+
+ /* If accepted, we aren't going to do any further tests below.
+ Otherwise, cache a real negative response, and get back to the right
+ state to send RCPT. Unless there's some problem such as a dropped
+ connection, we expect to succeed, because the commands succeeded above.
+ However, some servers drop the connection after responding to an
+ invalid recipient, so on (any) error we drop and remake the connection.
+ XXX We don't care about that for postmaster_full. Should we?
+
+ XXX could we add another flag to the context, and have the common
+ code emit the RSET too? Even pipelined after the RCPT...
+ Then the main-verify call could use it if there's to be a subsequent
+ postmaster-verify.
+ The sync_responses() would need to be taught about it and we'd
+ need another return code filtering out to here.
+ */
+
+ /* Remember when we last did a random test */
+ new_domain_record.random_stamp = time(NULL);
+