That makes sense. Thanks for the update.

After upgrading php from 4.x to 5.1.2 the problem disappeaded. As php 
4.x installations are slowly fading so maybe we can ignore the 
problem. This also suggest the problem isn't in libc but in php.

So, does anybody have an idea what or if we could do something about 
this? Or do we have to accept that Finnish locales seem to be broken 
in libc?

Using the examples the resulting separator is (in hexcode):

locale fi_FI: 0xa0

locale fi_FI.UTF-8: 0xc2 0xa0



The character is non-breakable space.

Try this instead (in a console):



<?php

setlocale(LC_ALL, 'fi_FI.UTF-8');

$localeinfo = localeconv();

var_dump($localeinfo['thousands_sep']);



and add header('text/plain; charset="utf-8"') if viewing in a browser.

Weird - I can't reproduce this on my system.  Running this PHP script:



<?php

setlocale(LC_ALL, 'fi_FI');

$localeinfo = localeconv();

var_dump($localeinfo['thousands_sep']);



Gives me:



string(1) " "

It looks like the system's localeconv is returning separators in the 
wrong charset. I can reproduce it, but there is probably nothing we 
can do about this, because we already explicitely request the 
separator characters in the interface's charset.

I was not able to determine which charset the separator has in fact, 
definitely not iso-8859-1.

Start a new message and add an attachment that is at least 1 MB in 
size. The English interface shows the size of the file correctly but 
in Finnish interface the character between 1 and 024 is a non-utf-8 
character.



English:

1. 1M  (application/x-executable-file)   Size: 1,024.00 KB   [Preview]

Finnish (Suomi):

1. 1M  (application/x-executable-file)   Koko: 1Â024,00 KB   [Esikatselu]





I think it is meant to be 1'024,00