Add WebGraphicsContext support for readbacks from any framebuffer.

Add WebGraphicsContext support for readbacks from any framebuffer.

This is needed for https://bugs.webkit.org/show_bug.cgi?id=65658 .

BUG=55927
TEST=


Committed: http://src.chromium.org/viewvc/chrome?view=rev&revision=95898

[Ken Russell (switch to Gerrit), 2011-08-04 22:55:07]

LGTM

Given that we're passing in an externally allocated FBO I wonder whether this
could live entirely in upstream code, reducing dependencies between Chromium and
WebKit. It looks fine however and it's good to share code.

http://codereview.chromium.org/7566046/diff/1/content/renderer/gpu/webgraphic...
File content/renderer/gpu/webgraphicscontext3d_command_buffer_impl.cc (right):

http://codereview.chromium.org/7566046/diff/1/content/renderer/gpu/webgraphic...
content/renderer/gpu/webgraphicscontext3d_command_buffer_impl.cc:274: }
Do you think we might want or need checks about the completeness of the FBO
given that it's now coming from outside code?

http://codereview.chromium.org/7566046/diff/1/content/renderer/gpu/webgraphic...
File content/renderer/gpu/webgraphicscontext3d_command_buffer_impl.h (right):

http://codereview.chromium.org/7566046/diff/1/content/renderer/gpu/webgraphic...
content/renderer/gpu/webgraphicscontext3d_command_buffer_impl.h:73: size_t
buffer_size, WebGLId framebuffer);
I would have organized these arguments differently to keep the related ones
together. For example, pixels, buffer_size, framebuffer, width, height.

[jbauman, 2011-08-04 23:18:33]

On 2011/08/04 22:55:07, kbr wrote:
> LGTM
> 
> Given that we're passing in an externally allocated FBO I wonder whether this
> could live entirely in upstream code, reducing dependencies between Chromium
and
> WebKit. It looks fine however and it's good to share code.
> 
It probably could be moved out, but I preferred to share the exact same code, so
behavior is the same both when we get the data from the compositor and when we
get it from the drawing buffer directly.
>
http://codereview.chromium.org/7566046/diff/1/content/renderer/gpu/webgraphic...
> File content/renderer/gpu/webgraphicscontext3d_command_buffer_impl.cc (right):
> 
>
http://codereview.chromium.org/7566046/diff/1/content/renderer/gpu/webgraphic...
> content/renderer/gpu/webgraphicscontext3d_command_buffer_impl.cc:274: }
> Do you think we might want or need checks about the completeness of the FBO
> given that it's now coming from outside code?
> 
>
http://codereview.chromium.org/7566046/diff/1/content/renderer/gpu/webgraphic...
> File content/renderer/gpu/webgraphicscontext3d_command_buffer_impl.h (right):
> 
>
http://codereview.chromium.org/7566046/diff/1/content/renderer/gpu/webgraphic...
> content/renderer/gpu/webgraphicscontext3d_command_buffer_impl.h:73: size_t
> buffer_size, WebGLId framebuffer);
> I would have organized these arguments differently to keep the related ones
> together. For example, pixels, buffer_size, framebuffer, width, height.

[jbauman, 2011-08-04 23:20:21]

http://codereview.chromium.org/7566046/diff/1/content/renderer/gpu/webgraphic...
File content/renderer/gpu/webgraphicscontext3d_command_buffer_impl.cc (right):

http://codereview.chromium.org/7566046/diff/1/content/renderer/gpu/webgraphic...
content/renderer/gpu/webgraphicscontext3d_command_buffer_impl.cc:274: }
On 2011/08/04 22:55:07, kbr wrote:
> Do you think we might want or need checks about the completeness of the FBO
> given that it's now coming from outside code?
Those can be expensive, so I'd rather avoid them (even if they aren't as
expensive as the actual readpixels). I'll let the calling code take care of
that, so they'll be done less often.

http://codereview.chromium.org/7566046/diff/1/content/renderer/gpu/webgraphic...
File content/renderer/gpu/webgraphicscontext3d_command_buffer_impl.h (right):

http://codereview.chromium.org/7566046/diff/1/content/renderer/gpu/webgraphic...
content/renderer/gpu/webgraphicscontext3d_command_buffer_impl.h:73: size_t
buffer_size, WebGLId framebuffer);
On 2011/08/04 22:55:07, kbr wrote:
> I would have organized these arguments differently to keep the related ones
> together. For example, pixels, buffer_size, framebuffer, width, height.
Yeah, I guess that organization makes more sense. Fixed.

[Ken Russell (switch to Gerrit), 2011-08-08 18:54:15]

LGTM

[commit-bot: I haz the power, 2011-08-08 22:53:25]

Change committed as 95898