Empty vector::push_back()?

[Gabriel Dos Reis]

Benjamin Kosnik <bkoz@redhat.com> writes:

| > I dunno :-)
| 
| > It isn't standard.
| 
| Maybe it should be removed then?

I contemplated that possibility, but I can't anticipate the amount of
code that will break and the impact it will have. On the other hand, I
find Phil's idea of having a hook into the diagnostic machinery
attractive -- even though I think we should not abuse of it.  

What is your take?

-- Gaby