Start using RenderFrameProxyHost objects.

content/browser/frame_host/navigation_controller_impl.cc

Index: content/browser/frame_host/navigation_controller_impl.cc
diff --git a/content/browser/frame_host/navigation_controller_impl.cc b/content/browser/frame_host/navigation_controller_impl.cc
index a25aeba2e7496d540593fba2622b239992547ad9..f676b61f7292ef09a497392c3d4f22c2248e9cbf 100644
--- a/content/browser/frame_host/navigation_controller_impl.cc
+++ b/content/browser/frame_host/navigation_controller_impl.cc
@@ -878,6 +878,13 @@ NavigationType NavigationControllerImpl::ClassifyNavigation(
     RenderFrameHost* rfh,
     const FrameHostMsg_DidCommitProvisionalLoad_Params& params) const {
   if (params.page_id == -1) {
+    // FIXME (nasko, creis):  An out-of-process child frame has no way of

[Charlie Reis, 2014/07/24 22:36:29]

We use TODO(...) rather than FIXME in the Chrome repo.

[kenrb, 2014/07/25 23:42:05]

Fixed.

+    // knowing the page_id of its parent, so it is passing back -1. The
+    // semantics here should be re-evaluated during session history refactor.

[Charlie Reis, 2014/07/24 22:36:29]

Can you add a mention of http://crbug.com/236848 to this sentence?

[kenrb, 2014/07/25 23:42:05]

Done.

+    // For now, we assume this means the child frame loaded and proceed.

[Charlie Reis, 2014/07/24 22:36:29]

"Note that this may do the wrong thing for cross-process AUTO_SUBFRAME
navigations."

[kenrb, 2014/07/25 23:42:05]

Done.

+    if (rfh->IsCrossProcessSubframe())
+      return NAVIGATION_TYPE_NEW_SUBFRAME;
+
     // The renderer generates the page IDs, and so if it gives us the invalid
     // page ID (-1) we know it didn't actually navigate. This happens in a few
     // cases: