (smfr) 2015-03-03 22:05:18 PST 
Marks as flakey image failure in https://trac.webkit.org/changeset/180991
Comment 2 Alexey Proskuryakov 2015-03-04 00:38:39 PST 
This is a recent regression, but I don't know how to identify the culprit.
Comment 3 Alexey Proskuryakov 2015-03-04 09:48:46 PST 
The first failure was on 2/28 (bug 140855 comment 8).
Comment 4 Alexey Proskuryakov 2015-03-04 09:50:11 PST 
Maybe <http://trac.webkit.org/changeset/180815>?
Comment 5 Alexey Proskuryakov 2015-03-04 09:52:32 PST 
No, it was flaky before that, just didn't happen to fail in a pattern that would make EWS post a comment.
Comment 6 Alexey Proskuryakov 2015-03-04 10:12:07 PST 
The first failure was when processing bug 142078.

Maybe <http://trac.webkit.org/changeset/180790>?
Comment 7 Alexey Proskuryakov 2015-03-04 16:58:09 PST 
Created attachment 247910 [details]
a diff from my machine

I can reproduce with a lot larger diff on a hidpi machine. That started somewhere between 180791 and 180801, so it's very likely the same issue.
Comment 8 Alexey Proskuryakov 2015-03-04 17:14:07 PST 
Wait, it's 1600x1200?!

There is positively nothing in this range that could cause this; maybe I got it wrong due to flakiness.
Comment 9 Alexey Proskuryakov 2015-03-04 20:04:38 PST 
My searching skills failed me. The test started to flakily fail when an expectation that it flakily fails was removed in <http://trac.webkit.org/changeset/180796>!

It's still an issue that we should investigate soon though - WebKitTestRunner getting confused about the window size is very much not OK.
Comment 10 Alexey Proskuryakov 2015-03-12 11:22:53 PDT 
Looks like there are two issues here.

1. The test legitimately fails with small pixel differences on some machines. Looks like an ATI&Nvidia vs Intel graphics at a first glance.

2. When I run the test on my Mavericks machine, I get 1600x1200 results, and one of these looks like it's upscaled from 800x600.