database corruption

[David Relson]

At 12:18 PM 5/28/03, Dave Lovelace wrote:
>Fred Yankowski wrote, in part:
> >
> > I too use "bogofilter -u" from procmail scripts without any explicit
> > locking.  Is that wrong?
> >
>Fred,
>I think I confused you.  (It sounds as though David understood me, though.)
>I think bogofilter does adequate locking (or what is hoped to be adequate).
>I was wondering whether updates might occur during the period when I
>ran db_verify, updates which might have caused the very errors I saw.
>(Which would make them entirely spurious, I hope.)
>
>I need to do some more testing ...
>--

I missed the procmail part of Fred's message...

Bogofilter does its own locking.  It seems to work quite well, though 
obviously _something_ is causing trouble.  After repairing your wordlists, 
create a cronjob to snapshot it.  If problems occur again, you'll have an 
easy way to fix them.

If problems _keep_ happening, then adding procmail locks are worth 
trying.  If that cures the problem, it's information that would be very 
good to know.

By the way, as part of the source distribution, the Makefile supports "make 
check".  Tow of the tests is for locking.  Test src/tests/t.lock2 runs 16 
simultaneous copies of bogofilter trying to cause trouble.  We've not had 
reports of problems with "make check", though it's not clear how widely 
it's been run.