Service CreateNewWindow on the UI thread with a new mojo interface

content/renderer/render_thread_impl.cc

Index: content/renderer/render_thread_impl.cc
diff --git a/content/renderer/render_thread_impl.cc b/content/renderer/render_thread_impl.cc
index 6c478d38a81d4348994fe085ac90c5ea0902f634..e4bf055815389130907b33a1abf6e6b2971a2406 100644
--- a/content/renderer/render_thread_impl.cc
+++ b/content/renderer/render_thread_impl.cc
@@ -537,6 +537,13 @@ mojom::RenderMessageFilter* RenderThreadImpl::current_render_message_filter() {
 }
 
 // static
+mojom::RenderMessageFilterUI*
+RenderThreadImpl::current_render_message_filter_ui() {
+  DCHECK(current());
+  return current()->render_message_filter_ui();
+}
+
+// static
 void RenderThreadImpl::SetRenderMessageFilterForTesting(
     mojom::RenderMessageFilter* render_message_filter) {
   g_render_message_filter_for_testing = render_message_filter;
@@ -2055,6 +2062,15 @@ mojom::RenderMessageFilter* RenderThreadImpl::render_message_filter() {
   return render_message_filter_.get();
 }
 
+mojom::RenderMessageFilterUI* RenderThreadImpl::render_message_filter_ui() {
+  if (!render_message_filter_ui_) {
+    blink_platform_impl_->GetInterfaceProvider()->GetInterface(

[Charlie Harrison, 2017/04/13 19:27:58]

This is weird and wrong. Probably should just use the connector directly?

[dcheng, 2017/04/13 21:00:14]

Maybe this should be an associated interface to keep the same behavior?

[Charlie Harrison, 2017/04/13 21:48:12]

Hm... How would that work? Isn't it fundamentally against the spirit of an
associated interface (same underlying pipe) for endpoints to be on two different
threads?

+        mojo::MakeRequest(&render_message_filter_ui_));
+  }
+
+  return render_message_filter_ui_.get();
+}
+
 gpu::GpuChannelHost* RenderThreadImpl::GetGpuChannel() {
   if (!gpu_channel_)
     return nullptr;