HISTORY

This file contains brief descriptions of changes made to the pseudo-tty
layers in chronological order.  This file was created 4/19/93 so dates
before that time are approximate.  Unless otherwise noted, all changes were
made by Dave Dykstra.

1991
====
10/1	Keith Muller's version of layers retrieved from anonymous ftp ucsd.edu 
	at pub/dmdlayers.tar.Z.

12/1	Modified openchan() libwindows call to return a pseudo-tty instead of
	a socket; this supports starting a program in a window from another
	process such as "agent !".  Also removed openrwchan() as it is no
	longer necessary.

1992
====
2/1	Added calls to signal() at the end of signal handling routines to
	reset signal handlers in case layers is compiled with /usr/5bin/cc
	on a sun.

4/23	Strip off eigth bit of TIOCUCNTL calls because the R&D Unix dmdld
	had turned the uppermost bit on for some reason.

5/1	Addition of one second delay after switching to layers mode to
	allow time for the terminal to initialize.

7/1	Allowed layers to come up even if not setuid-root.  When not setuid-
	root, can't write /etc/utmp unless it is world writable, and can't
	change ownership of pseudo-ttys.

9/1	Strip off 8th bit when inquiring the terminal for it's escape
	sequence identifier.  This allows getting into layers no matter
	what the parity setting.  Also allowed for missing square bracket
	in the returned escape sequence in case the wproc terminal controls
	option is set to 8 bits.  Will print out the id strings in a
	readable format if they are not recognized.

10/1	No longer requires empty files in $DMD/lib/layersys for each terminal
	id string to get into layers.

12/1	Added NO_TIOCUCNTL to support machine types on which the TIOCUCNTL
	ioctl on pseudo-ttys is not supported.

12/9	Added getagentanswer() libwindows call to get access to the return
	buffer from an agent call.  This is used by the agent(1) program to
	display the return buffer in ascii for agent calls like A_ROMVERSION.

12/14	Added '-V' option to print out version.
	Made lyrscfg.h to contain defines for different operating system
	configurations.
	Added support for mips-based OS's, specifically StarServer-FT,
	thanks to many changes by John Ampe.  This OS is a combination of
	System V and BSD.
	One notable addition is USE_NAMED_PIPE to use named pipes to
	communicate with support programs rather than unix-domain sockets.
	Another is support for system V utmp file format, although it is
	known to be buggy.
	Changed default shell to /bin/sh instead of /bin/csh because csh is
	not available on all systems.
	Added support for AT&T SVR4, USE_PTMX.

1993
====
2/1	Added -e option to force on hex encoding mode the same way that
	the environment variable DMDLOAD=hex does.

3/1	Increased number of pseudo-ttys supported from 96 to 256 on
	machines using the original bsd naming conventions
	(/dev/[pt]ty[pqrstuvwxyzPQRSTU][0-f]).

	Change to pass unrecognized agent calls on to the terminal rather
	than rejecting them immediately.

3/10	Added more portability defines for HP-PA thanks to Mark Larsen.

4/19	Added beginning of support for Solaris from Bob Pietkivitch.
	Added John Lash's '-m' option for larger packet size support on 730s.
	Got to compile without warnings on NCR 3000 SVR4 with both the AT&T
	and MetaWare compilers.
	Removed old sunos pty driver patches from misc directory.

4/21	Change to always return the result from an agent call in the agent
	answer buffer.
	Add flow control on agent calls to make sure all data from the
	pseudo-tty is drained before processing the agent call.
	Add support for A_JBOOT, A_JTERM, and A_JZOMBOOT even if on a system
	that has TIOCUCNTL.  The purpose is to be able to use the above
	flow control and to provide for transition to operating system
	versions that don't support TIOCUCNTL (such as sun -> solaris).
	Change libwindows to send the $DMDCHAN with the request if set
	before looking up ttyname() because it's much faster.

