System notes file for CTIO AOS/VS IRAF installation.
9-17 April 1987

System installed by Gary Lee Webb on 9 April.
CL came up and ran with minor problems.

Did not start keeping this notes file until 16 April (D.Tody).
May have missed some important mods.

Extant Bugs
------------

1. OS escapes.

	Interrupting an OS escape often causes the escape mechanism to
	get in a state where subsequent escapes will return immediately,
	even though the command has been sent to the cli, sh, or whatever.
	This leads to the CL and the host task reading from the terminal
	at the same time, seizing alternate lines of input.

2. Magtape i/o problem

	CTIO has a standard DG tape drive and two Cipher tape drives.
	The standard tape drive will accept transfers up to 32768 bytes
	(or maybe 32767 bytes), which is what the zfiomt.c driver code
	is written for.  The Cipher drives, however, have a maximum transfer
	size of 8192 bytes, hence the max transfer size is device dependent
	rather than just system dependent.  As a kludge fix for this I
	added some site dependent logic to zfiomt.c, but a real fix will
	probably require addition of the max transfer size to the dev$devices
	file, and changes to the VOS code to deal with this.

3. Path and script problems

	There were a number of problems with missing directories and
	confusion between executables and macro script tasks, leading to
	failure of mkpkg, xc, etc.  I will try to get Gary to document these
	better.
	    One of the problems (glw) was the use of directories C_DIR and
	F77_DIR to find the executable C and FORTRAN compilers.  Unfortunately,
	there are no standard directories for these (other than they should be
	in :UTIL), perhaps the standard C and F77 macros should be used rather
	than having IRAF write its own?

4. File dates

	After the system installation, a mkpkg on any package results in
	compilation of all of the files therein.  Evidently the file modify
	dates are not being restored properly when a backed up system is
	restored from tape.

5. Installation (glw)

	   The current link initialization macro will delete /TMP and link it
	to /DEV. I see no reason to combine these and did not do so.
	   For a user to be able to generate any processes under IRAF, he also
	had to be a valid MV/UX user (i.e., have an entry in /ETC/PASSWD).  If
	this is not a bug, it should be documented in the installation guide!
	   There are !SOLPLs all over the place: a list of files to be changed
	would be nice.

6.  Worries... (glw)

	   I note that the editor description files end with a .ed, just like
	the SED temporary files, making them likely to be deleted accidentally.


System Revisions
--------------------

aosvs/os/zfiomt.c
	Did a kludge fix to set the max transfer size down to 8192 bytes
	for devices other than mtb0.  (4/15 dct)

dev/vi.ed
	Changed the OS escape for VI from "vi" to "!vi" to cause the command
	to be executed by CSH rather than CLI.  Also added an entry for the
	vi500 to /etc/termcap, but I could never get VI to find the entry
	for the device.  Tried setting TERMCAP in .cshrc to point to another
	file, or directly to the vi500 termcap entry, but none of this made
	any difference.  (4/16 dct)

aosvs/os/prwait.c
	The include file <sys/wait.h> was not being found; it turns out to
	be in /usr/include on this system.  I had to reference the file as
	"/usr/include/wait.h" rather than the expected <wait.h> to get the
	module to compile.  It appears that the real include files are
	coming from some other place (maybe a text library?) and that the
	wait.h file is missing.  (4/16 dct)

aosvs/boot/rtar/mkpkg
	Added a "-B" flag to the $link call.  Without this the link fails
	with a multiple reference to main.o, and unresolved externals for
	all the bootlib routines.  It looked like this was missing from
	most of the other boot packages too.  (4/16 dct)

dev/slate.ed  (glw)
	Added this file (guessing a lot!) to provide Dan's favorite editor.
	It must work -- that's how I'm adding this comment!
	Where is the documentation for the *.ed format?

dev/sed.ed  (glw 16 IV 87)
	Modified the command line to be SED/NOED to avoid leaving *.ed files
	all over the place.

dev/termcap (glw 16 IV 87)
	Added lpt1, lpt4, and lpt5 printers, a/k/a adservs, diablo, laser.

pkg/images/imdebug/mktest.x
	Would integer overflow when creating a large image.  This is harmless
	on the other systems, but it has to be guarded against on the MV as
	it causes the task to abort.  (4/17)

pkg/images/iminfo/t_imstat.x
	Installed an optimized version of the IMSTAT task.  (4/18)

aosvs/boot/mkpkg/*.[ch]
	AOS/VS evidently cannot restore the modify dates of files when reading
	a tape onto disk.  This causes mkpkg to try to recompile everything
	when run for the first time on a newly installed system.  Modified the
	mkpkg program to add a new flag "-u".  This flag, if present, causes
	the dates of library modules to be forced to be no less than the date
	of a magic file (currently hlib$iraf.h).  It is assumed that the date
	of the magic file is about the same as the date at which the system
	was installed.  To be precise, the file should be touched after the
	tape is read in (already done for the CTIO system), and the first
	time mkpkg is run on a package the -u flag should be used to forcibly
	update the library module dates.  (4/18)

