Don't update subframes on parent frame commits.

content/browser/frame_host/navigation_controller_impl.cc

 // Copyright 2013 The Chromium Authors. All rights reserved.
 // Use of this source code is governed by a BSD-style license that can be
 // found in the LICENSE file.
 
 /*
  * Copyright (C) 2006, 2007, 2008, 2009 Apple Inc. All rights reserved.
  * Copyright (C) 2008 Nokia Corporation and/or its subsidiary(-ies)
  * Copyright (C) 2008, 2009 Torch Mobile Inc. All rights reserved.
  *     (http://www.torchmobile.com/)
  *
@@ skipping 1860 matching lines @@
   // sequence number in the same SiteInstance. Newly restored items may not have
   // a SiteInstance yet, in which case it will be assigned on first commit.
   if (!old_item ||
       new_item->item_sequence_number() != old_item->item_sequence_number() ||
       (new_item->site_instance() != nullptr &&
        new_item->site_instance() != old_item->site_instance())) {
     if (old_item &&
         new_item->document_sequence_number() ==
             old_item->document_sequence_number()) {
       same_document_loads->push_back(std::make_pair(frame, new_item));
+      // This is a bug; we should not return here. Rather, we should continue on

[Charlie Reis, 2016/05/05 17:45:42]

nit: Blank line before, and start with TODO(avi, creis).  (It's not necessarily
something you personally have to fix, but you're knowledgable about it.)

[Avi (use Gerrit), 2016/05/05 19:06:21]

Done.

+      // and navigate all child frames which have also changed. This bug is the
+      // cause of <https://crbug.com/542299>.
+      //
+      // However, this bug masks a deeper and worse problem. Doing a pushState
+      // immediately after loading a subframe is a race, one that no web page
+      // author expects. If we fix this bug, many large websites break. For
+      // example, see <https://crbug.com/598043> and the spec discussion at

[Charlie Reis, 2016/05/05 17:45:42]

Let's also mention 600534.

[Avi (use Gerrit), 2016/05/05 19:06:21]

That's a Google-only bug; do we want to mention it?

[Charlie Reis, 2016/05/05 19:34:22]

Ah.  No, just listing it in the CL description should be sufficient then.

+      // <https://github.com/whatwg/html/issues/1191>.
+      //
+      // For now, we accept this bug, and hope to resolve the race in a
+      // different way that will one day allow us to fix this.
+      return;
     } else {
       different_document_loads->push_back(std::make_pair(frame, new_item));
       // For a different document, the subframes will be destroyed, so there's
       // no need to consider them.
       return;
     }
   }
 
   for (size_t i = 0; i < frame->child_count(); i++) {
     FindFramesToNavigate(frame->child_at(i), same_document_loads,
@@ skipping 162 matching lines @@
     }
   }
 }
 
 void NavigationControllerImpl::SetGetTimestampCallbackForTest(
     const base::Callback<base::Time()>& get_timestamp_callback) {
   get_timestamp_callback_ = get_timestamp_callback;
 }
 
 }  // namespace content