fix botch in computing e_oldstyle

fix bug in scanner state machine that never let you out of quote state.

increase log level needed to print "entered" message.

user SMTP fixes to talk to ISI

install new state-driven scanner; make everyone use it, thus fixing
problems of quoted commas, etc.

change $g processing from a macro substitution per mailer to two sets
of rewriting rules per mailer -- one each for sender and recipient
fields; convert to NBS standard on Return-Receipt-To: and Prec

change $g processing from a macro substitution per mailer to two sets
of rewriting rules per mailer -- one each for sender and recipient
fields; convert to NBS standard on Return-Receipt-To: and Precedence:
fields; clean up From: processing code and generalize it to all
sender fields; tune debugging code; clean up canonname.
**** This delta invalidates previous configuration files ****

show more ...

add $* to match zero or more and $> to make a "subroutine" call; stick
in initial hooks for per-mailer rewriting; improve diagnostics in readcf,
including number lines; increase the number of rewriti

add $* to match zero or more and $> to make a "subroutine" call; stick
in initial hooks for per-mailer rewriting; improve diagnostics in readcf,
including number lines; increase the number of rewriting sets.

show more ...

log entering uid & pid; allow CANONUSER ($:) to abort a rewriting set.

increase MAXMAILERS to 25

clean up reply code processing some more; in particular, give more
detail in many messages. process connection failures correctly.

handle comments and quotes in headers; still doesn't deal with
backslashes however -- prescan should be cleaned up to deal with
this case. it turns out there are many bugs in prescan; it should
be e

handle comments and quotes in headers; still doesn't deal with
backslashes however -- prescan should be cleaned up to deal with
this case. it turns out there are many bugs in prescan; it should
be extensively rewritten.

show more ...

fix serious botch in SMTP reply code delta; change a bunch of
"Internal error" codes to a new "Remote protocol error" -- so that
I don't get blamed for things that other systems do.

do a better job at interpreting SMTP reply codes

give 554 message on syserr (instead of 451) if errno == 0; as it
was some permanent errors appeared transient

hack crackaddr to strip blanks off the end of a cracked address; this
happens because "u at h" becomes "$g " (with two spaces at the end).
this doesn't solve the general case (e.g., "u at h (me)" co

hack crackaddr to strip blanks off the end of a cracked address; this
happens because "u at h" becomes "$g " (with two spaces at the end).
this doesn't solve the general case (e.g., "u at h (me)" comes out as
"$g (me)", with three spaces), but does handle the ugliest case.
besides, the "at" syntax is supposed to go away.

show more ...

simplify timeout code; allow multiple simultaneous queue runs so that
large messages don't freeze things up; fix EINTR problem in sfgets;
clean up canonname to be really correct; lots of misc. cleanup

release e_id after fork in daemon code; ignore events scheduled by
another process

assign a new queue id to all jobs coming in via TCP

sigh.... more fun and games in the daemon code to try to guess right.

recreate a socket after every failed accept()

apply ruleset 4 to rewrite addresses in the body of the message; fix
a clock.c bug that caused it to lose events; more time cleanup.

clean up time manipulation to always be current; assign job
id's earlier to make sure there is always one with a message.

log locked files in queuer; don't mistakenly start up two runqueue's

put queueup log messages on a higher logging level

log more info (on log level 11); try to detect wild accept loops;
some minor cleanup and debugging checks; fix NoConnect option to;
only apply to expensive mailers (regardless of sendqueue order)