bogofilter/bogoutil concurrency crash?

[David Relson]

At 07:45 PM 4/6/03, Jim Correia wrote:

>On Sunday, April 6, 2003, at 09:37  PM, David Relson wrote:
>
>>Interesting...  I think another Mac OS-X person will need to confirm this 
>>problem.  A quick test on my Linux box worked properly.  In the grand 
>>tradition of pointing the finger, my guess is that it's either a locking 
>>problem or a berkeleydb problem.
>>
>>Have you run "make check"?
>
>I hadn't before you asked.
>
>Hopefully the output means something to whomever wrote the check tests :-) 
>(see below)

It does.  All tests should PASS.  As you can see, t.lock2 (lock test #2) 
FAILed with a whole slew of Segmentation Faults.  There should be _no_ seg 
faults.

>>In actual usage, it's rarely necessary to run bogoutil.
>
>In practice it isn't necessary to purge old words or any of the other 
>maintenance functions?

It depends on how busy your site is.  I've been running bogofilter for 6 
months for my small domain.  Over that period of time, approx 40,000 
messages have come in.  The current size of my wordlists are 9.9M for 
goodlist.db and 3.7M for spamlist.db.  I haven't needed to do maintenance.

>>For those times when you need to dump the database, you _could_ make a 
>>copy of it and dump the copy.  That would avoid any delays caused by 
>>bogoutil locking the production database.
>
>Ok. I'm not a procmail expert - is there a way I should write my 
>bogofilter recipe so that if it happens to crash, things go on reasonably?

At this point, I'd say your system isn't ready to put bogofilter into 
production.  We need to hear from (1) Matthias, who wrote the locking code 
and knows it inside and out (which I don't); and from (2) other OS-X users 
to hear what they have to say.

David