Sun-Spots-Request@RICE.EDU (William LeFebvre) (12/16/87)
SUN-SPOTS DIGEST Tuesday, 15 December 1987 Volume 5 : Issue 70
Today's Topics:
Re: Has anyone ported timed to the Sun
Re: 1 Mbyte 9 bit SIMMS for Sun3/60's
Sun-4 compiler stuff and deliveries
1/2 inch dumps on Sun
Hardware to run Inference-ART
cc Optimizer problem (bug?)
SCSI disk label gets wiped on reboot
rows & columns in stty(1)
incomplete loading of images
DNI 5.0 problems
Quota problems
Need help with printer spooler
VT240 emulator? (2)
Confining users to shells?
Redisplay All?
disk versus memory?
Text font in Sun3?
----------------------------------------------------------------------
Date: 4 Dec 87 18:55:33 GMT
From: lou@aerospace.aero.org (Lou Nelson)
Subject: Re: Has anyone ported timed to the Sun
I received a "timed" that works on the Suns from Matt Landau. Special
thanks! Also I received helpful messages from Jason Venner, Steve Miller,
and Eric Johnson. I thank you all.
Lou Nelson
lou@aerospace.aero.org
...!trwrb!aero!lou
------------------------------
Date: Fri, 4 Dec 87 17:25:07 EST
From: "William R. Pearson" <wrp@biochsn.acc.virginia.edu>
Subject: Re: 1 Mbyte 9 bit SIMMS for Sun3/60's
Regarding 1 Mbyte 9 bit SIMMS for Sun3/60's. Check out the memory chip
supplier in Begg's Oklahoma (IC's Unlimited? I just forgot their names,
they advertize in Byte, PC-Week, etc). Apparently they sell 100ns 9 bit 1
Meg SIMMS for about $260 ea.
------------------------------
Date: Thu, 3 Dec 87 18:26:12 PST
From: hoptoad.UUCP!gnu@cgl.ucsf.edu (John Gilmore)
Subject: Sun-4 compiler stuff and deliveries
I finally got a printed price list that includes Sun-4's (dated October 5,
arrived December 1) and found a few interesting things:
* SunOS for the Sun-4 does NOT include Fortran or Pascal compilers.
They are an extra cost option at $1000 apiece. (Any volunteers for
GNU CC front ends for these languages? I'm sure we'll see a SPARC
code generator pretty soon, and we can probably use much of
Berkeley's Fortran library.) See page 25 of the price list.
* When you order any Sun-4 compiler (including the C compiler that
comes with SunOS), you can get a cross-compiler to Sun-2 or Sun-3 at
no charge, if you order it at the same time. These are on page 36 of
the price list.
* The SunOS feature list mentions the assembler for Sun-3, but not for
Sun-4, though there is no separately priced assembler, so this is
probably just an oversight.
By the way, has anybody actually received a Sun-4 yet? [[ Apparently
so---see the next article. --wnl ]] My Sun-3 to Sun-4 upgrade was ordered
the day after the announcement (July 8), which said "shipments begin in
late August". Originally I was promised September delivery, but it's been
slipping and slipping and slipping. Currently expected sometime in
January, running SunOS 3.5, with 4.0 deferred until early summer '88. I'd
rather have it late and have it work, than the reverse, but am wondering
what other customers are seeing.
------------------------------
Date: 5 Dec 87 19:22:18 GMT
From: km@emory.UUCP (Ken Mandelberg)
Subject: 1/2 inch dumps on Sun
We just tried doing a dump to our Sun 1/2 inch 6250bpi tape drive from our
Sun4. As reported earlier by others in this newsgroup, you can get about
twice the througput out of the 4.3BSD version of dump as opposed to the
Sun provided version.
I was kind of hoping that the Sun4 would be fast enough that with the 4.3
dump program the tape would stream. This is not the case. The tape makes
about one rotation and then backups a little. The overall throughput seems
no better than that reported for the Sun 3/280. o
If anyone knows any further tricks I would like to hear it. I'm not eager
to change the kernel to speed things up right now. I also wonder if Suns
VME based controllers (if they ever come) will make a difference.
