broken i18n printing in gedit and gnumeric

[Tomas Pluskal <plusik pohoda cz>]

Hello,

I was trying to make gnumeric and gedit print Czech characters on FreeBSD
with no success, and after few hours of investigation I found out, that
both rely on mbstowcs() functions, which don't work as expected on
FreeBSD. In the i18n patch from Vlad Harchev, the function mbstowcs is
used to convert a string from current locale's encoding to unicode. This
works on linux (glibc), but I believe it is not correct (in the mean of
portability), because multibyte(3) functions are supposed to be used to
convert multibyte encodings (like Japanese) and not ISO8859-x charsets
etc.  In fact, on FreeBSD it doesn't convert current locale to unicode,
but ascii to unicode. I was consulting it on freebsd mailing lists and
this was the answer:

====cut================================================================

> I know that ISO8859-2 is not multibyte encoding, but in fact gnumeric,
> gedit (and I believe a lot of other software) expect the multibyte
> functions to work anyway (and to work as "translate characters from
> current locale's encoding to UNICODE" and reverse).

They are broken.  (Or perhaps they are relying on new behavior in C99,
but I rather doubt it.)

> Why does this work in linux ?

Because Linux (or rather, glibc) implements it the way these broken
programs expect.

-GAWollman

===cut=================================================================

Wouldn't it be possible to use another method to convert a string
in current locale's encoding to utf8 ? For example glib.h function:
g_locale_to_utf8() ?

Bye

Tomas Pluskal