[cc] Fix CompositorFrameSinkSupport BeginFrameAck interface.

cc/surfaces/direct_compositor_frame_sink.cc

Index: cc/surfaces/direct_compositor_frame_sink.cc
diff --git a/cc/surfaces/direct_compositor_frame_sink.cc b/cc/surfaces/direct_compositor_frame_sink.cc
index a915b941c1e6ada8c17fff3cefbfd350d8e76c1a..d2bc2c861125cb53bdec6139c0939808c5ee568e 100644
--- a/cc/surfaces/direct_compositor_frame_sink.cc
+++ b/cc/surfaces/direct_compositor_frame_sink.cc
@@ -151,7 +151,9 @@ void DirectCompositorFrameSink::OnNeedsBeginFrames(bool needs_begin_frame) {
 }
 
 void DirectCompositorFrameSink::OnDidFinishFrame(const BeginFrameAck& ack) {
-  support_->DidFinishFrame(ack);
+  // If there was damage, SubmitCompositorFrame includes the ack.

[piman, 2017/03/17 18:12:01]

nit: is this something we should DCHECK in SubmitCompositorFrame?

Are there cases where there is no damage but we still we submit a frame (e.g. to
update metadata)? In that case is the BeginFrameAck included in the
CompositorFrame?

[Eric Seckler, 2017/03/20 11:54:03]

Added a DCHECK.

Yes, whenever there's a CompositorFrame, the ack is attached to it.

I think there may be CompositorFrames without (eventually visible) surface
damage, but we consider any CompositorFrame as "has_damage" for BeginFrameAck
purposes. "has_damage" may not be the most fitting description (it's rather
"client sent updates as a result of the BeginFrame"), see discussion in
https://codereview.chromium.org/2591013004/ and
https://codereview.chromium.org/2527283003/ for why we arrived at the field's
name.

+  if (!ack.has_damage)
+    support_->BeginFrameDidNotSwap(ack);
 }
 
 }  // namespace cc