Test for BrowserPlugin-based WebView Focus

content/browser/frame_host/render_widget_host_view_child_frame.h

Index: content/browser/frame_host/render_widget_host_view_child_frame.h
diff --git a/content/browser/frame_host/render_widget_host_view_child_frame.h b/content/browser/frame_host/render_widget_host_view_child_frame.h
index 67f11be261aa0c7ad143e576233cf0bc74f63178..c46ee65d8a58a93723c2ffb20924fc406e8efaa0 100644
--- a/content/browser/frame_host/render_widget_host_view_child_frame.h
+++ b/content/browser/frame_host/render_widget_host_view_child_frame.h
@@ -189,6 +189,8 @@ class CONTENT_EXPORT RenderWidgetHostViewChildFrame
   void RegisterSurfaceNamespaceId();
   void UnregisterSurfaceNamespaceId();
 
+  uint32_t surface_frame_count() const { return surface_frame_count_; }
+
  protected:
   friend class RenderWidgetHostView;
   friend class RenderWidgetHostViewChildFrameTest;
@@ -216,6 +218,7 @@ class CONTENT_EXPORT RenderWidgetHostViewChildFrame
   float current_surface_scale_factor_;
   gfx::Rect last_screen_rect_;
   uint32_t ack_pending_count_;
+  uint32_t surface_frame_count_;

[Charlie Reis, 2016/07/06 18:28:49]

Seems mildly unfortunate to have to track this in all builds just to support
tests.  That said, I'd much rather have tests than not add this.  :)

Just thought I'd check if there's another way to do the waiting without adding
this counter.  If not, I'm ok with it.

[wjmaclean, 2016/07/06 19:49:31]

Agreed. kenrb@ and I haven't thought of another way, and even if it was, it
would likely be an observer or some such thing that would add at least as much
state/overhead as this.

[Charlie Reis, 2016/07/06 20:17:13]

Can we make it a bool instead of a counter?  The only use of it is to compare it
to zero, so something like has_received_surface_frame_ might be simpler.  (Or
are there plans to compare against other values soon?)

   cc::ReturnedResourceArray surface_returned_resources_;
 
   // frame_connector_ provides a platform abstraction. Messages