Enable swapping a frame back in to its parent process

content/browser/frame_host/render_frame_host_manager.cc

Index: content/browser/frame_host/render_frame_host_manager.cc
diff --git a/content/browser/frame_host/render_frame_host_manager.cc b/content/browser/frame_host/render_frame_host_manager.cc
index 39799b9c099cbc9297567e14ea61886f554c1ae8..18f01180fcce5ea8d895e817af3c7563cef4f583 100644
--- a/content/browser/frame_host/render_frame_host_manager.cc
+++ b/content/browser/frame_host/render_frame_host_manager.cc
@@ -932,7 +932,7 @@ SiteInstance* RenderFrameHostManager::GetSiteInstanceForURL(
   // to open a new tab to an interstitial-inducing URL, and then navigates
   // the page to a different same-site URL.  (This seems very unlikely in
   // practice.)
-  const GURL& current_url = (current_entry) ? current_entry->GetURL() :
+  const GURL& current_url = (current_entry) ? frame_tree_node_->current_url() :

[Charlie Reis, 2014/10/28 16:38:29]

I'm a little nervous about this because I don't trust
frame_tree_node_->current_url() as much as current_entry->GetURL().  I think
they should be the same (for main frames), though, and I can see that this is
needed for subframes until we get FrameNavigationEntries.

It's independent of the rest of this CL, so let's split it out into a different
CL with a CHECK_EQ of the two URLs in the main frame case, just to give
confidence that there's no behavior change there.  (Otherwise this is a
dangerous place to introduce regressions.)

[Nate Chapin, 2014/10/28 19:59:19]

Ok, will split out and land that first.

       current_instance->GetSiteURL();
 
   // View-source URLs must use a new SiteInstance and BrowsingInstance.