Ken Mandelberg | {decvax,sun!sunatl,gatech}!emory!km UUCP
Emory University | km@emory BITNET
Dept of Math and CS | km@emory.ARPA ARPA,CSNET
Atlanta, GA 30322 | Phone: (404) 727-7963
------------------------------
Date: 2 Dec 87 15:29:23 GMT
From: mcvax!dnlunx!lippolt@uunet.uu.net (Ben Lippolt)
Subject: Hardware to run Inference-ART
Is there any one out there who can answer one or more of the following
questions:
We have a number of Sun-workstations and use Inference-ART as our
development tool.
* Is it possible to use ART on a Sun 3/60 with 16 Mbyte of memory?
* What kind of disk do we need for a standalone system? What size
must the disk have? Is a SCSI disk fast enough, or do we need a
SMD disk?
We plan to use a Sun 3/160 with 380 Mbyte disk as a fileserver for
a number of clients.
* Is it possible to use ART on a client? If so, does the client need
a private swap-disk and how large should this swap-disk be? Is a
71 Mbyte SCSI disk enough as a swap-disk?
* Is it possible to use ART on the fileserver (a 3/160 with 16 Mbyte
memory and still serve one or two (small) diskless clients?
Thanks.
Please respond via e-mail.
------------------------------
Date: Fri, 04 Dec 87 14:51:23 EST
From: sdo@purdue.edu
Subject: cc Optimizer problem (bug?)
I have discovered what appears to be a bug in the optimizer for Sun's C
compiler. We're using sun 3-50's, with the standard C compiler for Sun OS
3.4.
We have been porting our Xinu operating system to a Sun 3-50. Xinu is a
small operating system used for classroom experimentation. It runs on a
bare Sun-3 in standalone mode. Everything had been working fine until we
noticed that register variables were being corrupted. I traced it to a
harmless procedure that was running with interrupts disabled. It seems
that the optimizer generated the following code to set a global variable
to 1 (TRUE):
moveq #1,d7
movl d7,_globalvar
without ever saving or restoring the value already in d7. This was
destroying the value already stored in d7 as a register int by the
calling procedure. The code generated by the compiler without the -O
option is:
movl #0x1,_globalvar
which works fine. It seems that the problem stems from the macros
which we are using to disable and restore interrupts in that they are
confusing the optimizer. Is this reasonable? Is there some rule
about when it is safe to use the optimizer in the presence of
asm("...") statements??
Our temporary solution is to compile without the optimizer, but I
would really like to know what is going on, and if it can be prevented.
Can anyone lend me an explanation?
Shawn Ostermann
_______
The following short program fragment and Makefile demonstrate the
problem. Since the code generated is intended to run as standalone,
the disable and restore statements will not operate when run under
Unix, but work fine for our purposes. Notice the code generated in
bug.s and nobug.s in the bomb procedure.
================================================
bug.c
================================================
int globalvar;
/* macro to store the SR onto the stack and disable interrupts */
#define disable() asm("movw sr, sp@-"); asm("movw #0x2700, sr");
/* macro to restore previous SR from the stack */
#define restore() asm("movw sp@+, sr");
main()
{
register int i;
i = 100;
printf("i is %2d\n",i);
bomb();
printf("i is %2d\n",i);
}
bomb()
{
disable();
globalvar = 1;
restore();
}
==============================================
Makefile
==============================================
bug: bug.c
cc -S -O bug.c
cc -O -o bug bug.s
nobug: bug.c
cc -S bug.c -o nobug.s
cc -o nobug nobug.s
--------------------------------------------------------
Shawn Ostermann ARPA: sdo@gwen.cs.purdue.edu
UUCP: ...!purdue!sdo
--------------------------------------------------------
------------------------------
Date: Fri, 4 Dec 87 15:36:10 EST
From: dcornutt%murphy@gswd-vms.gould.com (Dave Cornutt)
Subject: SCSI disk label gets wiped on reboot
Help! I have a 3/160 with two SCSI drives. Both are the Fuji 70M drives.
