Learning bayes with spamc

[Jeff Morton <io...@gmail.com>]

I've set up spamassassin so that the bayes files are owned by 'nobody'
and are centrally located, and spamc is used to report spam or ham to
train the filter.  I have no problem training the filter by piping a
single file at a time into spamc.

The local.cf file contains:

bayes_path /etc/mail/spamassassin/bayes/bayes
bayes_file_mode 0700

Unfortunately, with this setup I have no way of using sa-learn to
train the filter, even using sudo to call sa-learn as nobody is a
problem because then I need to make sure user nobody can get to the
files being trained with.

Is there a way, then, to cause spamc to learn from an entire folder
without writing a bash for loop to pipe the files in one at a time?
If I do

  cat * | spamc -L spam

then I get

  Message successfully un/learned

Which doesn't tell me how many messages were un/learned so I'm not
sure that I can pass it a chunk of files that way.

Linux FC6
SpamAssassin Client version 3.1.7

Re: Learning bayes with spamc

[Theo Van Dinter <fe...@apache.org>]

On Wed, Jan 31, 2007 at 12:03:00PM -0500, Jeff Morton wrote:
> Is there a way, then, to cause spamc to learn from an entire folder
> without writing a bash for loop to pipe the files in one at a time?

No.  spamc only knows how to deal with single messages.  If you want mbox,
mbx, or dir handling, you have to do it externally.

-- 
Randomly Selected Tagline:
I'll just sit back and smirk smugly now...