Make UseCounter take a LocaFrame instead of any Frame

third_party/WebKit/Source/core/page/FrameTree.cpp

Index: third_party/WebKit/Source/core/page/FrameTree.cpp
diff --git a/third_party/WebKit/Source/core/page/FrameTree.cpp b/third_party/WebKit/Source/core/page/FrameTree.cpp
index 29d2a49089d485622075f03d89e80fc5cb4d1aeb..4091caebdf28cfa50a6123b0bf121c3e23911c3b 100644
--- a/third_party/WebKit/Source/core/page/FrameTree.cpp
+++ b/third_party/WebKit/Source/core/page/FrameTree.cpp
@@ -51,8 +51,14 @@ FrameTree::~FrameTree() {}
 const AtomicString& FrameTree::GetName() const {
   // TODO(andypaicu): remove this once we have gathered the data
   if (experimental_set_nulled_name_) {
-    UseCounter::Count(this_frame_.Get(),
-                      UseCounter::kCrossOriginMainFrameNulledNameAccessed);
+    const LocalFrame* frame =
+        this_frame_->IsLocalFrame()
+            ? ToLocalFrame(this_frame_)
+            : (Top().IsLocalFrame() ? ToLocalFrame(&Top()) : nullptr);
+    if (frame) {
+      UseCounter::Count(frame,
+                        UseCounter::kCrossOriginMainFrameNulledNameAccessed);

[dcheng, 2017/05/12 22:15:18]

I think we might want to keep this: can we cheat by using Page::GetUseCounter()
directly?

(I'm actually surprised at how few direct references there are to GetUseCounter)

[Rick Byers, 2017/05/25 15:32:31]

Unfortunately no, in order to migrate UseCounter handling to the browser process
I really need to fix all this "cheating" and associate a real LocalFrame with
each call to Count.  Fundamentally there's always some local frame responsible
for triggering usage, and we might want to treat the usage differently depending
on which frame triggered it.

That said, it's not that big of a deal at the moment to cheat by picking _some_
LocalFrame in the page, but I figured we wouldn't want to search the whole frame
tree to find one.  So in practice I figured it was probably fine to cheat by
picking either the current or top frame here when one was local.  That should
handle most of the cases I think.  

This especially didn't seem worth worrying about too much when the TODO implied
the measurement would likely go away soon.

+    }
   }
   return name_;
 }