Unoffical Patch #17	(larry@garfield.cs.mun.ca Nov 19, 1991)

	- core dump due to selecting last_article+1 on screen

	- core dump due to consolidating (closing) last subject in a group
  	  under certain circumstances.

	- selecting consolidated articles with = command did not highlight
  	  the selected consolidated articles correctly.

	- fixed unconditional CLEAR TO END OF LINE that erased the next line on
  	  some terminals.

Unofficial Patch #18	(larry@garfield.cs.mun.ca;  Dec  1, 1991)

	- possible core dump/strange behavior when using '=' command to
	  return to the menu from article reading mode.
	  (Reported by Olaf Schlueter [olaf@oski.toppoint.de])

	- articles marked as A_LEAVE_NEXT ('=') are not given any special
	  attention in consolidated menu mode.  I considered this incorrect
	  (I leave stuff around all the time :) and modified nn's behavior
	  to notice them.

	- unable to de-select an auto-selected subject when using
	  consolidated-menu mode.

Patch #17 and #18 made official (peter@DIALix.oz.au)

Patch #19		(peter@DIALix.oz.au; 28 Nov 92)

Title:	Reduction in startup time
From:	Geoff Collyer <geoff@world.std.com>
	The initial startup time can be cut in half.

Title:	A/UX 3.0 support
From:	Jim Jagielski <jim@jagubox.gsfc.nasa.gov>
	Support for Apple A/UX 3.0, originally by Alexis Rosen

Title:	NeXT-2.2 support
From:	Lele Gaifax <lele@itnsg1.cineca.it>

Title:	Misc "static" function declarations corrected in prototypes.
From:	Lele Gaifax <lele@itnsg1.cineca.it>

Title:	NNTP name canolicalization in /usr/lib/nntp_server vs gethostname()
From:	Mark R. Levinson <chicken@minerva.npac.syr.edu>
	The code now converts the name from nntp_server to the official name
	before comparing it to the gethostname() result.  This is useful if
	nntp_server points to an alias (ie: news.your.domain.name)

Title:	Count lines if no "Lines:" field present.
From:	Paul Eggert <eggert@twinsun.com> and Luc Rooijakkers <lwj@cs.kun.nl>
	If there is no "Lines:" field, NN will count them.

Title:	Correct line counting algorithm conflict with NNTP and folders/digests.
From:	Peter Wemm <peter@DIALix.oz.au>
	The code would attempt to fetch a sub-article from a folder or
	digest from the NNTP server - which is clearly not possible...

Title:	Make line counting optional
From:	Peter Wemm <peter@DIALix.oz.au>
	#define DONT_COUNT_LINES option to config - it is undesirable to
	fetch all articles from a remote NNTP server just to display a 
	line count on the menu.

Title:	print a "??" instead of "-1" for lines count in some display modes
From:	J. Porter Clark <jpc@avdms8.msfc.nasa.gov>
	If you change the "layout", it is possible for NN to display a -1
	in the lines count field.

Title:	Timezone misalculations
From:	Paul Eggert <eggert@twinsun.com>
	NN mishandled non-GMT timezones in sorting resulting in articles being 
	presented out of order.

Title:	NNGRAB speedups
From:	Paul Eggert <eggert@twinsun.com>
	nngrab generated unnecessary files and processes (especially on
	BSD'ish machines that do not have a built-in "[" in the shell

Title:	NN wastes time seeking to find the file size
From:	Paul Eggert <eggert@twinsun.com>
	Make NN count bytes as they are read rather than seeking all over the 
	place.

Title:	Correct NN's header parser to RFC compliant behavior
From:	Paul Eggert <eggert@twinsun.com>
	NN gets confused on some variations of legitimate header continuation
	lines.

Title:	Remove defunct comp.ai.neural-nets digest special handling
From:	Paul Eggert <eggert@twinsun.com>
	The comp.ai.neural-nets digests are now in standard form.

Title:	Improve ":decode" to recognize more formats
From:	Mark Rawling <Mark.Rawling@mel.dit.csiro.au>
	Fix for fault in the chtbl code

Title:	Add a "ignore-formfeed" variable
From:	Romain King <romain@pryamid.com>
	Special request: display "^L" instead of a page break.  Not on
	be default.

Title:	Fix that darn "NO DIRECTORY (ok)" message that nnmaster is so fond of.
From:	Steve Simmons <scs@lokkur.dexter.mi.us>
	If nnmaster is going to complain about a missing directory, at least
	say which one it is....

