https://bz.apache.org/SpamAssassin/show_bug.cgi?id=7191

--- Comment #6 from Andrew <[email protected]> ---
Thank you for the response, yea sure, I can try it, jsut patched by commenting
this line, truncated the txrep db. Some time will be needed until the db is
filled with some data to test.

By the way, I have also reported here another problem wit TxRep
(https://bz.apache.org/SpamAssassin/show_bug.cgi?id=7269) maybe you can take a
look at it or know something about it..

And a little thing about correct user processing to SA: when incoming mail
arrives we detect the local user by the RCPT address when calling spamc from
MTA. 
And what correct to do when this is outgoing message (that is sent to MTA
thourgh SASL AUTH session from our users) ?
Default scripts that described in the official manual do not process this and
the user for outgoing mail will not be "OUR" user, it is still RCPT user, that
is some foreign address.. so the record of outgoing mail have username and
email fields identical. This is correct or not?

I have modified my script a little by providing the SASL AUTH name to it and:
postfix SA calling arg:
argv=/usr/local/bin/mail-scanner.sh ${sasl_username} ${sender} ${recipient}

and the script part:

SASL_USERNAME=$1
shift

SENDER=$1
shift

SPAMC_USER="vmail"

if [ "${SASL_USERNAME}" != "" ]; then
    SPAMC_USER=${SASL_USERNAME}
else
    SPAMC_USER=$1
fi

#... bla bla bla AV part...
#...and

${SPAMC} --u ${SPAMC_USER} -E <$INPUT >$OUTPUT
if [ "$return" = 1 ]; then
    logger -p mail.info "SpamAssassin found spam, but we going on"
elif [ "$return" != 0 ]; then
    logger -s -p mail.warning -t scanner "Temporary SpamAssassin failure
$return, delivering too, cause we do not want to loose mail"
    OUTPUT=$INPUT
fi

${SENDMAIL} -oi -f $SENDER "$@" <$OUTPUT

as you see this modified script checks if that was sasl'ed session and if yes -
uses the spamc user from sasl login. so after that we have outgoing mail
processed by userprefs of local sender.
I think this is correct to make txrep_whitelist_out function working when the
database has separate users.. I am right or... ? If yes, I think that it should
be described in the manual that for correct processing of outgoing mail there
should be a special case to detect the local user from sasl or sender (sasl is
most accurate I think).. You can use my script as an example then.. ? and of
course This is related not only for TxRep but for any part of SA where the
local username is used (userprefs, bayes, etc)

-- 
You are receiving this mail because:
You are the assignee for the bug.

Reply via email to