Convert HTMLFrameOwnerElement and FocusController to use Frame.

Source/bindings/v8/V8WindowShell.cpp

Index: Source/bindings/v8/V8WindowShell.cpp
diff --git a/Source/bindings/v8/V8WindowShell.cpp b/Source/bindings/v8/V8WindowShell.cpp
index 17e19be2a8dcf1e1a94c412dbe2a4eec65163c99..25a298c1bd0a775cc46eb625a929d9be92f7e40c 100644
--- a/Source/bindings/v8/V8WindowShell.cpp
+++ b/Source/bindings/v8/V8WindowShell.cpp
@@ -432,7 +432,7 @@ static v8::Handle<v8::Value> getNamedProperty(HTMLDocument* htmlDocument, const
     if (items->hasExactlyOneItem()) {
         Element* element = items->item(0);
         LocalFrame* frame = 0;
-        if (element->hasTagName(HTMLNames::iframeTag) && (frame = toHTMLIFrameElement(element)->contentFrame()))
+        if (element->hasTagName(HTMLNames::iframeTag) && (frame = toLocalFrame(toHTMLIFrameElement(element)->contentFrame())))

[eseidel, 2014/03/07 22:42:53]

Confused.  This isn't always true, is it?  Shouldn't we check to see that the
iframe has a real local frame?  I presumed that would be the point of using the
compiler to do this, that it would force us to audit all callers?

[kenrb, 2014/03/07 23:56:45]

Sorry about that, I had added a bunch of checks but they were on the wrong
branch.

In general, my intent was to just use toLocalFrame() on code paths where we
should only get LocalFrames, and do a check on paths where we might have to
handle both. toLocalFrame() contains an implicit assertion that we are right.
Unfortunately we aren't far enough along to instantiate a RemoteFrame and test
to see which asserts fire.

Also, in the places where we check isLocalFrame() but have to handle both types,
the correct behavior for RemoteFrames is not always obvious.

             return toV8(frame->domWindow(), creationContext, isolate);
         return toV8(element, creationContext, isolate);
     }