Thanks for the suggestion, Sanford.

OK, I removed the -y for all the POSTIVE SA tests (SA1-SA10) and then
created the following rule in local.cf:

body MY_SPAMASSASSIN_RULE        /this is a test of spamassassin/
describe MY_SPAMASSASSIN_RULE    Test of my spamassassin rule
score MY_SPAMASSASSIN_RULE       11

Sending a message with " this is a test of spamassassin" in the body ends up
with this in maillog:

spamd: clean message (4.6/5.0)
and
spamd: result: .  4 -
AWL,BAYES_00,HTML_90_100,HTML_MESSAGE,MY_SPAMASSASSIN_RULE scantime=6.3

I wrote the rule differently, adding is to the end, and a score of 11, I'm
getting a failure on SA5 now.  This is weird.  The maillog for that one
shows:

spamd: clean message (4.9/5.0) for spamd:101 in 7.2 seconds, 2872 bytes.
And
spamd: result: .  4 -
AWL,BAYES_00,HTML_90_100,HTML_MESSAGE,MY_SPAMASSASSIN_RULE
scantime=7.2,size=2872

Is my custom rule being overridden somehow??  It shows up in the spamd
logging, but doesn't apply the score??

I'm using SA 3.1.0.

Next I turned of Bayes and Pyzor.  

Maillog:
spamd: identified spam (6.2/5.0) for spamd:101 in 1.8 seconds, 2872 bytes
and
spamd: result: Y  6 - AWL,HTML_MESSAGE,MY_SPAMASSASSIN_RULE
scantime=1.8,size=2872

Now it scores higher??

Tested from command line on SPAM that was not held:

SPAMC32: Loading C:\imail\Declude\filters\SPAMC32\Enjoy.eml
9.4/5.0

AWL,DCC_CHECK,HTML_FONT_BIG,HTML_IMAGE_RATIO_02,HTML_MESSAGE,INVALID_DATE,J_
CHICKENPOX_93,RCVD_IN_BL_SPAMCOP_NET,SUBJ_YOUR_DEBT,URIBL_OB_SURBL

(bayes, etc. still OFF)

Test on high scorer that didn't fail SA before:

SPAMC32: Loading C:\imail\Declude\filters\SPAMC32\cbn.eml
28.0/5.0

ADDR_NUMS_AT_BIGSITE,ADVANCE_FEE_1,ADVANCE_FEE_2,ADVANCE_FEE_3,ADVANCE_FEE_4
,AWL,FAKE_HELO_MAIL_COM,FORGED_MUA_OUTLOOK,FORGED_YAHOO_RCVD,FROM_HAS_ULINE_
NUMS,INVALID_DATE,J_CHICKENPOX_48,J_CHICKENPOX_74,MILLION_USD,MISSING_HEADER
S,MSGID_FROM_MTA_HEADER,RCVD_IN_BL_SPAMCOP_NET,RCVD_IN_SBL,RCVD_IN_SORBS_WEB
,RCVD_NUMERIC_HELO,TO_CC_NONE

Tested my rule test mail from command line:

6.3/5.0

AWL,HTML_MESSAGE,MY_SPAMASSASSIN_RULE

Hmm...

It appears this is at least partially on the SA side.  Any ideas??

Thanks,
Geoff

-----Original Message-----
From: [EMAIL PROTECTED]
[mailto:[EMAIL PROTECTED] On Behalf Of Sanford Whiteman
Sent: Thursday, April 27, 2006 10:31 AM
To: Geoff Varney
Subject: Re: [Declude.JunkMail] Problems with SPAMC32 scoring??

> While I can't *guarantee* that it was working right with higher than
> 10 returns from spamd, here is what I see now.

'-et  10'  means  'return an exit code of 10 if the SPAMD total weight
was  greater  than  10.' I like to look at it in terms of what SPAMC32
returns, not anything it can't control, so I won't refer to the weight
you assigned to the test in the Declude test definition.

What  happens  if  you  run  SPAMC32  from  the command line against a
message  known to get a SPAMD weight > 10 and check the exit code with
'echo  %ERRORLEVEL%'?  Can  you  add  a  regression test in SPAMD that
specifically adds a weight of 11 and post the results?

And I would eliminate the '-y' from the test def, or at least for this
round  of testing, because this is in there for backward compatibility
with  the  *nix  SpamC  and, I would say, is not guaranteed to work in
expected ways when paired with the newer-fangled '-e' switch. The '-y'
is  good  switch  to  use  for interactive debugging, but it will only
create  more  overhead (more data sent back across the wire) when used
under Declude, AFAICS.

--Sandy


------------------------------------
Sanford Whiteman, Chief Technologist
Broadleaf Systems, a division of
Cypress Integrated Systems, Inc.
e-mail: [EMAIL PROTECTED]

SpamAssassin plugs into Declude!
 
http://www.imprimia.com/products/software/freeutils/SPAMC32/download/release
/

Defuse Dictionary Attacks: Turn Exchange or IMail mailboxes into IMail
Aliases!
 
http://www.imprimia.com/products/software/freeutils/exchange2aliases/downloa
d/release/
 
http://www.imprimia.com/products/software/freeutils/ldap2aliases/download/re
lease/

---
This E-mail came from the Declude.JunkMail mailing list.  To
unsubscribe, just send an E-mail to [EMAIL PROTECTED], and
type "unsubscribe Declude.JunkMail".  The archives can be found
at http://www.mail-archive.com.

---
This E-mail came from the Declude.JunkMail mailing list.  To
unsubscribe, just send an E-mail to [EMAIL PROTECTED], and
type "unsubscribe Declude.JunkMail".  The archives can be found
at http://www.mail-archive.com.

Reply via email to