Peter Eisentraut <peter_e@gmx.net> writes:
> When initdb is given an invalid (possibly mistyped) locale name, it just
> prints a warning and proceeds with the default locale from the
> environment. Someone already wondered about this before:
> /* should we exit here? */
> if (!ret)
> fprintf(stderr, _("%s: invalid locale name \"%s\"\n"), progname,
> locale);
> I obviously think we should. Why shouldn't we?
That would make it impossible to install at all on a machine with broken
locale support, which seems a bit of an overreaction, especially for the
noncritical locale items like LC_MESSAGES (which also happens to be the
one that fails most often). I could go with erroring for LC_COLLATE and
LC_CTYPE, but I'd prefer to see us fall back to "C" for any of the other
LC_ items that fail. Those guys are easy to fix after the fact, if the
user cares enough (or at all).
BTW, it would also be a good idea if the message specified which LC_
item we failed to set.
regards, tom lane
