[Webkit-unassigned] [Bug 54921] [Chromium] Issue 60965: Layout test fast/canvas/setWidthResetAfterForcedRender.html is failing on GPU path

[bugzilla-daemon at webkit.org]

https://bugs.webkit.org/show_bug.cgi?id=54921





--- Comment #9 from James Robinson <jamesr at chromium.org>  2011-02-22 19:22:16 PST ---
(In reply to comment #7)
> James,
> 
> Thank you for your review.
> 
> As you know, requiresLayer() function is called bye
> RenderBoxModelObject::styleDidChange()
> RenderObject::updateFillImages()
> 
> If the returned values are true, the rendering tree becomes different from the non-accelerated rendering trees. The canvas layer order is different from non-accelerated.

Yes - this is expected.  An accelerated canvas requires a layer whereas a non-accelerated canvas may or may not need its own layer.  Different does not mean wrong.

> 
> For example, in setWidthResetAfterForce, the canvas element becomes the same layer of the root. That causes the test failure. That is why I change this part.

I'm not sure what you mean by 'becomes the same layer of the root' - could you expand?

> 
> I want to make sure the layer purpose. I thought this looked like the very special layer to overlay the rendered images in partial. Is this correct? Or is there another purpose for the layer?

The RenderLayer is required because accelerated canvas require their own composited layer.  See RenderLayerCompositor::requiresCompositingForCanvas() to see a similar check.

I'm pretty sure Stephen fixed this test with https://bugs.webkit.org/show_bug.cgi?id=54561 at http://trac.webkit.org/changeset/78922 - could you confirm if the test still fails after that revision?

-- 
Configure bugmail: https://bugs.webkit.org/userprefs.cgi?tab=email
------- You are receiving this mail because: -------
You are the assignee for the bug.