Watermarking action problem?

[John Wilcock]

Just received a genuine out-of-office reply that was sent with a null 
sender, but didn't quote the original message and hence the watermark.

MS correctly added 5 points to the spam score, but this should not have 
been enough for the message to be considered as spam. (-1.5 + 5 = +3.5, 
with a threshold of 5) However, it still took the spam action rather 
than the nonspam action for the message.

This is with MS 4.67.6 (the latest gentoo ebuild), and the changelog 
doesn't mention any watermarking-related changes since.

The relevant MailScanner.conf settings are:

Use Watermarking = yes
Add Watermark = yes
Check Watermarks With No Sender = yes
Treat Invalid Watermarks With No Sender as Spam = 5
Required SpamAssassin Score = 5
Spam Actions = %rules-dir%/spam.actions.rules
Non Spam Actions = %rules-dir%/non.spam.actions.rules

These rules files hit defaults for this message, which are
(spam)
FromOrTo:       default store deliver header "X-Spam-Flag: YES"
(nonspam)
FromOrTo:       default store deliver

Log extract:

May  5 10:09:43 ns0 MailScanner[940]: Message D1EE98084.515BE had bad 
watermark, added 5 to spam score
May  5 10:09:44 ns0 MailScanner[940]: Message D1EE98084.515BE from 
212.227.15.27 () to tradoc.fr is spam, SpamAssassin (not cached, 
score=-1.5, required 5, autolearn=not spam, BAYES_00 -1.50, 
SPF_HELO_PASS -0.00, UNPARSEABLE_RELAY 0.00)
May  5 10:09:44 ns0 MailScanner[940]: Spam Actions: message 
D1EE98084.515BE actions are store,deliver,header


John.

-- 
-- Over 3000 webcams from ski resorts around the world - www.snoweye.com
-- Translate your technical documents and web pages    - www.tradoc.fr