Sun-Spots-Request@RICE.EDU (William LeFebvre) (12/18/87)
SUN-SPOTS DIGEST Thursday, 17 December 1987 Volume 5 : Issue 72
Today's Topics:
Administrivia: year-end hiatus
Re: System with Ciprico controller crashing in Common Lisp (2)
Solution for inverted color suntools problem
Optical disk/driver/support
MacPaint to Rasterfile
FORTRAN Bug
Problems creating windows and loading fonts
Sendmail problem with leftover lock files
Redirecting printer messages with lpd
Sending mail between Suns and Symbolics Lisp Machines
Network configuration question
Partial Screen Capture Program for Suns?
How do I modify the background canvas on NeWS windows?
----------------------------------------------------------------------
Date: Thu, 17 Dec 87 17:35:54 CST
From: William LeFebvre <phil@Rice.edu>
Subject: Administrivia: year-end hiatus
I will be taking a two-week vacation, without my terminal, to visit
relatives for Christmas and New Years. This means that there will be no
Sun-Spots digests again until January 4, unless I manage to throw one
together tomorrow. I haven't had time to show Vicky all the neat things I
added to the making of a digest, and she is going to be gone for most of
that time anyway. I am already running about a week behind in getting
messages out. Thinking of the tremendous backlog that will greet me upon
my return makes me ill. So please, if you can hold off on submissions for
the digest until then, I would really appreciate it.
"Season's Greetings" (that's "Merry Christmas" to those of you who
celebrate it).
William LeFebvre
<phil@Rice.edu>
------------------------------
Date: Fri, 11 Dec 87 15:27:52 PST
From: umn-cs!meccts!srcsip!pavo!alarson@cs-gw.d.umn.edu (Aaron Larson)
Subject: Re: System with Ciprico controller crashing in Common Lisp (1)
For the past 4 months, we've been having the same behaviour as you
described. We have two sun 3/280's, one we built ourselves (has sun
standard cage and cards) and one we bought from Sun. Both of the 280's
have a Rimfire controller and a couple of CDC9772's. On the home built
one we do not have any Xylogics controlers, on the Sun purchased one we
have one Xylogics controller in addition to the Rimfire. We have been
having trouble with Franz and Kyoto common lisps crashing our systems, and
VADS Ada programs having mysterious run sometimes, not run the next time
bugs. We don't know if the two problems are related. We have also
noticed the crashing problem on a diskless 3/60 client.
We have contacted Sun and they don't think its their problem, further
stating that they have no swap/page related problems with 3.4. We have
tried disabling the cache on the Rimfire; no difference. Our next step
was to remove the Rimfire from our Sun purchased machine, and run a Sun
only system to see if we could duplicate the bug.
We are not certain of the cause, but the problem appears to be exacerbated
when we have a fairly large amout swap space (>100MB). The following
procedure appears to crash the server about 75% of the time:
1) Start as many common lisps running at the same time as possible.
2) Start the scheduler in all of them
3) kill all the lisp processes at once.
[Steps 1&2 can be interleaved]
I do this from within GNU emacs using a keyboard macro, and when one of
the lisp subprocesses complains about insufficient swap space, I exit
emacs. Usually 2 or 3 tries causes a crash.
I would be very interested in following this up with you, either via mail
or voice.
------------------------------
Date: Sat, 12 Dec 87 01:42:34 EST
From: hedrick@aramis.rutgers.edu (Charles Hedrick)
Subject: Re: System with Ciprico controller crashing in Common Lisp (2)
Aaron Larson reports crashes with Lisp and other large programs on Suns
using Ciprico controllers, just as we saw. We found the bug. In the
Ciprico device driver, rf.c, in rfstrategy, is the line
/* get the size of the transfer*/
sizeofxfer = btodb(bp->b_bcount);
This causes bytes to be converted to disk blocks by truncation, i.e.
rounding down. As far as we can tell, the only place the kernel does disk
I/O in units other than full disk blocks is in the swap code, when writing
out the pte's for the text segment. The above line will cause only even
multiples of 512 bytes to be written. The left-over bytes will be junk.
