On Wed, 2014-07-23 at 14:34 -0700, Asai wrote:
> The mail server is running as a different user than amavis, so I ran
> this under the amavis user:
>
> 0.000 0 3 0 non-token data: bayes db version
> 0.000 0 624 0 non-token data: nspam
> 0.000 0 11919 0 non-token data: nham
While that seems a little biased, that's sufficient training with
default thresholds of 200 each.
> 0.000 0 120783 0 non-token data: ntokens
> 0.000 0 1405862394 0 non-token data: oldest atime
> 0.000 0 1406151128 0 non-token data: newest atime
Good, just a few hours ago.
> sa-learn is running under amavis as well.
Yet, there are no BAYES_xx rules hit, so the Bayesian classifier somehow
has been disabled.
Check your SA conf with a fine-toothed comb. Grepping the .cf files for
'bayes' should get you most relevant options. In addition, check for the
use_learner option. Besides the obvious use_bayes, the use_bayes_rules
option seems a likely candidate for your issue.
If that still doesn't explain it, it's time for some debugging. Both
using the plain 'spamassassin' executable directly, as well as debugging
Amavis.
--
char *t="\10pse\0r\0dtu\0.@ghno\x4e\xc8\x79\xf4\xab\x51\x8a\x10\xf4\xf4\xc4";
main(){ char h,m=h=*t++,*x=t+2*h,c,i,l=*x,s=0; for (i=0;i<l;i++){ i%8? c<<=1:
(c=*++x); c&128 && (s+=h); if (!(h>>=1)||!t[s+h]){ putchar(t[s]);h=m;s=0; }}}