Practically every time I reboot this thing, I find that the label on drive
1 has been wiped out, and I have to boot the diag program and restore the
label from a backup. This happens on both voluntary and involuntary
(e.g., power hit) reboots; it happens about 3/4 of the time. I've tried
reformatting the drive and running the read/write tests on it, and I can't
find any bad blocks that weren't already in the bad block table. Swapping
out the SCSI board didn't help. Before I go attempt to disassemble the
cabinet and get at the drives (which is like trying to get into a bank
vault), does anyone have any ideas?
Dave Cornutt, Gould Computer Systems, Ft. Lauderdale, FL
[Ignore header, mail to these addresses]
UUCP: ...!{sun,pur-ee,brl-bmd,uunet,bcopen,rb-dc1}!gould!dcornutt
or ...!{ucf-cs,allegra,codas,hcx1}!novavax!gould!dcornutt
ARPA: dcornutt@gswd-vms.arpa or .gswd
------------------------------
Date: Tue, 1 Dec 87 14:37:27 EST
From: ihnp4!vax135!whb@ut-sally.UUCP (Wilson H. Bent)
Subject: rows & columns in stty(1)
When you give the command 'stty everything', you get the standard 4.2BSD
bunch of information, PLUS rows & columns:
$ stty everything
new tty, speed 9600 baud, 24 rows, 80 columns
[ etc. ]
It seems that wherever this information comes from, that's where 'more'
and 'vi' (and others?) look first, ignoring TERMCAP. Actually, when using
a regular terminal, these values are both set to 0, which probably means
"that information must be obtained from TERMCAP". The only problem is,
under 'screen' (the window manager for standard terminals), the row &
column values are often erroneous: 10 rows, 40 columns seems to occur
fairly often, and today I got 42 rows, 80 columns (all on this 24x80
terminal of mine). This makes 'more' and 'vi' unusable.
The problem is not inherently in 'screen' since this feature is
undocumented - at least it's not in tty(4), which is all 'screen' is
manipulating.
I'm hoping this is just another case of RTFM (yes, but WHICH manual page?).
Wilson H. Bent, Jr. ... ihnp4!vax135!whb
AT&T - Bell Laboratories (201) 949-1277
Disclaimer: My company has not authorized me to issue a disclaimer.
[[ Reading our on-line manual page for stty(1) reveals no mention
whatsoever of options "rows" and "columns". Yet I have personally
succeeded in changing the settings with "stty rows 24" and "stty columns
80". A brute force way to `set things right' is "stty columns 0 rows 0".
Turns out that they added a new structure to the tty driver: winsize. One
can get and set it with the ioctl functions TIOCGWINSZ and TIOCSWINSZ.
Where is this documented? Nowhere that I can find. I found it out only
by looking in /usr/include/sys/ioctl.h. It is not mentioned in stty(1) or
tty(4), which both claim to be for Sun 3.2. It is hinted at in
termcap(3x) but not mentioned in termcap(5). Anyone at Sun care to tell
me where this *is* documented? What's even worse, I have gotten pseudo
tty lines via an inward telnet (from a generic 4.3BSD system) where the
rows and columns field were not preset to 0. They still had (presumably)
whatever value was lingering from the last process. This was even before
I discovered the "rows" and "columns" settings. I had a hard time
figuring out why my paginator insisted on displaying 50 lines at a time!
--wnl ]]
------------------------------
Date: Fri, 04 Dec 87 14:05:21 PST
From: Vern Paxson <vern%lbl-helios@lbl-rtsg.arpa>
Subject: incomplete loading of images
We're experiencing a strange problem on a monochrome 3/160 running 3.3.
Sometimes large executables are not fully loaded - some of the code is
garbage rather than instructions. Often but not always the load aborts
with an EMT trap. The problem is quite sporadic, sometimes occurring only
if dbx or adb is used. Making one or more copies of the executable
usually produces a copy that consistently works. The hardware on the
machine is:
Interphase V/SMD 4200 controller
Fujitsu Model 2344 disk drive
FPA board
We have a color 3/160 with the same controller and disk drive but no FPA
which has not shown this problem. Pulling the FPA on the monochrome,
however, has not made the problem go away.
If anyone has had a similar experience, I'd appreciate hearing from them.
