Fix lifecycle-vs-threads of GpuVDAContext3D.

Fix lifecycle-vs-threads of GpuVDAContext3D.

This context3d is meant to be created on the render thread but can only be
dereferenced on the compositor thread.  Before this CL this was violated as
RenderThreadImpl::GetGpuVDAContext3D() checked for context loss to decide
whether to recreate the context.  Now context loss proactively notifies the
render thread, instead.

BUG=142047
Committed: http://src.chromium.org/viewvc/chrome?view=rev&revision=151352

[Ami GONE FROM CHROMIUM, 2012-08-13 17:14:06]

jamesr & piman: please review.
posciak: FYI.

[piman, 2012-08-13 17:23:35]

https://chromiumcodereview.appspot.com/10832264/diff/5001/content/common/gpu/...
File content/common/gpu/client/webgraphicscontext3d_command_buffer_impl.cc
(right):

https://chromiumcodereview.appspot.com/10832264/diff/5001/content/common/gpu/...
content/common/gpu/client/webgraphicscontext3d_command_buffer_impl.cc:1428:
DCHECK(!context_lost_callback_ || !cb);
why?

https://chromiumcodereview.appspot.com/10832264/diff/5001/content/renderer/re...
File content/renderer/render_thread_impl.cc (right):

https://chromiumcodereview.appspot.com/10832264/diff/5001/content/renderer/re...
content/renderer/render_thread_impl.cc:775: new GpuVDAContextLostCallback();
Why the need to leak it? Could it be a scoped_ptr field on RenderThreadImpl? The
compositor thread is shut down in the destructor, so it should be safe.

[Ami GONE FROM CHROMIUM, 2012-08-13 17:38:16]

https://chromiumcodereview.appspot.com/10832264/diff/5001/content/common/gpu/...
File content/common/gpu/client/webgraphicscontext3d_command_buffer_impl.cc
(right):

https://chromiumcodereview.appspot.com/10832264/diff/5001/content/common/gpu/...
content/common/gpu/client/webgraphicscontext3d_command_buffer_impl.cc:1428:
DCHECK(!context_lost_callback_ || !cb);
On 2012/08/13 17:23:35, piman wrote:
> why?

Vestigial; was there to make sure my use of sCLC wasn't clobbering another use. 
Gone.

https://chromiumcodereview.appspot.com/10832264/diff/5001/content/renderer/re...
File content/renderer/render_thread_impl.cc (right):

https://chromiumcodereview.appspot.com/10832264/diff/5001/content/renderer/re...
content/renderer/render_thread_impl.cc:775: new GpuVDAContextLostCallback();
On 2012/08/13 17:23:35, piman wrote:
> Why the need to leak it? Could it be a scoped_ptr field on RenderThreadImpl?
The
> compositor thread is shut down in the destructor, so it should be safe.

Done, at the cost of having to 
#include
"third_party/WebKit/Source/Platform/chromium/public/WebGraphicsContext3D.h"
from render_thread_impl.h

[piman, 2012-08-13 17:45:05]

lgtm

https://chromiumcodereview.appspot.com/10832264/diff/10001/content/renderer/r...
File content/renderer/render_thread_impl.h (right):

https://chromiumcodereview.appspot.com/10832264/diff/10001/content/renderer/r...
content/renderer/render_thread_impl.h:307: : public
WebKit::WebGraphicsContext3D::WebGraphicsContextLostCallback {
Note: I'm fine with this, but FYI you can declare the inner class here and
define it in the .cc file, which does let you get rid of the header.

[Ami GONE FROM CHROMIUM, 2012-08-13 17:55:38]

https://chromiumcodereview.appspot.com/10832264/diff/10001/content/renderer/r...
File content/renderer/render_thread_impl.h (right):

https://chromiumcodereview.appspot.com/10832264/diff/10001/content/renderer/r...
content/renderer/render_thread_impl.h:307: : public
WebKit::WebGraphicsContext3D::WebGraphicsContextLostCallback {
On 2012/08/13 17:45:06, piman wrote:
> Note: I'm fine with this, but FYI you can declare the inner class here and
> define it in the .cc file, which does let you get rid of the header.

Taking advantage of the scoped_ptr for auto-dtor goodness requires a full
declaration here, not just a fwd-declaration.

[piman, 2012-08-13 18:35:57]

https://chromiumcodereview.appspot.com/10832264/diff/10001/content/renderer/r...
File content/renderer/render_thread_impl.h (right):

https://chromiumcodereview.appspot.com/10832264/diff/10001/content/renderer/r...
content/renderer/render_thread_impl.h:307: : public
WebKit::WebGraphicsContext3D::WebGraphicsContextLostCallback {
On 2012/08/13 17:55:38, Ami Fischman wrote:
> On 2012/08/13 17:45:06, piman wrote:
> > Note: I'm fine with this, but FYI you can declare the inner class here and
> > define it in the .cc file, which does let you get rid of the header.
> 
> Taking advantage of the scoped_ptr for auto-dtor goodness requires a full
> declaration here, not just a fwd-declaration.

Nah, you only need it with the destructor of RenderThreadImpl

[Ami GONE FROM CHROMIUM, 2012-08-13 18:50:13]

On 2012/08/13 18:35:57, piman wrote:
> Nah, you only need it with the destructor of RenderThreadImpl

D'oh!  of course!
(I messed up my first attempt by missing a RenderThreadImpl:: qualification on
the actual definition in the .cc file)
Done.

[commit-bot: I haz the power, 2012-08-13 19:36:55]

CQ is trying da patch. Follow status at
https://chromium-status.appspot.com/cq/fischman@chromium.org/10832264/11

[commit-bot: I haz the power, 2012-08-13 21:21:42]

Change committed as 151352