Moves FrameHostMsg_SetCookie to a Mojo message

content/renderer/render_frame_impl.cc

Index: content/renderer/render_frame_impl.cc
diff --git a/content/renderer/render_frame_impl.cc b/content/renderer/render_frame_impl.cc
index d799759d73354b4e16cd9613064cb4b017c14789..eac3b7eb38846107b85e7efab7bc136044f6fb63 100644
--- a/content/renderer/render_frame_impl.cc
+++ b/content/renderer/render_frame_impl.cc
@@ -1451,6 +1451,14 @@ MediaStreamDispatcher* RenderFrameImpl::GetMediaStreamDispatcher() {
       web_user_media_client_->media_stream_dispatcher() : NULL;
 }
 
+mojom::RenderFrameMessageFilter*
+RenderFrameImpl::render_frame_message_filter() {
+  if (!render_frame_message_filter_) {
+    RenderThread::Get()->GetChannel()->GetRemoteAssociatedInterface(

[jam, 2016/07/26 19:13:32]

it's a bit odd that this mojo interface is registered on the process host object
on the browser side, but on the renderer it's fetched via the frame object. it
seems that there should be symmetry.

any reason why we can't make a getter for this on the render thread class? that
way we'd only create one MP for this, which I think is all we need.

[Ken Rockot(use gerrit already), 2016/07/26 21:00:34]

Done

+        &render_frame_message_filter_);
+  }
+  return render_frame_message_filter_.get();
+}
 bool RenderFrameImpl::Send(IPC::Message* message) {
   return RenderThread::Get()->Send(message);
 }