Bug#317524: bogofilter-bdb: seems to be stuck in an endless loop

[David Relson]

On Tue, 2 Aug 2005 15:35:02 -0400
Clint Adams wrote:

> > >Okay.  What if you comment out the charset_default line in your config?
> > 
> > When I do this, bogofilter doesn't end up in the endless loop anymore
> > and seems to work as it should.
> 
> [Additional information at http://bugs.debian.org/317524 ]
> 
> There is apparently a problem if charset_default is set to iso-8859-1 .
> I have not attempted to reproduce this yet.

Hello Torsten & Clint,

From the bug description, the problem should be demonstrable with a
script like the one attached and an appropriate message file.  The
combination of attached script & message work fine for me.  

Knowing the exact bogofilter command run at the time of the loop and
having the message being processed would be very helpfu.  Can you
provide missing pieces so I can reproduce the problem?  

Bogofilter's use of iconv() has some special handling for characters
that don't belong in the "from" charset. Early on in the development of
the iconv code there were problems with that special handling, though I
don't recall an infinite loop such as you've experienced.  My suspicion
is that you may have encountered a similar problem.

By the way, what does "bogofilter -V" say?

Thanks,

David
-------------- next part --------------
From: somebody at example.com   Tue Mar 25 20:11:29 2003
Subject: iso-8859-1 problem
To: whoever at example.com

this is a test
-------------- next part --------------
#!/bin/sh

cat > bogofilter.cf <<EOF
charset_default=iso-8859-1
EOF

rm -f ./wordlist.db

for LANG in C de_DE.UTF-8 de_DE at euro; do
    LC_CTYPE=de_DE at euro
    echo LANG: $LANG, LC_CTYPE: $LC_CTYPE
    export LANG LC_CTYPE
    bogofilter -c bogofilter.cf -d . -n -v < problem.msg
    bogofilter -c bogofilter.cf -d . -s -v < problem.msg
    bogofilter -c bogofilter.cf -d . -v < problem.msg
    ls -l *.db
    echo
done