Collation discrepencies between locales [was: Re: Bits from the release team: full steam ahead towards buster]

[Clément Hermann <nodens@nodens.org>]

On 19/04/2018 22:45, Holger Levsen wrote:
> I now wondered if it's not only en_GB.utf8 which is "different", but also
> the NZ and US variants sort like that (and so differently than C)... not 
> sure if en_FR.utf8 exist, but using it, it sorts differently / like C ;)
> 
> (probably because it doesnt exist, thus the default, C, is used.)

Indeed, it doesn't exist. At least , for fr_* locale, it seems to be
consistent both in the different charsets available (e.g. fr_FR and
fr_FR.UTF-8) and country (fr_BE, fr_CA, fr_CH, fr_FR and fr_LU).

Actually I thought the localization had been made consistently with the
apparition of unicode locales, that is, fr_* locale would all give the
same result regardless of the charset (but older fr_FR for instance
might give a different order than before the apparition of the unicode
variant). I may be wrong - one would probably have to check the code in
GNU libc to be sure.

However, Note that the generation of locale matters: at first I thought
fr_FR and fr_BE where behaving differently, but after uncommenting all
fr_* locales in /etc/locale.gen, everything became consistent.


Cheers,

-- 
nodens