Re: bug in the unix join command or maybe in the sort command

[Paul Eggert]

Robert Castelo <address@hidden> writes:

> address@hidden joinbug]$ uname -a
> Linux zapiron.imim.es 2.6.8-1.521 #1 Mon Aug 16 09:01:18 EDT 2004 i686
> i686 i386 GNU/Linux
> address@hidden joinbug]$ locale
> LANG=en_US.UTF-8
> LC_CTYPE="en_US.UTF-8"
> ...
> address@hidden joinbug]$ sort --version
> sort (coreutils) 5.2.1

I changed my environment to match yours as much as possible, by
setting my locale to en_US.UTF-8, and I could not reproduce the "sort"
bug that you reported.  I'm afraid you'll have to investigate further
on your end.  One possibility is that it is a bug in your en_US.UTF-8
locale, or perhaps in your C library, which causes strcoll("a", "aa")
to return a positive number, causing "sort" to decide that "a" > "aa".

Perhaps you can investigate this hypothesis by running "sort" under a
debugger, and seeing what arguments are passed to strcoll and what
results it returns.  That would verify my guess.

You can get the source code to the latest test version of coreutils
here:

ftp://alpha.gnu.org/gnu/coreutils/coreutils-5.3.0.tar.gz