On 1/15/19 8:02 PM, David B Funk wrote:
It's a bit tricky to implement a milter correctly because people often don't understand that the message which sendmail hands to a milter is as-received from the incoming network connection.Any locally added stuff (EG the "Received:" header) isn't in that milter stream.Thus the milter must completely/correctly synthesize all locally added headers.
I might have known that at some point in the past but I definitely forgot about it until today.
Actually the spamass-milter method (calling spamc) makes it easier to debug.
Wait. What‽I thought spamass-milter used the same protocol to communicate with spamd that spamc does. Somehow I was not aware that spamass-milter actually used spamc to communicate with spamd.
Just create a script which wraps spamc in-between a couple of "tee"s to capture stdin & stdout and you'll have everything you want to know.
<Face Palm>
A simple example which ignores signal handling:
#!/bin/sh
# 'spamc' debugging script
FILE_NAME="/var/tmp/spamc-transcript-$$"
echo "spamc args: $*" "" > ${FILE_NAME}.in
tee -a ${FILE_NAME}.in | /real/path/to/spamc "$@" | tee ${FILE_NAME}.out
Adjust paths as needed.
*nod*nod* Thank you for the information Dave. -- Grant. . . . unix || die
smime.p7s
Description: S/MIME Cryptographic Signature
