Connect mojom::DisplayController from ash to ui.

services/ui/service.cc

Index: services/ui/service.cc
diff --git a/services/ui/service.cc b/services/ui/service.cc
index 025112dd8df4889de220aa557664fd148e3312d9..4a2f85aebb2b22c5d649f4e845882b630177171e 100644
--- a/services/ui/service.cc
+++ b/services/ui/service.cc
@@ -212,6 +212,8 @@ bool Service::OnConnect(const shell::Identity& remote_identity,
   if (input_device_server_.IsRegisteredAsObserver())
     input_device_server_.AddInterface(registry);
 
+  platform_screen_->AddInterfaces(registry);

[dcheng, 2016/09/12 18:04:30]

Dumb question: how do we guarantee that platform_screen_ outlives the lifetime
of ServiceRegistry? InterfaceFactorys don't appear to have any lifetime
guarantees.

[kylechar, 2016/09/12 18:28:15]

In this case platform_screen_ gets destroyed when the ui::Service that owns it
gets destroyed.

I'm not totally sure how Mojo application shutdown works but it looks like
shell:ServiceRunner is what ensures the order is correct. The
shell::InterfaceRegistry is local to the Run() method (owned by
shell::ServiceContext) while the shell:Service gets destroyed in the
shell::ServiceRunner destructor, so shell::InterfaceRegistry will always be
destroyed first.

[dcheng, 2016/09/13 01:04:48]

This would be nice to have documented somewhere.
(Unrelated, but I'm not sure why we need both the InterfaceFactory and Callback
methods of registering interfaces: seems like it'd be better if we only had
one?)

[kylechar, 2016/09/13 14:09:53]

T'll add comments on platform_screen_ and input_device_server_ noting they need
to outlive shell::InterfaceRegistry so someone doesn't call .reset() early. With
regards to your second question, I'm not sure, but I'm sure someone on the Mojo
team would at chromium-mojo?

https://groups.google.com/a/chromium.org/forum/#!forum/chromium-mojo

+
 #if defined(USE_OZONE)
   ui::OzonePlatform::GetInstance()->AddInterfaces(registry);
 #endif