Establish mojo service between Canvas (blink) and SurfaceManager (browser)

third_party/WebKit/Source/platform/graphics/CanvasSurfaceLayerBridgeClient.h

Index: third_party/WebKit/Source/platform/graphics/CanvasSurfaceLayerBridgeClient.h
diff --git a/third_party/WebKit/Source/platform/graphics/CanvasSurfaceLayerBridgeClient.h b/third_party/WebKit/Source/platform/graphics/CanvasSurfaceLayerBridgeClient.h
new file mode 100644
index 0000000000000000000000000000000000000000..72265db89bf7de36201e96dd261929ac35732490
--- /dev/null
+++ b/third_party/WebKit/Source/platform/graphics/CanvasSurfaceLayerBridgeClient.h
@@ -0,0 +1,31 @@
+// 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.
+
+#ifndef CanvasSurfaceLayerBridgeClient_h
+#define CanvasSurfaceLayerBridgeClient_h
+
+#include "cc/surfaces/surface_id.h"
+#include "cc/surfaces/surface_sequence.h"
+#include "platform/PlatformExport.h"
+
+namespace blink {
+
+// This class is an interface for all mojo calls from CanvasSurfaceLayerBridge
+// to OffscreenCanvasSurfaceService.
+class PLATFORM_EXPORT CanvasSurfaceLayerBridgeClient {

[dcheng, 2016/09/15 20:45:11]

Apparently this class just exists as a layer of indirection for mojo vs tests?
Can we get rid of it? The test can just implement a mock version of the mojo
interface and use it directly. See
https://cs.chromium.org/chromium/src/url/mojo/url_gurl_struct_traits_unittest...
for an example of how to do this all in a unit test.

[xlai (Olivia), 2016/09/15 22:53:10]

This class was added since Patch Set 7 and it was suggested by danakj@ (See her
comments
on Patch Set 6 for details). Basically, the major purpose of this class is to
make it cleaner between
code and test such that the unit test will not interfere too much with the
original code; it's a better
way to do testing. 
Another benefit is that the include of offscreen_canvas_surface.mojom-blink.h
(which occurs in/CanvasSurfaceLayerBridgeClientImpl.h) will not interfere with
HTMLCanvasElement.cpp that
include CanvasSurfaceLayerBridge.

On 2016/09/15 20:45:11, dcheng wrote:

[danakj, 2016/09/15 23:48:47]

I think what Daniel is getting at is that we should just use a mojo interface
and make the implementation of that interface in the unit tests. That's what we
were going for but you ran into problems where you were unable to instantiate a
mojo service in the unit tests? So we went with an abstract class instead.
Daniel believes you should be able to do this with just a mojo interface and
have it work in tests tho, maybe his example would help explain how/.

[dcheng, 2016/09/18 17:24:42]

I have a CL for this (https://codereview.chromium.org/2345423002/) although
there's some weird issues with sync methods in unit tests. I'll figure out
what's going on and send it out for review.

+public:
+    virtual ~CanvasSurfaceLayerBridgeClient() {};
+
+    // Calls that help initial creation of SurfaceLayer.
+    virtual bool syncGetSurfaceId(cc::SurfaceId*) = 0;
+    virtual void asyncRequestSurfaceCreation(const cc::SurfaceId&) = 0;
+
+    // Calls that ensure correct destruction order of surface.
+    virtual void asyncRequire(cc::SurfaceId, cc::SurfaceSequence) = 0;
+    virtual void asyncSatisfy(cc::SurfaceSequence) = 0;
+};
+
+} // namespace blink
+
+#endif // CanvasSurfaceLayerBridgeClient_h