I am a little surprised nobody has any advice for me. Let me restate this as a yes/no question:
If sa-learn is relearning the same messages over and over again and can't forget them (details below), is my Bayesian database just hosed beyond repair? /k > Clearly something is broken, but I'm not sure what. > > I am using SpamAssassin 2.64 on SuSE Linux 9.0. My problems started when I > upgraded from the original SA that shipped with the system (2.55, I believe) > and the bayes path specification changed in the local.cf file. > > Once that was rectified and the Bayes database was re-located, I thought all > was well, but now I have confirmed that I can add and re-add and re-add > messages using sa-learn, but when I try to "sa-learn --forget" them, the > response is an error message. > > Here are the commands and responses: > > cornelius:/tmp # su spamfilter -c '/usr/bin/sa-learn --spam ./unham/1.' > Learned from 1 message(s) (1 message(s) examined). > > cornelius:/tmp # su spamfilter -c '/usr/bin/sa-learn --spam ./unham/1.' > Learned from 1 message(s) (1 message(s) examined). > > (same message clearly learned from twice. I can repeat this endlessly.) > > cornelius:/tmp # su spamfilter -c '/usr/bin/sa-learn --forget ./unham/1.' > Learned from 0 message(s) (1 message(s) examined). > ERROR: the Bayes learn function returned an error, please re-run with -D for > more information > > (the unhelpful error message) > > When re-running sa-learn with the -D option, the command hangs perpetually at > this message: > > debug: Syncing complete. > > .. until I issue a Ctrl-C. > > I don't see why listing the debug information would cause a hang, but it does. > > What should I do here? Trash my Bayesian database entirely and start over? > What do these symptoms mean?
