cc: Update Main RendererCapabilities on DeferredInitialize

cc/layers/texture_layer.cc

Index: cc/layers/texture_layer.cc
diff --git a/cc/layers/texture_layer.cc b/cc/layers/texture_layer.cc
index bc6b58a0342145bce3130c02d6983591f33855f9..4d51f6472d61d01303f48bedc4ec75314f3fda42 100644
--- a/cc/layers/texture_layer.cc
+++ b/cc/layers/texture_layer.cc
@@ -37,7 +37,8 @@ TextureLayer::TextureLayer(TextureLayerClient* client, bool uses_mailbox)
       rate_limit_context_(false),
       content_committed_(false),
       texture_id_(0),
-      needs_set_mailbox_(false) {
+      needs_set_mailbox_(false),
+      did_use_shared_memory_resources_(false) {
   vertex_opacity_[0] = 1.0f;
   vertex_opacity_[1] = 1.0f;
   vertex_opacity_[2] = 1.0f;
@@ -221,6 +222,13 @@ bool TextureLayer::Update(ResourceUpdateQueue* queue,
   bool updated = Layer::Update(queue, occlusion);
   if (client_) {
     if (uses_mailbox_) {
+      bool use_shared_memory_resources =
+          layer_tree_host()->UsingSharedMemoryResources();
+      if (did_use_shared_memory_resources_ != use_shared_memory_resources) {
+        holder_ref_.reset();

[danakj, 2014/02/14 19:38:50]

Now that we're here, I wonder.. the call to PrepareTextureMailbox will tell the
client about UsingSharedMemoryResources. If they set a new (or empty) TM as a
result, it'll reset the holder ref. So what is this buying us? Is it saving us
redundant code in other places?

[boliu, 2014/02/14 22:03:08]

Good point...removed all this code and just make the client a little bit
smarter.

+      }
+      did_use_shared_memory_resources_ = use_shared_memory_resources;
+
       TextureMailbox mailbox;
       scoped_ptr<SingleReleaseCallback> release_callback;
       if (client_->PrepareTextureMailbox(