Sun-Spots Digest, v6n230

William LeFebvre Sun-Spots-Request at Rice.edu
Wed Sep 21 00:40:04 AEST 1988


SUN-SPOTS DIGEST        Monday, 19 September 1988     Volume 6 : Issue 230

Today's Topics:
                       Re: Bigger icons for frames
                        Re: 4.0 NFS/yp bug (sun 4)
                     overriding shared libc routines
             rpc.yppasswdd problem: can't change passswords.
              Problems reading Multivolume IBM tapes on SUN4
                 Using hardwired ports for uucp and tip?
                         lots of users on a Sun?
                    4.0 versions of various utilities?
                         VM Statistics under 4.0?
                            Unipress Software?
                    Way to search Sun-Spots archives?
                           GRAPHICS on the SUN?
                            TCP/IP on Sun/Os?

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:    Thu, 15 Sep 88 09:16:21 EDT
From:    jwm at stdc.jhuapl.edu (Jim Meritt)
Subject: Re: Bigger icons for frames

> I was trying to assign bigger icons to frames, like two 64 x 64 icons
> joined together....

The pic2icon.c program I posted a while back makes any size rasterfile
into an icon the same size (little -> REAL BIG).  These icons already have
the header set up for the right size, and the resulting icons go nicely
with the -WI flag on opening windows.  You have to do the usual
corrections for size in programs....

Jim Meritt

(I have icon2pic that goes the other way, but it is definitely no big deal...)

------------------------------

Date:    Wed, 14 Sep 88 23:44:58 PDT
From:    Paul O'Neill <pvo3366 at neptune.oce.orst.edu>
Subject: Re: 4.0 NFS/yp bug (sun 4)
Reference: v6n228

trudel at caip.rutgers.edu (Jonathan D. Trudel) writes:

>Here's an interesting bug...We just got a Sun 4/110 with the 4.0 software.
>We have it running as a yellow pages client of a 3/160 running 3.2.  When
>we have it in this configuration, it cannot do NFS mounting of any
>machine, even itself.  I have found that if I turn off yp service, I can
>do the mounting.

This appears to be a 4.0 ypclient on a 3.x ypserver problem that is not
architecture-dependent.  We have the same problems w/ a 4/110 and a 3/60
running 4.0 trying to get ypservice from a 3/280 running 3.4.

When the 4.0 client executes ypbind in rc.local it kills the portmapper.
The following patch to rc.local works for us:
__________

if [ -f /usr/etc/ypbind ]; then
      if [ -f /etc/security/passwd.adjunct ]; then
              ypbind -s;      (echo -n ' ypbind')     >/dev/console
      else
              ypbind;         (echo -n ' ypbind')     >/dev/console
      fi
fi

> ###############################################
> # extra stuff to start ypbind ok
>       portmap;                (echo -n ' portmap')    >/dev/console
>       /usr/etc/yp/ypset 128.193.64.1;  (echo -n ' ypset') > /dev/console
> ################################################
if [ -f /usr/etc/keyserv ]; then
      keyserv;                (echo -n ' keyserv')    >/dev/console
fi
__________

The extra portmap restarts the portmapper (this doesn't seem to be needed
every time.)  The ypset only works with the hard-wired internet address of
the server.  You may also need to hard-wire your broadcast and netmask,
i.e.

< ifconfig ie0 netmask +
---
> ifconfig ie0 netmask 0xfffff800 broadcast 128.193.64.0 

If you're starting ypbind manually, then you can manually do the portmap
and ypset.  If you can get an answer to ypwhich, you're set!

I'd appreciate hearing if this works for other folks, or if anyone can
shed some more light on this problem.

Paul O'Neill                 pvo at oce.orst.edu
Coastal Imaging Lab
OSU--Oceanography
Corvallis, OR  97331         503-754-3251

------------------------------

Date:    Sat, 27 Aug 88 19:05:45 EDT
From:    Rayan Zachariassen <rayan at ai.toronto.edu>
Subject: overriding shared libc routines

The approach given by Charles Hedrick is quite workable, but we weren't
happy with doubling the overhead involved in starting up every shared
library program on the system.  We want our local resolver routines to
override Sun's /etc/hosts and YP approach, and we wanted to use
getnetgroups() without having to run YP.  Only a fraction of programs on
the system are affected by such things, and we'd much prefer just touching
those (and affecting their startup overhead) rather than indiscriminately
affecting every program.

Our first thought was to patch the shared object library dependency stored
within each binary.  We found it much easier to patch the startup code
(ld.so) instead since it is easily located.  Here is the recipe and code
in detail:

[[ The shar file is only 6390 bytes and has been placed in the archives
under "sun-source" as "sofix.shar".  It 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 at rice.edu".  --wnl ]]

------------------------------

Date:    14 Sep 88 14:10 -0700
From:    Steve Cumming <stevec%lccr.sfu.cdn at relay.ubc.ca>
Subject: rpc.yppasswdd problem: can't change passswords.

This is not actually a bug, just something easy to overlook. It <may> even
be mentioned in the manual, somewhere.

But <I> didn't see it - perhaps this will save someone some time.

It turns out that /etc and /etc/yp/ have to be in the same file system.
When rpc.yppasswdd gets a request from a client to change a password, it
builds a temporary password file in /etc/ptmp. This is not configurable -
it's hard coded into the source. The very last thing it does before
sending back a reply to the client is a 

	rename(/etc/ptmp, foo)

where "foo" is tha domain specific password file specified on the command
line. (e.g. in /etc/rc)

We had symlinked /etc/yp to /usr/etc/yp. This caused rename(2) to fail,
and an error code to be sent to the rpc client.  Not a particularly
informative one, either. (Password not changed)

The upshot was that none of our users were able to change passwords for a
while.

A few remarks to sun:

1) yppasswdd has good error messages: to stderr, which is invisible when
   the process daemonizes itself. 

2) the code distributed in 3.4 src does not compile.  The routine
   xdr_yppasswdd, mentioned in /usr/include/rpcsvc/yppasswdd.h as a int (),
   does not seem to exist anywhere (that i could find)

Steve Cumming
School of CS
Simon Fraser U, CANADA

stevec at lccr.cs.sfu.ca
ubc-cs!fornax!stevec

------------------------------

Date:    Wed, 14 Sep 88 15:53:12 CDT
From:    texsun!edsr!neptune!jcn at sun.com (Jim Niemann)
Subject: Problems reading Multivolume IBM tapes on SUN4

I created a no label (NL) 2 tape volume data set on an IBM mainframe with
OPTCD=Q (convert to ascii).  LRECL=173, BLKSIZE=32351.  The first tape had
5188 blocks (and hit end of tape), the second contained 1221 tape blocks.

Problem on a SUN4/280 with 6250 BPI tape drive:

When I use dd on the 1st of the NL tapes: 

dd if=/dev/rmt8 of=/dev/null bs=32351, I get
5187+0 (5187 blocks read, with 0 short blocks)
I missed a block.  There really are 5188 blocks on the tape.
No I/O errors occurred.

But if I do:
dd if=/dev/rmt8 of=/dev/null bs=65536, I get
0+5188 (5188 blocks were read, each of 32351 bytes)
This is the correct answer.

On the 2nd tape, where the data set didn't reach the end of tape,
a bs of 32351 or 65536 gets me the same answer of 1221 blocks
(correctly).

This is not a dd problem, because a sample test program, when doing reads
for 32351 bytes, only reads 5187 blocks.  But when I change it to read for
65536 bytes, it correctly reads 5188 blocks.

Could this be a problem with the XT device driver on a SUN4?  Does anyone
else have any experience with processing multivolume mainframe tape files
on a SUN?

Jimmy Niemann
EDS Research
214 661-6052
{uunet,sun!texsun}!edsr!jcn

------------------------------

Date:    Wed, 14 Sep 88 12:58:35 EDT
From:    kodak!cygnus!dennett at cs.rochester.edu (Charlie Dennett)
Subject: Using hardwired ports for uucp and tip?

(This might not strictly be a Sun subject, but I thought I'd start here.)

I have a Sun 3/260 running SunOS3.5.  I am trying to get both uucp and tip
to work over a hardwired line rather than a modem.  I almost have it
working 100%.  The hardwired port is actually an ALM-2.  This is connected
to a Sytek broadband system we have here.  The port is running at 9600
baud.  This is how it behaves:

uucp - I will poll the remote system by manually running uucico with debug
mode x5.  The connection is made and the two systems start talking to each
other.  My system starts sending the first file.  A whole bunch of bytes
go out the port but then the flow stops.  After many seconds, the terminal
running uucico shows "alarm 1" and then another bunch of bytes flow out.
This sequence of flow, alarm, flow continues until the file eventually
gets to the remote system.  On the other hand, if the remote system has
something to send to me, it comes across in one continuous burst.  The
broadband modem connected to the ALM-2 port has 3 choices for flow control
- none, xon/xoff, and EIA.  The only one that works is EIA (uses the
rs-232 control lines).  With the other two the file never successfully
gets to the other system.

tip - The flow control on the broadband modem must be set to xon/xoff and
the tip internal variable tandem must be true for tip to work.  Anything
else drops characters.  I tried to transfer a file using tip's ~p command.
The file went flying across.  When tip reached the end of the file, the
terminal where I was running tip started beeping at me and the system was
so busy that I had to halt it and reboot.  When I logged onto the remote
system, I discovered the file that I sent had not been received
completely.  About 1/2 of the character had been dropped.  This was true
for all combinations on tip's tandem and broadband's flow control.
Running the remote session through tip seems to work fine.

The ports seem to work fine for both outgoing tip and uucp connections and
incoming regular people sessions.  For tip and uucp I made several new
/dev entries by adding 128 to the minor device numbers just like the
manual says.

If anyone out there has done anything like this, I would appreciate
hearing from you.  As far as I can tell, I've set up the L-devices, L.sys,
and /etc/remote files just like the manuals say.  Maybe there is some
undocumented feature.  Of course the whole problem could be that the
broadband modems EIA flow control does not work as claimed.

Thanks for any and all help.

Charlie Dennett   Eastman Kodak Co.   Information Services  KAD   Rochester, NY
UUCP: ...rutgers!rochester!kodak!cygnus!dennett

------------------------------

Date:    Wed, 14 Sep 88 12:20:48 PDT
From:    weiser.pa at xerox.com
Subject: lots of users on a Sun?

Mark Horton writes: "However, I just heard a rumor that Suns have problems
with lots of users.  I also recalled a rumor that there's a 3 bit register
somewhere in the hardware, and that if the load average goes above 8 the
performance suffers."

I am one of the promulgators of this rumor, so let me try to give a
'final' word.  

1. Yes, there are only 8 hardware TLB contexts (16 on a sun-4) and going
beyond this requires software swapping of contexts.  

2. But no, there is usually NO PROBLEM AT ALL just from a load which is
higher than 8 (or 16).   Special circumstances are needed to generate the
problem.  I have never seen it in an actually application mix, although I
can generate it with 8 (or 16) forks of a 5 line C program.  The key
factor which can generate degradation is enough processes which sleep
(e.g. via select, not sleep) for only a few tens or hundreds of
milliseconds.  Something about this drives the software swapping code
bananas.

3. And, even with lots and lots of users, your load is not likely to go
above 8 (or 16).  On our Vax 8650 (Sun-3/260 speeds) back in my Maryland
days, if the load got that high I just stopped working.  And this was a
big timesharing machine.

4. Barry Shein at Boston uses Suns as big student timesharing machines,
and reports no problems.

-mark

------------------------------

Date:    Wed, 14 Sep 88 17:18:31 MDT
From:    sitongia at hao.ucar.edu (Leonard Sitongia)
Subject: 4.0 versions of various utilities?

I am looking for patches/upgrades to the following packages for SunOS 4.0
on a Sun-4: {mon,top,zap,assassin,tap,tcopy}.  Please send e-mail if you
have this.

Thanks,
-Leonard E. Sitongia    System Programmer
			(303) 497-1509
USPS Mail:              High Altitude Observatory
			P.O. Box 3000
			Boulder CO  80307
Internet:               sitongia at hao.ucar.edu
SPAN:			NSFGW::"hao.ucar.edu!sitongia"	[NSFGW=9580]

[[ Top version 2.5 now has support for SunOS 4.0.  It is available through
the Rice archive server under "public" as "top.shar.01" through
"top.shar.04".  --wnl ]]

------------------------------

Date:    14 Sep 88 15:53:18 GMT
From:    unido!tb at bilbo.irb (Torsten Beyer)
Subject: VM Statistics under 4.0?

