Add ChannelHandle::mojo_handle to mus' ChannelHandle mojom.

components/mus/common/gpu_type_converters.cc

Index: components/mus/common/gpu_type_converters.cc
diff --git a/components/mus/common/gpu_type_converters.cc b/components/mus/common/gpu_type_converters.cc
index 7733c7d3bf7e5d51ebd95afcdba9d865ae18541c..4a2615e1654a65c8a08a38e0926788dc97c336ea 100644
--- a/components/mus/common/gpu_type_converters.cc
+++ b/components/mus/common/gpu_type_converters.cc
@@ -31,6 +31,7 @@ TypeConverter<mus::mojom::ChannelHandlePtr, IPC::ChannelHandle>::Convert(
   if (platform_file != -1)
     result->socket = mojo::WrapPlatformFile(platform_file);
 #endif
+  result->mojo_handle.reset(handle.mojo_handle);

[dcheng, 2016/07/01 01:31:43]

It's hard for me to tell where the type converter is used. What are the chances
of getting rid of this type converter sooner than later? It's not at all clear
to me that the DCHECKs are OK to have here (they generally aren't).

[Anand Mistry (off Chromium), 2016/07/01 05:52:54]

So, I'm making this change in prep for switching GPU-renderer channel to
ChannelMojo. I think switching this type converter to StructTraits<> is outside
the scoped of this CL.

[dcheng, 2016/07/04 07:11:59]

Fair enough, I've filed a followup bug and assign it to someone on the mus team.

[Peng, 2016/07/06 15:19:06]

I think when GPU starts using ChannelMojo, we could just pass a mojo handle
around instead of defining a mojo ChannelHandle structure and having converters
for it.

   return result;
 }
 
@@ -40,6 +41,11 @@ TypeConverter<IPC::ChannelHandle, mus::mojom::ChannelHandlePtr>::Convert(
     const mus::mojom::ChannelHandlePtr& handle) {
   if (handle.is_null())
     return IPC::ChannelHandle();
+  if (handle->mojo_handle.is_valid()) {
+    IPC::ChannelHandle channel_handle(handle->mojo_handle.release());
+    channel_handle.name = handle->name;
+    return channel_handle;
+  }
 #if defined(OS_WIN)
   // On windows, a pipe handle Will NOT be marshalled over IPC.
   DCHECK(!handle->socket.is_valid());