Next time that text segment is swapped in, the last group of pte's will be
bad, and there is a good chance that the system will crash. The fix is
/* get the size of the transfer*/
/* sizeofxfer = btodb(bp->b_bcount); */
sizeofxfer = howmany(bp->b_bcount, DEV_BSIZE);
This causes it to round up. We have given this fix to Ciprico. Our
system has not crashed since we installed the fix. This is a very serious
bug. I don't understand how any system with a Ciprico controller could be
working correctly, unless it has a very large amount of memory, or nobody
runs any large programs.
When we reported the problem, both Sun and Ciprico told us that they had
never heard of such a problem. (Did you report it to Ciprico? You only
mentioned reporting it to Sun.) Ciprico eventually decided it couldn't
possibly be their problem, and left it to us to solve. Generally one
expects vendors of 3rd-party controllers to pursue reports of failures
that might be due to their devices or device drivers. (In general Ciprico
has been very helpful. I am sure they honestly believed that this problem
couldn't be theirs.)
We now regard the Ciprico controller as a solid product. Our unexplained
non-ECC data errors also vanished in the process of investigating the
problem described above. Some kind reader pointed us to a way to install
new software on a machine with a Ciprico controller, without having to
move around drives or controllers. (I had indicated this as a problem in
my review.) The trick is to boot the machine as a diskless client, from
some other server. Then you can run rfutil to format and partition the
drive, and use dd to copy the minifs into the b partition. At that point
you can boot from the minifs and proceed with the installation. However
we'd still like to get the diagnostics working, or Sun is going to have
some problems with supporting disk drives connected to these controllers.
We're probably going to go ahead converting to Ciprico controllers on a
few machines that are badly I/O-bound, but not do all of our machines
until either Ciprico gives us all the standalone utilities or we find out
more about Sun's rumored fast disk controller. Would anybody be willing
to give detailed reports on experience with any of the other new native
VME disk controllers on Suns?
------------------------------
Date: Wed, 9 Dec 87 09:24:45 PST
From: texsun!sun!nemo@sun.com
Subject: Solution for inverted color suntools problem
From the man pages (notice I didn't say RTFM) on suntools: Old pre-3.2
applications run on the 8bit_color_only desktop will not appear because
they will be writing to the overlay plane. I.e., don't run old pre-3.2
applications on an 8bit_color_only desktop.
There is an application called the switcher that is used as an alternative
to adjacentscreens for getting between desktops on the Prism. Clicking
the switcher icon gets you to another desktop using some amusing video
wipe type animation. The switcher can also be used to simply set the
enable plane to 0 or 1 if the enable plane get out of wack. See the man
page switcher(1) for details.
------------------------------
Date: 11 Dec 87 02:27:36 GMT
From: vipin%samsun@sun.com (Vipin Samar)
Subject: Optical disk/driver/support
I had seen an earlier request seeking information on optical
disks/driver/support etc. When I was a grad student, I had put together an
Archival system based on the optical disks. It was an sort of a file
system, but basically geared towards archiving directories and files on
it. Many versions of the same file could be archived on it, and it was
quite easy for the user to retrieve any one of them. Based on some
statistical studies done, We had calculated that for a daily incremental
back up of a Unix system, where the initial size occupied 35% of the
optical disk, that disk will last for approximately 145 days. This was
done on a SUN3/50 running 3.2. The drive used was Alcatel Thompsons
"Gigadisc". This work was done at SUNY, StonyBrook in this summer.
vipin
Disclaimers notice: I dont speak for nobody; and nobody speaks for me.
vipin@sun.com || sun!vipin || (415) 691-6694 (o) || (408) 749-0259 (r)
------------------------------
Date: Tue, 8 Dec 87 21:36:15 PST
From: strout@lll-lcc.llnl.gov (Robert E Strout)
Subject: MacPaint to Rasterfile
I just passed up a few requests for a program to allow MacPaint files to
be displayed on the Sun. I was given a short program ( <100 lines C) that
converts a MacPaint file into a Sun Rasterfile, compliments of a fellow at
Rice. If anyone is interested in getting a copy of this program, send me
mail, and I will send you it.
