Bring RFH/RVH unit tests closer to reality of how RF/RV are initialized

content/browser/frame_host/render_frame_host_impl.cc

Index: content/browser/frame_host/render_frame_host_impl.cc
diff --git a/content/browser/frame_host/render_frame_host_impl.cc b/content/browser/frame_host/render_frame_host_impl.cc
index cc0ad50ff53128a6702981667a49762ad371388b..d6abcbbfa10a5605dedbdb59b600fba978dad436 100644
--- a/content/browser/frame_host/render_frame_host_impl.cc
+++ b/content/browser/frame_host/render_frame_host_impl.cc
@@ -645,12 +645,13 @@ bool RenderFrameHostImpl::CreateRenderFrame(int parent_routing_id,
 }
 
 bool RenderFrameHostImpl::IsRenderFrameLive() {
-  // RenderFrames are created for main frames at the same time as RenderViews,
-  // so we rely on IsRenderViewLive.  For subframes, we keep track of each
-  // RenderFrame individually with render_frame_created_.
-  bool is_live = !GetParent() ?
-      render_view_host_->IsRenderViewLive() :
-      GetProcess()->HasConnection() && render_frame_created_;
+  bool is_live = GetProcess()->HasConnection() && render_frame_created_;
+
+  // If the process is for an isolated guest (e.g. <webview>), rely on the
+  // RenderViewHost liveness check. Once https://crbug.com/492830 is fixed,
+  // this can be removed.
+  if (GetProcess()->IsIsolatedGuest())
+    is_live = render_view_host_->IsRenderViewLive();

[ncarter (slow), 2015/05/29 00:01:18]

If we're preserving the behavior of this function only for <webview>, does this
need to consider !GetParent() too?

[nasko, 2015/05/29 00:03:20]

Not really. We don't support any cross-process navigations in guests, so the
RenderViewHost is one for all frames. Any frame in such a process should key off
of the RenderViewHost.

 
   // Sanity check: the RenderView should always be live if the RenderFrame is.
   DCHECK(!is_live || render_view_host_->IsRenderViewLive());