[Rspamd-Users] learn_spam seems not to

[Lucas Rolff]

It's quite common for Benny to simply vent air to point out he's an SA user or something else, he does it across different mailing lists.

Are you using per user bayes/classifier or are you using default (global bayes for all users), and more importantly, if you simply run `rspamc newspam/99` for example, do you see the actual bayes_spam/ham revealed there? And do you have enough learns already (minimum is 200 by default), rspamd logs will tell you if you're below, when email comes in.

Could you post your statistic.conf, classifier-bayes.conf (if it exists) and bayes_expiry.conf configs?

- Lucas

> On 30 Sep 2025, at 08:28, Randy Bush <randy at psg.com> wrote:
> 
>> man rspamc
> 
> where do you think i came up with that incantation?  :)
> 
>> wild guess here is you miss --user ?
> 
>    -u username, --user=username
>           Emulate that message was received from specified 
>           authenticated user
> 
> note "from."  the messages are coming _from_ a jillion, usually fake,
> users
> 
> i also looked at
> 
>    -d user at domain, --deliver=user at domain
> 	   Emulate  that  message  was  delivered  to  specified  user
> 	   (for LDA/statistics)
> 
> but the spam is being delivered to any one of half a dozen of my
> mailboxen, e.g. randy at psg.com, root at rg.net, randy at bogus.com, ...
> 
> and the messages being fed to `rspamc -P ioc8ieMa5eec learn_spam`
> were not delivered locally (LDA).  after filtering they were forwarded
> to a remote dovecot server.
> 
> please try again with heavier clue bat.
> 
> randy
> 
> 
> -- 
> Users mailing list
> Users at lists.rspamd.com
> https://lists.rspamd.com/mailman/listinfo/users