-=-=-=-=-=-=-=-=-
Rob Strout II (resii) strout@lll-lcc.arpa
Lawrence Livermore National Laboratory ...{ihnp4,pyramid}!lll-lcc!strout
------------------------------
Date: Fri, 11 Dec 87 14:28:30 est
From: mlijews@nswc-wo.arpa
Subject: FORTRAN Bug
Here is a short FORTRAN program illustrating a bug in the sign function
which I haven't previously seen reported. It seems the sign function
doesn't like mixed arguments.
program test
implicit real*8(a-h,o-z)
x=4./3.
dd=1.
r=sign(dd,x)
p=sign(1.,x)
q=sign(1.d0,x)
write(6,*) x,dd,1.,r,p,q
stop
end
1.33333333333333 1.00000000000000 1.000000 1.00000000000000
7.8125018608262d-03 1.00000000000000
This was run under SunOS 3.3.
Michael Lijewski (mlijews@nswc-wo.arpa)
Code R44, Bldg.427
Applied Math Branch
NSWC
Silver Spring, MD 20903
------------------------------
Date: Fri, 11 Dec 87 11:17:57 EST
From: csrobe@icase.arpa (Charles S. Roberson)
Notice: icase.arpa is at 128.102.23.51. Update host tables.
Subject: Problems creating windows and loading fonts
I have a problem that I can't track down. I transfered a PD program from
a SUN 3 running Sun OS 3.2 to a SUN 3 running Sun OS 3.4. The program is
called (monthtool or suncal). Anyway, the window has a subwindow
(notesDsply) that is supposed to come up in the same font as the window
above it (notesPanel). The second call to window_create immediately
follows the first, but only the first one loads the new font. The second
one doesn't. I tried switching the order in which the windows were
created and it didn't help. Therefore I think it has something to do with
the fact that notesDsply is a TEXT SubWindow as opposed to a PANEL. I
also tried changing the WIN_FONT to TEXTSW_FONT but that didn't help.
Ok, here is the cincher -- I took the code, which I had transfered, back
to the original machine and it worked. Does anyone have an idea of what
is going on and how I can remedy it? The relevent sections of code
follow. Assume FONTDIR and NORMALFONT contain a valid pathname and
filename, respectively, since the font gets loaded in the first window.
----
strcpy(fontString, FONTDIR);
strcat(fontString, NORMALFONT);
notesFont = pf_open(fontString); /* Open regular sized font. */
...
notesPanel = window_create(frame, PANEL,
WIN_X, 0,
WIN_BELOW, datesDsply,
WIN_HEIGHT, DATES_HEIGHT - 65,
WIN_WIDTH, WINDOW_WIDTH,
WIN_FONT, notesFont,
PANEL_BACKGROUND_PROC, notesBackgroundHit,
PANEL_EVENT_PROC, notesPanelEvent,
0);
notesDsply = window_create(frame, TEXTSW,
WIN_X, 0,
WIN_BELOW, notesPanel,
WIN_HEIGHT, DATES_HEIGHT - 50,
WIN_WIDTH, WINDOW_WIDTH,
WIN_FONT, notesFont,
TEXTSW_IGNORE_LIMIT, TEXTSW_INFINITY,
0);
----
Many thanks.
-chip
Chip Roberson ARPANET: csrobe@icase.arpa
1105 London Company Way BITNET: $csrobe@wmmvs.bitnet
Williamsburg, VA 23185 UUCP: ...!uunet!pyrdc!gmu90x!wmcs!csrobe
------------------------------
Date: Fri, 11 Dec 87 12:49:14 PST
From: Rich Wales <wales@cs.ucla.edu>
Subject: Sendmail problem with leftover lock files
We here in the UCLA CS Department have been having a problem for several
months with Sendmail leaving mail queue files locked.
