MojoCompositorFrameSinkPrivate should support copy requests

cc/output/copy_output_request.h

Index: cc/output/copy_output_request.h
diff --git a/cc/output/copy_output_request.h b/cc/output/copy_output_request.h
index 112daa40dcefe08ad0b4c567badf096b378e2778..1b3898359be798ca0f7d36741e787c813fcb2871 100644
--- a/cc/output/copy_output_request.h
+++ b/cc/output/copy_output_request.h
@@ -83,6 +83,14 @@ class CC_EXPORT CopyOutputRequest {
 
   void SendResult(std::unique_ptr<CopyOutputResult> result);
 
+  void set_result_callback(CopyOutputRequestCallback result_callback) {

[danakj, 2017/02/06 23:02:12]

This looks like what CreateRelayRequest is for, to make a new CopyOutputRequest
with a new callback?

[Saman Sami, 2017/02/06 23:09:54]

Yeah, you're right. I can fix that, but do we have to make an unnecessary copy?

[danakj, 2017/02/06 23:24:08]

It's not a copy of much, what are you concerned for? And it keeps the internals
private.

[Saman Sami, 2017/02/13 23:07:37]

Pretty much every private member has a setter though?

[danakj, 2017/02/13 23:11:15]

Yes but this the thing that ensures the callback is run, it's passed in the
constructor unlike the other things. If you replace the callback now errors and
what is getting replied to gets much more complicated. I think the relationship
that a CopyOutputRequest is tied to a fixed callback is valuable and this would
allow much more difficult to understand things to happen, even if in this case
it's not bad. I don't want to see these methods added here for those reasons.

[danakj, 2017/02/13 23:12:19]

This along with the fact there is already a way to do this, with a RelayRequest.
This adds new APIs that are not needed.

+    result_callback_ = result_callback;
+  }
+
+  const CopyOutputRequestCallback& result_callback() {
+    return result_callback_;
+  }
+
  private:
   friend struct mojo::StructTraits<mojom::CopyOutputRequestDataView,
                                    std::unique_ptr<CopyOutputRequest>>;