Title:	get aux script to clean up common editor backup files
From:	Minh Tran-Le <mtranle@paris.intellicorp.com>
	Emacs makes file~ backups, some make file.bak, etc. Try and clean
	up a few of them after launching an editor.

Title:	make aux script generate control messages that INN will accept.
From:	Peter Wemm <peter@DIALix.oz.au>
	INN's inews is pretty fussy in what it will accept..

Title:	Resolve conflict NN with INN on SVR4/386
From:	Peter Wemm <peter@DIALix.oz.au>
	SVR4/386 has an "interesting" feature: You can write to a file
	with mmap() **without** updating the timestamps....
	INN uses this for the active file, nnmaster now has a "-U" switch
	to force an unconditional scan of the active file after the timeout,
	rather than if the timestamps changed.

Title:	VMS ANU-NEWS compatability
From:	Matt Heffron <heffron@falstaff.css.beckman.com>
	VMS ANU-NEWS's NNTP returns a different error code on no-such-article

Title:	Log group that generated a nntp error
From:	Matt Heffron <heffron@falstaff.css.beckman.com>
	If a NNTP error is generated, log the group that caused it.

Title:	Support terminals without clr_eol capability
From:	Luc Rooijakkers <lwj@cs.kun.nl>
	Some ancient terminals cannot clear to end-of-line. Simulate it by
	outputting spaces. This means that printf/puts/putchar/stdout can
	no longer be used in visual mode, because term.c has to keep track
	of the cursor position. Use the new tprintf/tputc instead (but only
	from files that are always linked with term.o). A list of the files
	that may do printf/puts/putchar/stdout is in `noterm'.

Title:	Miscellaneous cleanup of term.c
From:	Luc Rooijakkers <lwj@cs.kun.nl>
	Use tputs correctly unless BROKEN_TPUTS is #defined (UNISYS?)

Title:	Allow backslash to escape semicolons in extra headers
From:	Luc Rooijakkers <lwj@cs.kun.nl>
	This is necessary to let the user add a MIME Content-Type header
	via the news-header or mail-header variables.

Title:	Fixed annoying cursor position during some messages
From:	Luc Rooijakkers <lwj@cs.kun.nl>
	Don't reposition the cursor to the prompt line after a message
	if the message has overwritten the prompt.

Title:	Fixed invisible error messages
From:	Luc Rooijakkers <lwj@cs.kun.nl>
	If terminal has a separate `visual mode' screen, printf/nn_exit
	sequences don't work. Replaced user_error by nn_exitmsg which has
	the exit status as additional first argument, and replaced the
	printf/nn_exit sequences by calls to nn_exitmsg

Title:	Added charset variable and 8-bit checking for posts
From:	Luc Rooijakkers <lwj@cs.kun.nl>
	The new charset variable can be set to the character set of the
	terminal. It also determines the default value of the
	data-bits variable.
	If a message is posted whose body contains 8-bit characters
	and charset is not set to "unknown", nn will add MIME-Version
	and Content-XXX headers indicating its character set.
	Such messages cannot be posted if the current character set
	is a 7-bit set.
	To obtain the previous behaviour, do `set charset unknown'.

Title:	Cleaned fill/nofill use in nn.1.* files
From:	Luc Rooijakkers
	Lots of examples in the manual were being filled by *roff.

Title:	Sped up aux.sh and made it more portable
From:	Luc Rooijakkers
	Using case instead of if is faster on shells that don't have if

Title:	Preview mode and menu.c
From:	Steven Grimm <koreth@hyperion.com>
	The symptom: previewing (usually via auto-preview-mode) an article on
	the next-to-last page of a group menu then going to the last page of
	the menu caused nn to dump core if the previewed article had a greater
	article number than the last article on the last screen.  For example,
	viewing article 'h' would cause the problem if the last screen only
	went up to 'e'.

Title:	Correct a small over-optimization in term.c
From:	Luc Rooijakkers
	My term.c contained one cursor position update too many, causing
	things to go wrong sometimes.

Title:	DecStation/Ultrix 4.2 support
From:	"stolen" from Toivo Pedaste <toivo@uniwa.uwa.edu.au> my Peter Wemm.
	conf/m-dec3100.h for mips machines running Ultrix.

Title:	Misc cleanups of bad code
From:	Paul Eggert <eggert@twinsun.com>
	The code is loaded with functions that return no value, etc.
	This is a start at fixing some of the more serious faults.

