[chromium] DRT crash in fast/canvas tests w/accel 2D canvas due to NULL DrawingBuffer

[Kenneth Russell]

The following test cases in fast/canvas are failing in DRT when run with --enable-accelerated-compositing --enable-accelerated-2d-canvas due to a NULL m_drawingBuffer in CanvasRenderingContext2D. The reason is that drawingContext() is NULL when the CanvasRenderingContext2D is created, so the drawing buffer is never initialized.

fast/canvas/access-zero-sized-canvas.html
fast/canvas/canvas-skia-excessive-size.html
fast/canvas/toDataURL-noData.html

[James Robinson]

Created attachment 69880 [details]
Patch

[James Robinson]

Patch flips around the logic so we don't touch m_context3d or m_drawingBuffer if there's no drawingContext(), which makes sense since if there is no drawingBuffer() we won't be doing any software draws either.  This also means that <canvas> elements that are zero sized or excessively large will not get composited layers since canvas layers are only set if CanvasRenderingContext2D::isAccelerated() is true which checks m_context3d.

[Stephen White]

Comment on attachment 69880 [details]
Patch

View in context: https://bugs.webkit.org/attachment.cgi?id=69880&action=review

Looks good.

> WebCore/html/canvas/CanvasRenderingContext2D.cpp:171
> +        if (m_context3D) {

Nit:  Could use && and a single if (although it'd probably need more parens around the =; your call).

[Kenneth Russell]

Comment on attachment 69880 [details]
Patch

Looks fine to me.

[James Robinson]

Comment on attachment 69880 [details]
Patch

I thought this had already landed - whoops!

[James Robinson]

Comment on attachment 69880 [details]
Patch

Clearing flags on attachment: 69880

Committed r69529: <http://trac.webkit.org/changeset/69529>

[James Robinson]

All reviewed patches have been landed.  Closing bug.