[PATCH] Fix PR 23525 - Separate header mass propagation in irregular loops.

[Duncan P. N. Exon Smith]

> On 2015-Jun-16, at 12:16, Diego Novillo <dnovillo at google.com> wrote:
> 
> I fixed up the patch and committed it at r239843.
> 
> I'll continue working on this to use incoming edge masses. Duncan, please let me know if you notice anything else you'd like me to change as well.

I'd rather you hadn't committed until I gave an LGTM (although I
realize you may have just misread my SGTM).  No need to revert --
the churn doesn't seem worth it since the code changes look correct
-- but have a look at my review comments in the thread for r239843.
Mostly documentation and testcase changes.