be sure to open a new transcript on every queue run; when returning
an SMTP transcript, only send the last half (i.e., that part which
describes the sending half).

allow user-setable error codes in "error" mailer

fix a botch in argument processing such that it used the last flag as
a recipient if there were no other recipients (e.g., with "-t")

fix (one more time!) long line wrapping.....

split up FULLSMTP flag -- now have "X" (M_XDOT, use hidden dot algorithm),
"p" (M_FROMPATH, use reverse-path in MAIL FROM:<> command), and "L"
(M_LIMITS, enforce SMTP line limits). I would like to c

split up FULLSMTP flag -- now have "X" (M_XDOT, use hidden dot algorithm),
"p" (M_FROMPATH, use reverse-path in MAIL FROM:<> command), and "L"
(M_LIMITS, enforce SMTP line limits). I would like to change the format
of the mail defn one more time to make named fields so that it would
be more extensible.

show more ...

fix folded line output

curiouser and curiouser..... a read error on the reply from a QUIT will
cause another QUIT command, and thus an fclose(NULL). Some sites seem
to be sending a 421 Shutting down and then closing with

curiouser and curiouser..... a read error on the reply from a QUIT will
cause another QUIT command, and thus an fclose(NULL). Some sites seem
to be sending a 421 Shutting down and then closing without waiting for
the QUIT, so this happens....

show more ...

improve error recovery for bizarre SMTP cases; unfortunately CMU has
decided to start working again, so the strangest parts are untested.

externalize $g macro using ruleset 4 also

Don't try to drop our controlling TTY except when we are starting up a
daemon; this can cause UUCP jobs to hang waiting for carrier on a dialin
line (completely contrary to the intent). Many thanks

Don't try to drop our controlling TTY except when we are starting up a
daemon; this can cause UUCP jobs to hang waiting for carrier on a dialin
line (completely contrary to the intent). Many thanks to Keith Sklower
for pointing this one out. However, that damned parameter to disconnect()
has reappeared!

show more ...

Ignore SIGPIPE early (in main rather than in deliver) so that rude hosts
that close the connection early don't cause us to die -- in particular,
CMU-CS-A (nee CMU-10A) was doing this after the DATA c

Ignore SIGPIPE early (in main rather than in deliver) so that rude hosts
that close the connection early don't cause us to die -- in particular,
CMU-CS-A (nee CMU-10A) was doing this after the DATA command; the problem
went away quite mysteriously, so I have no idea why it happened

show more ...

take EHOSTDOWN as a temporary failure; change the format of "deferred"
message to be consistent with other messages.

fix botch that causes sendmail to core dump when there are large
numbers of requests

Detect an internal error that seems to be occuring

give more useful error messages (including the errno that caused
a temporary failure); reflect these messages out to mailq

have canonname pass the address through ruleset four also

take ENETUNREACH (Network unreachable) as a transient error -- this
probably isn't usually right, but it seem safer in the long run....

print "request" instead of "requests" if there is only one request

fix bug in commaize so that it won't core dump on every message (sigh);
process timeouts like normal errors so that they will get delivered to
the correct address (i.e., the Errors-To: address).

fix botch in reading the queue that caused it to throw away the last
entry it read -- this was particulary obvious with only one entry in
the queue.

Take errno 65, "Host is unreachable", to be a recoverable error

clean up output of mailq mode

Put CRLF on all lines in user SMTP. Pass this info down to
innumerable routines.

Use a more resilient algorithm to handle the special case of an SMTP
connection that is going down spontaneously (with a 421 reply code)

tag syserr's specially in the log so they can be grep'ed out easily