Migrate GpuChildThread to use ConnectionFilter instead of the ChildThread's InterfaceRegistry to exp

content/gpu/gpu_child_thread.h

Index: content/gpu/gpu_child_thread.h
diff --git a/content/gpu/gpu_child_thread.h b/content/gpu/gpu_child_thread.h
index a0e2bd36d52727e375831d86085c21217f206fcf..93b62d818a51934567092449a3bdfa440197c66c 100644
--- a/content/gpu/gpu_child_thread.h
+++ b/content/gpu/gpu_child_thread.h
@@ -72,6 +72,8 @@ class GpuChildThread : public ChildThreadImpl,
   void Init(const base::Time& process_start_time);
 
  private:
+  class QueueingConnectionFilter;

[Ken Rockot(use gerrit already), 2017/04/20 17:41:44]

Does this really need to be a nested class declaration? I think it can just be
defined entirely in an anonymous namespace in the cc?

+
   GpuChildThread(const ChildThreadImpl::Options& options,
                  std::unique_ptr<gpu::GpuWatchdogThread> gpu_watchdog_thread,
                  bool dead_on_arrival,
@@ -130,6 +132,9 @@ class GpuChildThread : public ChildThreadImpl,
   std::unique_ptr<ui::GpuService> gpu_service_;
   mojo::AssociatedBinding<ui::mojom::GpuMain> gpu_main_binding_;
 
+  // Holds a closure that releases pending interface requests on the IO thread.
+  base::Closure release_closure_;
+
   DISALLOW_COPY_AND_ASSIGN(GpuChildThread);
 };