Added counters for pointer boundary events fired.

third_party/WebKit/Source/core/events/EventTarget.cpp

Index: third_party/WebKit/Source/core/events/EventTarget.cpp
diff --git a/third_party/WebKit/Source/core/events/EventTarget.cpp b/third_party/WebKit/Source/core/events/EventTarget.cpp
index d98bb138c288ff47c9a2f04c3a6e7e6343915351..c8de656eea1f2ad817b9e716246152ef448a1d04 100644
--- a/third_party/WebKit/Source/core/events/EventTarget.cpp
+++ b/third_party/WebKit/Source/core/events/EventTarget.cpp
@@ -588,6 +588,21 @@ bool EventTarget::fireEventListeners(Event* event, EventTargetData* d, EventList
             if (event->isPointerEvent() && static_cast<PointerEvent*>(event)->pointerType() == "touch")
                 UseCounter::count(executingWindow->document(), UseCounter::PointerDownFiredForTouch);
         }
+    } else if (checkTypeThenUseCount(event, EventTypeNames::pointerenter, UseCounter::PointerEnterLeaveFired)
+        || checkTypeThenUseCount(event, EventTypeNames::pointerleave, UseCounter::PointerEnterLeaveFired)
+        || checkTypeThenUseCount(event, EventTypeNames::pointerover, UseCounter::PointerOverOutFired)
+        || checkTypeThenUseCount(event, EventTypeNames::pointerout, UseCounter::PointerOverOutFired)) {
+        LocalDOMWindow* executingWindow = this->executingWindow();
+        Node* node = toNode();
+        if (executingWindow && node && node->getNodeType() == Node::kElementNode && event->isPointerEvent()) {
+            const Element* element = static_cast<Element*>(node);
+            const PointerEvent* pointerEvent = static_cast<PointerEvent*>(event);
+            const UseCounter::Feature feature = (event->type() == EventTypeNames::pointerenter || event->type() == EventTypeNames::pointerleave)
+                ? UseCounter::PointerEnterLeaveFiredWhileCaptured
+                : UseCounter::PointerOverOutFiredWhileCaptured;
+            if (element->hasPointerCapture(pointerEvent->pointerId()))
+                UseCounter::count(executingWindow->document(), feature);

[Navid Zolghadr, 2016/08/24 21:18:01]

This probably won't work as you expect. In the normal case (the way right now
Chrome implements) after setPointerCapture (which makes hasPointerCapture return
true) there might be a boundary event when then pointer capture is processed.
But that is different from the boundary events that we suppress while a pointer
is captured. Do we want to make a distinction there?

[mustaq, 2016/08/25 15:00:32]

Yes, good catch: this will trigger when set/release capture is called for an
element that is not currently under the pointer. And this is true in different
ways for both our model of PointerEvent (no boundary event while captured) and
the Edge model.

I will add an API "hasProcessedPointerCapture". I will mark it as temporary
because of the confusion it would add to the API surface.

[mustaq, 2016/08/25 15:45:54]

Done, ptal. I didn't mark the new API as temporary because it could have other
uses.

+        }
     }
 
     ExecutionContext* context = getExecutionContext();