Title:	Macro to call metamail
From:	mikew@kpc.com
	A simple example macro for passing MIME articles to metamail

Title:	Bug in folder code procuces core dump.
From:	Stefan Schwarz <stefans@bauv106.bauv.unibw-muenchen.de>

Title:	rfmail support
From:	Bo Kullmar <bk@kullmar.kullmar.se>

Title:	Clean up code part 2
From:	Peter Wemm <peter@DIALix.oz.au>
	Continued on from what Paul Eggert started... AARRGGHH!! It's a mess!

Title:	Update help screens
From:	Peter Wemm <peter@DIALix.oz.au>
	As suggested by Bill Wohler <wohler@sap-ag.de>, sort out the help
	pages to "clarify" the differences between a few commands, and to
	add the "(" and ")" keys.. [darn.. It's a tight fit in there.]

Title:	Domain/OS fixes and others
From:	Michael T Pins <mtpins@icaen.uiowa.edu>
	1> Makes nnmaster run much happier on HP/Apollo's running Domain/OS
	2> Allows nn to compile on varients of HP-UX, among others, that it
	   won't compile on otherwise.

Title:	aux.sh fixes
From:	Peter Wemm <peter@DIALix.oz.au>, bug reports from many...
	The changes to aux.sh to use "here documents" broke a lot of
	/bin/sh's Hack them out of the code, as not everybody has ksh.

Title:	SunOS 4.1 supports window siziing
From:	Paul Eggert <eggert@twinsun.com>
	Add a change to s-sunos4-0.h

Title:	Add support for PCS Cadmus with MIPS cpu
From:	Peter Pichol <pp@informatik.uni-kiel.dbp.de>)
	Create m-pcs.h

Title:	Add support for DEC Alpha's running OSF/1
From:	Mike Wexler <mikew@kpc.com>
	Create m-alpha.h and s-osf1.h

Title:	A way to read even less news!
From:	Keld.Simonsen@dkuug.dk
	It has been an old desire (viz. the FAQ) to be able to skip
	followups to subjects, one did not care to read. The following two
	variables implement a way to omit these followups, plus a way to
	auto-select the articles that you are most likely to be interested
	in, namely followups to articles that you have already read, or
	followups to articles you have posted yourself.

	The killing of uninteresting followups is done without entries in
	the kill file. The thread is killed if the root article of the
	thread is a followup itself, and the subject is not selected.

	The default of this patch to nn is no changed behaviour.

	Two new variables are introduced, can be set in the .nn/init file:

	ignore-re  (boolean, default false)
	If set, articles with subjects already seen in a previous
	invocation of nn or another newsreader - and not auto-selected -
	are automatically killed.

	auto-select-rw (boolean, default false)
	If set, a subject of an article read or posted is automatically 
	used for subsequent auto-selecting (if not already selected).

Title:	Fixes for non ANSI SunOS cc
From:	Paul Eggert <eggert@twinsun.com>

	NN 6.4.19gamma's term.c contains a cast that assumes that the
	compiler is an ANSI C compiler.  This doesn't work in K&R
	compilers (e.g. SunOS 4.1.2 with bundled cc).

Title:	void vs int mixups in return values
From:	Paul Eggert <eggert@twinsun.com>
	Here's a patch to NN 6.4.19gamma's term.c that fixes a problem
	where it returns a value in a void function.

Title:	After you `:set record xxx', `:set' misdisplays mail-record
From:	Paul Eggert <eggert@twinsun.com>
	The :set listing were incorrectly displaying mail-record and
	news-record as always unset.

Title:	Improvement/bug fix to nn: archiving full articles using NNTP
From:	Fred Hucht <fred@hal6000.thp.Uni-Duisburg.DE>
	If you set nnmaster to archive articles while in NNTP mode, it
	only archived the headers...

Title:	Add a flag to show Originator
From:	Jim Jagielski <jim@jagubox.gsfc.nasa.gov>
	Add's support for 'J' in news-header to display the "Originator:"

Title:	Catch short term error conditions to prevent nnmaster aborting.
From:	J.P.M. van der Jagt <jeanpaul@duteca.et.tudelft.nl>
	Prevent things like "400 Server load too high, try again later"
	from making nnmaster exit.

