bug#17056: dfa.c patch for systems with no locale support

[Aharon Robbins]

Me: 
> >> What if
> >> the system's setlocale() actually returns NULL?

Paul:
> > Then we don't know what the locale is.  Perhaps setlocale ran out of
> > memory and so returned NULL.  If so, subsequent calls might use some
> > weird Turkish locale with multibyte characters.  So it sounds safer to
> > assume the worst in that case.

Eric:
> This is in main().  POSIX guarantees that until setlocale() is called
> for the first time, we are in the C locale.  If you were in a library, I
> could buy the argument of setlocale() returning NULL as indicative of
> some rare error.  But in main(), where you are the first call, the only
> thing that a NULL return implies is that your attempt to change the
> locale had no effect, so it remains at the locale it was before, which
> is the C locale since all programs start in the C locale.

So, are we for or against my suggested change?  Sounds like "for",
but I'm not sure.

Thanks,

Arnold