cc: DelegatedRendererLayer creates a ChildId with the ResourceProvider

cc: DelegatedRendererLayer creates a ChildId with the ResourceProvider

When committing, create a child id if needed. If the context3d is lost
then the child id needs to be recreated.

Tests:
LayerTreeHostDelegatedTestCreateChildId.runSingleThread
LayerTreeHostDelegatedTestCreateChildId.runMultiThread

R=piman
BUG=123444
Committed: https://src.chromium.org/viewvc/chrome?view=rev&revision=184844

[piman, 2013-02-26 22:49:21]

LGTM, but for one question.

https://codereview.chromium.org/12328118/diff/12/cc/test/layer_tree_test_comm...
File cc/test/layer_tree_test_common.cc (right):

https://codereview.chromium.org/12328118/diff/12/cc/test/layer_tree_test_comm...
cc/test/layer_tree_test_common.cc:83: m_testHooks->treeActivatedOnThread(this);
Should this be before LayerTreeHostImpl::commitComplete? The assumption is that
activation happens in finishCommitOnImplThread, which is called before
commitComplete.

[danakj, 2013-02-26 22:56:54]

https://codereview.chromium.org/12328118/diff/12/cc/test/layer_tree_test_comm...
File cc/test/layer_tree_test_common.cc (right):

https://codereview.chromium.org/12328118/diff/12/cc/test/layer_tree_test_comm...
cc/test/layer_tree_test_common.cc:83: m_testHooks->treeActivatedOnThread(this);
On 2013/02/26 22:49:21, piman wrote:
> Should this be before LayerTreeHostImpl::commitComplete? The assumption is
that
> activation happens in finishCommitOnImplThread, which is called before
> commitComplete.

It could be commit complete - they're equal at the moment. But I think this will
remain true forever, whereas commit complete is true only if we block commit?

[piman, 2013-02-26 23:19:20]

https://codereview.chromium.org/12328118/diff/12/cc/test/layer_tree_test_comm...
File cc/test/layer_tree_test_common.cc (right):

https://codereview.chromium.org/12328118/diff/12/cc/test/layer_tree_test_comm...
cc/test/layer_tree_test_common.cc:83: m_testHooks->treeActivatedOnThread(this);
On 2013/02/26 22:56:54, danakj wrote:
> On 2013/02/26 22:49:21, piman wrote:
> > Should this be before LayerTreeHostImpl::commitComplete? The assumption is
> that
> > activation happens in finishCommitOnImplThread, which is called before
> > commitComplete.
> 
> It could be commit complete - they're equal at the moment. But I think this
will
> remain true forever, whereas commit complete is true only if we block commit?

This part is only relevant for non-impl-side-painting I guess - I assume when we
remove non-impl-side-painting, this code would go away too. I'm just trying to
see if we can preserve the order of the callbacks so that if we inspect state
from either of the callbacks, it's consistent.
If it doesn't matter, then feel free to land as is.

[danakj, 2013-02-26 23:32:54]

https://codereview.chromium.org/12328118/diff/12/cc/test/layer_tree_test_comm...
File cc/test/layer_tree_test_common.cc (right):

https://codereview.chromium.org/12328118/diff/12/cc/test/layer_tree_test_comm...
cc/test/layer_tree_test_common.cc:83: m_testHooks->treeActivatedOnThread(this);
On 2013/02/26 23:19:21, piman wrote:
> On 2013/02/26 22:56:54, danakj wrote:
> > On 2013/02/26 22:49:21, piman wrote:
> > > Should this be before LayerTreeHostImpl::commitComplete? The assumption is
> > that
> > > activation happens in finishCommitOnImplThread, which is called before
> > > commitComplete.
> > 
> > It could be commit complete - they're equal at the moment. But I think this
> will
> > remain true forever, whereas commit complete is true only if we block
commit?
> 
> This part is only relevant for non-impl-side-painting I guess - I assume when
we
> remove non-impl-side-painting, this code would go away too. I'm just trying to
> see if we can preserve the order of the callbacks so that if we inspect state
> from either of the callbacks, it's consistent.
> If it doesn't matter, then feel free to land as is.

Oh sorry I replied about a completely different line of code... Yes this goes
away when implSidePainting is always true.

You're absolutely right. However, with impl painting on, the tree is activated
after commit complete. Creating a test environment where different settings call
things in different order is really horrible, so I prefer to lie here and do it
after commitComplete - same as with impl painting. It should be safe because it
just delays the activate notification, the tree is definitely activated when
this is called.

(For example theres lots of callbacks thread proxy does but single thread proxy
doesn't which makes testing both modes a lot trickier than it should be. I'm
scarred from this.)

[commit-bot: I haz the power, 2013-02-26 23:33:37]

CQ is trying da patch. Follow status at
https://chromium-status.appspot.com/cq/danakj@chromium.org/12328118/6031

[commit-bot: I haz the power, 2013-02-27 02:50:47]

Change committed as 184844