Bug 46412 – [CHROMIUM] <canvas> clearRect() doesn't respect active clipping region, when --enable-accelerated-2d-canvas is on

• Component: Canvas
• Status: RESOLVED FIXED

[Stephen White]

According to the spec, canvas's clearRect() should respect the active clipping region.  Chrome with --enable-accelerated-2d-canvas doesn't.

See layout test http://philip.html5.org/tests/canvas/suite/tests/2d.clearRect.clip.html.

[Stephen White]

Created attachment 68672 [details]
Patch

[James Robinson]

Comment on attachment 68672 [details]
Patch

View in context: https://bugs.webkit.org/attachment.cgi?id=68672&action=review

> WebCore/platform/graphics/skia/GraphicsContextSkia.cpp:351
> +    if (platformContext()->useGPU() && !platformContext()->canvasClipApplied()) {
>          platformContext()->prepareForHardwareDraw();
>          platformContext()->gpuCanvas()->clearRect(rect);
>          return;
>      }
>  
> +    // Temporarily force the mode to "clear" to trigger a readback

I'd much prefer doing a syncSoftwareCanvas() here rather than depending on side effects of the Xfer mode being set to do the readback.

[Stephen White]

Created attachment 68903 [details]
Patch

[Stephen White]

(In reply to comment #2)
> (From update of attachment 68672 [details])
> View in context: https://bugs.webkit.org/attachment.cgi?id=68672&action=review
> 
> > WebCore/platform/graphics/skia/GraphicsContextSkia.cpp:351
> > +    if (platformContext()->useGPU() && !platformContext()->canvasClipApplied()) {
> >          platformContext()->prepareForHardwareDraw();
> >          platformContext()->gpuCanvas()->clearRect(rect);
> >          return;
> >      }
> >  
> > +    // Temporarily force the mode to "clear" to trigger a readback
> 
> I'd much prefer doing a syncSoftwareCanvas() here rather than depending on side effects of the Xfer mode being set to do the readback.

Good idea.  Done.

[James Robinson]

Comment on attachment 68903 [details]
Patch

R=me

[Stephen White]

Committed r68402: <http://trac.webkit.org/changeset/68402>