From: "Thomas Arend" <[EMAIL PROTECTED]>

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

Am Dienstag, 28. Dezember 2004 15:34 schrieb jdow:
> From: "Thomas Arend" <[EMAIL PROTECTED]>
>
> -----BEGIN PGP SIGNED MESSAGE-----
> Hash: SHA1
>
> Am Montag, 27. Dezember 2004 22:01 schrieb jdow:
> > From: "Morris Jones" <[EMAIL PROTECTED]>
> >
> > > Kevin Curran wrote:
> > > > Tests show that an email will get a different score depending on
> > > > whether spamassassin or spamc is called.
> > > >
> > > > What's up with that?
> > > >
> > > > Thanks!
> > >
> > > You probably need to stop spamd and restart it so it rereads the .cf
> >
> > files.
> >
> > > Cheers,
> > > Mojo
> >
> > Do remember that just before Christmas break I characterized a vaguely
> > similar problem with spamd. With per user rules enabled any given
> > spamd instance works perfectly the first time. The second time it will
> > appear to pick up the user rules but not the user scores. This is run
> > as the user with "DROPPRIVS" in the .procmailrc or as the user running
> > spanc. It is 100% repeatable here. Fortunately there is at the moment
> > only one user of the two here moved over to the new installation. So
> > moving to a direct spamassassin call seems to have eliminated the
> > problem, for now. I am waiting for someone to say they also can see
> > this effect. Then I'll go to the web (yuck) and file a BK report on it.
> > (I don't trust or like web user interfaces. {^_-})
> >
> > {^_^}
>
> I'm using SuSE 9.1 (latest updates) SA 3.0.2 with postfix, /etc/procmail
> and spamd/spamc. I get exactly the same scores (disregarding the AWL) for
> spamassassin and spamc/spamd.
>
> ---- my comments
> 1) Are you setup for per user rules in the ~/.spamassassin/user_prefs
file?
>    If not set up to do that. And setup a few simple rules and scores you
>    can test with text included in a test file.
> 2) Cut down the -m option for spamd to 1.
> 3) Restart spamd
> 4) Run spamassassin <testfile|more to get baseline scores.
> 5) Run spamc <testfile|more should be same as baseline scores.
> Now the kicker
> 6) Run spamc <testfile|more again. All scores picked up from user_prefs
>    will be 1 rather than the score in the user_prefs file.
>
> For reference I am using postfix not in its customary chroot jail,
> procmail with per user .procmailrc files, and spamd in the .procmailrc.
> But I do not have to send a mail through the whole system to see the
> effect. The above steps bypass most of the mail system and still show
> the effect. I make sure the test file includes strings designed to kick
> off rules. (I have a "JD_CHERRY_POPPED" rule and included "cherry popped"
> in the text I tested. I took a known spam for headers and put in my own
> text to force the user_prefs scores and rules.)
>
> On thinking this over from the description above I wonder if this is
> in some way connected with the growing spamd memory usage. Spamd does
> grow after the first run. I didn't look after the second. (I could if
> it's important.) It acted as if it thought it already had my scores
> and rules memorized. Yet it had forgotten the scores. It should have
> forgotten my rules, too. Then a second user would not have his mail
> contaminated by my rules. (Boys aren't as bothered by porn. {^_-})
>
> {^_^}

Hello again,

I can't reproduce this effect. For me all works fine. Scores are the same at
any time. But maybe I have not so much spam. So I have definitly on memory
shortage.

Thomas

< OK, that suggests something, I'm not sure what. I have a gigabyte of
< memory with a lot of it free. So it isn't a memory problem.

{O.O}


Reply via email to