Thanks, Vern
Vern Paxson vern@lbl-csam.arpa
Real Time Systems ucbvax!lbl-csam.arpa!vern
Lawrence Berkeley Laboratory (415) 486-6411
------------------------------
Date: Fri, 4 Dec 87 15:23:26 mst
From: modular!earley@arizona.edu (Joe Earley)
Subject: DNI 5.0 problems
Today we put up the 5.0 DNI software on our Sun 3/60. Within a span of
about 10 minutes we discovered that file transfers worked great but we
couldn't edit files on a 780 running VMS 4.6 without seeing the cursor
movement characters echoed back to our terminal screens. We suspect that
the 780 crash which followed the discovery of the editor problem was
caused by the DNI software. We quickly retreated back to the old version
of DNI. Has anyone else had similar problems? Did we miss something in
the manuals in getting it loaded? We've seen the note about pass through
mode not working. Does this mean that DNI does all echoing of characters?
Surely not.
We probably will not attempt to try it again until we can have some free
time on the 780. In the meantime any help or suggestions would be greatly
appreciated.
Joe Earley, Modular Mining Systems
USENET: {ihnp4,allegra,cmcl2,noao}!arizona!modular!earley
INTERNET: modular!earley@arizona.edu
------------------------------
Date: 5 Dec 1987 17:05-EST
From: Ralph.Hyre@ius2.cs.cmu.edu
Subject: Quota problems
Judging from recent traffic, not many people seem to use quotas on their
Suns, or maybe nobody is having any problems.
I also notice the apparently bogus entries in /etc/mtab on a client which
claims to mount remote filesystems with the quotas option in /etc/fstab.
The entries are correct on the server.
/etc/fstab says a0:/usria0 /usria0 nfs rw,hard,bg,quota 0 0
BUT
/etc/mtab says a0:/usria0 /usria0 nfs rw,hard,bg,noquota 0 0
Quotas are enabled in the kernels and turned on both the client and
server. What could be wrong here? Is it an undocumented 'Quotas and The
NFS' sort of problem related to the diffeciencies in NFS quotas?
The other problem is that the 'quotacheck -a -v' in /etc/rc.local
incorrectly computes quotas for some users on a various filesystems,
typically doubling their blocksize figures. After the system is rebooted
I typically have to log in and run quotacheck -v <filesystem> manually to
fix it.
[relevant section of rc.local]
/etc/mount -vat nfs >/dev/console
(echo -n 'check quotas: ') >/dev/console
/usr/etc/quotacheck -a >/dev/console
(echo 'done.') >/dev/console
/usr/etc/quotaon -a
------------------------------
Date: Sun, 6 Dec 87 13:44:59 PST
From: John Bossert <bossert@thebes.thalatta.com>
Subject: Need help with printer spooler
Having beaten my head against the wall, I'm turning (again) to a higher
authority...
I'm using the devps package on a Sun3/50 running SunOS 3.3 with a
Postscript printer. The "native" devps attempts to bypass the 4.2/3 line
printer spooler to a large degree. I'm trying to integrate it into the
lpd spooler package and am running into some problems. If I only define
the "if" filter and use "lpr -h", I'm ok. I've hacked the devps
plain-text printer into a filter that puts a Postscript envelope around
the file to be printed, and defined it as "if".
Where I run into troubles is if I don't print things as "lpr -h". The
information that would go on the burst page doesn't seem to get caught
within the envelope I build in "if". It thus disrupts the Postscript
stream and I get nothing.
So, reading the manual, I interpret it to say that if I define an "of"
filter, it will process the burst page, leaving "if" to handle the file to
be printed. Ok, I copy the "if" filter to "of" thinking that I will
essentially get two jobs sent to the printer. Then, I can redefine "of"s
output to /dev/null and my problem is solved. Well...
This not only doesn't work, but seems to hang lpd. So, after trying many
combinations of things (including looking at how Adobe handles things with
Transcript), I'm turning to you. My question:
How can I trap and process (send to /dev/null) the header/burstpage
information gotten when one Does Not use "lpr -h"?
As a workaround, how can I set up mailtool to use "lpr -h" as the command
it uses to print a file? Doing the obvious, (setting "printmail" to
"/usr/ucb/lpr -h" in mail defaults) doesn't do it.