---------------------------------------------------------------------
From SKIP@SOLPL.AS.ARIZONA.EDU Thu May 28 12:14:58 1987
Received: from noao.arpa by noao-lyra.arpa.noao (5.51/SAG.7)
	id AA08983; Thu, 28 May 87 12:14:53 MST
Received: from solpl.as.arizona.edu by noao.arpa (5.51/SAG.7)
	id AA03426; Thu, 28 May 87 12:14:46 MST
Received: by SOLPL.AS.ARIZONA.EDU (1.00/1.0)
	  id AA00065; Thu, 28 May 87 12:14:20 mst
Date: Thu, 28 May 87 12:14:20 mst
From: Skip Schaller <SKIP@SOLPL.AS.ARIZONA.EDU>
Message-Id: <8705281714.AA00065@SOLPL.AS.ARIZONA.EDU>
To: chile@noao, tody@noao
To: Dan@SOLPL.AS.ARIZONA.EDU, Smith@SOLPL.AS.ARIZONA.EDU,
        Gary@SOLPL.AS.ARIZONA.EDU, Webb@SOLPL.AS.ARIZONA.EDU
Fm:  Skip Schaller

	I will be leaving Tucson on June 18 for Chile.  I will call you
on Monday morning June 22.  I expect to be able to pitch in immediately
if you so desire.  I will be in Chile until August 16.  Let me know as
soon as you know what it is exactly that you want me to do for you.  If
I can prepare anything here ahead of time, so much the better.

	Please send me any AOSVS/IRAF bug reports as soon as possible.  It
will be much easier for me to fix them here.


	The following are my responses to the CTIO AOSVS/IRAF installation
notes that I got from Doug Tody:

1)  Keyboard interrupt during OS escape.
	I duplicated this problem at solpl.  I will try to fix during
this next update.  It may be an AOSVS problem and not fixable.

2)  Magtape maximum record size for certain drives.
	Doug changed the IRAF VOS just the other to deal with this problem.

3) Pathname problems.
	As far as I can tell (by looking at other AOSVS systems),
F77_DIR and C_DIR are the standard DG directories for those languages.
In any case, the installation manual tells you which scripts to check
and modify to agree with your system.  (I had to do this for the
Tenerife installation.  It was trivial.)  Unfortunately, the standard
scripts cannot be used due to their lack of functionality and interface
to mkpkg.

4) File dates.
	Hopefully by the next release, Doug's changes to mkpkg to update
file dates for library members without recompiling, will be incorporated.
Many DG sites have complained about AOSVS not restoring the original
file modification times.

5) Link installation problems.
	The link installation script does NOT delete /TMP and link it to /DEV.
Read it again.  With MV/UX installed, that part of the script does nothing.

6) Problems with /etc/passwd.
	I could not reproduce this problem at solpl.  I did reproduce it
during the Tenerife installation.  I found out that the minimum needed
was to have this file present with one entry for user "iraf".
	Sometime ago I tried to eliminate IRAF dependence on this file
by avoiding the use of those C subroutines given in the DG documentation
that access this file.  Apparently, there is at least one more, execl ().
In any event, the DG C subroutines should do something more graceful when it
can't access the information it wants.  I will take this matter up with
DG.  In the meantime, all current AOSVS/IRAF sites now have MV/UX so they
should really keep this file up to date for all users.

7) Node name changes.
	There are NOT solpl!'s all over the place.  They are confined to
the files which may contain site dependence.  In any event, since CTIO
does not have any networking software, the IRAF networking is automatically
turned off and the solpl!'s are harmless and need not be changed.  If there
are any files that particularly bother you, give me their names and I'll
see what I can do.

8) Editor descriptor files (.ed).
	The editor descriptor 'edit' was already provided so that SED
does not generate files with conflicting extensions.

9) Vi.
	We execute vi directly from the CLI.  We use a slightly different
entry for the vi500 in /etc/termcap than the one used by IRAF, so as to
get around some DG bugs.  The user needs a .exrc file in his home directory
to set the terminal type.

10) Wait.c
	The include file wait.c is missing from the DG C release and
should be copied from :usr:include:wait.h to :util:c_dir:sys.

11) Mkpkg -B flag.
	The -B flag was present in the mkpkg file in the immediately superior
directory, but I will put it in all the subdirectory mkpkg files as it
should be.

12) No documentation for dev$*.ed files.
	I agree with you, Gary.  However it turns out that most of it
is not necessary to change.

13) image$imdebug/mktest.x
	I reported the integer overflow problems to Doug some time ago
when I ran the benchmarks.


NNN