Remove CHECKs from WebContentsImpl::OnDocumentLoadedInFrame

content/browser/web_contents/web_contents_impl.cc

Index: content/browser/web_contents/web_contents_impl.cc
diff --git a/content/browser/web_contents/web_contents_impl.cc b/content/browser/web_contents/web_contents_impl.cc
index ff003ddb3c808f2ecd7ce45abb86207620d9d2bd..9b443fc7460fcff4e7e2f93154fd09989d991012 100644
--- a/content/browser/web_contents/web_contents_impl.cc
+++ b/content/browser/web_contents/web_contents_impl.cc
@@ -2763,8 +2763,16 @@ void WebContentsImpl::OnDidRunInsecureContent(
 }
 
 void WebContentsImpl::OnDocumentLoadedInFrame() {
-  CHECK(render_frame_message_source_);
-  CHECK(!render_view_message_source_);
+  if (!render_frame_message_source_) {
+    RecordAction(base::UserMetricsAction("BadMessageTerminate_RVD2"));

[Charlie Reis, 2015/01/27 00:18:03]

Why RVD2?  I thought we usually abbreviated the class name that was triggering
this.  (I'm wondering if we should fix the case below as well or just leave
it...)

[nasko, 2015/01/27 01:29:37]

Fixed both. I forgot it on the first iteration.

+    // TODO(nasko): With --site-per-process, subframes can be in a different
+    // process from the main frame and this will kill the wrong process.
+    // Once RenderViewHost is not used to route messages, this should not
+    // happen.

[Charlie Reis, 2015/01/27 00:18:03]

I don't think this TODO is necessary.  "!render_frame_message_source_" implies
that render_view_message_source_ is valid (due to the DCHECK at the start of
OnMessageReceived), so the message would have come from the RVH.  That implies
that GetRenderProcessHost is the right thing to kill.

We can probably just drop the comment.

[nasko, 2015/01/27 01:29:36]

Done.

+    GetRenderProcessHost()->ReceivedBadMessage();
+    return;
+  }
+
   RenderFrameHostImpl* rfh =
       static_cast<RenderFrameHostImpl*>(render_frame_message_source_);
   FOR_EACH_OBSERVER(