Title:	make nn 6.4 PL18 compile (and work ;-)) for hp_ux 8.0.
From:	Bart Muyzer <bartm@cv.ruu.nl>

Title:	`nn -xXm' does't clear to end of line (NN 6.4.19gamma)
From:	Paul Eggert <eggert@shadow.twinsun.com>
	NN wasn't properly clearing to end-of-line after each output
	line.

From:	Paul Eggert <eggert@twinsun.com>
Title:	infelicities in NN 6.4.19gamma term.c
       	The term.c code had some really bizzare conditional tests that
	were uncessesary, and some were obvious typos.  This patch doesn't
	fix any real bugs, but it removes the needless code.  It also
	introduces a few optimizations in critical places: (This is a
	particularly big win on most Sparcs, which don't have integer
	division.)

From:	Peter Wemm <peter@zeus.DIALix.oz.au>
Title:	silly bug in nn-use-elm script
	contrib/nn-use-elm had a leading  #/bin/sh instead of #!/bin/sh
	This causes invocation of csh on many systems... (including mine :-)

From:	Peter Wemm <peter@zeus.DIALix.oz.au>
Title:	long standing bug in sort.c
	non-related articles were often being associated in
	subject-match-parts mode.  This tries to make it less likely.

From:	Steve Grimm <koreth@hyperion.com>
Title:	prefix.c fix, HAVE_TRUNCATE for SunOS.
	prefix.c change #ifdef NNTP to #ifdef INEWS_PATH
	SunOS has truncate.

From:	Michael Masterson <mjm@thumper.progcons.com>
Title:	account.c is broken
	The code uses a structure before defined if accounting is enabled.
	Update s-sys5-4.h.  m-i80386.h needs work.

From:	Peter Wemm <peter@DIALix.oz.au>
Title:	improve m-i80386.h
	m-i80386.h didn't deal with network byte order very well.  As far as
	I can tell, if you want network byte order, then you probably want it
	for NFS sharing.  In which case you have TCP/IP.  Use it's routines.

From:	Steven Grimm <koreth@hyperion.com>
Title:	Patch for preview mode display glitch
	the 'xn' termcap flag is conflicting with clrline() and long
	subject lines, causing the following line to be cleared.

From:	Peter Wemm <peter@DIALix.oz.au>
Title:	fix aux.sh
	aux.sh was failing when using the nn_elm script from the FAQ.
	In the process of reformatting it, I found the bug, a missing
	semicolon.  It would fail whenever mailer-pipe-input was unset.

From:	Peter Wemm <peter@DIALix.oz.au>
Title:	The Biggie!  Merge in NOV code!
	Well, this is the big one.  #ifdef NOV in config.h, and it even works
	over NNTP.  The some of the hacks to get NN to use the libnov code is
	pretty baroque, but necessary in order to maintain compatability.
	NN's internals are *very* dependant on the nnmaster database and it
	does not particularly appreciate the change.  If you enable NOV, all
	bets are off. :-)  Most of the code is from Geoff Collyer
	<geoff@world.std.com> although the NNTP mods and other tweaks are
	from others whom I cannot identify..
	This is very much a Stop-Gap solution.

From:	Michael Masterson <mjm@thumper.progcons.com>
Title:	auto-select-rw and maybe ignore-re kill folder mode
	NN takes a SEGV while reading a folder with these enabled.
	[Peter: I'm unable to reproduce the bug....]

From:	Steven Grimm <koreth@hyperion.com>
Title:	Patch for ignore-re and -x
	Here's a tiny patch so that ignore-re is turned off when you're
	using the -x flag, or when you go into a group to read old
	articles.  I think this is better than the way it currently works;
	currently, if you have ignore-re set on a group, there's no way
	(that I can discover) to get nn to show you an article whose
	thread's initial article has expired.

From:	Peter Wemm <peter@DIALix.oz.au>
Title:	folder move vs. auto-select-rw
	Basically, patch up Michael Masterson's problem.  Although I
	couldn't core dump (lucky me..), I did notice that if I was
	reading a folder, the articles I read were added to the kill file.
	I've added a temporary override to auto-select-rw that makes it
	ignored in a folder.  This should fix Michael's problem.

From:	Peter Wemm <peter@DIALix.oz.au>
Title:	6.4.19 is no more.
	So as to distinguish out butchering from Kim's work, we've decided
	on a revision "bump".  It's NN-6.5 now.

From:	Peter Wemm <peter@DIALix.oz.au>
Title:	gcc -Wall cleanup
	It now compiles a lot more cleanly under -Wall, but not 100%
	clean..  :-)  I found some sneaky bugs in the menu code and the
	init file parser in the process.