4/22	Add matchtty() call to match pty name on major/minor number if a
	match is not found by name.  This is for systems that have two /dev
	names for the same ptys: the one that layers sees may not be the
	same one that ttyname() sees, and having layers() call ttyname() is
	much slower.
	Change to print error message to opened window if can't allocate
	pty; previously the window was silently deleted.
	Made a more reasonable algorithm for determining timeout times
	based on the baud rate.  If the user does not specify '-b' baud
	rate, the maximum is assumed to be 4800 baud to allow for network
	delays and other delays; the timeout for 4800 baud is now 6 seconds
	for transmit and 4 seconds for receive, double the preceding minimum
	(and the mimimum was usually selected).
	Eliminated PXTIMEOUT and RXTIMEOUT defines and instead use the value
	for 1200 baud in the table at initialization time.
	Put in a one second delay when forking a process to create a new
	window to allow time for the process to open the pty before layers
	tries to read it.

4/27	Ignore leading blanks in a '-f' file.
	Fixed a problem with starting up multiple dmdld's from a '-f' file
	(only for libwindows programs that use A_JBOOT, not those using the
	JBOOT TIOCUCNTL ioctl): too many packets get sent on channel 0
	before being acked so some of them fail allocating a buffer to send
	the JBOOT message.  The fix is to delay agent messages if there are
	unacked messages on channel 0, using the agentblock flag.

4/28	Got rid of wtinit.  It was just a simple program that sent the
	startup string for layers, with hex encoding on or off.

5/4	Removed setenv.c and getenv.c.  Just use standard putenv/getenv instead.

5/5     Created 'grantpt' and 'movelogin' library programs.  These are small
	programs that are to be installed setuid root to grant ownership
	of pseudo ttys and to move entries around in the utmp file, respect-
	ively.  Now layers itself no longer needs to be setuid root.
	'Grantpt' is not needed on SVR4 because a similar program comes
	with the system, accessed through a library call that is called,
	not coincidentally, 'grantpt()'.
	In addition, layers (via 'movelogin') correctly handles System V
	format utmp files.
	There is no longer a utmp entry for every window created.  Only one
	window gets an entry, just like on the XT-driver layers.  The entry
	is initially given to channel 1, but the user can move it to any
	channel via a new agent call A_RELOGIN.  If the window that has the
	utmp entry is deleted, layers reallocates a pty reserved for that
	that channel and uses it when/if the window is reopened.

5/7	Finished support for larger packet sizes.  Send the A_XTPROTO message
	when starting up.  When response is received, send JXTPROTO if larger
	packet sizes are to be used.  If no '-m' or '-b' option is used and
	larger packet sizes are supported, 64 byte packets will be used unless
	the baud rate claims to be 1200 baud or less; then 32 byte packets
	are used.  Don't want to go higher than 64 because sometimes the stty
	baud rate can't be believed.  The user can choose a larger packet size
	with '-m'.  Also, if '-b' is 9600 baud or higher, 128 byte packets will
	be used.
	Added agentcom(), A_JXTPROTO, A_JTIMO, and A_JTIMOM to send these
	control messages to the terminal in a flow-controlled manner.

5/11	Added '-s' option to print protocol statistics to stderr on exit.
	No longer fclose stdout and stderr upon startup.

5/12	Removed checking for "fast" in rcfile(); it wasn't having any effect
	anyway.
	Renamed 'bsdagent.h' to 'hostagent.h'.  This will cause a problem for
	the few support programs that had to include bsdagent.h, but most of
	them are probably going to need modification anyway to use the
	new "ioctlagent(...A_J..." equivalents to the "ioctl(...J..." calls.
	'Bsdagent' is now really a misnomer since pseudo-tty layers works
	on many other machine types.
	Removed include/sys/sccsid.h from package.
	Made libwindows that works with both pseudo-tty layers and XT-driver
	layers.
	Added getagentansresult() and getagentanschan() libwindows calls to
	return the first and second shorts respectively from an agent call
	response.  These are now used by the agent program.
	Added xtagent, xts, xtt, and relogin programs to the package.  Xtagent
	is the same as the popular agent AT&T exptool with the name changed
	for machines that have both installed.  The latter 3 programs are
	equivalent to programs in the XT-driver layers package.

