Aura: fix full-screen Flash with delegated renderer

Aura: fix full-screen Flash with delegated renderer

During the fullscreen transition, on some sites Flash destroys and recreates its
context, which causes the RenderWidgetFullscreenPepper to switch back and forth
between composited and non-composited modes, recreating its compositor. That
causes it to send 2 successive frames with resources that have the same id (1)
but from different compositors (and different mailboxes).
Make sure we disambiguate those by recreating the DelegatedRendererLayer.

BUG=284821
Committed: https://src.chromium.org/viewvc/chrome?view=rev&revision=221341

[danakj, 2013-09-04 13:07:51]

LGTM

I want to ask for a test but I have no idea how to test RWHVA in a reasonable
way for corner cases.

https://codereview.chromium.org/23513012/diff/1/content/browser/renderer_host...
File content/browser/renderer_host/render_widget_host_view_aura.cc (right):

https://codereview.chromium.org/23513012/diff/1/content/browser/renderer_host...
content/browser/renderer_host/render_widget_host_view_aura.cc:1465: // indicates
the renderer's compositor may have been recreated, in which
This mentions the compositor being recreated, but really each output surface
should have its own id namespace. It could mean a new 3d context, or could also
mean a new shm handler etc. Can you also mention that a new output surface
should mean new id namespace so future us doesn't think this is just about
compositor lifetimes?

https://codereview.chromium.org/23513012/diff/1/content/browser/renderer_host...
File content/browser/renderer_host/render_widget_host_view_aura.h (right):

https://codereview.chromium.org/23513012/diff/1/content/browser/renderer_host...
content/browser/renderer_host/render_widget_host_view_aura.h:592: //
disambiguate resources with the same id coming from different compositors.
s/compositors/output surfaces/.

[commit-bot: I haz the power, 2013-09-04 21:44:38]

CQ is trying da patch. Follow status at
https://chromium-status.appspot.com/cq/piman@chromium.org/23513012/12001

[piman, 2013-09-04 21:45:02]

https://codereview.chromium.org/23513012/diff/1/content/browser/renderer_host...
File content/browser/renderer_host/render_widget_host_view_aura.cc (right):

https://codereview.chromium.org/23513012/diff/1/content/browser/renderer_host...
content/browser/renderer_host/render_widget_host_view_aura.cc:1465: // indicates
the renderer's compositor may have been recreated, in which
On 2013/09/04 13:07:51, danakj wrote:
> This mentions the compositor being recreated, but really each output surface
> should have its own id namespace. It could mean a new 3d context, or could
also
> mean a new shm handler etc. Can you also mention that a new output surface
> should mean new id namespace so future us doesn't think this is just about
> compositor lifetimes?

Done.

https://codereview.chromium.org/23513012/diff/1/content/browser/renderer_host...
File content/browser/renderer_host/render_widget_host_view_aura.h (right):

https://codereview.chromium.org/23513012/diff/1/content/browser/renderer_host...
content/browser/renderer_host/render_widget_host_view_aura.h:592: //
disambiguate resources with the same id coming from different compositors.
On 2013/09/04 13:07:51, danakj wrote:
> s/compositors/output surfaces/.

Done.

[commit-bot: I haz the power, 2013-09-05 04:20:50]

Change committed as 221341