Browser side OZONE_NATIVE_BUFFER allocation

content/common/gpu/client/gpu_memory_buffer_impl.h

Index: content/common/gpu/client/gpu_memory_buffer_impl.h
diff --git a/content/common/gpu/client/gpu_memory_buffer_impl.h b/content/common/gpu/client/gpu_memory_buffer_impl.h
index 2565c02653a71a4ac1a0b46a991dd4154839d95b..362b0bd2c9bd8b20b6d72d98371501a8885fe1f0 100644
--- a/content/common/gpu/client/gpu_memory_buffer_impl.h
+++ b/content/common/gpu/client/gpu_memory_buffer_impl.h
@@ -11,6 +11,7 @@
 #include "ui/gfx/size.h"
 
 namespace content {
+class GpuMemoryBufferFactoryHost;
 
 // Provides common implementation of a GPU memory buffer.
 class GpuMemoryBufferImpl : public gfx::GpuMemoryBuffer {
@@ -29,10 +30,14 @@ class GpuMemoryBufferImpl : public gfx::GpuMemoryBuffer {
   // Allocates a GPU memory buffer with |size| and |internalformat| for |usage|
   // by |child_process|. The |handle| returned can be used by the
   // |child_process| to create an instance of this class.
+  // The buffer is being allocated for |child_id| and |factory_host| can
+  // be used to talk to the allocator.

[reveman, 2014/08/13 08:03:11]

Can you merge the mention of |child_id| with the text above instead? ie. "for
|usage| by |child_process| with |child_id|"...

[achaulk, 2014/08/13 15:53:19]

Done.

   static void AllocateForChildProcess(const gfx::Size& size,
                                       unsigned internalformat,
                                       unsigned usage,
                                       base::ProcessHandle child_process,
+                                      int child_id,
+                                      GpuMemoryBufferFactoryHost* factory_host,

[reveman, 2014/08/13 08:03:11]

It's weird that we pass a factory to this function. Makes it seem like we might
need to be able to pass different factory instances to this function. The use of
the factory is supposed to be an implementation detail of this function. I would
prefer if we add a GpuMemoryBufferFactoryHost::GetInstance() function that can
be used when necessary instead. If we want to avoid global variables, then we'd
need to add a GpuMemoryBufferImplFactory and pass the FactoryHost to the ctor of
this new factory instance. Adding GpuMemoryBufferFactoryHost::GetInstance()
function is fine with me for now though.

[achaulk, 2014/08/13 15:53:19]

Sure, I can do that. The only implementation is a singleton anyway.

                                       const AllocationCallback& callback);
 
   // Notify that GPU memory buffer has been deleted by |child_process|.