5/13	Added A_CHANGEPROC agent call to change the owner process of a channel.
	This is sent by the Runlayer() libwindows function to notify layers
	that a new process has been started in the window so layers can kill
	it when the window is deleted.
	Added inlayers() call to libwindows; returns 0 if not in layers, 1
	if in xt-driver layers, or 2 if in pseudo-tty layers.
	Added Runwaitlayer() libwindows function.

5/18	Added relogin, xtagent, xts, xtt, and xtproto man pages.

5/19	Added Jboot(), Jzomboot(), Jterm(), and Jwinsize() libwindows calls.

5/20	Increased the sleep time before starting to talk XT protocol from
	1 second to 3 seconds.  The 5620 needs 3 seconds.  Modified
	recvagent() to recognize the length of the received agent packet
	from the terminal.

5/24	Removed "-i" parameter to forked shells.
	Allowed autocurrent to be turned of on any terminal, not just 630s.
	Look for all layers support files under $DMDSYS if they're not found
	under $DMD.  $DMDSYS defaults to "/usr".  This allows separation of
	the software development package from the layers package.

5/26	Split the openchan() libwindows routine into findchanname() and
	openchanname().  This is needed in Runlayer() to get obtain the
	channel name synchronously in the parent process and still have
	"/dev/tty" work properly for the child process.

5/27	Added USE_IGNORESIGCHLDGRANTPT for Solaris.
	Removed usages of dup2.  They weren't needed, and it's not available
	on the current version of Amdahl uts.

5/28	Modified recvagent() to indicate agent call failure only when the
	return code is -1, not when it's non-zero.  The 730windows agent calls
	currently return the agent number (non-zero) when the call succeeds.

6/3	Added $DMDAGENTDIR environment variable passed from layers to 
	libwindows.  This is the directory that layers uses for the socket
	or named pipes.  It takes precedence over $DMD and the compiled-in
	default for $DMD.  This eliminates problems caused by either a
	libwindows compiled with different defaults from layers or the user
	change $DMD after getting into layers.

6/4	Added setagenterrf() libwindows call for programs that don't want
	errors sent to stderr.
	Made libwindows open "/dev/tty" if none of stdin, stdout, and stderr
	are connected to a tty.

6/7     Added USE_SYSVLIKE and USE_SYSVR4LIKE defines to lyrscfg.h.  This
	makes it easier for systems on which layers is compiled on both
	SVR4 versions and non-SVR4 versions of the OS.  On those systems
	either set EXTRACFLAGS=-DUSE_SYSVR4LIKE in the top level makefile
	or edit lyrscfg.h to put "#define USE_SYSVR4LIKE" at the top to
	compile SVR4 versions.

6/8	Fixed writes to the terminal to re-try write if write returns EINTR
	error; implemented check_write().  This solves the mystery of many
	lost packets, NAKS, and timeouts.
	Added buffered writes; implemented buf_write() and flush_write().
	If there is data waiting to be read from any source, layers buffers
	up data to the terminal and only does a write() when either the
	buffer gets full or there is no more data to be read.  This allows
	larger writes() to the terminal which is more efficient and allows
	better packing of data over some networks for better performance.

6/18	Added DEFDMD make variable to allow compilation in a different
	location from the default $DMD variable.

6/21	Added onloginlayer() call to libwindows.  Routine determines whether
	or not the application is running on the logged-in window.  Many
	applications get confused by broadcast messages that are sent to
	windows that show up in the utmp (who) file.  'Xtagent -l' returns
	the same information for shell scripts.

