On 13-06-13 20:28, David Rees wrote: > On Wed, Jun 12, 2013 at 6:06 AM, Chris Moules <christop...@gms.lu> wrote: >> NOTE: the patch that I sent was only for the single (repeated) error >> that I was receiving. This was while hitting the DB via dspam_train. >> >> I have been evaluating all the storage backends and found Hash to be the >> fastest, but seems buggy with TOE as trainingMode (it crashes, looks >> like when it tries to 'extend' the storage). > > Is training actually working for you on PostgreSQL? It does not appear > to be working for me... > > See: http://sourceforge.net/mailarchive/message.php?msg_id=30708939 > > Haven't had time to dig into the issue further yet. > > As a testament to dspam, on established accounts it still works very > well despite the lack of training, though it consistently misses some > newsletters because of it. >
Yes I saw those messages, but unfortunately did not have anything to add to it. Training works here, I'm calling dspamc (i.e. client-server mode) from dovecot-antispam but otherwise I see no huge differences with your setup. Postgresql version is 9.1.9. I don't have any useful C knowledge to dig into the actual code and find out where the magic breaks for you... WHat does raise questions is that the last we saw similar trouble, STevan actually removed the escape string syntax, as you can see in the commit I referred to. Also the everlasting question comes bck: if you trained other accounts successfully with the postgres backend, what did you change to make it break? -- Tom ------------------------------------------------------------------------------ This SF.net email is sponsored by Windows: Build for Windows Store. http://p.sf.net/sfu/windows-dev2dev _______________________________________________ Dspam-devel mailing list Dspam-devel@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/dspam-devel
