Re: New uncompatible locale system with glibc2.2 ?
Raphael Hertzog <rhertzog@hrnet.fr> writes:
> Le Thu, Oct 12, 2000 at 05:50:38AM +0900, Changwoo Ryu écrivait:
> > > I can't believe recompiling makes them work.. It's not such a
> > > problem.
>
> Recompiling doesn't help, you're right.
>
> > > I don't know which one is buggy, glibc or the programs. glibc does
> > > not handle this situation. But anyway LC_MESSAGES depends on LC_CTYPE
> > > in some way.
> >
> > Now I believe glibc is buggy...
>
> Yeah, and I reopened #72547 http://bugs.debian.org/72547 since the hack
> mentionned there does work... it seems that glibc misinterpret some
> information contained in the headers of the .mo files (in particular
> Content-Type).
It's not a misinterpretation. glibc *exactly* knows the translated
messages are ISO-8859-[12] encoded, by the "Content-Type" header. But
glibc (a) converts them into Unicode texts once, then (b) converts
them again into the charset of the current locale.
The only problem is that glibc doesn't know about the current charset
at (b), when setlocale() is not called. "C" locales is assumed in
this case. So all non-ASCII Unicode characters are printed as
question marks.
--
Changwoo Ryu
Reply to: