Sun-Spots-Request@RICE.EDU (William LeFebvre) (08/18/88)
SUN-SPOTS DIGEST Tuesday, 16 August 1988 Volume 6 : Issue 188
Today's Topics:
Re: Microsoft Windows on 386i
Re: Xylogics 753 timing
Sun at SIGGRAPH '88
More places that mail RFCs
are y'all quite bored of this "set prompt" business yet?
SunOS4.0EXPORT & -Bstatic
gprof bombs on Sun-4/SunOS4.0
"Permanent" screenblank on 3/60's
Booting problems when netmask set
Problems with "yppasswd"
Ultrix YP problem (mail.aliases)
More problems with TOPS on a SUN
Selection service question
UPS and surge protector recommendations sought
Garnering info on 386i
Programming my Sun keyboard?
Send contributions to: sun-spots@rice.edu
Send subscription add/delete requests to: sun-spots-request@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@rice.edu" or mail the word "help" to the same address
for more information.
----------------------------------------------------------------------
Date: 10 Aug 1988 23:32-PDT
From: Robert Ling <rling@june.cs.washington.edu>
Subject: Re: Microsoft Windows on 386i
As explained to me by an engineer at SUN, the 386i can run Microsoft
Windows in a "DOS window". It does not require any additional video card.
Although a DOS window can be sized, it does not have any effect with
normal DOS applications; i.e. you don't get more lines. Microsoft Windows
is the exception to this; you can enlarge the DOS window while running
Windows and be able to use the full DOS window.
It should be noted that with the EGA card for the 386i, you can have a
maximum of 4 DOS windows at the same time. There is no such limitation
with the CGA/Hercules emulation that's available with a stock 386i.
- Robert Ling <rling@june.cs.washington.edu>
------------------------------
Date: Thu, 11 Aug 88 14:44:19 PDT
From: speck@vlsi.caltech.edu (Don Speck)
Subject: Re: Xylogics 753 timing
Doing sequential reads and writes on the raw disk device is very different
from what the filesystem will do. Sequential I/O will involve a complete
revolution of the disk between transfers (unless the controller has
readahead). The filesystem doesn't allocate sequentially, it tries to
pick rotationally optimum blocks, as determined by the "rotdelay"
parameter.
Notice that in most tests you got a little less than 60 transfers per
second. What you are really measuring is the rotation rate of the disk.
Any disk controller can finish up its DMA (or start DMA for writing) in
the time of one disk revolution. What you want to measure instead is how
long it takes to be ready for the next transfer. To do this, lseek
forward by N sectors between transfers, and increase N until you get a
sharp increase in throughput (usually 20 < N < 30, but on readahead
controllers it can be zero). Record *that* throughput, in transfers per
second.
My interpretation of your numbers:
I'm surprised that read throughput on the readahead controllers depends at
all on transfer size. This implies a high fixed overhead on each read,
about 5ms (comparable to a vax-780 with an Emulex controller). Either the
driver has to do an extraordinary amount of setup, or the controller has a
long "think" time. The amount of system cpu time used indicates how much
of each effect.
Writing is about the same speed on all controllers, indicating that the
caching controllers are write-through, very important for data integrity.
The slower controllers have to blow about two revolutions in each 63K
write to refill their buffers, and one to three revolutions in each 63K
read to empty their buffers. This is consistent with my observations that
the XY451 and SCSI controller DMA does not keep up with the transfer rate
of the drives.
Don Speck speck@vlsi.caltech.edu {amdahl,ames!elroy}!cit-vax!speck
------------------------------
Date: Thu, 11 Aug 88 08:17:33 EDT
From: Chuck Musciano <chuck@trantor.harris-atd.com>
Subject: Sun at SIGGRAPH '88
While at SIGGRAPH '88 in Atlanta last week, I visited the Sun booth and
saw the following interesting things:
There was a big picture on the side of the booth that looked like this:
4.3 BSD--------\
ATT System V---->----System V
Open Look------/
I didn't get a chance to query a sales rep about this, but it implied to
me that System V is the future Sun operating system. Needless to say, I
was alarmed. Any comment from Sun? Can anyone shed more light on this?
4.1 (and we're not even up on 4.0 yet...) will ship with a combined
SunView/X11/NeWS window system. The SunView is really SunView 2, which is
old SunView reimplemented on top of X. I was told that there would be
some modifications required to move existing SunView applications to
SunView 2, but that "well-behaved" applications should port easily.
Sun announced the 4/110TC (true color) 24-bit color system. This looked
really good. They also announced the 4/150CGX graphics system. The 150
is a 6 slot box.
Open Look (OpenLook? Openlook? OpEnLoOk?) was interesting, but seemed no
better or worse than SunView. One thing I did not like was that buttons
have single pixel shadows directly against the button image, while menu
labels (click and a menu appears) have a single pixel shadow !one! pixel
away from the image. I couldn't figure out what was going on until I got
very close (3 inches) from the monitor. I'm picking nits here, but having
two different gadgets look so similar is not a good idea.
To other SIGGRAPH attendees: did you get your red Sun bag? I've put mine
away in the hope that it will appreciate in value until SIGGRAPH '89.
Chuck Musciano
Advanced Technology Department
Harris Corporation
(407) 727-6131
ARPA: chuck@trantor.harris-atd.com
------------------------------
Date: Thu, 11 Aug 88 16:42:26 EDT
From: tower@bu-it.bu.edu (Leonard H. Tower Jr.)
Subject: More places that mail RFCs
There are three mail servers for rfc's I'm aware of. The concisest
summary I've seen is appended. There is also an anonymous uucp server at
ohio-state, ask postmaster@tut.cis.ohio-state.edu .
enjoy -len
__________
CSNET CIC Info Server (many things -- comp.sources.unix archives, CSNET info,
RFCs, etc.)
address: info@sh.cs.net
body: request: info
topic: help
request: end
BITNET NIC Server (BITNET information and RFCs)
address: nicserve@bitnic.bitnet
body: help
SRI-NIC File Server (official source for RFCs and Internet documents)
address: service@sri-nic.arpa
body: help
------------------------------
Date: Thu, 11 Aug 88 15:50:09 EDT
From: tower@bu-it.bu.edu (Leonard H. Tower Jr.)
Subject: are y'all quite bored of this "set prompt" business yet?
Use basename:
set host = `hostname`
set unixname = "`basename ${host} .bu.edu`"
if (`whoami` != "root") then
a setprompt 'set prompt = "${unixname}:`pwd` \\!>"'
else
a setprompt 'set prompt = "${unixname}:`pwd` \\!#"'
endif
setprompt
enjoy -len
[[ That's what I do. --wnl ]]
------------------------------
Date: Thu, 11 Aug 88 15:15:22 +0930
From: Kevin J. Maciunas <munnari!cs.flinders.oz.au!kevin@uunet.uu.net>
Subject: SunOS4.0EXPORT & -Bstatic
Fellow (non-US) Spots,
Sun have a real GOTCHA in 4.0-EXPORT. Attempts to link with -Bstatic (GNU
emacs 18.51 for example) FAIL with multiple definitions of things like
"_edata" and "_etext". The reason is simple!
Run nm on libc.a (the static version). You'll soon see des_crypt.o and
friends are **DYNAMIC**. ar(1) is useful at this point.
Hope this saves someone else considerable hassles...
Sun will be told when the local office call back.
[[ I linked undump with "-Bstatic"...no problem. Maybe it depends
somewhat on the program? --wnl ]]
Cheers,
/Kevin
Kevin J. Maciunas ACSnet: kevin@cs.flinders.oz
Discipline of Computer Science, Internet: kevin@cs.flinders.oz.au
Flinders University, UUCP: ..uunet!munnari!flinders!kevin
South Australia, 5042. 'phone: +61 08 275 2129
------------------------------
Date: Thu, 11 Aug 88 07:42:30 CDT
From: Scott Guthery <spar!ascway!guthery@decwrl.dec.com>
Subject: gprof bombs on Sun-4/SunOS4.0
When running gprof on SunOS 4.0 on a 4/110 I get:
[nllookup] binary search fails???
Segmentation fault (core dumped)
gmon.out is 76500 bytes. Any suggestions? Anybody know the format of the
gmon.out file?
Cheers, Scott
------------------------------
Date: Thu, 11 Aug 88 02:37:04 PDT
From: portal!jel@sun.com (John Little)
Subject: "Permanent" screenblank on 3/60's
We also had a problem with screenblank permanently blanking the display on
mono 3/60's. Typing on the keyboard, mouse motion, and L1A all failed to
do anything -- the only fix was turning the machine off and back on. This
struck me as a pretty extreme method of saving the phosphor :-).
Upon further investigation, I found that the diagnostic LED's were either
off or not moving. After talking to Sun, an engineer there confirmed that
this was a problem with 3/60's, and advised running lockscreen instead.
Admittedly, this is not a great solution, but we have had no further
problems of this type.
We are running 3.5.
John Little
jel@portal.com
------------------------------
Date: 11 Aug 88 23:40:31 GMT
From: pyramid!leadsv!laic!darin@decwrl.dec.com (Darin Johnson)
Subject: Booting problems when netmask set
Since our Suns had been getting "Setting subnet mask to 0xffff0000" when
booting (we just were attached to an ethernet gateway), I put in an
explicit "netmask 0xffffff00" on the ifconfig lines in rc.boot (as
suggested in a June message). However, after doing this, one of the Suns
failed to boot the next day, saying that the yp server was not responding.
On closer investigation, it turned out that the /usr partition didn't
mount, and mount was retrying forever. After setting the netmask to
0xffff0000, the machine booted fine.
Does anyone have any idea why this was happening? I would like to have
the netmask back at 0xffffff00.
Also, since the Sun administration manual is very vague on this point (I
am running 3.4) could anyone enlighten into all the mysteries of subnets,
netmasks, ifconfig, /etc/networks, etc...
(email would be nice...)
Darin Johnson (...pyramid.arpa!leadsv!laic!darin)
(...ucbvax!sun!sunncal!leadsv!laic!darin)
------------------------------
Date: Thu, 11 Aug 88 14:16:56 -0100
From: Jean Pierre Schiltz <mcvax!capsogeti.fr!schiltz@uunet.uu.net>
Subject: Problems with "yppasswd"
We run Unix 4.2 version 3.5, on a network with yellow pages passwords.
The command 'yppasswd' works fine on the clients, but on the server the
answer is 'you don't have a login name'.
What we have to do is type e.g. 'yppasswd schiltz', in which case it works
fine. Is this a result of a bad installation or a bug in 3.5 ?
Jean-Pierre Schiltz - Cap Sogeti Innovation - Centre de Recherche de Paris
118 rue de Tocqueville - 75017 Paris - FRANCE - +33 (1) 46 22 60 27
schiltz@crp.capsogeti.fr ...uunet!mcvax!inria!csinn!schiltz
------------------------------
Date: Thu 11 Aug 1988 15:03:23 EST
From: Jim Evins <evins@nrl-radar.arpa>
Subject: Ultrix YP problem (mail.aliases)
We are bringing up a uVAX 3600 running Ultrix, and are trying to make it
our master YP server. All the data bases seem to work fine from the uVAX,
except mail.aliases. I can't make use of these aliases from our client
machines (Suns), including the slave server (also a Sun). We are running
Ultrix V2.2, and SunOS 3.2. This database worked fine, when a Sun was our
master YP server.
The makefile as supplied by DEC used makedbm on /usr/lib/aliases, this
didn't work because aliases was not in a format compatible with it. The
makefile supplied by Sun used sendmail on /usr/lib/aliases to build the
data base. This approach didn't seem to work because the sendmail program
left off the special YP entries from the resulting database ( I am not
sure if this prevented the database from working, but it did prevent the
database from propagating to the slave server ). I then wrote a filter
using awk, which converted /usr/lib/aliases to a form usable by makedbm.
Now when I do a 'makedbm -u mail.aliases', I get a list identical to the
one created by Sun's sendmail, except the order of the aliases. However,
I still can't make use of these aliases on a client machine.
