Sun-Spots Digest, v6n134
William LeFebvre
Sun-Spots-Request at RICE.EDU
Tue Jul 12 12:19:14 AEST 1988
SUN-SPOTS DIGEST Monday, 11 July 1988 Volume 6 : Issue 134
Today's Topics:
Re: NFS bug
Re: Calctool problem
SunOS umount and single-process deadlock (open letter to Sun)
Sun 3/160 question: display separate from CPU?
Reading large blocks off 1/2" tape?
Problems with big swap space on Sun 3/140 (SunOS 3.4)?
Any advice on using Fujitsu 2372's?
cmdtool under OS 4.0?
What does this error mean?
dumpregion -> postscript?
Tek 41xx emulator?
help with gdb and dbx using FORTRAN
Need information regarding optical disk.
Send contributions to: sun-spots at rice.edu
Send subscription add/delete requests to: sun-spots-request at rice.edu
Bitnet readers can subscribe directly with the CMS command:
TELL LISTSERV AT RICE SUBSCRIBE SUNSPOTS My Full Name
Recent backissues are available via anonymous FTP from "titan.rice.edu".
For volume X, issue Y, "get sun-spots/vXnY". They are also accessible
through the archive server: mail the request "send sun-spots vXnY" to
"archive-server at rice.edu" or mail the word "help" to the same address
for more information.
----------------------------------------------------------------------
Date: Tue, 28 Jun 88 15:57:59 EDT
From: dff at morgan.com (Daniel F. Fisher)
Subject: Re: NFS bug
Reference: v6n116
Rob Nagler writes:
> We are running Sun OS 3.2 on Sun 3 machines. Given the "right"
> circumstances, we have seen seriously corrupted data written by an NFS
> client to a network partition. . . .
Last January, I observed similar behavior when using NFS to write a log
file from a Sun 3/110 running 3.4 to a Sun 3/280 also running 3.4.
As Rob describes, data written to the file by the workstation did not make
it onto the disk when a program such as "tail -f" is simultaneously
reading the same file. I spent sometime experimenting with this
phenomenon and determined that the "right" circumstances are that the
writer be writing on an NFS mounted partition with the O_APPEND flag set
and that the reader be reading the same file simultaneously--either from
the server, the workstation or other machine. This bug did not occur if
the file was either ND or 4.2 mounted. Nor did is happen if the O_APPEND
flag was off. A C program that demonstrates this bug is included below.
I reported the bug to Sun Support. On 29 Jan 88 I was informed that it
had been assigned SS# 138427 and had been reported to sun engineering with
severity 1. On 19 Feb 88 I was informed that the bug had been fixed in
SunOS 4.0 Beta-1. I have not had occasion to check this though I think
that I will soon.
Hope this helps, though given the migration schedule for SunOS 4.0, I
think it would have been better to see this bug addressed in SunOS 3.X.
Disclaimer: The opinions and conclusions expressed here are my own and do
not necessarily reflect those of my employer. They represent my
understanding of these issues at this time.
Daniel F. Fisher (dff at morgan.com)
-------------
/* bug.c: This program demonstrates an NFS bug related to the
* O_APPEND flag.:
*
* When this program finishes, ./out should have 5 lines
* that look like this:
* "01234567890123456789012345678901234567890123456789".
* But when the file is written over NFS and simultaneously
* tailed, as in "bug & tail -f out", some characters are lost.
*
* In this test program, the output is slowed down by not
* writing more than 1 character at a time, and by pausing,
* counting to 10000, between characters. While slowing writes
* increases the effect of bug, it does not introduce the bug.
* As a trivial modification would demonstrate.
*
* The bug goes away when the O_APPEND flag is taken away--demonstrated
* by yet another trivial modification. This bug also goes away when
* the bug program and the out file are all hosted by the same machine,
* either directly on a server or through ND on a workstation.
