[cc] Fix CompositorFrameSinkSupport BeginFrameAck interface.

cc/ipc/mojo_compositor_frame_sink.mojom

 // Copyright 2016 The Chromium Authors. All rights reserved.
 // Use of this source code is governed by a BSD-style license that can be
 // found in the LICENSE file.
 
 module cc.mojom;
 
 import "cc/ipc/begin_frame_args.mojom";
 import "cc/ipc/compositor_frame.mojom";
 import "cc/ipc/copy_output_request.mojom";
 import "cc/ipc/frame_sink_id.mojom";
@@ skipping 17 matching lines @@
   // to screen the next time frames from all CompositorFrameSinks are aggregated
   // to produce a display CompositorFrame. If a client wishes to allocate a new
   // surface (e.g. during resize), then it can simply allocate a new
   // |local_surface_id|.
   // For successful swaps, the implementation must call
   // DidReceiveCompositorFrameAck() asynchronously when the frame has been
   // processed in order to unthrottle the next frame.
   SubmitCompositorFrame(cc.mojom.LocalSurfaceId local_surface_id,
                         cc.mojom.CompositorFrame frame);
 
+  // Notifies the frame sink that a BeginFrame was completed, but that no
+  // CompositorFrame was produced as a result of it.
+  BeginFrameDidNotSwap(cc.mojom.BeginFrameAck ack);

[Fady Samuel, 2017/03/17 13:16:07]

Just a question, maybe this can be addressed in a separate CL. What do we expect
to ultimately call this API? LayerTreeHostImpl? If so, shouldn't
cc::CompositorFrameSink also be updated?
https://cs.chromium.org/chromium/src/cc/output/compositor_frame_sink.h?q=cc::...

[Eric Seckler, 2017/03/17 13:40:22]

It is called by users of MojoCompositorFrameSink and CompositorFrameSinkSupport
only. Everything that uses cc::CompositorFrameSink, including
cc::LayerTreeHostImpl+cc::Scheduler, reports ACKs directly to a BeginFrameSource
(via BFS::DidFinishFrame), rather than to the frame sink.

[Fady Samuel, 2017/03/17 13:45:25]

Ohh I see, then you should update a couple of TODOs then. Thanks!

[Eric Seckler, 2017/03/17 14:01:11]

Ok, sure while we're at it - I guess this shouldn't be regression-risky :)

Also done in exo::CompositorFrameSinkHolder.

+
   // Notify that the surface is no longer in use (and is okay to be evicted) so
   // that its resources gets returned in time.
   EvictFrame();
 };
 
 interface MojoCompositorFrameSinkClient {
   // Notification that the previous CompositorFrame given to
   // SubmitCompositorFrame() has been processed and that another frame
   // can be submitted. This provides backpressure from the display compositor
   // so that frames are submitted only at the rate it can handle them.
@@ skipping 18 matching lines @@
 interface MojoCompositorFrameSinkPrivate {
   // Claims this FrameSinkId will embed |surface_id| so it should own the
   // temporary reference to |surface_id|.
   ClaimTemporaryReference(SurfaceId surface_id);
 
   // Requests that the renderer send back a copy of the surface that this
   // CompositorFrameSink submits to. The result can be in form of a bitmap
   // or a texture. See cc::CopyOutputRequest.
   RequestCopyOfSurface(cc.mojom.CopyOutputRequest request);
 };