use accelerated compositing in fullscreen pepper

Enable accelerated compositing in fullscreen pepper.

The compositing itself is trivial since there is just a fullscreen plugin instance (that we blit full-screen), but it allows the use of the 3D contexts in pepper when fullscreen is on (they require accelerated compositing).

Accelerated compositing is only triggered when a Surface3D is bound to the instance. Otherwise the regular 2D painting path is used.

This CL also refactors a bit RenderWidgetFullscreenPepper, merging the trivial WidgetFullscreenContainer into RenderWidgetFullscreenPepper - it's simpler that way.

BUG=none
TEST=Flash

Committed: http://src.chromium.org/viewvc/chrome?view=rev&revision=70636

[apatrick, 2011-01-06 02:59:31]

http://codereview.chromium.org/5994002/diff/2001/chrome/renderer/render_widge...
File chrome/renderer/render_widget_fullscreen_pepper.cc (right):

http://codereview.chromium.org/5994002/diff/2001/chrome/renderer/render_widge...
chrome/renderer/render_widget_fullscreen_pepper.cc:355: float kTexCoords[] = {
const float?

http://codereview.chromium.org/5994002/diff/2001/chrome/renderer/render_widge...
chrome/renderer/render_widget_fullscreen_pepper.cc:365: program_ =
gl->CreateProgram();
You're relying on the context being destroyed to destroy this right? Just
checking :) Ditto buffer_. If this context might be part of a share group is it
possible these might leak across. I don't recall the semantics of sharing
programs but I'm pretty sure buffers are shared.

[piman, 2011-01-06 04:27:46]

http://codereview.chromium.org/5994002/diff/2001/chrome/renderer/render_widge...
File chrome/renderer/render_widget_fullscreen_pepper.cc (right):

http://codereview.chromium.org/5994002/diff/2001/chrome/renderer/render_widge...
chrome/renderer/render_widget_fullscreen_pepper.cc:355: float kTexCoords[] = {
On 2011/01/06 02:59:31, apatrick wrote:
> const float?

Done.

http://codereview.chromium.org/5994002/diff/2001/chrome/renderer/render_widge...
chrome/renderer/render_widget_fullscreen_pepper.cc:365: program_ =
gl->CreateProgram();
On 2011/01/06 02:59:31, apatrick wrote:
> You're relying on the context being destroyed to destroy this right? Just
> checking :) Ditto buffer_. If this context might be part of a share group is
it
> possible these might leak across. I don't recall the semantics of sharing
> programs but I'm pretty sure buffers are shared.

Done.

I was relying on the context destruction, and it has a separate namespace
currently but I want to eventually share with the main renderer's context
(because currently this requires the pepper module to re-create the context when
switching to full-screen and back), at which point the cleanup will be
necessary. Might as well get it done now.

FYI, all "names" are shared: textures, buffers, FBOs, shader/programs, fences,
display lists (originally) etc.

[apatrick, 2011-01-06 16:43:03]

LGTM

[brettw, 2011-01-06 17:52:32]

I didn't check the 3D stuff, but the plumbing LGTM.

http://codereview.chromium.org/5994002/diff/8001/webkit/plugins/ppapi/ppapi_p...
File webkit/plugins/ppapi/ppapi_plugin_instance.cc (right):

http://codereview.chromium.org/5994002/diff/8001/webkit/plugins/ppapi/ppapi_p...
webkit/plugins/ppapi/ppapi_plugin_instance.cc:944: if (fullscreen_container_) {
To match the surrounding style, don't use {} for single-line conditionals.

[piman, 2011-01-06 20:18:25]

http://codereview.chromium.org/5994002/diff/8001/webkit/plugins/ppapi/ppapi_p...
File webkit/plugins/ppapi/ppapi_plugin_instance.cc (right):

http://codereview.chromium.org/5994002/diff/8001/webkit/plugins/ppapi/ppapi_p...
webkit/plugins/ppapi/ppapi_plugin_instance.cc:944: if (fullscreen_container_) {
On 2011/01/06 17:52:32, brettw wrote:
> To match the surrounding style, don't use {} for single-line conditionals.

It has an else clause - it's required in that case, isn't it ?

[brettw, 2011-01-06 20:22:12]

http://codereview.chromium.org/5994002/diff/8001/webkit/plugins/ppapi/ppapi_p...
File webkit/plugins/ppapi/ppapi_plugin_instance.cc (right):

http://codereview.chromium.org/5994002/diff/8001/webkit/plugins/ppapi/ppapi_p...
webkit/plugins/ppapi/ppapi_plugin_instance.cc:944: if (fullscreen_container_) {
Nope, the only thing different about this is if one of the arms has it, all of
them must.

[piman, 2011-01-06 20:38:58]

http://codereview.chromium.org/5994002/diff/8001/webkit/plugins/ppapi/ppapi_p...
File webkit/plugins/ppapi/ppapi_plugin_instance.cc (right):

http://codereview.chromium.org/5994002/diff/8001/webkit/plugins/ppapi/ppapi_p...
webkit/plugins/ppapi/ppapi_plugin_instance.cc:944: if (fullscreen_container_) {
On 2011/01/06 20:22:13, brettw wrote:
> Nope, the only thing different about this is if one of the arms has it, all of
> them must.

Aaight, done.

[brettw, 2011-01-06 20:42:02]

(LGTM before if it wasn't clear, I'm happy with Al's review).