BUG in spanish (es_ES) locale date

[German Gomez Garcia]

Hello,

We were having problems with PostgresSQL ordering of strings, and
tracing it back we have found that there is a bug in the spanish (es_ES)
locale data. 

Extracted from localedata/locales/es_ES

% Ordering algorithm:
%  1. Spaces and hyphen (but not soft
%     hyphen) before punctuation
%     characters, punctuation characters
%     before numbers,
%     numbers before letters.                                                   
                                                                                
                                  
%  2. Letters with diacritical marks are                                        
                                                                                
                                  
%     members of equivalence classes
%  3. A distinction is made with regards
%     to case as noted below.
%  4. Special characters are ignored
%     when comparing letters, but then
%     they are considered
%  5. The alphabets are sorted in order
%     of first appearance in ISO 10646:
%     Latin, Greek, Cyrillic, etc.

But when in fact spaces are not considered where ordering, it seems that
they are just ignored. Although the collation scheme is clearly about
spaces < punct. < number < letters. The attached program is an example
of what I mean, it uses strcoll that is supposed to use locales when
sorting. I've compiled it both in Debian Woody, Debian Sid, Redhat 7.2
and Redhat 9.0, the locales are correctly generated but it isn't
working.

This is a really anoying bug, I've checked glibc-2.3.2, maybe in the CVS
is already fixed?

I have doublechecked it in Microsoft Visual C++ 6.0 with the same
program, and it is working correctly there.

Regards,

-- 
German Jose Gomez Garcia  | Mundinteractivos - El Mundo      |
Area de Internet          | Pradillo, 42                     |
address@hidden       | 28002 - Madrid (SPAIN, EU)       |
http://www.elmundo.es/    | Tel: (+34) 915864800 (Ext: 4971) |