Store the last committed origin in RenderFrameHostImpl.

content/browser/frame_host/navigator_impl.cc

Index: content/browser/frame_host/navigator_impl.cc
diff --git a/content/browser/frame_host/navigator_impl.cc b/content/browser/frame_host/navigator_impl.cc
index 44873594afe061d98414abd3f993b87c5ff7d9de..9d14f3417000c23886df9aa068235ae058c64eff 100644
--- a/content/browser/frame_host/navigator_impl.cc
+++ b/content/browser/frame_host/navigator_impl.cc
@@ -640,10 +640,12 @@ void NavigatorImpl::DidNavigate(
         site_instance);
   }
 
-  // Keep track of each frame's URL in its FrameTreeNode, whether it's for a net
-  // error or not.
-  // TODO(creis): Move the last committed URL to RenderFrameHostImpl.
+  // Keep track of the last committed URL and origin in the RenderFrameHost
+  // itself.  These allow GetLastCommittedURL and GetLastCommittedOrigin to
+  // stay correct even if the render_frame_host later becomes pending deletion.
+  // The URL is set regardless of whether it's for a net error or not.
   render_frame_host->frame_tree_node()->SetCurrentURL(params.url);
+  render_frame_host->set_last_committed_origin(params.origin);

[alexmos, 2016/12/06 17:22:40]

I originally tried to do this inside SetCurrentOrigin, which matches what we did
for SetCurrentURL.  But that didn't work -- SetCurrentOrigin is called above
this, intentionally before CommitPending when oopifs are on, so that the
replication state already has the latest origin when it's sent with
FrameMsg_SwapOut.  But at that point, current_frame_host() is still wrong, so
FTN::SetCurrentOrigin can't use it to update the last committed origin.

I didn't do this next to SetCurrentOrigin above, since I thought it's better to
set this on the RFH after it really becomes the current RFH.

[Charlie Reis, 2016/12/08 19:51:47]

Interesting.  Yes, I agree with your decisions.

 
   // Separately, update the frame's last successful URL except for net error
   // pages, since those do not end up in the correct process after transfers