Dumped version 2.99.19

[Matti Aarnio <mea@nic.funet.fi>]

Hello folks,

	My lattest source dump is available at:
ftp://ftp.funet.fi/pub/unix/mail/zmailer/zmailer-2.99.19-951018.tar.gz


ZMailer highlights:
	- The new scheduler works in such a manner that I begin
	  believe in it -- all loose ends on process accounting
	  are now fixed (now I can believe it ;^) )

	- Optionally the new scheduler can use a bi-directional
	  pipe -- socketpair() -- and thus save one fd per child.
	  Most hostenv/ -descriptors have been added a socketpair(),
	  IF I knew the host has it, or is likely to have it...
	  ("likely" == "feels like having BSD networking in the box..")

	- proto/cf/canon.cf, proto/cf/aliases.cf
	    contain a fix that lets system to use aliases
	    (in the $MAILVAR/db/aliases, or .forward, or..)
	    with contents like:
		foo-test: "|fooproc foo:+foo.foo"
	    ( Basically the same should be done for
		foo-test: "/foopath/foo:+foo.foo"
	      but it is late at night, and I have an early wakeup.. )

	    This contains the FIRST case of  "ssift"  statement in
	    the scripts.  It simply was not doable with "tsift" stmt.

/Matti Aarnio <mea@nic.funet.fi>

(Hmm... I just pulled a 3 MB file from tsx-11.mit.edu at 65 kB/s into
 here to see how fast the our new lines are -- unbelievable! I suppose
 our outbound FTP has some speed limiters so USA people won't get quite
 these rates...  We upgraded our international 4 Mbps link to 8 Mbps
 some 30 hours ago, and already after 2 h use, the long-term average of
 the flow had jumped to 5 Mbps out from Finland. Inbound average indicated
 3 Mbps at the same time...  You can guess, why such radical difference ;-)
 Anyway, USA is closer to us, than the rest of the Europe!  NORDUnet lines
 are roughly a decade faster, than anything south from here... :)  )


---------------- from ChangeLog ----------------
Wed Oct 18 09:10:07 1995  Matti Aarnio  <mea@nic.funet.fi>

	* Released 2.99.19

	* hostenv/*, scheduler-new/pipes.c, scheduler-new/transport.c,
	  scheduler-new/threads.c, scheduler-new/Makefile.in:
	    If the system can do a bi-directional pipe, at least
	    a socketpair(), we need only ONE fd per child.  Enable
	    by hostenv define  SOCKETPAIR=
	    (TODO: SysV has STREAMS-pipes, add support for them into
	           the  scheduler-new/pipes.c, then all three existing
		   methods are covered... )
	    ( scheduler-new/pipes.c  is a new module.. )

	* scheduler-new/threads.c, scheduler-new/transports.c:
	    Finally found (?) the reason, why occasionally running
	    childs are lost, and mis-calculated.  It turned out to
	    occur most frequent with  error-channel !
	    Anyway, sometimes a child is killed, and the recovery
	    was mis-treated, but the reporting-fd still had data!
	    Thus every now and then, while a process was gone, but fd
	    (and thus a pid-slot) was still in use, and the process
	    was thought to be valid for "feeding a child", for example..

	* transports/smtp/smtp.c:
	    When doing close() due to change of host, or a final close, don't
	    just close(), do "QUIT" + close()!

Fri Oct 13 17:21:27 1995  Matti Aarnio  <mea@nic.funet.fi>

	* hostenv/IRIX53, router/prototypes.h, scheduler*/mailq.c:
	    IRIX-5.3 diffs from  Darryl Miles <dlm@r-net.u-net.com>

	* scheduler-new/
	    - Still more hammering at it with DEC ATOM-tools Third Degree :)
	      Now only memory leakages around are in the system libc! (Brr...)
	    [ well, new leaks introduced latter, but that is another story.. ]

Thu Oct 12 20:23:26 1995  Matti Aarnio  <mea@nic.funet.fi>

	* smtpserver/smtpserver.c:
	    Fixed an error in SMTP EXPN, and VRFY.

	* scheduler-new/
	    - fixed expiry processing (when multiple recipients exist for
	      the message)
	    - fixed process counters -- idle_cleanup() leaked pids..

Mon Oct  9 11:07:25 1995  Matti Aarnio  <mea@nic.funet.fi>

	* scheduler{,-new}/transport.c:
	    Improved (?) support for IBM AIX systems
	    (select() system call.)

	* include/hostenv.h, hostenv/*:
	    STDLIB= -> USE_STDLIB -> include <stdlib.h> in  hostenv.h
	    UNISTD= -> USE_UNISTD -> include <unistd.h> in  hostenv.h

	* router/rfc822.c, router/rfc822hdr.c:
	    Hide away (effectively removed) "resent"-processing, so that
	    if there are headers in there, no new "Resent-" headers are
	    added just because some "Resent-Message-Id:" -exists, but no
	    "Resent-To:"...