7/2	Changed all sleeps to use select() with no file descriptors and
	a timeout (mysleep function) to avoid apparent interactions of the
	sleep() and alarm() calls on some versions of unix (in particular
	Mips RISC/os).
	Added DMDMAXPKT environment variable to set default XT packet size
	just like -m option.

7/7	Added comments to almost all source files stating that the source
	is freely distributable.

7/27	Network XT support added.  If the terminal supports network XT and
	it's connected to a network, network XT will automatically be used.
	Override with 'DMDLOAD=regular' to drop it down to regular XT or
	'DMDLOAD=hex' to drop down to regular XT with hex encoding.  The
	same thing can be done with the '-r' and '-x' options, respectively.

7/28	Added '-d' to selectively enable debugging messages at run-time.
	Added '-o' to select name of the file that messages go into.

7/28	Expanded usage message to provide short description of each option.

7/29	Trace information enhanced to decode protocol information similar
	to XT-driver layers' 'xtt' program.

7/30	Added '-h' flag and '$DMDHIGHWATER' environment variable to set the
	high water mark for network xt.

8/16	Changed the 'ifdef USE_TERMIO's into 'ifndef NO_TERMIO's so that the
	default on SunOS systems is to use termio rather than sgtty.  I was
	getting trouble with sgtty on dmdld on SunOS systems and termio works
	fine so I decided to change the SunOS systems to use termio like every
	other supported system.

8/16	Added 730 and 730+ to the 630 termcap entries in misc/MTG.termcap and
	added MTG.terminfo for corresponding terminfo entries.

9/3	Allow zero-length "lsys" file parameters when not on 5620s.  The 705
	has a built-in function key that runs "layers /dev/null".
	Added '-l' option to allow running nested layers within a layers
	session.  Previously I had allowed it without an option and just
	printed a warning.  I know of no terminal that currently supports
	running layers within layers but theoretically it could work.

9/7	Clear c_lflag completely on termio-based systems because there can
	be new flags that should be turned off.  Formerly it only turned off
	certain known settings.  In particular, 'IEXTEN' is bad and it was on
	by default on Solaris systems.
	Wait some time for child processes to die after sending them a SIGHUP 
	because at least on SVR4 systems closing the pseudo-tty sends another
	SIGHUP and that can cause editors such as 'vi' to abort while saving
	files.
	Add parameter to crc() function to control whether or not to write
	crc back out to the buffer.  This allows displaying the actually
	received crc in the debugging info when a bad crc is received rather
	than displaying the expected crc.

9/9	Added printing out of system error messages in libwindows when binds
	failed on systems that use unix domain sockets (those that used
	named pipes already had the messages).

9/17	Aack! Had forgotten to add the parameter to the crc() function when
	calculating crcs for ACKs sent to the terminal.
	Added some XT statistics which were missing from the network XT
	protocol: transmit packets, transmit bytes, and receive bytes.

9/21	Hadn't properly set up the externs in process.c to go with the
	middle change of 9/7.  Compiled on some systems but not on others.

9/23	Modified mysleep() to hold SIGALRM signals while sleeping.  This 
	prevents it from getting interrupted by the XT timeout alarms.
	That was causing a very bad problem with the middle change of 9/7:
	the mysleep() delays there were always getting interrupted and
	the delay code was designed to keep retrying mysleep while it is
	interrupted so it got into an infinite loop.  Also modified that
	retry to not retry for ever.

9/24	If the user specifies regular xt or hex encoded regular xt and 
	a max packet size of 32, don't bother to ask the terminal if it
	supports any better with A_XTPROTO.  This is mostly to support
	terminal-side XT implementations that don't respond at all to 
	the A_XTPROTO request like they're supposed to.
	Fixed bug where exitagent() deletes sockets/named pipes belonging
	to a different layers session; was incorrectly looking for
	ttyname*.  For example, layers session on pts/1 would delete
	sockets for layers session on pts/10, etc.  Thanks to Paul Connors.

