Bug 101911 – [Cairo] fillRectWithColor with Color::transparent doesn't perform anything

• Status: RESOLVED
• Resolution: FIXED
• Component: Layout and Rendering

[Hurnjoo Lee]

fillRectWithColor with Color::transparent doesn't perform anything
because fillRectWithColor does early-return if the alpha value of
color is zero. But we expect that fill the rect with transparent color                                                                                                                                                                
in case the cairo_operator is CAIRO_OPERATOR_SOURCE.

[Hurnjoo Lee]

Created attachment 173591 [details]
proposed patch

[Dominik Röttsches (drott)]

Comment on attachment 173591 [details]
proposed patch

View in context: https://bugs.webkit.org/attachment.cgi?id=173591&action=review

> Source/WebCore/ChangeLog:13
> +        Covered by existing tests.

Do you have an example? Which test is passing with this change, or in other words: What was broken before?

[Hurnjoo Lee]

(In reply to comment #2)
> (From update of attachment 173591 [details])
> View in context: https://bugs.webkit.org/attachment.cgi?id=173591&action=review
> 
> > Source/WebCore/ChangeLog:13
> > +        Covered by existing tests.
> 
> Do you have an example? Which test is passing with this change, or in other words: What was broken before?

I wrote it in meaning that do not affect to existing test. If there is any problem in this expression, I will revise it.
Actually, the problem is occurred in case using TextureMapper on WebKit2 EFL port.
In UpdateAtlas::beginPaintingOnAvailableBuffer, in case of using GraphicsContextCairo (EFL or GTK), doesn't fill the reused area with transparent color.

[Dominik Röttsches (drott)]

(In reply to comment #3)

> I wrote it in meaning that do not affect to existing test. If there is any problem in this expression, I will revise it.
> Actually, the problem is occurred in case using TextureMapper on WebKit2 EFL port.
> In UpdateAtlas::beginPaintingOnAvailableBuffer, in case of using GraphicsContextCairo (EFL or GTK), doesn't fill the reused area with transparent color.

Thanks, I understand. If there is such an actual problem - as you describe -please provide a layout test exactly reproducing this problem or provide convincing reasoning why it's not possible to create such a regression test.

[Kenneth Rohde Christiansen]

Comment on attachment 173591 [details]
proposed patch

This fixes 103081!

[Dominik Röttsches (drott)]

Please consider creating a regression test for this.

[Kenneth Rohde Christiansen]

Btw I would love if someone could performance test fillRect (transparent) vs clearRect!

[WebKit Review Bot]

Comment on attachment 173591 [details]
proposed patch

Clearing flags on attachment: 173591

Committed r135737: <http://trac.webkit.org/changeset/135737>

[WebKit Review Bot]

All reviewed patches have been landed.  Closing bug.

[Yael]

(In reply to comment #5)
> (From update of attachment 173591 [details])
> This fixes 103081!

Also https://bugs.webkit.org/show_bug.cgi?id=102989