cc: Free Mailbox on TextureLayerImpl::ReleaseResources

cc: Free Mailbox on TextureLayerImpl::ReleaseResources

Add unit tests to run through this code path

BUG=239864
Committed: https://src.chromium.org/viewvc/chrome?view=rev&revision=246690

[boliu, 2014-01-22 16:43:25]

Originally left out cleaning mailboxes because webview didn't need it and this
comment in the bug:
https://code.google.com/p/chromium/issues/detail?id=245935#c6

In chrome, ReleaseResources is called on OutputSurface loss, at which point the
ResourceProvider is recreated and the mailbox would already be released back to
TextureLayerClient. So this CL should have make no difference there.

On webview, this can of course also happen with DeferredInitialze and ReleaseGL,
where the ResourceProvider is not recreated.

One problem that applies to both situations: TextureLayerClient using mailboxes
have optimizations to have PrepareMailbox return false if the mailbox has not
changed. I think this means that, eg, a static canvas would not be redrawn again
after a context loss. This needs to be fixed at each individual
TextureLayerClient though. And this patch should not make this any worse.

So far I've found that WebExternalTextureLayerImpl (Canvas2DLayerBridge +
DrawingBuffer) and PepperPluginInstanceImpl (PepperGraphics2DHost) will need to
be fixed.

[danakj, 2014-01-23 18:06:43]

LGTM w/ 2 things.

https://codereview.chromium.org/140693015/diff/40001/cc/layers/texture_layer_...
File cc/layers/texture_layer_unittest.cc (right):

https://codereview.chromium.org/140693015/diff/40001/cc/layers/texture_layer_...
cc/layers/texture_layer_unittest.cc:1962:
layer_tree_host()->SetRootLayer(texture_layer);
Add this as a child of the root instead of making it the root. You can do this
by calling LayerTreeTest::SetupTree() first, then it will make a root layer and
you can attach it to that.

https://codereview.chromium.org/140693015/diff/40001/cc/layers/texture_layer_...
cc/layers/texture_layer_unittest.cc:1975: virtual void
DidActivateTreeOnThread(LayerTreeHostImpl* host_impl) OVERRIDE {
This belongs only in the subclasses right?

[boliu, 2014-01-23 18:54:09]

https://codereview.chromium.org/140693015/diff/40001/cc/layers/texture_layer_...
File cc/layers/texture_layer_unittest.cc (right):

https://codereview.chromium.org/140693015/diff/40001/cc/layers/texture_layer_...
cc/layers/texture_layer_unittest.cc:1962:
layer_tree_host()->SetRootLayer(texture_layer);
On 2014/01/23 18:06:43, danakj wrote:
> Add this as a child of the root instead of making it the root. You can do this
> by calling LayerTreeTest::SetupTree() first, then it will make a root layer
and
> you can attach it to that.

Done. Of course also had to update the code to get the layer impl.

https://codereview.chromium.org/140693015/diff/40001/cc/layers/texture_layer_...
cc/layers/texture_layer_unittest.cc:1975: virtual void
DidActivateTreeOnThread(LayerTreeHostImpl* host_impl) OVERRIDE {
On 2014/01/23 18:06:43, danakj wrote:
> This belongs only in the subclasses right?

Doh! Good catch, removed.

[commit-bot: I haz the power, 2014-01-23 18:55:05]

CQ is trying da patch. Follow status at
https://chromium-status.appspot.com/cq/boliu@chromium.org/140693015/130001

[commit-bot: I haz the power, 2014-01-23 21:30:50]

Change committed as 246690