REGRESSION (r105401-105403): Blue flash on css border transition

• Component: Layout and Rendering
• Status: RESOLVED FIXED

[Vangelis Kokkevis]

To reproduce:
1. Visit http://www.webkit.org/blog/138/css-animation/
2. Hover over the div that reads: "This div will acquire a slightly thicker blue border when hovered.". Mouse out.

Notice how the entire page flashes blue .  

This is happening both with the latest WebKit nightly (Version 5.1.2 (7534.52.7, r106392)) and Chrome Canary . 

Reported also in the chromium issue tracker as:
http://code.google.com/p/chromium/issues/detail?id=111304

[Vangelis Kokkevis]

Bisected to the following WebKit revision range:
http://trac.webkit.org/log/trunk/?rev=105427&stop_rev=105390&verbose=on

From a quick scan, here's a couple of possible suspects:
* [CSSRegions]Add support for background-color in region styling https://bugs.webkit.org/show_bug.cgi?id=71488
http://trac.webkit.org/changeset/105426/trunk [[e-mail]]

* Strange Result for getComputedStyle on borderWidth set in em
 https://bugs.webkit.org/show_bug.cgi?id=18294
http://trac.webkit.org/changeset/105403/trunk [[e-mail]]

* Differentiate between SVG/CSS width/height attributes/properties
 https://bugs.webkit.org/show_bug.cgi?id=76447 
 http://trac.webkit.org/changeset/105402/trunk [[e-mail]]

* Cache RenderStyle pointer as a method to avoid performance regression for region styling https://bugs.webkit.org/show_bug.cgi?id=76265
http://trac.webkit.org/changeset/105394/trunk [[e-mail]]

[Vangelis Kokkevis]

Correcting the title as the issue is really happening on css border width animation transition.  The page is not triggering the compositor at all.

Maybe that makes: http://trac.webkit.org/changeset/105403/trunk a more likely suspect.

[Mihnea Ovidenie]

Hi,

The first change mentioned:

* [CSSRegions]Add support for background-color in region styling https://bugs.webkit.org/show_bug.cgi?id=71488
http://trac.webkit.org/changeset/105426/trunk [[e-mail]]

contains changes that are no longer in the trunk.

[Alexey Proskuryakov]

Further reduced the range to r105401-105403.

[Alexey Proskuryakov]

<rdar://problem/10789804>

[Simon Fraser (smfr)]

Almost certainly http://trac.webkit.org/changeset/105402

[Antti Koivisto]

A reduced test case would be cool.

[Alexis Menard (darktears)]

(In reply to comment #7)
> A reduced test case would be cool.

https://gist.github.com/1718650

seems to reproduce an ugly blue rectangle on the Qt port.

[Alexis Menard (darktears)]

(In reply to comment #8)
> (In reply to comment #7)
> > A reduced test case would be cool.
> 
> https://gist.github.com/1718650
> 
> seems to reproduce an ugly blue rectangle on the Qt port.

Clicking on item2 then item1. Ok the test case is a bit weird but it was for another bug.

[Alexis Menard (darktears)]

(In reply to comment #9)
> (In reply to comment #8)
> > (In reply to comment #7)
> > > A reduced test case would be cool.
> > 
> > https://gist.github.com/1718650
> > 
> > seems to reproduce an ugly blue rectangle on the Qt port.
> 
> Clicking on item2 then item1. Ok the test case is a bit weird but it was for another bug.

http://trac.webkit.org/changeset/105403/trunk is the culprit. Reverting it fixes the issue. I'm looking at it right now as I did the original patch.

[Alexis Menard (darktears)]

Created attachment 125314 [details]
Patch

[WebKit Review Bot]

Comment on attachment 125314 [details]
Patch

Clearing flags on attachment: 125314

Committed r106672: <http://trac.webkit.org/changeset/106672>

[WebKit Review Bot]

All reviewed patches have been landed.  Closing bug.