New bogofilter TXN snapshot available.

[Matthias Andree]

Torsten Veller <ml-en at veller.net> writes:

> * Matthias Andree <matthias.andree at gmx.de>:
>> Please test this to help us make 0.93.0 a good release.
>> 
>> Be sure to read the RELEASE.NOTES-0.93 and doc/README.db files before use!
>
> While upgrading I had to dump and reload my wordlist. I read the documents but
> couldn't find anything saying, that this is necessary.

I wasn't aware this is necessary, and have just tried Berkeley DB
versions 3.3, 4.0, 4.1 and 4.2, and only 4.1 shows this behaviour, all
other versions are fine when it comes to bogofilter 0.92.99 reading a
data base that 0.92.8 has created and written. DB 4.1 even segfaults
when trying to write...

> Is it necessary, am i blind, or is/was there something wrong with my old
> database?

db_verify can tell if the DB is still intact.

*shrug* I don't know what the issue is. On first run of the new version,
it starts off as checksum error:

checksum error: catastrophic recovery required
PANIC: DB_RUNRECOVERY: Fatal error, run database recovery
wordlist.db: pgin failed for page 0
fatal region error detected; run recovery
fatal region error detected; run recovery
fatal region error detected; run recovery
fatal region error detected; run recovery
fatal region error detected; run recovery
fatal region error detected; run recovery
fatal region error detected; run recovery
fatal region error detected; run recovery
PANIC: DB_RUNRECOVERY: Fatal error, run database recovery
fatal region error detected; run recovery
Can't open file 'wordlist.db' in directory '/tmp/old41'.
error #22 - Invalid argument.

and on subsequent runs just:

wordlist.db: page 1: reference count overflow
bogofilter: (db) DB->get(TXN=134886032,  'delay' ), err: 22, Invalid argument

Apparently DB 4.1 doesn't like a change of the checksum flag, AFAIR, the
non-TXN versions didn't set the checksum bit. Will investigate.

-- 
Matthias Andree