From:	Peter Wemm <peter@DIALix.oz.au>
Title:	Merge in Article search from Andy Burgess <aab@cichlid.com>
	There are numerous collisions with the NOV code, and it does not
	work with digests, etc, etc.  It is *not* enabled in config.h by
	default, as It has not revieved enough testing by the group.
	It's also not documented in the manual... (Volunteers?)
	Also, don't enable this if you read via nntp... Your server will
	hate you...

-- 6.5.0 beta goes out Aug 31 1993 --

From:	Paul Eggert <eggert@twinsun.com>
Title:	The code still says 6.4 in various files, including the documentation.

From:	Paul Eggert <eggert@twinsun.com>
Title:	Bogus prototypes for syslog() functions are wrong.

From:	Paul Eggert <eggert@twinsun.com>
Title:	NO_X_NEWSREADER should be defined by default.
	[Peter: Awww!  If Tin can do it, why can't I? :-) ]

From:	Paul Eggert <eggert@twinsun.com>
Title:	Redeclaration of catch_winch on machines with RESIZING

From:	Paul Eggert <eggert@twinsun.com>
Subject: nn-6.5.0 mishandles time zone changes near short month boundaries
	NN 6.5.0 mishandles non-GMT time zones near short month
	boundaries, which causes articles to be incorrectly sorted by
	date.  For example, pack_date("30 Sep 1993 23:00 -0700") and
	pack_date("1 Oct 1993 06:00 GMT") should yield the same value, but
	in NN 6.5.0 the former yields 3614760 while the latter yields
	3616200.  This is becoming more and more of a problem as more
	people use non-GMT time zones in their news.  Here is a patch.

From:	Toivo Pedaste <toivo@uniwa.uwa.edu.au>
Title:	Missing 'static' on some functions
	db_init_group(), catch_winch(), grep_article()

From:	Steve Davies <srd@peora.sdc.ccur.com>
Title:	signal prototypes misdeclared
	There were a few places where signal handlers were declared void,
	but in fact were sig_type. This was OK on machines that had the
	same definition, but not on those with sig_type == int.
	Also, truncate() is messed up..

From:	Steve Grimm <koreth@hyperion.com>
Title:	tidy up config.h
	There were a few typos and errors in config.h-dist

From:	Peter Wemm <peter@DIALix.oz.au>
Title:	fix up prototype problems with truncate and gethostname
	It seems every system has a different idea on what they are supposed
	to be.  I give up...  NN has not got internal functions called
	nn_truncate and nn_gethostname, which are prototyped, and will call 
	"the real thing" if it's available.  It seems the safest way.

From:	Steve Davies <srd@peora.sdc.ccur.com>
Title:	6.5 NOV + NNTP cleanups
	NN + NNTP + NOV : db.c, global.h, nn.c, nntp.c
	- This patch fixes problems when NN is run on the same machine as
	the NNTP server.
	- When compiled with NNTP and NOV, a temp file was being created
	when the active, active.times, and .overview files were read from
	the NNTP server.  This patch eliminates the creation of that
	temporary file.  In db.c (readactfile, readtimfile,
	db_read_group), instead of reading from the temp file, the data is
	read directly from the NNTP server.  The idea for this comes from
	Peter Orbaek (poe@daimi.aau.dk).  This requires a simple patch to
	the nov sources (which probably should be applied by hand rather
	than running this file through patch):
	[NOTE: NN-6.5 contains a 'nov-lite' distribution with this patch
	applied and everything non-essential for NN removed. -Peter ]

--- nov/libnov/oldlib.c	Tue Sep  7 13:47:31 1993
+++ nov/libnov/lib.c	Tue Sep  7 13:47:42 1993
@@ -213,6 +213,7 @@
 	}
 
 	while ((line = fgetms(gp->g_stream)) != NULL) {
+		if (line[0] == '.') break;
 		art = (struct novart *)malloc(sizeof *art);
 		if (art == NULL ||
 		    (prssts = prsovline(line, gp, art, prevart)) == PRSFAIL) {


	- The code in db_read_art for handling crossposts when NOV is
	defined does not handle the (hypothetical?) case where there are
	extra headers besides Xref: in the .overview file.  It also
	doesn't work when cnews is used to maintain the .overview files,
	since cnews appears to use xref: instead of Xref: in the .overview
	file.  A new routine "db_fixup_cross_postings" is added to db.c .
	This routine is derived from code by Charles Henrich and
	rmtodd@servalan.servalan.com.

