I finally fix it. There was another bayes db, leave by the admin I have
replaced, under the home directory of the amavis user with 20 spam messages
learned and 30.000 ham learned. I dont know if they have even been learned
properly. I replaced it with the db of the root user and now is catching
almost every spam message.

I have also disable local test in amavis config file and now I run every
learn/test process as the amavis user. 

Thanks for all great support!!

Jaime.

-----Mensaje original-----
De: Matt Kettler [mailto:[EMAIL PROTECTED] 
Enviado el: viernes, 27 de enero de 2006 16:06
Para: Jaime Aguado
CC: users@spamassassin.apache.org
Asunto: Re: Configuration issue

Jaime Aguado wrote:
> Hi all, I have spamassassin 2.64 running on a SuSE Enterprise 9 
> running postfix as smtpd.
>
> My spamassassin system is leaking most spam messages (60-80%) without 
> tagging them.
> I have done some configuration tests:
>
> # spamassassin --lint gives no output error.
>
> # spamassassin < /tmp/spam-message
>
> [...]
> Subject: *****SPAM***** Special Alert to Investors
> Date: Sat, 28 Jan 2006 00:22:23 +0000
> Message-Id: <[EMAIL PROTECTED]>
> X-Spam-Flag: YES
> X-Spam-Checker-Version: SpamAssassin 2.64 (2004-01-11) on 
> trueno.fresh-it.com
> X-Spam-Level: *********
> X-Spam-Status: Yes, hits=9.7 required=4.0 
> tests=BAYES_99,DATE_IN_FUTURE_12_24,
>         MIME_BASE64_TEXT autolearn=no version=2.64
> X-Spam-Report: 
>         *  5.4 BAYES_99 BODY: Bayesian spam probability is 99 to 100%
>         *      [score: 1.0000]
>         *  1.0 MIME_BASE64_TEXT RAW: Message text disguised using 
> base64 encoding
>         *  3.3 DATE_IN_FUTURE_12_24 Date: is 12 to 24 hours after
Received:
> date
> [...]
>
> # spamc -R < /tmp/spam-message
> [...]
> Content analysis details:   (8.6 points, 4.0 required)
>
>  pts rule name              description
> ---- ----------------------
> --------------------------------------------------
>  5.4 BAYES_99               BODY: Bayesian spam probability is 99 to 100%
>                             [score: 1.0000]
>  1.0 MIME_BASE64_TEXT       RAW: Message text disguised using base64
> encoding
>  2.2 DATE_IN_FUTURE_12_24   Date: is 12 to 24 hours after Received: date
>
> The args passed to spamd are: SPAMD_ARGS="-d -c -a -L"
>
> Finally, the message reaches the inbox mail client untagged as spam. 
> What I am doing wrong?

Your tests are being run as the root user, what user is your mail being
scanned as? Unless you're using a bayes_path declaration, the bayes
databases could be different, resulting in different scores.

Also, for what it's worth, it's generally bad form for spamd to scan mail as
root. SA 3.x actualy refuses to do so and if you try will wind up using the
"nobody" user.



Reply via email to