Mitchell Hudson wrote:
I did actually pull out the number of tokens and I have quite a few in
there:
0.000 0 3 0 non-token data: bayes db version
0.000 0 23930 0 non-token data: nspam
0.000 0 8304 0 non-token data: nham
0.000 0 200096 0 non-token data: ntokens
0.000 0 1175826856 0 non-token data: oldest atime
0.000 0 1201535593 0 non-token data: newest atime
0.000 0 0 0 non-token data: last journal
sync atime
0.000 0 1201492913 0 non-token data: last expiry atime
0.000 0 91902 0 non-token data: last expire
atime delta
0.000 0 111416 0 non-token data: last expire
reduction count
So I'm not worried about not having any training. And the spamassassin
-D bayes <message.txt I assumed you meant spamassassin -D bayes >
message.txt, but in any case I let it run for about 30 minutes and it
didn't return any data, which seemed very strange.
I meant spamassassin -D bayes <message.txt. You want to feed a message
in to spamassassin, not over-write it with spamassassin's output.
You got no output, because you gave it no message.
I also tried reinstalling the perls bayes modules from cpan, which did
update, but didn't correct the problem. I've sat there and watched
/var/log/current for quite a while, and watched the debug info. It's
so strange bayes seems to load, as in it will learn and accesses the
database, adds new tokens, etc. But doesn't actually score. I've
checked all my configuration files and can't find anything weird. I'm
at a loss.
Try a manual message scan, as above, and see how that does.