Hi SUN folks,

We have just installed our first 4.0 Machine here and encountered several
(admittedly minor) problems. Where have all those virtual memory
statistics (like in <sys/vmmeter.h> et.al) gone ? We'd be interested in
obtaining information about virtual memory utilisation under 4.0 therefor
we NEED this data. Any 4.0 guru out there to help us ?

	thanx in advance
	-Torsten

Torsten Beyer                               e-mail : tb at unido.uucp
University of Dortmund                               ..uunet!unido!tb
IRB                                                  tb at unido.bitnet
P.O.Box 500500
D-4600 Dortmund 50                          voice  : +49 231 7552422
West-Germany

[[ Id' like to find out, too, so I can put the capability back into "top".
--wnl ]]

------------------------------

Date:    14 Sep 88 21:07:07 GMT
From:    felix!arcturus!holly at hplabs.hp.com (Holly Guthrey)
Subject: Unipress Software?

Does anyone have any experience, positive or negative, using spreadsheets
or project management software packages on SUN workstations ?  I am
looking into Unipress Software's Master Plan package for project
management and Q-Calc for spreadsheeting and have also seen offerings from
Quadratron (Q-Plan) and Access (20/20).  I have been using SMART on an
IBM-PC and am looking for something with similar amount of flexibility and
features.  Thanks in advance.                     

Holly Guthrey
Rockwell International
Send replies to: ...!sun!sunkist!arcturus!holly

------------------------------

Date:    14 Sep 88 11:23:53 PDT (Wednesday)
From:    Hamilton.OsbuSouth at xerox.com
Subject: Way to search Sun-Spots archives?

There's a lot of great info coming over this distribution, but I have no
immediate need for 99% of it.  I know I can request back issues from the
Archive Server, but more important, is there some way to do keyword or
other searches over the collected Tables of Contents?

Thanks,

--Bruce
CSNet: Hamilton.osbuSouth at Xerox.COM
UUCP: xerox.com!hamilton.OSBUSouth
213/333-8075

[[ There is a file that is kept more-or-less up to date containing the
issue number and subject line for every article in volume 6.  There is a
similar file for volume 5.  Both files are stored in the archives under
"sun-spots" and are called "v5.index" and "v6.index".  A sample line would
be:
v5n17:   'textedit' tool? (reply)
The subject line reads "'textedit' tool?", the message is contained in
v5n17, and this particular message is a reply.  The file is sorted
alphabetically by the subject field.  Once you have this file, you can do
your own searches over the subject fields.  I realize that this is not a
complete or thorough solution, but I believe that it is a good compromise
(especially given that I make sure a message's subject field is
representative of the message's contents).  --wnl ]]

------------------------------

Date:    14 Sep 88 22:52:17 GMT
From:    ut-emx!jezebel at cs.utexas.edu (Jim Rull)
Subject: GRAPHICS on the SUN?

Netters,

I am doing a survey on the various ways one can generate graphics on the
sun.  Any users of movie.byu, wavefront and NCAR graphics packages ?  What
do you think of these ?

Which, in your opinion, is the most commonly used graphics package on the
Sun for general 2-D graphics ?  What about 3-D graphics ?  Or is there no
such thing ?

If you were to choose graphics packages to do general purpose graphics,
(like contours, wireframes, shaded plots) which packages would you choose
and why ?

Jim Rull

------------------------------

Date:    14 Sep 88 16:01:06 PDT (Wednesday)
From:    Kang.ESAE at xerox.com
Subject: TCP/IP on Sun/Os?

Can anyone tell me where I could look into to find out more about how
TCP/IP functions and is utilized on the Sun's version of Unix.  Other
questions are:

	What is the relationship between NFS and TCP/IP?
	Is Sun's version of TCP/IP different from other implementations?

	What commands/system calls are available to utilize TCP/IP
	capabilities on the Sun?

	Where can I obtain documentation on the above mentioned items
	and please forward me the address/phone numbers.

Thanks a lot in advance,
Steve

p.s. Apologies to Xerox DL members for duplicate message...Didn't think.

------------------------------

End of SUN-Spots Digest
***********************



More information about the Comp.sys.sun mailing list