On Sun, 2004-02-22 at 15:18, jdow wrote:
[...]
> report_header 1
> 
> There's no such option of which I am aware.
> 
> > debug: Failed to parse line in SpamAssassin configuration, skipping:
> defang_mime 0
> 
> There's no such option of which I am aware.

I didn't edit /etc/mail/spamassass/local.cf so if these options are
deprecated and according to man Mail::SpamAssassin::Conf it looks as if
they are than they must have been leftovers from my previous install... 
I'll remove these.

> > debug: Failed to parse line in SpamAssassin configuration, skipping:
> RCVD_IN_BL_SPAMCOP_NET    3.00
> 
> Betcha it would not skip "score RCVD_IN_BL_SPAMCOP_NET    3.00" instead.
[...]

Thx, no more errors but still no bayes...

> See comments above. Is this an imported local.cf from a much older version
> of spamassassin?

AFAIK those files must have been created by the previous stable version
of SA.

[...]
> I hope you fed the ham and spam through on two separate runs. Else you
> have to delete the bayes databases and do it over correctly. The
> spamassassin faq (wiki) can help you understand this all a little better,
> I suspect. It looks for all the world like you fed all "+8000 messages"
> on the ham track with none fed through as spam. That WILL leave your
> bayes database hosed and useless. If this is indeed what you did then
> delete ~/.spamassassin/bayes*. Then start over using the --ham and --spam
> options with separated ham and spam mailbox files. In common with all other
> software of which I am aware to date spamassassin cannot read your mind for
> you when it learns your preferences for ham from spam.

I used a script that looks like this:
---- train-sa ---
#!/bin/bash
LOGFILE=/data/logs/train-sa.log
PREFS=/home/m8ram/.spamassassin/user_prefs
(
cd
evolution --force-shutdown
sa-learn --prefs-file=$PREFS --ham --showdots --no-rebuild --mbox 
evolution/local/ces/mbox
[...]
sa-learn --prefs-file=$PREFS --spam --showdots --no-rebuild --mbox evolution/loc
al/SPAM/mbox
sa-learn --prefs-file=$PREFS --rebuild
) >> $LOGFILE 2>&1
---- train-sa ---

With one sa-learn command for every mbox file in the ~/evolution
directory...

(When I first posted I didn't have the --prefs-file option specified but
I reran the script in the mean time.)

But still 'sa-learn --dump magic' and 'sa-learn
--prefs-file=/home/m8ram/.spamassassin/user_prefs --dump magic' give:
0.000          0          0          0  non-token data: bayes db version
0.000          0          0          0  non-token data: nspam
0.000          0          0          0  non-token data: nham
0.000          0          0          0  non-token data: ntokens
0.000          0          0          0  non-token data: oldest atime
0.000          0          0          0  non-token data: current scan-count
0.000          0          0          0  non-token data: last expiry atime

> Modulo where you stored the ham and spam, the commands would look something
> like this.
> sa-learn --ham --showdots --mbox ~/ham
> sa-learn --spam --showdots --mbox ~/spam

I guess my script does this?

TIA
-- 
# Mertens Bram "M8ram"   <[EMAIL PROTECTED]>     Linux User #249103 #
# SuSE Linux 8.2 (i586) kernel 2.4.20-4GB      i686                256MB RAM #
#  7:43pm  up 30 days 23:21,  11 users,  load average: 0.09, 0.11, 0.04 #

Reply via email to