[Mimedefang] Mimedefang with Anomy and SpamAssassin

[Rich West]

I'm using SpamAssassin 2.01, actually...

I had to disable SpamAssassin because I started getting user complaints 
that email was not going through for them.  From what the sendmail logs 
showed, most messages were going through just fine (and I was happy that 
inbound mail was properly getting tagged as SPAM).

However, some MIME messages, or perhaps some messages with MIME 
attachments, were generating an error similar to this one:
...
Feb  1 17:11:38 cranium sm-mta[31597]: g11MBGOC031597: 
from=<Marleen_Young at njb.uscourts.gov>, size=3109438, class=0, nrcpts=1, 
msgid=<OF36EC3804.AEA39F2C-ON85256B53.004EF45C at uscmail.dcn>, 
proto=ESMTP, daemon=MTA, relay=canal.ca5.uscourts.gov [207.41.16.202]
Feb  1 17:11:39 cranium mimedefang-multiplexor: Slave 1 stderr: 
Premature end of base64 data at 
/usr/lib/perl5/site_perl/5.6.1/Mail/SpamAssassin/PerMsgStatus.pm line 1218.
Feb  1 17:11:39 cranium mimedefang-multiplexor: Slave 1 stderr: 
Premature end of  base64 data at 
/usr/lib/perl5/site_perl/5.6.1/Mail/SpamAssassin/PerMsgStatus.pm line 1218.
Feb  1 17:13:38 cranium sm-mta[31597]: g11MBGOC031597: Milter 
(mimedefang): time out before data read
Feb  1 17:13:38 cranium sm-mta[31597]: g11MBGOC031597: Milter 
(mimedefang): to error state
Feb  1 17:13:38 cranium sm-mta[31597]: g11MBGOC031597: Milter: data, 
reject=451 4.7.1 Please try again later
Feb  1 17:13:38 cranium sm-mta[31597]: g11MBGOC031597: 
to=<hbuck at helenereynolds.com>, delay=00:02:21, pri=30837, stat=Please 
try again later
...

I can live with the problem with Anomy.. it's no big deal.  I figured 
that it was due to either $3, $4, or $5 being null which caused the 
concat to fail, but it is just a warning.. I thought I would mention it 
on the list, though. :)

And SpamAssassin is a great add-on to mimedefang, but the fact that in 
multiple places (two sites, multiple servers) it barfed in the same way 
has me a little gun-shy.  I'm not 100% sure that it is entirely a 
SpamAssassin problem, but with the fact that I simply disabled the call 
in filter_begin and things began to work, I am pretty close to convinced 
that it is SpamAssassin and not MimeDefang. :)

Just a thought.. could the base64 barf be caused by SpamAssassin trying 
to decode (or defang) the already decoded (or defang'ed) MIME message?

-Rich

Richard West                            mailto:richard.west at divatv.com
Sr. Systems Administrator
Diva - Princeton, NJ                    http://www.divatv.com



On 1 Feb 2002 at 11:42, Rich West wrote:

> Also, just as an FYI, when SpamAssassin was installed and I turned on 
> the use of spamassassin from within MimeDefang, I got the following error:
> Feb  1 10:30:28 enigma mimedefang[18337]: [ID 946266 mail.debug] Perl 
> stderr: Ca
> n't locate object method "new" via package 
> "Mail::SpamAssassin::MyMailAudit" (pe
> rhaps you forgot to load "Mail::SpamAssassin::MyMailAudit"?) at 
> /opt/local/bin/m
> imedefang.pl line 2051.
> 
> I added Mail::SpamAssassin::MyMailAudit to the list of modules within 
> mimedefang.pl to make it look like:
> ...
> use Digest::SHA1;
> use Mail::SpamAssassin;
> use Mail::SpamAssassin::MyMailAudit;
> use Anomy::HTMLCleaner;

Are you using SpamAssassin 1.5, or 2.0?  The default installation of 
SpamAssassin 2.0 replaces the use of ::MyMailAudit with 
::NoMailAudit, which is purportedly more efficient, etc. than 
::MyMailAudit (which is just a subclass of Mail::Audit).

According discussions I've been involved with on the SA mailing list, 
the NoMailAudit library doesn't (yet) replicate the data constructor 
which MIMEDefang depends on in order to interface with SA properly.

Specifically including Mail::SpamAssassin::MyMailAudit as you've done 
allows 'Fang to work fine with version 2.0.  Presumably ::NoMailAudit 
will eventually be updated so that it will work, too.