Re: Re: NULL context in mlfi_close ?

[mhrivnak]

As an update, I've tried three different versions of spamass-milter and I've
also recompiled sendmail myself, all with the same exact result as below. 
Could this possibly be a result of security settings in Mandrake?

Thanks again!

Michael

>On Thu, 03 Jul 2003 22:08:53 -0500 Dan Nelson <address@hidden>
wrote.
>In the last episode (Jul 03), mhrivnak said:
>> I used version 1.2 with Slackware 8.0 for quite some time, but have
>> recently done a fresh installation of mandrake 9.1.  The system
>> serves as firewall/home router/mail server/web server etc.
>> 
>> I believe I have everythng configured, but I get these errors in syslog:
>> 
>> Jul  3 18:33:14 gso167-140-251 sendmail[16161]: h63MXEWU016161: Milter
(spamassassin): to error state
>
>There should be a line above this one with a better error message.
>
>> Jul  3 18:33:14 gso167-140-251 spamass-milter[16162]: mlfi_close
>> Jul  3 18:33:14 gso167-140-251 spamass-milter[16162]: NULL context in
mlfi_close! Should not happen!
>> Jul  3 18:33:14 gso167-140-251 sendmail[16161]: h63MXEWU016161:
from=<address@hidden>, size=483, class=0, nrcpts=1,
msgid=<address@hidden>, proto=ESMTP,
daemon=MTA, relay=internal-gw.uncg.edu [152.13.2.73]
>> Jul  3 18:33:14 gso167-140-251 sendmail[16163]: h63MXEWU016161:
to=<address@hidden>, delay=00:00:00, xdelay=00:00:00, mailer=local,
pri=30700, dsn=2.0.0, stat=Sent
>> 
>> I have all the updates applied in Mandrake 9.1, so it's sendmail
>> 8.12.9.  I have the latest spamd configured and running.  I've
>> recreated sendmail.cf as instructed in the README, and continuously
>> get an error that sendmail.cf has a duplicate milter entry.  Sure
>> enough, every time I recreate sendmail.cf, there are two identical
>> entries relating to the milter.  I've been simply commenting-out one
>> of them.
>
>You positive you don't have two INPUT_MAIL_FILTER lines in your .mc
>files?  m4 is sort of weird and I've found that even commented macros
>sometimes get expanded.
>
>As for the NULL context error, from my reading of the milter docs, it
>really shouldn't ever happen.  If the context value was NULL, the
>milter would have segfaulted long before getting to mlfi_close :)  I
>suppose it could be a bounds overflow just before mlfi_close is called.
>
>The NULL context thing shouldn't stop the milter from working.
>Something is definitely wrong, but since the connection is closing
>anyway, I don't make it a fatal error.  Are messages getting tagged?
>
>-- 
>       Dan Nelson
>       address@hidden