Break circular dependency between InitializeDeviceDisablingManager and DeviceDisabledScreen

chrome/browser/chromeos/login/screens/device_disabled_screen.cc

Index: chrome/browser/chromeos/login/screens/device_disabled_screen.cc
diff --git a/chrome/browser/chromeos/login/screens/device_disabled_screen.cc b/chrome/browser/chromeos/login/screens/device_disabled_screen.cc
index 66d644c1c6fb0bf47631bf5d51a946636647962c..4ba8b71038b166ab35628b8ff50afb7be922c334 100644
--- a/chrome/browser/chromeos/login/screens/device_disabled_screen.cc
+++ b/chrome/browser/chromeos/login/screens/device_disabled_screen.cc
@@ -24,6 +24,7 @@ DeviceDisabledScreen::DeviceDisabledScreen(
   DCHECK(view_);
   if (view_)
     view_->SetDelegate(this);
+  CHECK(device_disabling_manager_);

[emaxx, 2017/04/12 17:05:49]

I think DCHECK is more suitable there. See the Style Guide:
https://chromium.googlesource.com/chromium/src/+/master/styleguide/c++/c++.md...

[achuithb, 2017/04/12 19:45:14]

Personally I don't see the value of DCHECKing on this line when the next line
will crash anyway.

[Ivan Šandrk, 2017/04/13 13:37:59]

It wasn't really crashing on the next line, it was crashing deeper down the call
stack. I got a bit confused by this at first, and it took me some time to figure
it out (C++ doesn't crash on calling into nullptr object member functions if
they're not virtual, it crashed only when it tried to touch some data).

But if that's what you actually meant by your comment, I will remove it (I still
don't have a gut feeling for (D)CHECKs).

[achuithb, 2017/04/13 18:40:03]

It's fine if you want to leave in the DCHECK. In general, it's not hard to
figure out there's a null ptr dereference when you get a crash stack, and we
don't want to litter our code with DCHECKs before using any pointer.

   device_disabling_manager_->AddObserver(this);
 }