Add hittest mode for Touch-action which ignore inline elements and svg elements

Source/core/rendering/RenderObject.h

Index: Source/core/rendering/RenderObject.h
diff --git a/Source/core/rendering/RenderObject.h b/Source/core/rendering/RenderObject.h
index 7fd2dfebb7a2ebf2c8ebef682eeffd473a9b1c3f..3500aafbfd7ea0539c1db42011a6d9c53af31bef 100644
--- a/Source/core/rendering/RenderObject.h
+++ b/Source/core/rendering/RenderObject.h
@@ -922,7 +922,13 @@ public:
     void remove() { if (parent()) parent()->removeChild(this); }
 
     bool isInert() const;
-    bool visibleToHitTestRequest(const HitTestRequest& request) const { return style()->visibility() == VISIBLE && (request.ignorePointerEventsNone() || style()->pointerEvents() != PE_NONE) && !isInert(); }
+    virtual bool visibleToHitTestRequest(const HitTestRequest& request) const
+    {
+        if (request.touchAction())

[leviw_travelin_and_unemployed, 2014/02/28 18:14:39]

Instead of making this virtual and adding a branch for touchAction to every
implementation, why not keep the single action and add a method like
"visibleForTouchAction"? Then you just need 4 virtual methods that just return
true or false with no logic, and the logic here would just be "if
(request.touchAction() && visibleForTouchAction())" That seems way clearer than
multipl visibleToHitTestRequest methods with logic in them.

[gnana, 2014/03/03 07:42:31]

Done.

+            return false;
+        return style()->visibility() == VISIBLE && (request.ignorePointerEventsNone() || style()->pointerEvents() != PE_NONE) && !isInert();
+    }
+
     bool visibleToHitTesting() const { return style()->visibility() == VISIBLE && style()->pointerEvents() != PE_NONE && !isInert(); }
 
     // Map points and quads through elements, potentially via 3d transforms. You should never need to call these directly; use