aw: Block child compositor if frame is not consumed by parent

aw: Block child compositor if frame is not consumed by parent

Pipeline onDrawHardware, kModeSync, and kModeDraw. Each step will block
the previous step in the pipeline. This is the avoid the parent
compositor getting ahead of child.

BUG=

Committed: https://crrev.com/9b78e178cd29a7b2cf9788aad922c34ebaf78ef0
Cr-Commit-Position: refs/heads/master@{#295866}

[boliu, 2014-09-20 00:03:15]

boliu@chromium.org changed reviewers:
+ hush@chromium.org

[boliu, 2014-09-20 00:03:16]

This is the "pipeline and drop" patch. It's not too bad. And it will
automatically handle the case if/when kModeSync also disappears.

[hush (inactive), 2014-09-20 02:15:53]

https://codereview.chromium.org/585093003/diff/1/android_webview/browser/brow...
File android_webview/browser/browser_view_renderer.cc (left):

https://codereview.chromium.org/585093003/diff/1/android_webview/browser/brow...
android_webview/browser/browser_view_renderer.cc:307:
ReturnUnusedResource(shared_renderer_state_->PassDrawGLInput());
why is this removed?

https://codereview.chromium.org/585093003/diff/1/android_webview/browser/hard...
File android_webview/browser/hardware_renderer.cc (right):

https://codereview.chromium.org/585093003/diff/1/android_webview/browser/hard...
android_webview/browser/hardware_renderer.cc:109: SetFrameData();
what is the purpose of this?
setFrameData() is only changing a bunch of states of HardwareRenderer which will
be destructed soon

https://codereview.chromium.org/585093003/diff/1/android_webview/browser/hard...
android_webview/browser/hardware_renderer.cc:203: DCHECK(current_context) <<
"DrawGL called without EGLContext";
this means in production code we will continue without EGLContext?

[boliu, 2014-09-20 02:20:21]

https://codereview.chromium.org/585093003/diff/1/android_webview/browser/brow...
File android_webview/browser/browser_view_renderer.cc (left):

https://codereview.chromium.org/585093003/diff/1/android_webview/browser/brow...
android_webview/browser/browser_view_renderer.cc:307:
ReturnUnusedResource(shared_renderer_state_->PassDrawGLInput());
On 2014/09/20 02:15:53, hush wrote:
> why is this removed?

If we are consuming the input, we can't detect when parent doesn't consume it

https://codereview.chromium.org/585093003/diff/1/android_webview/browser/hard...
File android_webview/browser/hardware_renderer.cc (right):

https://codereview.chromium.org/585093003/diff/1/android_webview/browser/hard...
android_webview/browser/hardware_renderer.cc:109: SetFrameData();
On 2014/09/20 02:15:53, hush wrote:
> what is the purpose of this?
> setFrameData() is only changing a bunch of states of HardwareRenderer which
will
> be destructed soon

All resources *must* be returned to child. This puts the data into the
compositor, then we destroy the compositor which ends up calling
UnusedResourcesAreAvailable below.

https://codereview.chromium.org/585093003/diff/1/android_webview/browser/hard...
android_webview/browser/hardware_renderer.cc:203: DCHECK(current_context) <<
"DrawGL called without EGLContext";
On 2014/09/20 02:15:53, hush wrote:
> this means in production code we will continue without EGLContext?

Yes. This shouldn't happen anymore since we tear down hardware in trim.

Also the next line shouldn't happen anymore either...

[hush (inactive), 2014-09-20 02:31:46]

The CQ bit was checked by hush@chromium.org

[hush (inactive), 2014-09-20 02:31:46]

lgtm

cool.
thanks!

[commit-bot: I haz the power, 2014-09-20 02:32:23]

CQ is trying da patch. Follow status at
 https://chromium-cq-status.appspot.com/patch-status/585093003/20001

[commit-bot: I haz the power, 2014-09-20 13:18:12]

Committed patchset #2 (id:20001) as a5303ad162d283ff567eb68c6a78a262fd482989

[commit-bot: I haz the power, 2014-09-20 13:19:18]

Patchset 2 (id:??) landed as
https://crrev.com/9b78e178cd29a7b2cf9788aad922c34ebaf78ef0
Cr-Commit-Position: refs/heads/master@{#295866}