MacViews: Fix ViewTest.HandleAccelerator by faking window activation

ui/views/test/views_test_helper_mac.mm

Index: ui/views/test/views_test_helper_mac.mm
diff --git a/ui/views/test/views_test_helper_mac.mm b/ui/views/test/views_test_helper_mac.mm
index 8ef73d27487d7106d4acc9b2bac3b4c38eab21ea..7a7df511dfd5c2949ddc5ca282b56a2b0342db8d 100644
--- a/ui/views/test/views_test_helper_mac.mm
+++ b/ui/views/test/views_test_helper_mac.mm
@@ -7,6 +7,7 @@
 #import <Cocoa/Cocoa.h>
 
 #import "base/mac/scoped_nsautorelease_pool.h"
+#include "ui/base/test/scoped_fake_nswindow_focus.h"
 #include "ui/compositor/scoped_animation_duration_scale_mode.h"
 #include "ui/views/test/event_generator_delegate_mac.h"
 #include "ui/views/widget/widget.h"
@@ -46,4 +47,8 @@ void ViewsTestHelperMac::TearDown() {
     DCHECK(!Widget::GetWidgetForNativeWindow(window)) << "Widget not closed.";
 }
 
+void ViewsTestHelperMac::EnableFakeWindowActivation() {
+  faked_focus_.reset(new ui::test::ScopedFakeNSWindowFocus);

[sky, 2016/03/10 18:59:16]

Is there a reason not to do this always? By that I mean move this code to the
constructor?

[tapted, 2016/03/11 09:56:19]

Done (see what you think).

This has actually crossed my mind a few times already.. I don't think
interactive_ui_tests should get it so there's some non-fake coverage of
activation flows. Also at least one  test --
NativeWidgetMacInteractiveUITest.ParentWindowTrafficLights -- would regress
since there's some native behaviour that we can't fake. But apart from that, I
think I was just being conservative.

So I've explored how this would look, including logic to prevent the faker
getting installed for interactive_ui_tests - ptal.

+}
+
 }  // namespace views