*/
#include <sys/file.h>
#include <stdio.h>
#define FILEFLAGS (O_CREAT | O_APPEND | O_TRUNC | O_WRONLY)
char *str = "01234567890123456789012345678901234567890123456789\n";
int main()
{
int fd, z, i, n, k;
if ((fd = open("out", FILEFLAGS, 0666)) < 0)
{
perror("open"); exit(1);
}
for (z = 0; z < 5; z++)
{
for (i = 0; i < 51; i += n)
{
k = 1;
if ((n = write(fd, str + i, k)) < 0 || n > k)
{
perror("write"); exit(1);
}
for (k = 0; k < 10000; k++);
}
}
exit(0);
}
------------------------------
Date: Wed, 29 Jun 88 08:46:04 EDT
From: Chuck Musciano <chuck at trantor.harris-atd.com>
Subject: Re: Calctool problem
Ted Schroeder writes:
> HELP! What'd I do wrong? I got calctool from the archives and (I think)
> followed the instructions for installation properly. But instead of a
> nicely layed out keyboard I get something that looks like it has one too
> few rows (like the main window wasn't opened big enough or something).
Steve Cole <steve at hanauma.stanford.edu> had this same problem, using the
correct fonts (the superscripts worked, and the special characters like
the root symbol were there). He made the following change:
> When I installed it the keyboard layout was quite
> strange - 11 keys/row, and things obviously not lined up right. I poked
> around and changed WIN_COLUMNS on line 148 of screen.c from 45 to 28
> (since it looked like 7 keys/row was correct) and everything works great.
> I'm wondering if my hardware (3/110) is somehow different that 45 worked
> for you, or if this was truly a bug.
Ted Schroeder shows six keys to a line, rather than eleven, but fooling
with the WIN_COLUMNS attribute is the fix. I would suggest that Ted
change 45 to 48 or 50, which gets the "wrap" right when SunWindows starts
putting in the buttons. I apologize for this behavior, and have no idea
what is causing it. Different SunWindows releases? If anyone can shed
some light, I'd be very glad to hear their ideas.
Chuck Musciano
Advanced Technology Department
Harris Corporation
(407) 727-6131
ARPA: chuck at trantor.harris-atd.com
------------------------------
Date: Wed, 29 Jun 88 13:50:15 EDT
From: dae at shire.cs.psu.edu
Subject: SunOS umount and single-process deadlock (open letter to Sun)
Work-Phone: +1 814 865 9505
Home-Phone: +1 814 862 4811
Silly Sun triviality #237: Under 3.2 and 4.0, if umount -a fails once, it
may fail forever!
[boot system single-user]
# mount -a
# cd /sys/sundev
# ...edit files, re-compile a kernel
# umount -a
/usr: Device busy
[silly me, of course it's busy--that's my current directory!]
# cd /
# umount -a
/usr: Device busy
[Huh?]
At this point a ps shows a "umount -a" running in the background, off in
/usr, trying furiously to unmount the file system it can't unmount because
it's trying to umount it.
Come on guys. If you're going to change the behavior of a command, don't
make the new behavior the default, ok? Especially if it's something
that's been the same since, say, Version 7, so that people think they know
how it works?
Perusing the man page suggests that an option will turn this behavior on,
but only for NFS file systems, which this wasn't. I didn't see any
mention that -a forces this.
I like Sun boxes. They're fast. I used to like them more, because they
used to run Berkeley Unix. I don't object to innovations, but when I bump
into some silly POINTLESS triviality every time I try to get some work
done, I begin to wish somebody would start porting 4.3 to Sun hardware.
To be fair, this is small stuff compared to the butchery DEC is doing with
Ultrix.
--Daemon
------------------------------
Date: Tue, 28 Jun 88 13:48:51 EDT
From: Sean Marrett <neuro at moe.mcrcim.mcgill.edu>
Subject: Sun 3/160 question: display separate from CPU?
Happy Days !
Well, our Pixar image computer has finally arrived, and we are ready to
install it, along with it's 3/160 host. We've hit a snag though. The
system as supplied comes in a large (actually two) 72 inch rack. We would
like to have the display and keyboard in a nice quiet room about 50 feet
away, but our local Sun office says that this is "not supported" (Sound
familiar ?).
