I wrote:
> (I do agree that that message isn't the most helpful thing. It looks
> like chklocale.c is overoptimistically assuming that what it's handed
> is valid, even if GetLocaleInfoEx says it isn't.)
After further digging, that optimism isn't *completely* without
foundation; it seems to be based on the fact that we know setlocale()
thought the locale string was OK. Which is interesting --- apparently
Microsoft is more willing to accept Unix-style locale names than
I thought. But they couldn't be bothered to make GetLocaleInfoEx()
and setlocale() take the same set of strings ...
regards, tom lane