compositor-worker: Initialize CW machinery plumbing to compositor and fire CW rAF callbacks.

third_party/WebKit/Source/web/CompositorProxyClientImpl.h

Index: third_party/WebKit/Source/web/CompositorProxyClientImpl.h
diff --git a/third_party/WebKit/Source/web/CompositorProxyClientImpl.h b/third_party/WebKit/Source/web/CompositorProxyClientImpl.h
new file mode 100644
index 0000000000000000000000000000000000000000..777e5058edb5f50232b1dbd76711debda424fcfe
--- /dev/null
+++ b/third_party/WebKit/Source/web/CompositorProxyClientImpl.h
@@ -0,0 +1,42 @@
+// Copyright 2015 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 CompositorProxyClientImpl_h
+#define CompositorProxyClientImpl_h
+
+#include "core/dom/CompositorProxyClient.h"
+#include "wtf/Noncopyable.h"
+
+namespace blink {
+
+class CompositorMutatorImpl;
+class CompositorWorkerGlobalScope;
+class WorkerGlobalScope;
+

[jbroman, 2016/04/29 19:42:22]

Class comment here, along the lines of the ones given in
go/compositorworker-design? Something along the lines of:

// Mediates between one CompositorWorkerGlobalScope and the associated
CompositorMutatorImpl.
// There is one CompositorProxyClientImpl per worker, but there may be may be
multiple for a
// given mutator, e.g. if a single document creates multiple CompositorWorker
objects.
//
// Should be accessed only on the compositor thread.

While I can infer this by reading both associated classes, I'd rather not. :)

(If my description is bad or wrong, even more reason for you to write a better
one!)

[jbroman, 2016/04/29 19:42:22]

Class comment here, along the lines of the ones given in
go/compositorworker-design? Something along the lines of:

// Mediates between one CompositorWorkerGlobalScope and the associated
CompositorMutatorImpl.
// There is one CompositorProxyClientImpl per worker, but there may be may be
multiple for a
// given mutator, e.g. if a single document creates multiple CompositorWorker
objects.

While I can infer this by reading both associated classes, I'd rather not. :)

(If my description is bad or wrong, even more reason for you to write a better
one!)

[flackr, 2016/05/03 22:24:55]

Your description is concise and accurate - I've lifted it pretty much exactly.

+class CompositorProxyClientImpl final : public GarbageCollectedFinalized<CompositorProxyClientImpl>, public CompositorProxyClient {

[jbroman, 2016/04/29 19:42:22]

nit: does this have any finalization logic? could it just be
GarbageCollected<...>?

[flackr, 2016/05/03 22:24:54]

For now, no. Switched to no finalization - I'll leave it to the patch that adds
necessary finalization to change it.

+    USING_GARBAGE_COLLECTED_MIXIN(CompositorProxyClientImpl);
+    WTF_MAKE_NONCOPYABLE(CompositorProxyClientImpl);
+public:
+    CompositorProxyClientImpl(CompositorMutatorImpl*, double referenceMonotonicTime);
+    DECLARE_VIRTUAL_TRACE();
+
+    bool mutate(double monotonicTimeNow);

[jbroman, 2016/04/29 19:42:22]

Please comment what the return type means. From reading the rest of this CL, I
know it means "should reinvoke", but it's not obvious from here.

[flackr, 2016/05/03 22:24:54]

Done.

+
+    // CompositorProxyClient:
+    void setGlobalScope(WorkerGlobalScope*) override;
+    void requestAnimationFrame() override;
+
+private:
+    bool executeAnimationFrameCallbacks(double monotonicTimeNow);
+
+    CompositorMutatorImpl* m_mutator;
+
+    WeakMember<CompositorWorkerGlobalScope> m_globalScope;
+    double m_referenceMonotonicTime;

[jbroman, 2016/04/29 19:42:22]

Why plumb this in separately? Isn't it guaranteed to be equal to
m_globalScope->timeOrigin()?

[flackr, 2016/05/03 22:24:54]

So it is, thanks! Removed.

+    bool m_requestedAnimationFrameCallbacks;
+};
+
+} // namespace blink
+
+#endif // CompositorProxyClientImpl_h