You are viewing a plain text version of this content. The canonical link for it is here.
Posted to dev@spamassassin.apache.org by bu...@bugzilla.spamassassin.org on 2004/07/15 07:42:10 UTC
[Bug 3483] bayes _journal not syncing to _toks
http://bugzilla.spamassassin.org/show_bug.cgi?id=3483
felicity@kluge.net changed:
What |Removed |Added
----------------------------------------------------------------------------
Summary|[review] bayes _journal not |bayes _journal not syncing
|syncing to _toks |to _toks
------- Additional Comments From felicity@kluge.net 2004-07-14 22:42 -------
reverting out of review state for now.
ok, so the last patch is good, but it doesn't actually fix the problem. the main problem is that the
opportunistic calls are made from scan(), but scan() isn't called _at all_ if is_scan_available() fails.
I need to think about this some more, but I'm thinking that the autolearn code ought to call the
opportunistic code as well, and it really ought to be both, here's the thoughts:
- scan() will cause journal updates with timestamps, but autolearning may be disabled. so if the opp
calls are only made from autolearning, the journal can grow huge from the timestamp updates and
never sync.
- autolearning can occur when scanning doesn't (this ticket), and autolearning actually also adds tokens
which is what causes the need for an expire.
a possibility is to remove the call from scan() and put it in PMS::check() instead, which can call it post
scan and autolearn. the problem is that autolearn, and check, don't require the bayes db to be tied at
all, which the opportunistic calls do, as does scan, which is why it's there now... hrm. what a bugger.
ok, new thought:
- move it from scan() to is_scan_available(). the function is called a lot, so the opportunistic calls will
occur multiple times per message which isn't good but it's not terrible. if we add in a flag to
opportunistic calls that specifies journal_only, that'll let it exit really quick if there's nothing to be done.
thoughts?
------- You are receiving this mail because: -------
You are on the CC list for the bug, or are watching someone who is.