Kyle Wheeler
2005-06-17 18:01:11 UTC
Hello,
I've noticed some odd behavior in spamc/spamd, and I'm curious if it's
the "expected" behavior or if something is wrong. Specifically, in my
email headers, I see all the usual tags, and then autolearn=disabled.
Now, I want autolearning, I like autolearning, and it's done
autolearning in the past with no problem.
The reason I think it might be expected behavior is because of my setup,
which is a little odd. My mail gets scanned twice: once by spamassassin,
once by spamc, with a "spamassassin -d" in the middle. (I'm in the
middle of tentatively migrating from spamassassin to spamc/spamd --- I
used to have problems with spamc/spamd locking up and not processing
email, but with newer versions, I want to give it another shot.) I'm
using SpamAssassin 3.0.3 (stuck here until Debian decides it's worth
their while to release 3.0.4 into stable :P). I run the original
spamassassin with no flags, via procmail, like so:
:0 fw: spamassassin.lock
| env HOME=${HOME} /usr/bin/spamassassin
HOME is constructed, because this procmail is called by vpopmail's
delivery mechanism.
That's been working flawlessly, including auto-learning, but it's eating
lots of CPU. Now I'm adding this:
:0 fw
| spamassassin -d
:0 fw: spamassassin.lock
| spamc -U /var/lib/spamassassin/sock -u ***@mydomain.com
Spamd is running like this:
/usr/bin/spamd \
--syslog=stderr \
--max-children 5 \
--socketpath=/var/lib/spamassassin/sock --socketowner=vpopmail \
--socketgroup=vchkpw \
-H --create-prefs --vpopmail -u vpopmail 2>&1
Is it disabling the autolearning because it already autolearned it,
maybe? Or is it disabling the autolearning because of something else
that might be going wrong?
~Kyle
--
I am against religion because it teaches us to be satisfied with not
understanding the world.
-- Richard Dawkins
I've noticed some odd behavior in spamc/spamd, and I'm curious if it's
the "expected" behavior or if something is wrong. Specifically, in my
email headers, I see all the usual tags, and then autolearn=disabled.
Now, I want autolearning, I like autolearning, and it's done
autolearning in the past with no problem.
The reason I think it might be expected behavior is because of my setup,
which is a little odd. My mail gets scanned twice: once by spamassassin,
once by spamc, with a "spamassassin -d" in the middle. (I'm in the
middle of tentatively migrating from spamassassin to spamc/spamd --- I
used to have problems with spamc/spamd locking up and not processing
email, but with newer versions, I want to give it another shot.) I'm
using SpamAssassin 3.0.3 (stuck here until Debian decides it's worth
their while to release 3.0.4 into stable :P). I run the original
spamassassin with no flags, via procmail, like so:
:0 fw: spamassassin.lock
| env HOME=${HOME} /usr/bin/spamassassin
HOME is constructed, because this procmail is called by vpopmail's
delivery mechanism.
That's been working flawlessly, including auto-learning, but it's eating
lots of CPU. Now I'm adding this:
:0 fw
| spamassassin -d
:0 fw: spamassassin.lock
| spamc -U /var/lib/spamassassin/sock -u ***@mydomain.com
Spamd is running like this:
/usr/bin/spamd \
--syslog=stderr \
--max-children 5 \
--socketpath=/var/lib/spamassassin/sock --socketowner=vpopmail \
--socketgroup=vchkpw \
-H --create-prefs --vpopmail -u vpopmail 2>&1
Is it disabling the autolearning because it already autolearned it,
maybe? Or is it disabling the autolearning because of something else
that might be going wrong?
~Kyle
--
I am against religion because it teaches us to be satisfied with not
understanding the world.
-- Richard Dawkins