We have three Sun 3/280's running SunOS 3.4. We are *not* running Sun's
Sendmail; rather, we are running the Berkeley 5.58 Sendmail as adapted for
the Suns by Lawrence Berkeley Labs (LBL). Each Sun has its own mail queue
directory (i.e., we are not NFS'ing /usr/spool/mqueue).
The problem we see is that some files in an outgoing mail queue direc-
tory (/usr/spool/mqueue) get locked (with "lf..." and "xf..." files
present) and stay locked, even though the Sendmail daemon is clearly not
processing them at the time.
Removing the offending "lf..." and "xf..." files allows everything to go
through without difficulty.
In at least some cases, it appears that the files get locked right from
the start, as the mail gets initially submitted. Our users use at least
three different front-ends to compose and send mail (Berkeley Mail, MH,
and a UCLA-specific program); the problem seems to be plaguing users of
all three of these front-end systems. However, I'm not sure whether all
the abnormally locked files are locked right from the start (though I'm
going to start checking for this from now on).
There are never any core dumps around (neither in /usr/spool/mqueue, nor
in the directories of users whose mail files are found to be locked). The
Sendmail daemon spun off from /etc/rc ("/usr/lib/sendmail -bd -q1h") is
always still around.
Has anyone else ever seen this problem? Any ideas for a solution?
-- Rich Wales // UCLA Computer Science Department // +1 (213) 825-5683
3531 Boelter Hall // Los Angeles, California 90024-1596 // USA
wales@CS.UCLA.EDU ...!(ucbvax,rutgers)!ucla-cs!wales
"Could you please continue the petty bickering? I find it most intriguing."
------------------------------
Date: Fri, 11 Dec 87 15:42 PST
From: <JON@UCLASTRO.BITNET>
Subject: Redirecting printer messages with lpd
To all,
We have a apple Laserwriter as the default printer on our local Sun
network. Error messages that get returned from the printer are placed in
the error log file specified by lf= in /etc/printcap. However, I would
rather these messages either be sent directly to the user's terminal, or
if not logged on, mailed to the user who initiated the job. My conception
of how this would work would be to tell the lpd deamon to send the
messages to a filter instead of the file. However, from the man pages for
lpd, there does not seem to be a way of doing this. Am I mistaken (as
usual)? If not mistaken, how would I go about doing this? Presumably,
lpd would need to be modified. How can I get source for lpd? Or is there
a public domain version of lpd? System info: the print server is a Sun
running SunOS 3.2 (4.2 unix).
Thanks,
Jonathan Eisenhamer
UCLA Astronomy
jon@uclastro.bitnet
bonnie::jon (SPAN 5828)
(213) 206-8596
[[ As described, this is not possible with the standard version of lpd.
You would need to modify lpd, thus you would need the source to it. What
I did for the DVI driver for our Imagen printers was as follows: I
created my own shell file which acts as the DVI filter. This file runs
the DVI to imPress converter and redirects standard error to a temporary
file. If the converter did not write any output to the printer (which is
the ccase when it cannot find one of the fonts), it exits with a special
return code. The shell file checks for the return code and sends a
"diablo" document to the printer containing the lines written to standard
error. Thus the failure messages appear on the printer in lieu of the
queued document. Rather effective. --wnl ]]
------------------------------
Date: Sat, 12 Dec 87 11:29:35 PST
From: trwrb!smpvax1!sbb@ucbvax.berkeley.edu
Subject: Sending mail between Suns and Symbolics Lisp Machines
Recently I have been attempting to get sendmail to forward mail from our
Sun network to some of our Symbolics Lisp machines. With the verbose flag
turned on, I can see that we are successfully initiating a connection to
the Symbolics, but when we try to send the HELO SMTP line, things just
hang. We've heard some reports about sendmail only terminating the lines
that it sends out with a LF, instead of CRLF which apparently is what the
Symbolics expects. So I tried using the -ba and -bs flags to sendmail to
get it to CRLF terminate its lines, but with no luck.