So, does anyone know the trick to installing a Sun 3/160 display and
keyboard 50 feet away from the computer (out of sight and ear) ? Please
email any responses to me, and if there is a lot of interest, I will post
a summary to the net. This SHOULD be trivial.
Sean Marrett
Internet: neuro at moe.mcrcim.mcgill.edu
: sean%petvax%medcor.mcgill.ca at cornellc.ccs.cornell.edu
Bitnet: sean%petvax at medcor.mcgill.ca
McConnell Brain Imaging Unit
Montreal Neurological Institute
3801 University St.
Montreal, Que.
Canada H3A 2B4
Voice: (514)-284-5830
------------------------------
Date: Tue, 28 Jun 88 21:17:57 +0300
From: leonid at TAURUS.BITNET
Subject: Reading large blocks off 1/2" tape?
Has anyone got experience with reading large blocks (512K and up) off a
1/2" tape drive connected through a Xylogics Pertec board ? My attempts
all ended up with hard error 33 - Data Late. Any info will be appreciated.
Leonid
@cunyvm.cuny.edu:leonid at math.tau.ac.IL
leonid at taurus.BITNET
------------------------------
Date: 28 Jun 88 21:11:01 GMT
From: declarer/Karl B./dummy <umb!candy at husc6.harvard.edu>
Subject: Problems with big swap space on Sun 3/140 (SunOS 3.4)?
We've repartitioned a disk here (a 141MB Micropolis, if it matters) to
have a 32MB swap space. (b partition, in other words.) /etc/pstat -s
complains about a ridiculous number of procs/texts; is this a problem with
pstat, or with the disk? Also, we wanted to make a new kernel with
maxusers increased to 16 or so. That new kernel can't be booted on,
however -- the screen goes dead as soon as it loads, and doesn't return
until the Sun is power cycled.
Anyone have any experience with swap sizes other than 17.1?
This is a Sun 3/140 running 3.4.
Karl. karl at umb.edu ...!harvard!umb!karl
------------------------------
Date: Wed, 29 Jun 88 15:57:15 EDT
From: dela at ee.rochester.edu
Subject: Any advice on using Fujitsu 2372's?
We are thinking of hanging a Fujitsu 2372 on a 3/260 and another on a
3/280. Both machines have a Xylogics 451 controller.
Has anybody else used this drive? Did it work OK? What are the answers to
init's questions about how to format the drive? How did you determine the
answers to init's questions? What else should I be worried about?
Del Armstrong
Internet : dela at ee.rochester.edu
UUCP : ...allegra!rochester!ur-valhalla!dela
Twisted pair: (716) 275-5342
Last resort : Hopeman 407
Electrical Engineering
University of Rochester
Rochester, N.Y. 14627
------------------------------
Date: Tue, 28 Jun 88 11:00:22 EDT
From: Harold Pomeranz <pomeranz at cs.swarthmore.edu>
Subject: cmdtool under OS 4.0?
Has anyone else running 4.0 noticed the distinct tendancy of cmdtool to
die? Most common times for the process to quit are when a window
obscuring the cmdtool window is resized or moved so that it no longer
obscures the cmdtool. Does anyone know what's going on? Does SUN know?
Hal Pomeranz
System Administrator
pomeranz at cs.swarthmore.edu
..rutgers!bpa!swatsun!pomeranz
------------------------------
Date: 28 Jun 88 19:59:55 GMT
From: ssc-vax!ray3rd at beaver.cs.washington.edu (Ray E Saddler III)
Subject: What does this error mean?
Hey, when I come into the office each morning, both of my 3/60's have this
message:
le0: trailer error
I've since noticed it appearing when the machine has been logged in and
sitting idle.
What does this error mean? Is it serious? Is it something I can ignore
without worry?
Ray E. Saddler III | Path: ..!ssc-vax!ray3rd
Boeing Aerospace | From: ray3rd at ssc-vax.UUCP
P.O. Box 3999 m.s. 3R-05 |
Seattle, Wa. 98124 USA | VoiceNet: (206) 657-2824
------------------------------
Date: Tue, 28 Jun 88 11:07:48 MDT
From: stc-auts!kak at stcvax.UUCP (Kris Kugel)
Subject: dumpregion -> postscript?
