Blink Compositor Animation: Introduce CompositorAnimationHost correctly.

third_party/WebKit/Source/web/WebFrameWidgetImpl.h

Index: third_party/WebKit/Source/web/WebFrameWidgetImpl.h
diff --git a/third_party/WebKit/Source/web/WebFrameWidgetImpl.h b/third_party/WebKit/Source/web/WebFrameWidgetImpl.h
index 38148a15914faee3c566e737ef234d69b8dcaa2a..efda4edcaf871f0be62df26441cd7b6d3689d8dc 100644
--- a/third_party/WebKit/Source/web/WebFrameWidgetImpl.h
+++ b/third_party/WebKit/Source/web/WebFrameWidgetImpl.h
@@ -47,6 +47,8 @@
 #include "wtf/HashSet.h"
 
 namespace blink {
+
+class CompositorAnimationHost;
 class Frame;
 class Element;
 class InspectorOverlay;
@@ -218,6 +220,7 @@ class WebFrameWidgetImpl final
   WebLayerTreeView* m_layerTreeView;
   WebLayer* m_rootLayer;
   GraphicsLayer* m_rootGraphicsLayer;
+  std::unique_ptr<CompositorAnimationHost> m_compositorAnimationHost;

[jbroman, 2016/12/13 16:38:32]

Up to you, but if having a compositor animation host is the usual case then I'd
lean towards using WTF::Optional<> (or making CompositorAnimationHost support
nullability itself, but I gather we're leaning away from that). This is
approximately a pointer-to-a-pointer, which is a little much.

[loyso (OOO), 2016/12/13 22:23:35]

Yup. I was thinking about base::Oprional here. Will do that.

[loyso (OOO), 2016/12/19 04:03:12]

Since we want to replace it with a raw weak pointer (similar to
WebLayerTreeView* m_layerTreeView;) I prefer to continue with
std::unique_ptr<CompositorAnimationHost>.
Added a todo note in
https://bugs.chromium.org/p/chromium/issues/detail?id=584551

   bool m_isAcceleratedCompositingActive;
   bool m_layerTreeViewClosed;