On Tue, Feb 18, 2025 at 07:59:49AM +0000, Mark Delany via mailop wrote:
> I'm seeing a curious submission failure with the latest macOS Mail.app
> (Sequoia 15.3.1)
> and wondering whether others are seeing it also. The submission mechanism is
> via an
> stunnel into an SMTP server and stunnel/openssl was also recently upgraded so
> it may be a
> server-side issue thus my survey here to help me disambiguate.
>
> A trace of the traffic shows that when Mail.app sends an email with an
> attachment it fails
> to send the final "." <CRLF> sequence to terminate the DATA and thus the
> Submission server
> discards the email when the socket subsequently closes.
Is this regardless of the type of attachment sent? I just tested a
message with an attached ".txt" file, and it got through submission into
Postfix just fine.
System Software Overview:
System Version: macOS 15.3.1 (24D70)
Kernel Version: Darwin 24.3.0
Feb 19 01:14:04 amnesiac postfix/submission/smtpd[2791163]: connect from
unknown[...]
Feb 19 01:14:04 amnesiac postfix/submission/smtpd[2791163]: Anonymous TLS
connection established from unknown[...]: TLSv1.2 with cipher
ECDHE-RSA-AES256-GCM-SHA384 (256/256 bits)
Feb 19 01:14:04 amnesiac postfix/submission/smtpd[2791163]: AAC1193558A:
client=unknown[...], sasl_method=GSSAPI, sasl_username=...
Feb 19 01:14:04 amnesiac postfix/cleanup[2791167]: AAC1193558A:
message-id=<...>
Feb 19 01:14:04 amnesiac postfix/qmgr[1110399]: AAC1193558A: from=<...>,
size=5183, nrcpt=1 (queue active)
Feb 19 01:14:04 amnesiac postfix/virtual[2791168]: AAC1193558A: to=<...>,
orig_to=<...>, relay=virtual, delay=0.14, delays=0.13/0.01/0/0, dsn=2.0.0,
status=sent (delivered to maildir)
Feb 19 01:14:04 amnesiac postfix/qmgr[1110399]: AAC1193558A: removed
Feb 19 01:15:04 amnesiac postfix/submission/smtpd[2791163]: disconnect from
unknown[...] ehlo=2 starttls=1 auth=1 mail=1 rcpt=1 data=1 quit=1 commands=8
> When Mail.app sends an email without an attachment, the final "." <CRLF> is
> presented and
> of course the Submission server accepts the submission.
>
> First question. Is anyone else seeing this? It seems like such a huge
> regression that
> Apple could not have possibly made a release with this bug so I'm looking
> elsewhere.
It seems there are other factors involved.
> Second question. Do submission servers ever accept a submission if the "."
> <CRLF> sequence
> is not presented prior to the socket closing? I can't imagine they do but it
> might be a
> rationale for why Mail.app let this behaviour slip thru.
None that I know of. That would result in transmission of truncated
messages. The only way to skip "." is to use BDAT.
--
Viktor.
_______________________________________________
mailop mailing list
[email protected]
https://list.mailop.org/listinfo/mailop