10/18	Removed checking in ptyioctlagent() for channel numbers between
	0 and 7.  This is because some of the 730windows agent calls use
	negative numbers in the channel number slots.

10/21	Fixed bug that prevented "grantpt -r" from properly releasing the
	slave-side pty permissions (i.e. giving it back to root) on machines
	with BSD ptys (e.g. SunOs).   Had to clear the close-on-exec flag
	of file descriptor for the master side of the pseudo-tty before trying
	to do that.

11/9	Changes to support compiling with an XT-driver-style jioctl.h: include
	lyrscfg.h in layers files that had only included jioctl.h, define
	the XTJ*3 defines inside of hostagent.c, and only define USE_HOSTAGENT
	in lyrscfg.h if it isn't already defined.

11/10	Incorporated some changes I had been saving up:
	  - The Makefiles now use "$DMDSYS" instead of "$DMD" as the base
	    of installation.  "$DMD" is the default value of "$DMDSYS" for
	    backward compatibility.
	  - Reorganized lyrscfg.h to have "major option groups" and "machine
	    specific type options".
	  - Incorporated some of the changes from Paul Connors to support
	    Mips Risc/OS /EPIX operating system. 

12/2	Aargh! I had forgotten to take out a hack which made an 8;7;5 5620
	(version 2.0 roms) be treated like an 8;7;3 5620 (version 1.1).
	This hack made 8;7;5 5620s unusable, causing layers to attempt to
	use 32ld to download set_enc.j and lsys.8;7;3.

	Also, to be more consistent with xt-driver layers, make last window
	created in a layers startup file current, unless a user explicitly
	specifies a "current" command in the startup file.

12/3	Undid the extra inclusions of lyrscfg.h done on 11/9 because it turns
	out it was already included via common.h which includes hostagent.h
	which includes lyrscfg.h.  Also, I forgot to mention that sometime
	before 12/2 I turned off the default define of "FAST" in lyrscfg.h
	because it only took effect if layers was setuid-root anyway.

12/22	Fix in grantpt thanks to Mark Larsen for Mips machines that implement
	the /dev/ptc clone pty device.  Need to compare the master pseudo-tty
	major number to the /dev/ptc minor number rather than the /dev/ptc
	major number.

12/23	Minor improvement to libwindows in findchanname() (and openchan()
	since it calls findchanname()) to automatically call closeagent()
	under pseudo-tty layers if it has to call openagent() because the
	user hadn't done it.  This prevents requiring the user program to
	call closeagent() if openchan() is the only libwindows call that
	the user makes (like in dmdpi).

1994
====

1/10	Added a function "setagenttermiof()" in libwindows to allow a user
	program to define a function that returns a pointer to a termio
	structure that defines non-raw mode tty settings.  This is for
	programs that use Runlayer() but have their tty set into raw mode.

1/25	Solaris has a problem with unlinking a socket before the socket
	file descriptor is closed; sometimes later binds then fail.
	Changed agent.c and hostagent.c to compensate.

2/7	Discovered that "termios" functionality should be used on operating
	systems that provide it; "termios" is an extension to "termio" with
	extra stty settings.  In order to transfer all stty settings to
	another tty as layers does, "termios" is needed.  Added the code
	to use it and have it turned on by default unless NO_TERMIOS is
	defined.  Also added support in libwindows for systems that have
	neither "termio" or "termios" (e.g. BSD systems) because that was
	missing.
	At the same time, redefined the interface added on 1/10 to accomo-
	date different types of tty handling on different operating systems.
	Instead of "setagenttermiof()" in libwindows there is now instead
	"setagentnonrawf()" to pass in a pointer to a different user-defined
	function.  The user-defined function gets a file descriptor as a
	parameter, sets that file descriptor into non-raw mode, and returns
	-1 upon failure or something other than -1 upon success.

4/25	Added support for linux in the top level Makefile and lyrscfg.h.