From:	Peter Wemm <peter@DIALix.oz.au>
Title:	include a minimal nov distribution.
	Because we need a trivial patch to the libnov code which Geoff
	Collyer doesn't want to put in his code (and quite understandably
	because it's a "hack"), and I dont want 10,000 email messages
	saying "NOV + NNTP doesn't work" when they build it without
	patching because they never read intructions that tell them to
	patch it. :-)  If anybody does this to me now because they've
	removed the patched version, I'll reply with a uuencoded 16 Meg
	kernel crash dump. (ie: a bunny dropping..) :-)

From:	Peter Wemm <peter@DIALix.oz.au>
Title:	Fix "Out of memory: cannot allocate -4 bytes"
	If NN could not determine the window size from the termcap or the
	TIOCGWINSZ ioctl's, it would try and malloc a negative buffer..

From:	Peter Wemm <peter@DIALix.oz.au>
Title:	typo in nn_gethostname()
	hostname.c would not compile if HAVE_HOSTNAME defined.

From:	Jim Jagielski <jim@jagubox.gsfc.nasa.gov>
Title:	nn -a0 botches display on some hosts
	There is some interaction between curses, tprintf, and terminfo.

From:	Jim Jagielski <jim@jagubox.gsfc.nasa.gov>
Title:	Add new conf/s-aux3-0.h

From:	Robert.Stockton@ELROND.GANDALF.CS.CMU.EDU
Title:	nov code overrunning '.' on XOVER data when active is incorrect.
	If the active file says there is articles, but the .overview is
	empty, the novall() and novnext() miss the "eof" dot and hang
	forever.

From:	Toivo Pedaste <toivo@uniwa.uwa.edu.au>
Title:	Explicit support for mouse under xterm
	This code can detect if running under xterm, and will enable mouse
	events if so.  The mouse clicks are used for menu selection,
	reading, previewing and so on.  It makes NN "almost" hands-free
	for most news reading tasks.

From:	Peter Wemm <peter@DIALix.oz.au>
Title:	Minor tweaks for mouse-code
	Because of a "feature" apparently in the SYSV termio driver and
	possibly the POSIX termios, NN was disabling raw mode during
	screen updates in order to enable xon-xoff flow control to work.
	This was causing character echoing of mouse events during screen
	redraws and looked rather ugly.  This problem never occurs on
	older BSD systems with the sgtty style driver.  This tweak causes
	NN to not switch off raw mode on Xterms when mouse is enabled..
	Heck, if xterm needs flow control, something is very wrong...

From:	Tom Limoncelli <tal@warren.mentorg.com>
Title:	Possible memory leaks
	Purify is not at all happy with some of the things that the NOV
	code does.  I dont think there is much of a problem, although I
	agree that it's messy.  Basically, NN is allocating buffers for
	the hash tables and not freeing them at exit.  As these are reused
	while NN travels from group to group, this is mostly static.
	NN does not free() these on exit and Purify complains.  The
	invitation to fix this is open to all.. :-)

From:	Peter Wemm <peter@DIALix.oz.au>
Title:	major reorganisation of NOV code.
	Shift files from nov/* to the top level directory for radical
	simplification of compilation.  Rationalise include files.
	Substitute some simple routines to reduce extra files needed from
	the NOV distibution.

From:	Peter Wemm <peter@DIALix.oz.au>
Title:	major repairs of above reorganisation.
	Well, the "simple" reorganisation turned out to be rather dramatic
	in it's affects.  This cost weeks of debugging and nearly caused
	me to give up and reverse the last upmteen patches (I had
	neglected to do a CVS checkin before starting).  It turned out to
	me a malloc problem.  I *suspect* there is still a memory leak
	there, but it doesn't seem to show up on the process size after a
	lengthy reading session..

From:	Ian Nicholls <iann@storesys.coles.oz.au>
Title:	NCR tower 32 tweaks.
	nn-6.5.0.b2 had troubles with include files.  This helps a bit.

From:	Marc Boucher <marc@cam.org>
Title:	additions to conf/s-sys5-4.h
	This adds some omissions to the config file.

From:	Marc Boucher <marc@cam.org>
Title:	bzero/memset are out, clearobj is in...
	Following Kim's coding conventions, change a couple of memset()'s
	that I introduced to clearobj's.  This looks a lot nicer now..
