Fix for the NavigationControllerTest.BackSubframe test failures with PlzNavigate

content/browser/frame_host/frame_navigation_entry.cc

Index: content/browser/frame_host/frame_navigation_entry.cc
diff --git a/content/browser/frame_host/frame_navigation_entry.cc b/content/browser/frame_host/frame_navigation_entry.cc
index bf78bef31f702fbaab573adefe8f1bf4ea08f3ec..71974aab665ed67e22eb596aa6449c1d54d8f312 100644
--- a/content/browser/frame_host/frame_navigation_entry.cc
+++ b/content/browser/frame_host/frame_navigation_entry.cc
@@ -90,8 +90,18 @@ void FrameNavigationEntry::SetPageState(const PageState& page_state) {
   if (!DecodePageState(page_state_.ToEncodedData(), &exploded_state))
     return;
 
-  item_sequence_number_ = exploded_state.top.item_sequence_number;
-  document_sequence_number_ = exploded_state.top.document_sequence_number;
+  // If the item_sequence_number_ field is already populated, then set it only
+  // if the item sequence number decoded from the page state is valid. This
+  // typically happens in tests.
+  if (exploded_state.top.item_sequence_number > 0 ||

[nasko, 2016/09/07 20:56:02]

I don't think this is the correct approach. The whole reason for this method is
to avoid mismatches between item_sequence_number_ and
exploded_state.top.item_sequence_number, where this CL is adding ways for this
to happen. I would prefer we fix the unit testing code to behave closer to
reality than to add the code here.

[ananta, 2016/09/07 22:50:06]

Added the functionality to the test code. PTAL

+      item_sequence_number_ < 0) {
+    item_sequence_number_ = exploded_state.top.item_sequence_number;
+  }
+  // Please refer to the comment above for explanation.
+  if (exploded_state.top.document_sequence_number > 0 ||
+      document_sequence_number_ < 0) {
+    document_sequence_number_ = exploded_state.top.document_sequence_number;
+  }
 }
 
 scoped_refptr<ResourceRequestBodyImpl> FrameNavigationEntry::GetPostData()