On Tue, Oct 13, 2015 at 11:55:16AM -0400, Phil Stracchino wrote:
> > That release also has no smtputf8 support, so it is difficult to
> > see why that might be better than 3.0.x with smtputf8 disabled.
>
> You make a good, but mystifying, point.... if smtputf8 was not
> supported in 2.11, and redelivery and miss retraining were both working
> in 2.11, then smtputf8 cannot be a requirement either for redelivery or
> for miss retraining. So why does miss retraining work on 3.0.x with
> smtputf8 enabled, yet fail if I turn it off...?
I think you're reading causality into coincidence.
> If I turn off smtputf8 on 3.0.x and attempt a FP redelivery, the log
> looks as though it then runs into problems with opendkim:
I don't see an OpenDKIM problem, I see a problem with an invalid
recipient address:
> Oct 4 14:01:00 epsilon3 postfix/smtpd[11430]: connect from
> localhost[127.0.0.1]
> Oct 4 14:01:00 epsilon3 postfix/smtpd[11430]: 66BC5157019:
> client=localhost[127.0.0.1]
> Oct 4 14:01:00 epsilon3 postfix/cleanup[11433]: 66BC5157019:
> message-id=<20151004180100.66bc5157...@epsilon3.caerllewys.net>
> Oct 4 14:01:00 epsilon3 opendkim[3414]: 66BC5157019: s=mail d=houzz.com SSL
> Oct 4 14:01:00 epsilon3 opendkim[3414]: 66BC5157019: bad signature data
> Oct 4 14:01:00 epsilon3 postfix/qmgr[11405]: 66BC5157019: from=<>,
> size=132690, nrcpt=1 (queue active)
> Oct 4 14:01:00 epsilon3 postfix/smtpd[11430]: disconnect from
> localhost[127.0.0.1] helo=1 mail=1 rcpt=1 data=1 quit=1 commands=5
> Oct 4 14:01:00 epsilon3 postfix/local[11434]: 66BC5157019:
> to=<????@caerllewys.net>, orig_to=<????>, relay=local, delay=0.56,
> delays=0.46/0.01/0/0.09, dsn=5.1.1, status=bounced (unknown user: "????")
The local(8) delivery agent is processing a message with an invalid
(non-printable) recipient.
The message in question had no "Message-ID", so one was added by
cleanup(8) (date.queue-id). So we can't correlate by message-id
or by the sender (<>). Perhaps you could put some such messages
on "HOLD", and report what envelope addresses they really have.
--
Viktor.
