cc: Re-enable GPU rasterization after content veto.

cc/trees/layer_tree_host.h

Index: cc/trees/layer_tree_host.h
diff --git a/cc/trees/layer_tree_host.h b/cc/trees/layer_tree_host.h
index 26560e7facc54de59aeac0c710b117231a05baf0..752a9f8fa75e96b9ae53830e25d6079cc880761d 100644
--- a/cc/trees/layer_tree_host.h
+++ b/cc/trees/layer_tree_host.h
@@ -490,6 +490,13 @@ class CC_EXPORT LayerTreeHost : public MutatorHostClient {
  private:
   friend class LayerTreeHostSerializationTest;
 
+  // Maximum amount of times to re-enable gpu rasterization after it's been
+  // turned off due to content.
+  enum { kMaxTimesToEnableGpuRasterization = 5 };

[enne (OOO), 2016/06/30 01:02:24]

Why do you want to have a limit on how many times you reeanble? If you have some
long running page that changes state, I feel like we should change an indefinite
number of times?  If you're worried about changing too much, then change the
number of commits you need before swapping?

[ericrk, 2016/06/30 18:46:00]

+1

+  // The number of commits to waitch before attempting to re-enable gpu

[enne (OOO), 2016/06/30 01:02:24]

<_<

+  // rasterization.
+  enum { kFramePeriodToReeanbleGpuRasterization = 10 };

[enne (OOO), 2016/06/30 01:02:24]

Reeanble?

[ericrk, 2016/06/30 18:46:00]

There's a significant cost to switching I'd imagine (re-decode all images,
etc...)... Is 10 frames too little? I'd be more tempted to do this on the order
of at most once a second?

+
   void InitializeProxy(
       std::unique_ptr<Proxy> proxy,
       std::unique_ptr<BeginFrameSource> external_begin_frame_source);
@@ -616,6 +623,7 @@ class CC_EXPORT LayerTreeHost : public MutatorHostClient {
 
   uint32_t surface_id_namespace_;
   uint32_t next_surface_sequence_;
+  int num_times_gpu_rasterization_reenabled_ = 0;
 
   DISALLOW_COPY_AND_ASSIGN(LayerTreeHost);
 };