Delete background tab IOSurfaces on Mac.

content/common/view_messages.h

Index: content/common/view_messages.h
diff --git a/content/common/view_messages.h b/content/common/view_messages.h
index 7a5433958531b5b364fb2cb3391b779d60f6cd59..adc7c1162671e9be3821b10f388b326b893ebbb3 100644
--- a/content/common/view_messages.h
+++ b/content/common/view_messages.h
@@ -499,13 +499,20 @@ IPC_STRUCT_BEGIN(ViewHostMsg_UpdateRect_Params)
   // The size of the RenderView when this message was generated.  This is
   // included so the host knows how large the view is from the perspective of
   // the renderer process.  This is necessary in case a resize operation is in
-  // progress.
+  // progress. If auto-resize is enabled, this should update the corresponding
+  // view size.
   IPC_STRUCT_MEMBER(gfx::Size, view_size)
 
   // New window locations for plugin child windows.
   IPC_STRUCT_MEMBER(std::vector<webkit::npapi::WebPluginGeometry>,
                     plugin_window_moves)
 
+  // GPU SwapBuffers params:
+  IPC_STRUCT_MEMBER(int32, gpu_surface_id)
+  IPC_STRUCT_MEMBER(uint64, gpu_surface_handle)
+  IPC_STRUCT_MEMBER(int32, gpu_route_id)
+  IPC_STRUCT_MEMBER(int32, gpu_process_host_id)

[piman, 2012/04/18 20:02:00]

Ok, these fields look completely arbitrary.
In GpuProcessHost, you're creating this with setting no field except those 4.
For a message coming from the renderer you set all the fields except those 4.
In RenderWidgetHostImpl::OnMsgUpdateRect you have a big if (gpu_surface_id) {
do_something(); } else { do_something_different(); }

It really calls for 2 separate messages.

RenderWidgetHelper doesn't actually care if the msg is a ViewHostMsg_UpdateRec
or not, it always goes through the dispatch mechanism.

I'll argue, why not simply forward the
GpuHostMsg_AcceleratedSurfaceBuffersSwapped message through that mechanism?

[jbates, 2012/04/18 23:01:58]

I think that would be cleaner with this CL, but it would take us farther from
the goal of merging UpdateRect and BuffersSwapped (this CL is a stepping stone
towards that). If we eventually merge the two, then even if this UpdateRect is a
SwapBuffers, it would still contain the rest of the UpdateRect data from the
renderer and the OnMsgUpdateRect code would execute in both cases.

[piman, 2012/04/18 23:32:22]

I think we should do that change (or whatever is required then) once we get to
that.
For the merging, I don't think you need to merge the GPU-initiated and the
Renderer-initiated messages, they just need to carry some common data (which can
be extracted in a common struct).

[jbates, 2012/04/19 01:24:18]

Sounds good, I'll put up a new CL tomorrow-ish.

+
   // The following describes the various bits that may be set in flags:
   //
   //   ViewHostMsg_UpdateRect_Flags::IS_RESIZE_ACK