Gratefully,
John Bossert
In-Real-Life: John Bossert, Thalatta Corporation, (+1 206 643 7187)
Domain: bossert@Thalatta.COM Path: uw-beaver!uw-entropy!thebes!bossert
[[ Have you tried setting the boolean printcap capability "sh" for that
printer (i.e.: in the /etc/printcap file)? It is supposed to tell lpr to
always suppress the burst page. --wnl ]]
------------------------------
Date: Thu, 3 Dec 87 16:45:35 EST
From: Bob Chatham x8677 11/331 <bchatham@bfly-vax.bbn.com>
Subject: VT240 emulator? (1)
Does anyone know of a VT240/340 emulator for the Sun?
Thanks,
Bob Chatham
------------------------------
Date: Mon, 7 Dec 87 11:11:55 EST
From: David B. Leonard <dbl@cs.duke.edu>
Subject: VT240 emulator? (2)
Anyone know of a free (or real cheap) VT240 emulator tool for Sun 3's
(we're not picky as to whether it's implemented for X-windows or SunView)?
Alternatively, how about a tool similar to tektool, but which also does
VT100 escape sequences like the popular PC VT100/Tek 4014 emulator
programs and emulator terminals such as the Selenar Hirez 100XL?
If this has been covered ad infinitum on this list (I am not a frequent
reader), then I would appreciate anyone who could respond to this
information request directly.
David B. Leonard -- National Biomedical Simulation Resource
Duke University Medical Center, Box 3709, Durham, NC 27710
(919)681-3048,3049 - US WATS (800)334-2083 - NC WATS (800)672-2543
{decvax,mcnc}!duke!dbl dbl@duke.cs.duke.edu
------------------------------
Date: 5 Dec 87 03:26:09 GMT
From: sundc!potomac!potomac.ads.com!jtn@uunet.uu.net (John T. Nelson)
Subject: Confining users to shells?
Has anyone experience with confining certain user accounts to shells which
let them perform only certain functions in a manner that is reasonablly
secure (i.e. that can't just break out of the shell with a ctrl
character)?
I would like to write such a shell but haven't had much luck so far.
John T. Nelson UUCP: sun!sundc!potomac!jtn
Advanced Decision Systems Internet: jtn@potomac.ads.com
1500 Wilson Blvd #512; Arlington, VA 22209-2401 (703) 243-1611
Sine Visa Ars Nihil Est
------------------------------
Date: 2 Dec 87 10:16:05 GMT
From: mcvax!swivax!huub@uunet.uu.net (Huub Knops)
Subject: Redisplay All?
Is there a simple way to "Redisplay All" from a C program ?
Huub Knops, (huub@swivax.UUCP)
SWI Universiteit van Amsterdam
------------------------------
Date: 4 Dec 87 23:59:00 GMT
From: sun!rtech!llama!sid@ucbvax.berkeley.edu (Sid Shapiro)
Subject: disk versus memory?
Hi folks,
I'd like to ask one of those question for which there may not be an
answer, or at least not an easy one. So.
Given a development environment --- compiles, loads, debugs --- which
would you rather have - a 3/60 with lots of memory (say in the 16-24 MB
range) or a local disk. Assume that storage of your sources is on remote
disks. Just on performance alone - startup of programs and paging
activity - which is better 16 MB of memory or a local disk.
Thanks,
Sid Shapiro -- Relational Technology, Inc
{mtxinu,cpsc6a}!rtech!sid
(415)748-3470
[[ How much does the 3/60 have *with* the local disk? If it's anything
less than 4Mb, go for the memory. --wnl ]]
------------------------------
Date: 5 Dec 87 00:53:50 GMT
From: quintus!nitin@sun.com (Nitin Rai)
Subject: Text font in Sun3?
Does anyone know of a rotated text font (90 deg counter clockwise) on a
Sun? I use a Sun 3 that runs SunView.
Nitin
e-mail address:
...!sun!quintus!nitin OR nitin@quitus.UUCP, nitin%quintus@SUN.COM
------------------------------
End of SUN-Spots Digest
***********************