5/12	Added a hint message that $DMD/xt or $DMDSYS/xt has to be world
	writable if the socket or named pipe can't be created.  Also,
	for systems that use sockets, it wasn't properly trying to bind
	$DMDSYS/xt if $DMD/xt failed even though the error message said
	it did; fixed that so it tries both directories.

5/16	Changed libwindows/hostagent.c to always write a couple characters
	to the window after opening a remote channel for XT driver-based
	layers.  This had already been done for streams XT, and now it was
	also needed for character XT on amdahl 2.1.6a.  Made characters
	written out to be space-return so hopefully they shouldn't be
	noticed.

1995
====

3/9	Discovered that an "xtagent E" would result in the C_JAGENT response
	packet in getting sent to the shell after layers exitted rather than
	being read by layers, at least on a 630.  Added istermagentpending()
	to agent.c and added the places in control.c that use it to make
	sure that the response packet gets read up.  There was already logic
	in place to read an extra C_UNBLK from a 5620, and I reused that.
	I am not sure what would happen if "xtagent E" on a 5620 results in
	both a C_UNBLK and a C_JAGENT; I have no way of testing it.

3/23	Backed out the change from 3/9 because discovered a better solution
	that was more robust.  It turned out that the TIOCFLUSH ioctl that
	was called in preference to a TCFLSH ioctl did not drain standard
	input like it was supposed to on a Sunos 4* system just before
	layers exits.  Changed boot.c and utmp_tty.c to try both TIOCFLUSH
	and TCFLSH if both are defined.

5/31	Added required second parameter to gettimeofday in misc/movelogin.c 
	for the case of USE_UTXENT.  This is needed especially on Unixware
	2.01 but it should have been there all along.

6/23	This is not a change to layers, but a change to the software
	development package I want to note: I changed pcc2/common/manifest.h
	to add extern void exit() and #include "string.h" to get sty and
	therefore mc68ccom to compile on Solaris 2.4 and Unixware 2.01.

6/26	Added USE_ALT_PUTUTLINE and USE_ALT_PTYALLOC for use on HP-UX 9.0.

6/28	Added NO_WINSIZE option to be able to compile on a system that 
	doesn't have TIOCSWINSZ, for example HP-UX 8.0.

7/10    Added "-Xt" to the EXTRACFLAGS comment in Makefile for Unixware
	2.01, to keep it from complaining about redeclaring "malloc" in
	sty.c in pcc2/common in the 630 software development package.

11/20   Expanded EXTRACFLAGS comment in Makefile for Solaris 2.0 to include
	IRIX 5.3 or later.  Also, for the record, I removed the following
	from the README file for the combined ptlayers/sdp630 because I
	don't currently know of any cases that it applies to:

	Also see the comments on the "EXTRACFLAGS" variable in the
	ptlayers/Makefile because for some operating systems it is better
	to only set the EXTRACFLAGS for ptlayers and not sdp630; if there
	is a comment in ptlayers/Makefile for your operating system and not
	in the top level Makefile, edit the ptlayers/Makefile AND invoke
	the ptlayers and sdp630 Makefiles directly instead of going through
	the top level makefile.  This will avoid using the same EXTRACFLAGS
	for both packages.

1996
====
7/26	Minor changes to support compiling with gcc:
	- lib/libwindows/Makefile explicitly used 'cc' instead of $(CC).
	- This is not a change to layers, but to dmdld and 32ld: there is a
	  default optimization in gcc that makes memcpy a builtin function
	  and assumes that structures are aligned on 4-byte boundaries.
	  Dmdld and 32ld were using structure pointers to point to relocation
	  entries which are 10 bytes each and are sometimes not aligned.
	  Changed one of the pointers to char * so memcpy wouldn't make
	  any assumptions.
	- Also, gcc had a problem with mktemp("string_constant") in 32ld.
	  Apparently the string constant is placed in a read-only section.
	  Changed the calls to mktemp(strdup("string_constant")).