I tried explicitly setting the Y option in the sendmail.cf file to
mail.aliases, on a client ( I refroze the configuration file and started
up a new sendmail daemon. ). The aliases were still unknown.
-Jim Evins
<evins@nrl-radar.arpa>
------------------------------
Date: 11 Aug 88 22:06:14 GMT
From: harvard!ll-xn!adelie!munsell!uucp@gatech.edu
Subject: More problems with TOPS on a SUN
As a followup to a previous posting where I was unable to get
"connectivity" between a Macintosh and my SUN the problem was that the
version of TOPS that I was running under UNIX didn't support multiple
ethernet cards(2). I successfully got it running on a machine with a
single ethernet card.
Now for the next ripple. Sun sent me a beta release 2.1 version which was
supposed to have corrected the problem. After installing it on my machine
with the multiple ethernet cards(one for the main network, the other for
the subnet) I'm seeing ie1 no carrier. Any answers, ideas, suggestions
would be gratefully appreciated.
Mike Harris, Atex Inc., A Kodak Company 1-617-276-7392
32 Wiggins Ave., Bedford, MA
------------------------------
Date: Thu, 11 Aug 88 18:54:34 CDT
From: olson%antares@anl-mcs.arpa
Subject: Selection service question
I am attempting to use the selection service in SunView under release 3.4.
My goal is to allow the user to select a single character in a tty
subwindow, and have the program retrieve the entire line containing the
selected letter. I have attempted this by sending a query with the
attribute SELN_REQ_FAKE_LEVEL with value SELN_LEVEL_LINE, which the
documentation claims is "a text line bounded by newline characters,
including only the terminating newline (exactly what I want!).
The call looks like this:
static Seln_result read_proc();
Seln_holder holder;
char context = 0;
holder = seln_inquire(SELN_PRIMARY);
seln_query(&holder, read_proc, &context,
SELN_REQ_FAKE_LEVEL, SELN_LEVEL_LINE,
SELN_REQ_CONTENTS_ASCII, 0,
0);
where read_proc is the callback, which receives a reply of type
SELN_REQ_UNKNOWN. The code works correctly, in that it receives just the
selected area, when I eliminate the SELN_REQ_FAKE_LEVEL attribute.
Any ideas?
Bob Olson (olson@anl-mcs.arpa)
Argonne National Laboratory
------------------------------
Date: 11 Aug 88 15:26:45 GMT
From: oliveb!stpstn!aad@ames.arc.nasa.gov (Anthony A. Datri)
Subject: UPS and surge protector recommendations sought
I've got a Sun 3/180 and a Sun 3/280 that I really want to put on a UPS,
and I'd like some recommendations/discouragements as to vendors. We'd
also like to put a surge protector on the main feed onto our floor, as
opposed to buying a few dozen smaller ones and putting one on each
machine.
Any advice appreciated.
Anthony A. Datri,SysAdmin,StepstoneCorporation,stpstn!aad
------------------------------
Date: Thu, 11 Aug 88 09:37:14 CDT
From: paul gilna <pgil@sphinx.uchicago.edu>
Subject: Garnering info on 386i
This is one of those boring "please compress your n years of experience
into a four line response" requests, so if you can't hack it, `just say
n'.
I Should like to hear ( directly please, less net clutter) from anyone who
is currently using a SUN 386i as I am trying to assess the merits of
buying one as a small LAN server. Bear in mind when deciding to reply
that I shall probably pester you with inane queries but hey, what's a net
for?
Initially I am interested in how big an administrative chore running one
would appear to a UNIX novice.
cheers, paul gilna
UUCP: ...!ihnp4!gargoyle!sphinx!pgil, BITNET: pgil@sphinx.UChicago.edu
Analogue: (312) 702-6971 VOICE: Hullo, um, is Paul there?
------------------------------
Date: Thu, 11 Aug 88 14:55:21 EDT
From: "Craig M. Kanarick" <kanarick@prophet.bbn.com>
Subject: Programming my Sun keyboard?
Is there some way, to reprogram the function keys using the eeprom on my
Sun 3/160?
-- Craig
------------------------------
End of SUN-Spots Digest
***********************