Cleanup GpuMemoryBuffer

content/common/gpu/client/gpu_memory_buffer_impl_mac.cc

Index: content/common/gpu/client/gpu_memory_buffer_impl_mac.cc
diff --git a/content/common/gpu/client/gpu_memory_buffer_impl_mac.cc b/content/common/gpu/client/gpu_memory_buffer_impl_mac.cc
index e0fe58da0e9ff250f638ef491408b9da9b1598df..4df34ed0071da810bfac7df0ae99090353b0d094 100644
--- a/content/common/gpu/client/gpu_memory_buffer_impl_mac.cc
+++ b/content/common/gpu/client/gpu_memory_buffer_impl_mac.cc
@@ -78,4 +78,10 @@ scoped_ptr<GpuMemoryBufferImpl> GpuMemoryBufferImpl::CreateFromHandle(
   }
 }
 
+// static
+bool GpuMemoryBufferImpl::NeedsDestructionSyncPoint(
+    gfx::GpuMemoryBufferType type) {
+  return false;

[piman, 2014/10/31 01:24:18]

true for IO_SURFACE_BUFFER?

Do we really need a per-platform implementation of this? What really matter is
what we do relative to the handle type (whether it's an OS handle that we dup,
or it's a reference to a key in the GPU process), and that doesn't vary per
platform, so a single implementation with a single switch statement seems the
most reasonable.

IO_SURFACE_BUFFER, X11_PIXMAP_BUFFER, OZONE_NATIVE_BUFFER,
SURFACE_TEXTURE_BUFFER need a sync point.
SHARED_MEMORY_BUFFER doesn't.

+}
+
 }  // namespace content