1997
====
8/27	Changed Makefile to show preferred flags for IRIX: remove -lsocket
	  and -lnsl and add -cckr to keep it from complaining about redeclaring
	  "malloc" in pcc2/common in the 630 software development package.
	  These are especially needed for IRIX 6.x but they work on IRIX 5.x.
10/29	Attempted to port to amdahl's UTS system:
	  - Added USE_POLL conditional compile to use poll(2) instead of select
	  - Added USE_PTY_FOR_PIPE because UTS's poll can't poll from a
	    named pipe
	  - Added setraw() function as a subset of setcntrl() which only
	      sets raw mode, for use in USE_PTY_FOR_PIPE
	  - Changed calls to fchmod to chmod (and added ttyname parameter)
	      in setcntrl, resetcntrl, and slavettyset because UTS doesn't
	      have a fchmod.  This slightly changes the behavior on
	      the "C_BRAINDEATH" control because it will change the mode
	      on the slave side of the pty instead of the master side,
	      which is the right thing to do anyway.
	  - Changed termio_setcntrl() and termio_resetcntrl() to use the
	      "saveterm" tty settings structure that was saved in savemode()
	      rather than using their own "ttysave" structure.
	  Unfortunately, the port did not succeed because of a bug for which
	    I could not find a workaround, at least in release 2.1.2: when
	    more data is available in a pty than is read, data disappears
	    in the queue until another write occurs to the slave side of
	    the pty, either with a poll or a read with O_NDELAY on or a
	    regular read without O_NDELAY.  This causes dmdld downloads
	    and xtagent "!" commands to fail.
	  Other remaining problems: I once saw the retried write in 
	    check_write after an EINTR hang.  May need to reorganize things
	    to avoid alarms going off inside that write.  Also, there's always
	    one defunct process owned by layers.

	Changed the exitagent() sequence to take advantage of a new Solaris
	    (at least) feature of being able to move utmp entries without
	    needing to make 'movelogin' setuid-root: the new pututxline()
	    will set a utmp entry if it has write access to the tty that is
	    being set, and the layers control channel is set to exclusive
	    access mode (TIOCEXCL) until after resetcntrl().  Moving
	    moveutmp() until after resetcntrl() solves the problem and
	    allows movelogin to not be setuid-root.  Since SVR4 already had
	    a grantpt() system call to avoid the need for the grantpt
	    setuid-root support program, this allows layers and all support
	    programs to be completely non-root on Solaris.


1998
====
2/11    Found that Linux 2.0 has "getutline" and "pututline" so removed the
	  '#undef USE_UTENT' from lyrscfg.h for linux and instead added two
	  new defines NO_UT_EXIT and USE_VOID_PUTUTLINE and implemented the
	  minor differences in misc/movelogin.c

	Also made these additional changes to the 630 software development
	  package for Linux at this time:
	    1. added "-fno-asm" to the recommended flags for Linux to avoid
		clashes with the keyword "inline" in building mc68ld.
	    2. changed ccs/pcc2/common/scan.c to cast NULL to an (int) because
		apparently its default type is something else on Linux.
	    3. changed a line ccs/as/common/pass0.c and in ccs/ar/common/ar.c
		from 'static char *seed="AAA"' to 'static char seed[4]="AAA"'
		to make gcc put the string into writable memory.
	    4. changed some format strings "0x%l.8x" to "0x%.8lx" in
		ccs/pcc2/mc68/local.c.
	    5. changed variable called "abort" to "getout" in
		dmdversion/dmdversion.c.
	    6. added putenv("POSIXLY_CORRECT=yes") before the call to getopt
		in ccs/dmdcc/mc68/dmdcc.c to make getopt behave properly.
	    Also, yacc bombed on a few ".y" files so had to make
	    ccs/pcc2/mc68/(cgram.c|sty.c) and ccs/ld/mc68/y.tab.[ch]
	    on a Solaris system and copy the results over.