I need to make a user manual for a suntools application I've designed.
Because of the nature of the application, it's going to be a LOT easier to
explain if I can use pictures from the screen to show what I mean. I
think somebody or other here has ventura desktop publishing, and I figure
that through some method or other they can use postscript images if they
are on a dos formated floppy.
Ok, I've got dumpregion, and I've figured out how to move files onto a dos
formated disk. I've gotten several rasterfile-to-postscript programs, but
some blow up on the dumpregion output, and given the article (which?)
about printable postscript, and my general ignorance about postscript and
this whole procedure, I am not sure how to proceed.
So, my questions are:
1. Is anybody doing this kind of thing and can tell me how to do it?
2. If not, can anybody at least suggest what might work? Which versions
of what tools to use?
Thanks In Advance,
Kris A. Kugel
Storage Tek: { uunet!nbires ncar decvax boulder }!stcvax!stc-auts!kak
------------------------------
Date: 28 Jun 88 14:40:53 GMT
From: botticel at magnus.steinmetz.ge.com (David J. Botticello)
Subject: Tek 41xx emulator?
Here's a question for net-land.
Does anyone know of a program for the Sun 3 (Color) that provides
emulation of a Tekronics 4107 or higher terminal. I'm interested in either
PD or commercial.
Please E-mail if possible. Thanks.
Dave Botticello - GE CRD - phone(voice) 518-387-6065 Home 518-393-8671
E-mail - botticel at orion.steinmetz.ge.com -modern domain address or
uunet!steinmetz!orion!botticel -archiac source routing version or
botticel%orion.tcpip at ge-crd.arpa -thru the Arpanet or
botticel%orion at steinmetz.UUCP -You guessed it...thru UUCP
land - 1 River Road KWC-1604
Schenectady, N.Y. 12301
------------------------------
Date: Wed, 29 Jun 88 14:17:48 edt
From: Brett Harper <harperb at eecae.ee.msu.edu>
Subject: help with gdb and dbx using FORTRAN
I am trying to find a source level debugger to use with FORTRAN. (don't
ask me why!) But I've been having troubles with both dbx, and gdb. The
problem I am having with dbx is that the line that it says you are
currently at is off by as many lines as are INCLUDEded into the code
(either by the standard INCLUDE statement, or using the C preprocessor).
What this means is that the pointers that dbx is using are with reference
to the file before the include statement is processed.
The problem that I have with gdb is that when I type:
'gdb program' , where program is compiled with -g (above was also)
The message I get when gdb comes up is:
Reading symbol data from program...Invalid symbol data: Range type
spec r8;16;0; at symtab pos 9.
The problem seems to be in dbxread.c, although I have no idea what the
problem actually is, what am I doing wrong ?
(plea for help) If anybody knows anything about these two problems I
would greatly appreciate hearing from you. (Even if you have had the same
problem and still don't have it fixed !)
Sincerely, Brett Harper
Case center for CAE
MSU
------------------------------
Date: Fri, 24 Jun 88 11:59:03 edt
From: Tom <mnetor!syntron!jtsv16!tom at uunet.uu.net>
Subject: Need information regarding optical disk.
Can anyone provide me with information on "canned" optical disk
sub-systems to run on the sun 3/XX ? I have information on all kinds of
optical disk sub-systems, but they all seem to be geared towards the AT
bus, and the drivers tend to be for DOS rather than 4.2BSD. I would
prefer to stick with the 5 1/4" devices, the most interesting so far is
the Maxtor, but they don't have a unix driver for the controller board.
Tom Ott
JTS Computer Systems Ltd. UUCP: {geac|utzoo|utgpu}!syntron!jtsv16!tom
Toronto +1 416 665 8910 or: suncan!jtsv16!tom
------------------------------
End of SUN-Spots Digest
***********************
More information about the Comp.sys.sun
mailing list