[Imc-tech] Perl and SA

Arun Bhalla bhalla at uiuc.edu
Thu Jun 10 18:32:30 CDT 2004


On a whim, I tried db4.0_upgrade, and it seems to work.  (You're right
that db_dump/db4.0_load doesn't work -- I imagine the versions dump
in slightly different formats, and db4.0_load never complains.)

My work is at /tmp/bayes/old.  I imagine you could drop them in place.

\u@\h:\w\$ sa-learn -D --dbpath `pwd`  --dump all |& less
[...]
debug: bayes: 26649 tie-ing to DB file R/O /tmp/bayes/old/bayes_toks
debug: bayes: 26649 tie-ing to DB file R/O /tmp/bayes/old/bayes_seen
debug: bayes: found bayes db version 2
0.000          0          2          0  non-token data: bayes db version
0.000          0     117688          0  non-token data: nspam
0.000          0      64911          0  non-token data: nham
0.000          0     143062          0  non-token data: ntokens
0.000          0 1086293280          0  non-token data: oldest atime
0.000          0 1086732234          0  non-token data: newest atime
0.000          0 1086731609          0  non-token data: last journal sync atime
0.000          0 1086638818          0  non-token data: last expiry atime
0.000          0     345600          0  non-token data: last expire atime delta
0.000          0      25528          0  non-token data: last expire reduction count
0.002          4       1058 1086727778  H*r:66.218.67
0.551       2077        932 1086724589  largest
0.016        122       4096 1086721186  N:H*x:FN.N
0.696        129         31 1086725653  anti
[...]


Looks good!

Arun


"Zachary C. Miller" writes:
> Another unexpected victim of the Perl upgrade was Spamassassin's
> bayesian filtering.
> 
> Perl 5.6.1 used libdb2.
> Perl 5.8.4 uses libdb4.0. 
> 
> Files created by these versions are completely incompatible. 
> 
> If you find that you have some problems with Perl and any other old
> DB_File files please let me know and I'll see what I can do about
> fixing them.
> 
> I made some attempt to use db4.0_upgrade or db_dump/db4.0_load but
> neither worked. So in the end I had to just delete the
> /home/spamfilt/.spamassassin/bayes_* files and let the autolearner
> start from scratch.
> 
> This is why there has been increased spam volume over the past several
> days (since I upgraded SA). Spam volume should slowly but surely go
> down over the next few weeks as the bayesian filter relearns about all
> the variations of spelling for viagra and such. 
> 
> Unrelatedly, since yesterday there was a typo in
> /etc/spamassassin/local.cf which was my fault and which I have now
> fixed. That one invalidating some (but not all) of SA's rules and the
> errors caused by it filled up /var. Doh!
> 
> -- 
> Zachary C. Miller - @= - http://wolfgang.groogroo.com/
> IMSA 1995 - UIUC 2000 - Just Another Leftist Muppet - Ya Basta!
>  Social Justice, Community, Nonviolence, Decentralization, Feminism,
>  Sustainability, Responsibility, Diversity, Democracy, Ecology
> _______________________________________________
> Imc-tech mailing list
> Imc-tech at urbana.indymedia.org
> http://lists.cu.groogroo.com/cgi-bin/listinfo/imc-tech
> 

--
Arun Bhalla



More information about the Imc-tech mailing list