spamassassin-users mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Karsten Br├Ąckelmann <guent...@rudersport.de>
Subject Re: Inconsistent spam scores between spam headers and rewritten subject line.
Date Tue, 16 Aug 2011 15:10:35 GMT
On Tue, 2011-08-16 at 22:29 +0930, Rodney Baker wrote:
> On Tue, 16 Aug 2011 05:02:20 John Hardin wrote:

> > Just as a test, if you comment that bit out of your personal .procmailrc
> > does everything work they way you'd expect (i.e. one SA pass, the correct
> > score in the X- headers)?
> 
> Yep,that was the first thing that I did. Somehow spamassassin is still 
> checking the messages, even though they're not being piped through spamc via 
> procmail. I'm sure that fetchmail isn't doing it, so that leaves sendmail, 
> dovecot or kmail. So begins the process of elimination (or maybe I just leave 
> it out of procmailrc and be done with it...).

If you don't use Delivery Control Options with fetchmail (see that
section in the man pages) like an explicit MDA or SMTP, this should not
be where SA gets invoked. You don't, do you? The default is to pass it
on to port 25, which should just be your Sendmail.

A site-wide procmail configuration doesn't exist, as you mentioned in
another reply to this thread.

Dovecot will not filter messages. It's an IMAP server that serves what
has been delivered already. The dovecot MDA could, but you seem to use
procmail for direct delivery into the Maildir store. Another one to rule
out.

Kmail as an MUA must not modify delivered mail (and doesn't), so while
it could call SA again, you won't see SA headers. Both Dovecot and Kmail
are after the procmail recipe you initially showed anyway, so there's no
chance they could cause the matching issues you reported.

Leaves us with Sendmail in the chain to dig further...

After all, procmail already sees SA headers, without a filter. What
you're hunting for is before procmail in the chain.


Regarding "leaving it out of procmail" and being done with it -- maybe.
This is likely to bite later, though. If it is before procmail, odds are
it's using a site-wide user. Which implies Bayes training has to be done
as that user, not the recipient...


-- 
char *t="\10pse\0r\0dtu\0.@ghno\x4e\xc8\x79\xf4\xab\x51\x8a\x10\xf4\xf4\xc4";
main(){ char h,m=h=*t++,*x=t+2*h,c,i,l=*x,s=0; for (i=0;i<l;i++){ i%8? c<<=1:
(c=*++x); c&128 && (s+=h); if (!(h>>=1)||!t[s+h]){ putchar(t[s]);h=m;s=0;
}}}


Mime
View raw message