procmail
[Top] [All Lists]

RE: Procmail + spamassassin

2003-03-20 14:52:27
Dallman Ross [mailto:dman(_at_)nomotek(_dot_)com] wrote:

On Thu, Mar 20, 2003 at 11:05:15AM -0500, Matt Dunford wrote:

:0:
* ! ? $SPAMASSASSIN -e
spam

However I found that spamassassin's exit codes were unreliable.
Sometimes my recipe would exit 0, but when I ran spamassassin 
from the cmd line on the same msg, it would exit a non-zero 
number (ie, spam).  I could never reproduce this or track it down.  
But this happened rarely and perhaps it's fixed in newer versions.

The reason for your inconsistent results is that you are running your
test only on the message headers, not the body. [. . .]

Another thing I should have mentioned is that there was a buffer-
overrun issue in at least the 1.3x series of SpamAssassin, such that
mail piped to SA via the above test would occasionally fail (causing
SA to exit 0 on spam).  When this came to light (six months or so ago 
on my home shell system), I switched to a different style of SA recipe 
that didn't use the test call.  I meanwhile forgot about the issue 
until re-reading Matt's report.

I have no idea if the current (1.4x-1.5x) SA has repaired this
overflow issue or not.  As I posted, I use a different method to
call SA myself.  My algorithm developed partly out of the need
to avoid the test call, but then morphed significantly from that
into something else altogether.

Frankly, my calls to spamc, and previously to spamassassin, can
be measured on the fingers of one hand per week, because almost
all my spam and good mail is handled earlier on in my .procmailrc
by private recipes.  So I'm not a good person to say whether these
idiosyncrasies of SA are still around.  I do know that my method
of calling spamc works fine, though.

Dallman



_______________________________________________
procmail mailing list
procmail(_at_)lists(_dot_)RWTH-Aachen(_dot_)DE
http://MailMan.RWTH-Aachen.DE/mailman/listinfo/procmail

<Prev in Thread] Current Thread [Next in Thread>