Bug 107833 – REGRESSION (r135628-135632): Double box shadow failure to render

• Component: CSS
• Status: RESOLVED FIXED

[Nathan Crank]

I have found a strange bug in the nightlies of webkit and canary.

If you check out http://codepen.io/nathancrank/pen/JCFyk in release Safari or Chrome, you'll see the second box shadow render on the submit input. If you load it in nightlies it will only render the first shadow, unless you add a third shadow declaration to it.

[Alexey Proskuryakov]

Two potentially related revisions in the range:

<http://trac.webkit.org/changeset/135629> Adding occlusion detection to reduce overdraw in RenderBox background rendering

<http://trac.webkit.org/changeset/135632> [CSS3 Backgrounds and Borders] Implement new CSS3 background-position parsing.

[Alexey Proskuryakov]

<rdar://problem/13087228>

[Justin Novosad]

(In reply to comment #1)
Just verified that the regression was in fact caused by http://trac.webkit.org/changeset/135629

[Justin Novosad]

Created attachment 185066 [details]
Patch

[Simon Fraser (smfr)]

Comment on attachment 185066 [details]
Patch

View in context: https://bugs.webkit.org/attachment.cgi?id=185066&action=review

> Source/WebCore/ChangeLog:8
> +        Reviewed by NOBODY (OOPS!).
> +
> +        Test: fast/backgrounds/gradient-background-shadow.html

Please explain the bug and how you fixed it, identifying the revision that caused it.

> Source/WebCore/ChangeLog:12
> +        Changeing occlusion culling test to never cull background color

Changing

[WebKit Review Bot]

Comment on attachment 185066 [details]
Patch

Attachment 185066 [details] did not pass chromium-ews (chromium-xvfb):
Output: http://queues.webkit.org/results/16201044

New failing tests:
fast/backgrounds/gradient-background-shadow.html

[Justin Novosad]

Created attachment 185243 [details]
Patch for landing

[Justin Novosad]

Created attachment 185245 [details]
Patch for landing

[WebKit Review Bot]

Comment on attachment 185245 [details]
Patch for landing

Rejecting attachment 185245 [details] from commit-queue.

New failing tests:
fast/backgrounds/gradient-background-shadow.html
Full output: http://queues.webkit.org/results/16195168

[Justin Novosad]

(In reply to comment #9)
> (From update of attachment 185245 [details])
> Rejecting attachment 185245 [details] from commit-queue.
> 
> New failing tests:
> fast/backgrounds/gradient-background-shadow.html
> Full output: http://queues.webkit.org/results/16195168

The failure in the new test looks like it is caused by a regression that must have landed since yesterday.  Looks like there are LSB discrepancies in gradient rendering.  This is probably limited to the skia port.

fmalita, you know anything about this?

[Justin Novosad]

Geez, looks like I was using a stale checkout.  Related regressions were added to test expectations 3 weeks ago with: http://trac.webkit.org/changeset/139512
We should fix skia gradients so that we can re-enable some ref tests.

[Justin Novosad]

The nature of the problem with skia gradients is understood and documented here: https://code.google.com/p/chromium/issues/detail?id=169602

Until that is fixed, I'll just put the new test in chromium TestExpectations. It passes on other WK ports.

[Justin Novosad]

Created attachment 185282 [details]
Patch for landing

[WebKit Review Bot]

Comment on attachment 185282 [details]
Patch for landing

Clearing flags on attachment: 185282

Committed r141160: <http://trac.webkit.org/changeset/141160>

[WebKit Review Bot]

All reviewed patches have been landed.  Closing bug.