Hello -

Just installed a 6.0 snapshot (from yesterday, though then rebuilt),
and noticed what seemed to be a possible regression (or otherwise PEBCAK)
in smtpd for my specific dumb-host configuration -

Basically, configuration is as follows:

- configuration:

  - Host is test VM behind private workstation NAT on company net
  - /etc/mail/mailname set to hostname of the physical workstation
    hosting the VM so RDNS matches when connecting to internal
    MX and mail is not rejected
  - /etc/mail/aliases has 'root: [email protected]'
  - /etc/mail/smtpd.conf : stock

The internal MX simply accepts/relays mail for internal IP's
after some basic sanity checks, so typically I just get the mails -

More specifically:

On 5.9 VM in same configuration (possibly 5.8? not sure when I switched
over from postfix here), and definately OpenSMTPD 5.9.1,
mails to root via e.g. nightly cron or simple test of:

echo 'test' |Mail -s test root

would result in mail flowing to personal inbox ([email protected])
with:

From: Charlie Root <[email protected]>
To: [email protected]

However, in OpenSMTPD 6.0, mails are delivered as:

From: Charlie Root <[email protected]>
To: root < >

Which seems a bit malformed / unexpected. Further,
though this is probably a problem with the internal MX,
CC's forwared to a personal offsite address vai inbound
sieve rules are filled in with the gateway hostname here
which is quite bizarre.

Replacing my /usr/src/usr.sbin/smtpd on the 6.0 snapshot box
with the 5.9 branch built an smtpd with 'expected' behavior -

After reading a bit of the enqueue.c code and other items which didn't
seem to differ on what I thought to be the important bits,
I did review logs from 'smtpcl trace all' against both smtpd versions on the
6.0 snapshot, and didn't see any obvious differences in the imporatnt parts,
e.g. envelope lines in log, SMTP chat transcripts, etc -

my next test would probably be to block the outbound SMTP and zap smtpd
before it can relay so I can check the queue files, and maybe try to
disable TLS and  grab a packet dump, but thought I might mention
this "as is" in case there are any obvious 'gotchas'/'eurekas' or
something that I could be mis-doing. Seems like the kind of
thing an expert could quickly fix or hit me with a cluestick about :)
For now, this snapshot is running on the 5.9.1 copy chugging on a
dpb and all is well.

Happy to provide more details as is needed.

Thanks,

- Chris

Reply via email to