Re: Coreutils 5.0.1: spurious error from uniq

[Paul Eggert]

Andreas Schwab <address@hidden> writes:

> LC_CTYPE=de_DE.UTF-8
> LC_TIME=POSIX
> LC_COLLATE=POSIX
> 
> This is fine-tuned to DTRT(tm) for my needs.  Maybe the LC_CTYPE and/or
> LC_COLLATE settings are confusing glibc.  I'll check that, and see if
> setting them to en_US.UTF-8 helps.

POSIX says:

  If different character sets are used by the locale categories, the
  results achieved by an application utilizing these categories are
  undefined.

  <http://www.opengroup.org/onlinepubs/007904975/basedefs/xbd_chap07.html>

So you're not allowed to mix LC_COLLATE=POSIX with LC_CTYPE=de_DE.UTF-8,
and the results are undefined if you do.  Similarly for LC_TIME.

Also, perhaps the "locale" command should warn if it discovers
incompatible character sets in the various locale categories.

Do you have a "POSIX.UTF-8" locale?  That should fix your problem too.

Anyway, it's still possible that 'uniq' has a bug, or perhaps
'strcoll', depending on your further investigation.