spamassassin-dev October 2011 archive
Main Archive Page > Month Archives  > spamassassin-dev archives
spamassassin-dev: [Bug 6421] Spamd (3.2.5) seems to cut all mess

[Bug 6421] Spamd (3.2.5) seems to cut all messages over 163770 (+/- 2) bytes

From: <bugzilla-daemon_at_nospam>
Date: Sat Oct 29 2011 - 14:29:22 GMT

--- Comment #10 from Bjørn Mork <> 2011-10-29 14:29:22 UTC ---
Created attachment 5000
Logs showing the bug triggered with --ssl and not triggered without it

I should have done this before, but better late than never...

And I obviously forgot to mention an important part of my setup: I am using
spamass-milter. The MTA is sendmail.

It seems that spamass-milter is the missing part of the puzzle. I am not able
to recreate the problem by merely calling spamc directly, even if I give it
input which has previously failed (when spamc was called via spamass-milter).

But resending the same mail via sendmail, having it call spamc through
spamass-milter, will reliably trigger the problem.

In fact, I can reliably trigger it using *any* sufficienly large enough mail as
long as it goes through spamass-milter. But --ssl also affects this. The
problem does not show up unless I use --ssl between spamass-milter and spamd.

I am attaching a completely dummy "test.mail" which does trigger the problem
for me, as long as it is passed via ssl to spamd by spamass-milter. The mail
contains nothing but a large enough file full of null bytes. My tests
indicates that the actual contents doesn't matter, just the size.

I am also attaching "spamd-spamc-sendmail-log.txt" with logs showing
"test.mail" received twice: first using --ssl and failing, and then without
--ssl and working.

Note that I sent it to an external address which is forwarded back to me, as my
setup won't normally scan anything received from a host on the local network.
This was just done so because it was simpler than modifying the setup to make
spamass-milter scan local mail.


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