On Saturday 02 May 2009, Gene Heskett wrote: >On Saturday 02 May 2009, Theo Van Dinter wrote: >>bayes_seen is rather irrelevant. > >To this problem, or generally? > >>bayes_toks is very binary-oriented, and uses lots of pack() calls. >> >>There is no SA-based "validity" check for the DB files/data. If you >>think the DB file itself is corrupt, you could try the appropriate DBM >>tools (db_verify, etc.) The dump/restore method really should have >>solved your issue. If you're still having the same problem, I would >>say either a) are you sure you're looking at the right DB file, b) do >>the dump/restore again and make sure to delete/move the DB file before >>restoring, c) make sure the data you're restoring is valid (gigo and >>all that). > >You all keep referring to the DB file. Where will I find it? And if I am > to delete it prior to the --restore, what file? Look at the script > attached, and if possible tell me whats wrong if it is. > >I am probably complicating the issue in that the attached script is run > every morning at 10am to process the ham/spam I have dragged and dropped > into the correct folders. This is because I run as root, but all mail > fetching and processing is done by the unpriviledged user gene. Hence the > tomfoolery with the script to isolate me from this old fart. I am not 100% > convinced that my sa-learn sessions are doing me any good, I have fed one > outfits daily messages to sa-learn --spam for a year now, and they are > still walking right on by SA. > >As I am not savvy on lock files, its possible that the addition of a lock or >two might be in order, something to stall my script if it fires off when > spamd is active, or vice-versa. > > >Something else odd here, running spamassassin --lint -D does not generate >those errors when run as root... Nor does it when run as gene. > >And occasionally it generates a different error number: >bayes: unknown packing format for bayes db, please re-learn: 73 at >/usr/lib/perl5/vendor_perl/5.10.0/Mail/SpamAssassin/BayesStore/DBM.pmline >1883. >bayes: unknown packing format for bayes db, please re-learn: 76 at >/usr/lib/perl5/vendor_perl/5.10.0/Mail/SpamAssassin/BayesStore/DBM.pmline >1883. >bayes: unknown packing format for bayes db, please re-learn: 73 at >/usr/lib/perl5/vendor_perl/5.10.0/Mail/SpamAssassin/BayesStore/DBM.pmline >1883. > >Note the 76 in the 2nd line. > >>On Sat, May 2, 2009 at 2:34 PM, Gene Heskett <gene.hesk...@verizon.net> > >wrote: >>> Greetings; >>> >>> 1. The suggestions to rebuild the bayes db didn't make any difference. >>> 2. The error complains about the packing format of the db, when as near >>> as I can tell, it isn't packed, its plain text, or at least the >>> bayes_seen file is. And its nearly 9 megabytes. >>> >>> bayes_toks, OTOH, is inscrutable. and over 2 megabytes. >>> >>> Is there a way to check this bayes_toks file for validity, & maybe even >>> fix it, or should I just nuke all bayes_* and retrain?
To answer the question, I mv'd all the bayes_toks to .back files. SA immediately started working, probably better than it has for quite some time. It is now generating new bayes_toks files. So the old ones have now been nuked. I added some more delays to my scripts to make sure there weren't any potentials for access clashes. They may have been too short. -- Cheers, Gene "There are four boxes to be used in defense of liberty: soap, ballot, jury, and ammo. Please use in that order." -Ed Howdershelt (Author) If a train station is a place where a train stops, what's a workstation?
