Hello,
as mentioned in the post from December 30th, 2022, (Subject: "Cyrus IMAP error
(?): 'Undefined error 0'..."), we've been seeing the following response from the
Cyrus imapd for successful delivery to a mailbox:
250 2.1.5 Undefined error: 0
SESSIONID=<imapd-1672619404-57025-2-15431856533157375633>
I started digging in the sources and found that between 3.0.8 and 3.4.4 the
following code was added to the top of lmtpengine.c:send_lmtp_error():
int code;
if (resp) {
int i;
for (i = 0; i < strarray_size(resp); i++) {
prot_puts(pout, strarray_nth(resp, i));
}
return;
}
Presumably the "prot_puts(pout, strarray_nth(resp, i));" would emit the correct
success notification ("250 2.1.5 Ok SESSIONID=<sess-id>"), but this depends on
"resp" being non-null, however, in lmtpd.c:deliver(), in near the bottom of the
loop which iterates through all the recipients, we see this line:
msg_setrcpt_status(msgdata, n, r, NULL);
Which causes msg_setrcpt_status() to always assign:
m->rcpt[rcpt_num]->resp = NULL;
Because resp is NULL, the "if (resp) {" block in send_lmtp_error() is not
invoked, and control continues at the switch statement, which tests (r). As r
contains zero (which indicates no error, I presume), the following matches:
case 0:
code = LMTP_OK;
break;
which causes this:
prot_printf(pout, error_message(code), error_message(r), session_id());
to emit the famous "250 2.1.5 Undefined error: 0 SESSIONID=..." line.
This would seem to be a bug.
The 3.0.8 send_lmtp_error() emits the success notification line when r == 0.
There seem to be some wires crossed about how to propagate success/failure status.
The 3.6.0 code seems to be the same as the 3.4.4 code.
-Rob
------------------------------------------
Cyrus: Info
Permalink: https://cyrus.topicbox.com/groups/info/T98f17a6ea7456fc1-M0902a38fac2c64d3f7fa3643
Delivery options: https://cyrus.topicbox.com/groups/info/subscription
