how to train spamc with spam?

Hi does anyone know how to test if spamc is learning spam ?
and if its not, what might be wrong? (ie what is the correct way for my situation)

  • I suspect the stuff I manually pipe to it with :
spamc --learntype=spam

is NOT being learnt - as the filtering is still so poor. Same spams keep getting through.

its beginning to drive me potty!

Can anyone advise ?


using centralised spamd user=spamd for filtering all local accounts (because only one account gets a significant amount of email). Incoming stuff is piped to it by postix etc. and then delivered to /var/mail/userid

To try to fix the problem, I’ve been running with cron hourly with:

5 * * * * /usr/bin/sa-learn --sync -u spamd

But the only files that ever change in the .spamassassin directory are:

and I’m pretty sure they are not modified straight away on hitting the learn-as-spam filter, but I don’t know if they’re changing anyway from other spamassassin activity

I’ve realised that this problem must be related to file ownership and/or process ownership…

spamd is running as a system-user with limited rights, but when it talks to spamc it tries to save the data for the local user of spamc (and fails) instead of in its usual centralised files.

I haven’t been able to figure out the right way to do this from the man pages - does anyone know how to correct this?