Re: procmail: Non-zero exitcode (1) from "/usr/bin/bogofilter"

On Wed, 12 Sep 2007 21:15:02 -0400
[email protected] wrote:

> Quoting David Relson <[email protected]>:
> 
> > On Wed, 12 Sep 2007 17:43:39 -0400
> > [email protected] wrote:
> >
> >> My bogofilter went south again.  I am getting procmail: Non-zero
> >> exitcode (1) from "/usr/bin/bogofilter"
> >>
> >> from procmail for every email.  This started happening on Monday,
> >> and I just noticed.  Nothing is getting classified as spam.  Ive
> >> had and fixed this problem before, but cant find my notes about
> >> bogoutil and checking my spam vs. ham.  Any bogofilter gurus out
> >> there that can point me the right way would be great.  Im running:
> >> bogofilter version 1.1.3
> >>      Database: Sleepycat Software: Berkeley DB 4.3.29: (September
> >> 12, 2006) AUTO-XA
> >> Copyright (C) 2002-2006 David Relson, Matthias Andree, Greg Louis
> >> Copyright (C) 2002-2003 Eric S. Raymond, Adrian Otto, Gyepi Sam.
> >>
> >>
> >> thanks,
> >>
> >> ddh
> >>
> >> --
> >> Dwayne Hottinger
> >> Network Administrator
> >> Harrisonburg City Public Schools
> >
> > Hi Dwayne,
> >
> > Version 1.1.3 is somewhat old, but that's not related to the
> > problem.
> >
> > FWIW, the exit codes are defined (in common.h) as:
> >    RC_SPAM	= 0,
> >    RC_HAM	= 1,
> >    RC_UNSURE	= 2,
> >
> > Does your installation depend on a procmail recipe testing for
> > "X-Bogosity: Spam/Ham/Unusre" or is it testing the exit code?
> >
> > Have you manually tested bogofilter, i.e.
> >
> >    bogofilter -e -p < msg.ham.txt
> >    bogofilter -e -p < msg.spam.txt
> >
> > One possibility is that your wordlist has become b0rked.  Has it's
> > size (in MB) changed (perchance)?  Have you tried getting a
> > wordcount, i.e. "bogoutil -d /PATH_TO_WORDLIST/wordlist.db | wc
> > -l", and is it a reasonably (large) number?
> >
> > HTH,
> >
> > David
> >
> > David
> >
> 
> David,
> bogoutil -d /PATH_TO_WORDLIST/wordlist.db | wc -l shows 690546.  My  
> original wordlist.txt that started my training from some years ago,  
> scores 0 with bogofilter -e -p < wordlist.txt.  I do use a procmail  
> script it hasnt changed in quite a while:
> # send through bogofilter
> :0HB:
> * ? bogofilter  -l
> /opt/bogofilter-spam/bogofilter-spam.
> 
> thanks,
> 
> ddh

Dwayne,

690546 is a reasonable number, though perhaps a bit low.

Running "bogoutil -d /.../wordlist.db | tail" will print the last few
words (alphabetically).  You should see words at the end of the
alphabet (or possibly utf-8 encoded asian language text).  If you see
mid-alphabet, that would be a sign that of b0rkedness.

What's the size of the database, i.e. "ls -lh /.../wordlist.db" and, if
you run "bogoutil -d .../wordlist.db > wordlist.txt" and "ls -l
wordlist.txt" what are those sizes.  If memory serves, wordlist.db can
legitimately be double the size of the text size.  A much bigger ratio
than 2 indicates that bogoutil couldn't read the full wordlist.

Also, the FAQ has some info on recovering a b0rked wordlist.

Lastly, are you using transactions?  They help maintain database
integrity.

David
_______________________________________________
Bogofilter mailing list
[email protected]
http://www.bogofilter.org/mailman/listinfo/bogofilter