Removed most calls to GraphicsContext3D from DrawingBuffer

Removed most calls to GraphicsContext3D from DrawingBuffer

Also includes some misc related refactors that were missed on previous CLs

BUG=326382
Committed: https://src.chromium.org/viewvc/blink?view=rev&revision=165020

[bajones, 2014-01-08 03:02:36]

kbr@: core/ and platform/ review please?

jamesr@ or adamk@: web/ review please?

[bajones, 2014-01-08 03:13:08]

https://codereview.chromium.org/127493002/diff/1/Source/platform/graphics/gpu...
File Source/platform/graphics/gpu/DrawingBuffer.cpp (right):

https://codereview.chromium.org/127493002/diff/1/Source/platform/graphics/gpu...
Source/platform/graphics/gpu/DrawingBuffer.cpp:81: RefPtr<GraphicsContext3D>
contextSupport(GraphicsContext3D::createContextSupport(context));
Rather than require that both the context and context support are passed the
pattern that I'd like to establish is that each class that uses
WebGraphicsContext3D instantiates it's own contextSupport if it needs it. As
refactoring continues the support class will become more lightweight, so this
shouldn't be a big deal. It does mean that we'll probably end up parsing the
extensions string multiple times, though.

https://codereview.chromium.org/127493002/diff/1/Source/platform/graphics/gpu...
File Source/platform/graphics/gpu/DrawingBuffer.h (right):

https://codereview.chromium.org/127493002/diff/1/Source/platform/graphics/gpu...
Source/platform/graphics/gpu/DrawingBuffer.h:176: blink::WebGraphicsContext3D*
m_context;
General design/style question: When using GraphicsContext3D we used RefPtr to
manage it, but that won't work for WebGraphicsContext3D, since it's actually
instantiated Chromium-side. In WebGLRenderingContext we're using OwnPtr to
manage the object lifetime, is there a preferred pattern for classes like this
one that the object is passed into?

[bajones, 2014-01-08 23:35:18]

I think https://codereview.chromium.org/127163003 is actually going to stick
this time, so this CL is valid again.

[bajones, 2014-01-09 22:57:35]

Reasonably big update to the CL. I've moved all painting-related functions off
of the GraphicsContext3D and onto the DrawingBuffer. This works well for a
couple of reasons:

 * It brings GraphicsContext3D to a point where it holds no state, which means
that you can create as many instances of it as you want and use them
interchangeably.
 * It makes the GC3D more lightweight, which facilitates the use previous use
case.
 * The paint functions (paintRenderingResultsToCanvas,
paintRenderingResultsToCanvas) required a DrawingBuffer anyway, so making them
members of DrawingBuffer simplifies things, including allowing a couple of
functions to become private.

[bajones, 2014-01-10 19:38:56]

Review ping? I need to context switch for a bit, and it would be nice to not
leave this hanging for too long.

[adamk, 2014-01-10 22:34:13]

I was going to let kbr have a first shot at this before rubber-stamping

[Ken Russell (switch to Gerrit), 2014-01-13 22:36:13]

LGTM as long as this passes all tests.

[bajones, 2014-01-13 22:45:27]

On 2014/01/13 22:36:13, Ken Russell wrote:
> LGTM as long as this passes all tests.

Thank you! And yes, it's passed everything I've thrown at it.
adamk@: Can I get that rubber stamp now?

[adamk, 2014-01-14 00:18:55]

rslgtm for web/

[commit-bot: I haz the power, 2014-01-14 00:34:26]

CQ is trying da patch. Follow status at
https://chromium-status.appspot.com/cq/bajones@chromium.org/127493002/240001

[commit-bot: I haz the power, 2014-01-14 03:39:50]

Change committed as 165020