[Webkit-unassigned] [Bug 70634] Mark GraphicsLayers as opaque when possible

[bugzilla-daemon at webkit.org]

https://bugs.webkit.org/show_bug.cgi?id=70634





--- Comment #76 from Dana Jansens <danakj at chromium.org>  2011-11-28 13:34:26 PST ---
(In reply to comment #75)
> (In reply to comment #70)
> > (From update of attachment 116550 [details] [details])
> > This patch has some bad effects with WebGL Canvas. In particular, when an image contents change, we need to determine its opaque status for the RenderLayer's opaque region. But in order to create an Image* for a generated image, the contents may change causing a loop where there should not be. This causes a synchronization issue in the HTMLCanvasElement/WebGL interaction, since it does not expect to generate an image in its contents changed listeners.
> 
> I haven't been watching this bug closely, but wanted to make it clear that it isn't feasible to determine whether WebGL content is completely opaque -- aside from seeing whether the WebGL context was created without an alpha channel. See http://www.khronos.org/registry/webgl/specs/latest/#5.2.1 . The assumption in the general case must be that WebGL content is not opaque. It would be fine to assume this all the time initially, and optimize it later. Please make sure that this patch does not introduce a CPU readback during WebGL rendering; this would kill performance.

Thanks for this. I will require some plumbing through the RenderImageResource to be able to tell if it is a WebGL image at all and avoid this readback.

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