Re: Converting old wordlist.db to Berkeley format?

Am 02.09.2016 um 10:28 schrieb Geoff:
> I have been using bogofilter since 2003, and have a large wordlist.db (154Mb),
> that has been regularly compacted, using the pre-0.93 bogofilter instructions
> in the FAQ. Negligently I have stayed with an old self-compiled version of
> bogofilter which has been perfectly adequate for my needs.  One reason for that
> is that I vaguely recall (it was ages ago), having problems when the change to
> the Berkeley DB was made.
>
> For reasons with which I need not bore you, I now want to use the current
> (1.2.4) bogofilter from my Arch distro repo.  If I run ..
>
> bogoutil -d wordlist.db | bogoutil -l wordlist.db.new
>
> .. the command generates: 573 log files, a 115.3 Mb wordlist.db.new, __db.001,
> __db.002, __db.003, lockfile-d and lockfile-p 
>
> The resulting wordlist (renamed to wordlist.db, of course) does not, however,
> appear to be usable:
>
> db_verify: /home/me/.bogofilter/wordlist.db: No such file or directory
> BDB5105 Verification of /home/me/.bogofilter/wordlist.db failed.
>
> Attempts to use bogoutil -w fail with :
> Error accessing file or directory '/home/me/.bogofilter/wordlist.db'.
> error #2 - No such file or directory.
>
> .. even though the directory and file are certainly present.

Geoff,

that's an interesting failure mode. It seems that something's goofed
with the __db.* files, and that can confuse Berkeley DB. The Berkeley DB
library is normally robust, but if it is made to operate on __db.* files
from an older bogofilter version it can feel quite screwed up even if
the database is still fine.

The log files are usually fine though, and also precious; if they are of
an older incompatible format, Berkeley DB will just start writing to a
new log file and go on, and if you are sure no bogofilter/bogoutil
program is running at the time, deleting the __db.* files is safe.

I don't know exactly if it's worth several round trips of my asking
questions and your answering them just for analysis purposes so let's
keep it short; the key questions are:

- Is the bogofilter version (along with the necessary libdb) that you've
used all the years still available to you, from a backup?
- what was the old bogofilter version?
- what was the Berkeley DB version that bogofilter used to use?
- what is the current Berkeley DB version that bogofilter is attempting
to use.

The canonical way to bring your old wordlist.db to a new installation
would be - if the old versions are still available to you:

0. stop your mail system and cron jobs that could call bogofilter and
related programs
1. put the old database (along with log.* files if any) and
executables/libraries back in place. Assume that it were called
bogoutil_old (no need to rename it).
2. delete the __db.* files (they will be recreated as needed): rm __db.*
3. using the old code, try to dump the old data: bogoutil_old -d
~/.bogofilter/wordlist.db >~/.bogofilter/wordlist.txt     ## if that
complains, we need something else, see below.
4. rename the .bogofilter directory: mv .bogofilter .bogofilter.old
5. create a new one: mkdir .bogofilter
6. if necessary, use cp -p to copy old configuration file from the old
into the new directory (if you've had any inside, which isn't too
common, might be a DB_CONFIG)
7. install the new bogofilter 1.2.4 executables and the libdb
8. import the dumped data: bogoutil_new -l ~/.bogofilter/wordlist.db
<~/.bogofilter.old/wordlist.txt
9. check if everything works, if needed, restart your mail system.
Note that command line options and output have changed since the old
days of bogofilter 0.92.X - be sure to read the RELEASE.NOTES file.
10. if so, remove the .bogofilter.old directory.

The "something else", for instance, if you don't have backup copies of
the old bogofilter/bogoutil executables or the database, that you've
used all the time, is described in the new bogofilter version's
doc/README.db file, it shows several recovery methods in order of
ascending level of desperation in section 3.2.  If the database itself
is still intact.


Hope that helps.

Cheers,
Matthias

_______________________________________________
bogofilter mailing list
[email protected]
https://www.bogofilter.org/mailman/listinfo/bogofilter