Re: Untranslated gettext UTF-8 strings and different platforms

Thomas Deniau <[email protected]> wrote:

> If I've understood correctly the docs, gettext does a binary comparison
> between the msgids, so if the files are in the same encoding, the 
> lookup should not fail. Does anyone know what differences between the
> two platforms causes these different behaviours ? Or does anyone see 
> any bug in the code ?

In case anyone has this problem in the future, I got the answer from
Bruno Haible on the Sourceforge translation-i18n mailing list : Debian
stable still has glibc 2.3.2, which has a comparison bug in
glibc/intl/hash-string.c for UTF-8 strings. This was fixed in glibc
2.3.3.

The solution is either to use Debian unstable, or to compile a new
gettext with the "--with-included-gettext" option, which uses gettext's
own implementation of the incriminated function (and make php use the
new gettext, of course).

(see the post by Bruno Haible on the
[email protected] mailing-list, on Thu, 4 Jan 2007
at 23:03:39 +0100)


-- 
Thomas Deniau