Notify OverscrollController of gesture events in plugins.

chrome/browser/renderer_host/chrome_render_widget_host_view_mac_history_swiper.mm

Index: chrome/browser/renderer_host/chrome_render_widget_host_view_mac_history_swiper.mm
diff --git a/chrome/browser/renderer_host/chrome_render_widget_host_view_mac_history_swiper.mm b/chrome/browser/renderer_host/chrome_render_widget_host_view_mac_history_swiper.mm
index ca8813508c85ed2c8627a4db5a416f9eb8ad4c6c..ccb342db85e495740bb28dc4488526c80877458d 100644
--- a/chrome/browser/renderer_host/chrome_render_widget_host_view_mac_history_swiper.mm
+++ b/chrome/browser/renderer_host/chrome_render_widget_host_view_mac_history_swiper.mm
@@ -131,6 +131,14 @@ BOOL forceMagicMouse = NO;
               blink::WebGestureEvent::kMomentumPhase) {
         return;
       }
+      // GestureScrollBegin and GestureScrollEnd events are created to wrap
+      // individual resent GestureScrollUpdates from a plugin. Hence these
+      // should not be used to indicate the beginning/end of the swipe gesture.
+      // TODO(mcnee): When we remove BrowserPlugin, delete this code.
+      // See crbug.com/533069
+      if (event.resending_plugin_id != -1) {

[erikchen, 2017/05/05 19:15:36]

this class is a state machine, and I'm a little bit worried about it ending in
an inconsistent state if there's another layer on top of it consuming and/or
generating events. Can you confirm that we correctly propagate state for the
gesture and touch APIs, e.g. beginGestureWithEvent: and touchesBeganWithEvent:

[Kevin McNee, 2017/05/16 16:24:23]

With a guest, the sequence of calls to touches*WithEvent and *GestureWithEvent
is the same as with a normal page. Looking at a stack trace, these events are
not coming up from the guest.

+        return;
+      }
       waitingForFirstGestureScroll_ = YES;
       break;
     case blink::WebInputEvent::kGestureScrollUpdate: