Add test for filtering of IPC messages when RenderFrameHost is swapped out.

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 d697e186d0c4dae004a901635c31d549514bffb9..33ef7fd8ac51ccccd0e5df51beb6785422e05e86 100644
--- a/content/browser/frame_host/render_frame_host_manager.cc
+++ b/content/browser/frame_host/render_frame_host_manager.cc
@@ -468,10 +468,14 @@ void RenderFrameHostManager::DidNavigateMainFrame(
     // The pending cross-site navigation completed, so show the renderer.
     // If it committed without sending network requests (e.g., data URLs),
     // then we still need to swap out the old RFH first and run its unload
-    // handler.  OK for that to happen in the background.
+    // handler.  OK for that to happen in the background, only if it hasn't
+    // happened yet.

[Charlie Reis, 2014/03/21 21:35:13]

Please move "if it hasn't happened yet" to the end of the previous sentence,
since it's about whether we need to run the unload handler, not whether it runs
in the background.

[nasko, 2014/03/24 17:48:11]

Done.

     if (pending_render_frame_host_->render_view_host()->
-            HasPendingCrossSiteRequest())
+            HasPendingCrossSiteRequest() &&
+        pending_render_frame_host_->render_view_host()->rvh_state() !=
+            RenderViewHostImpl::STATE_WAITING_FOR_COMMIT) {

[Charlie Reis, 2014/03/21 21:35:13]

What if it were STATE_WAITING_FOR_UNLOAD_ACK?  That means we've already called
SwapOut and don't need to again.  Presumably the same thing for PENDING_SWAP_OUT
and PENDING_SHUTDOWN, right?

Should this check be rvh_state() == STATE_DEFAULT?

(Edit: I think so, since that's what we're checking for in RVHI::SwapOut.)

[nasko, 2014/03/24 17:48:11]

Done.

       SwapOutOldPage();
+    }
 
     CommitPending();
     cross_navigation_pending_ = false;