Has anyone else attempted to do this? Is there some bit of magic that
isn't documented well (or at all) that we need to be using? Any and all
replies will be appreciated.
steve
Steve Byrne ...ucbvax!trwrb!smpvax1!sbb
Inference Corp. (213) 417-7997
5300 W. Century Blvd.
Los Angeles, CA 90045
------------------------------
Date: Fri, 11 Dec 87 18:22:29 EST
From: csed-1!csed-20!baldo@hc.dspo.gov (James Baldo)
Subject: Network configuration question
I would like to know if anyone could recommend a network configuration
with the following requirements and equipment:
1 sun3/180
1 sun3/280
20 sun3/50 70 mb disk
20 sun3/50 140 mb disk
20 sun3/50 diskless
The organization is a research group that spends 50% of its effort writing
reports and 50% building prototype software. The network will have a T1
link to a VAX running Wollongong TCP/IP which also is an ARPANET host.
Your comments on a network configuration for the above system would be
greatly appreciated.
Are current problem, based on the configuration we have is described
below:
We are using a ProNet-80 token ring backbone with the following
connections to the ring:
1 sun3/180 server
1 sun3/280 server
1 T1 link to the VAX
5 ethernet subnets
The problem is that we would like to place diskless nodes on our subnets.
Unfortunately the RARP from the diskless nodes will not pass through the
ProNet-80 gateways (they are not suppose to pass broadcast packets), same
problem with the ND boot. Does anyone have a suggestion besides placing a
server on each subnet?
Thanks in advance,
James Baldo Jr.
Please mail all responses to the following address
since I am not on the mailing list:
baldo!csed-1.uucp@hc.dspo.gov
------------------------------
Date: 11 Dec 87 01:40:50 EST (Fri)
From: keithw@lehi3b15.csee.lehigh.edu
Subject: Partial Screen Capture Program for Suns?
First, a thanks to the info-postscript people. In response to my
request for a public domain sun rasterfile to postscript conversion
program, I have obtained a filter (called suntops). If anyone needs it
an be FTPed from anonymous%dopey.csee.lehigh.edu@ibm1.cc.lehigh.edu (love
those addresses).
Next request, I am now looking for a programmable/parameter driven filter
that will allow sun rasterfile screen captures of a specified portion of
the screen. The standard sun screendump is just to darn big (1+ MB) and
the suntops conversion is even bigger (2+ MB). I would like to specify a
portion of the screen to capture and then convert (with my new suntops
filter) to postscript for printing. Anybody have such a parameter driven
sun screen capture program or code that is public domain?
Thanks in advance
Keith Werkman, CSEE Dept., Lehigh U., Bethlehem, PA 18015
================================================================
UKJWERK@VAX1.CC.LEHIGH.EDU
KJW1@LEHIGH.BITNET
ihnp4!c11ux!lehi3b15!keithw
(or maybe keithw%scarecrow.csee.lehigh.edu@ibm1.cc.lehgih.edu)
[[ Check out the program "dumpregion" in the Sun-Spots source archives.
It is called "sun-source/dumpregion.c" and can be retrieved via anonymous
FTP from the host "titan.rice.edu" or via the archive server. For more
information about the archive server, send a mail message containing the
word "help" to the address "archive-server@rice.edu". --wnl ]]
------------------------------
Date: 10 Dec 87 00:34:45 GMT
From: sundc!potomac!potomac.ads.com!jtn@uunet.uu.net (John T. Nelson)
Subject: How do I modify the background canvas on NeWS windows?
Are there tools available that will let me modify the background "desktop"
in Sun NeWS windows? Is there anything like a MacPaint or MacDraw colour
package that can do the job?
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
------------------------------
End of SUN-Spots Digest
***********************