Move window.close implementation to DOMWindow

Source/core/frame/RemoteFrame.cpp

Index: Source/core/frame/RemoteFrame.cpp
diff --git a/Source/core/frame/RemoteFrame.cpp b/Source/core/frame/RemoteFrame.cpp
index e2991b4b1bc4da0c691b993412b700e0d71e1361..f6256f4c3dd21d5ec92c9b6d794283d834073757 100644
--- a/Source/core/frame/RemoteFrame.cpp
+++ b/Source/core/frame/RemoteFrame.cpp
@@ -150,6 +150,13 @@ void RemoteFrame::createView()
     }
 }
 
+void RemoteFrame::close()
+{
+    // window.close only makes sense on a top-level frame.
+    if (!tree().parent())
+        remoteFrameClient()->close();

[dcheng, 2015/06/10 20:10:41]

LocalDOMWindow::close() has a bunch of other stuff. It seems like at least some
of these other things are relevant for RemoteDOMWindow too, right?

For example, respecting the "allow scripts to close windows" setting, setting
m_windowIsClosing, etc.

Also, can we just use ChromeClient::closeWindowSoon()?

[nasko, 2015/06/11 00:04:39]

I've added those to RemoteDOMWindow.
 
I'd rather not, as it goes over to the widget and for remote frames we don't
have a widget on the content/ side, just a proxy.

[dcheng, 2015/06/11 02:00:06]

Well... but we can change that to do what we need... the important part is
ChromeClient lets us get to the WebViewClient.

Logically, there's no real difference between closing a LocalDOMWindow and a
RemoteDOMWindow. Closing either one ultimately still requires the browser to do
things for us.

And if we do use ChromeClient::closeWindowSoon() here, I think we might be able
to just move the entire function up into DOMWindow, eliminating the duplicated
checks, etc.

+}
+
 RemoteFrameClient* RemoteFrame::remoteFrameClient() const
 {
     return static_cast<RemoteFrameClient*>(client());