mus/views: Call to WaitForIncomingMethodCall() can fail.

ui/views/mus/screen_mus.cc

Index: ui/views/mus/screen_mus.cc
diff --git a/ui/views/mus/screen_mus.cc b/ui/views/mus/screen_mus.cc
index cec007bfe6b5adf9bc2690578c776462dc9e37fb..d6d110647907717770bce05c4d35680721a6d9b7 100644
--- a/ui/views/mus/screen_mus.cc
+++ b/ui/views/mus/screen_mus.cc
@@ -96,12 +96,13 @@ void ScreenMus::Init(shell::Connector* connector) {
   //
   // TODO(rockot): Do something better here. This should not have to block tasks
   // from running on the calling thread. http://crbug.com/594852.
-  display_manager_observer_binding_.WaitForIncomingMethodCall();
+  bool success = display_manager_observer_binding_.WaitForIncomingMethodCall();
 
   // The WaitForIncomingMethodCall() should have supplied the set of Displays,
-  // unless mus is going down, in which case encountered_error() is true.
+  // unless mus is going down, in which case encountered_error() is true, or the
+  // call to WaitForIncomingMethodCall() failed.
   if (displays_.empty()) {
-    DCHECK(display_manager_.encountered_error());
+    DCHECK(display_manager_.encountered_error() || !success);

[sky, 2016/04/27 21:09:59]

Isn't encountered_error set in this case?

[sadrul, 2016/04/27 21:39:24]

Apparently not. (it's easy to hit this DCHECK() with
https://codereview.chromium.org/1920033005/ without this patch)

     // In this case we install a default display and assume the process is
     // going to exit shortly so that the real value doesn't matter.
     displays_.push_back(gfx::Display(0xFFFFFFFF, gfx::Rect(0, 0, 801, 802)));