Problems with bogofilter getting stuck

[Matthias Andree]

On Wed, 17 Mar 2004, Oded Arbel wrote:

> > you.  There's also a db_recover in the package.
> 
> I found db_recover, but while db_verify accepts a database file, db_recover 
> apparently wants to be run with the "home of the db environment" and doesnot 
> like to accept db files as a parameter. I tried to run it 
> with /var/lib/bogofilter (where my wordlist file is) as the "home", but it 

That was correct - bogofilter will use the parent directory of the
wordlist.db file as a Berkeley DB "home directory".

> just said :
> db_recover: Recovery complete at Wed Mar 17 04:14:52 2004
> db_recover: Maximum transaction id 80000000 Recovery checkpoint [0][0]
> 
> and didn't do anything. googling also got me nowhere. can you please explain 
> how can I fix the wordlist file ?

No released bogofilter version, stable or developer version, stores
information to recover from application crash, system crash or write
error events ("disk full") that happen at the wrong moment.

If you have your original spam and ham mailboxes still around, remove
the data base and re-build it from scratch.

If you don't have them around, you can try using db_dump with the -r
option added and load the data with db_load - but you'll have to check
if the data looks complete; i. e. if it doesn't suddenly stop at a
particular letter and still has the .MSG_COUNT magic token. While this
will fix the "getting stuck" condition, there are no guarantees this
will result in a data base that gives useful filtering results.

Code that addresses this data base corruption problem is under
development but not yet suitable for public consumption (development
takes place on the branch-db-txn branch in CVS) - there are unresolved
issues with concurrent access that are symptoms of bugs in the code that
need to be tracked down and fixed.

-- 
Matthias Andree

Encrypt your mail: my GnuPG key ID is 0x052E7D95