Re: how to test the new glibc-2.2's UTF-8 locales

[Edmund GRIMLEY EVANS <edmundo@xxxxxxxx>]

Bruno Haible <haible@xxxxxxx>:

>   - gettext automatically converts the translations to the current locale's
>     encoding.

Does the header, gettext(""), still contain the name of the original
encoding, or is that translated too?

If it is left unchanged, then presumably my program which applies
charset conversion to the output of gettext won't work with glibc-2.2.
What's the best way of making it work with any C library? An autoconf
test? Does anyone already have a recipe, or hints?

> > What happens when gettext tries to automatically convert, but conversion
> > fails (because transliteration would be needed)?  Does gettext fall back
> > to the untranslated string, or is there some other recovery?
> 
> It falls back to the translated, but un-iconv-ed string.

Presumably transliteration would be better. Can this be changed?

If a Russian message in UTF-8 happens to contain some fancy quotation
mark that KOI8-R doesn't have, then the result is likely to be fairly
unreadable if the message is spat out unconverted in a KOI8-R locale.

This reminds me of a question I haven't had answered yet: what does
iconv do in the worst case, e.g. if it has to transliterate a Chinese
character into ASCII? Does it convert to '?', or what?

According to the spec, iconv has to do some kind of conversion, even
when no sensible conversion is possible.

Edmund
-
Linux-UTF8:   i18n of Linux on all levels
Archive:      http://mail.nl.linux.org/lists/