Sun-Spots-Request@Rice.edu (William LeFebvre) (09/26/88)
SUN-SPOTS DIGEST Sunday, 25 September 1988 Volume 6 : Issue 236
Today's Topics:
Re: Can someone give me a hand with color canvases
Re: Music font for Sun
a slow Sun network
Problems with strtod in 4.0
NGROUPS problems on SunOS 4.0
csh '*' expansion is not consistent (SunOS 4.0 on Sun3)
NFS cache bug - reported to Sun
Problems with VME address spaces
Mail is lost if / fills up on SunOS 4.0 clients
wiring questions
Getting a dead 386i serviced?
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: Tue, 20 Sep 88 18:18:29 CDT
From: vixen!ronbo@cs.utexas.edu (Ron Hitchens)
Subject: Re: Can someone give me a hand with color canvases
I said:
> So far as I know, this rule was undocumented until the 4.0 SunView manual.
Several people here and in mail to me said (quoting Mike Khaw):
> Actually, it's documented in my 3.2 SunView Programmer's Guide (Revision A
> of 15 October 1986), Chapter 5 - Canvases, Section 5.8 Color in Canvases,
> subsection "Color in Retained Canvases", page 69
Sorry about that. I thought my trusty Sunview manual was a 3.2 version,
but I looked in the back at the revision info and it turns out I've been
using a 3.0 version all this time. I probably imprinted the idea that I
had a 3.2 manual because the first one I used was the early 3.0BETA draft,
and the one I have now is the "new one".
BTW, I think Sun's manuals are generally pretty good, considering the
volume of stuff they have to document and the rapidity with which things
change. Keep up the good work guys.
Ron Hitchens ronbo@vixen.uucp hitchens@cs.utexas.edu
------------------------------
Date: Wed, 21 Sep 88 07:50:44 EDT
From: gfr%wolfgang@gateway.mitre.org (Glenn Roberts)
Subject: Re: Music font for Sun
Reference: v6n229
> Can anyone help me find a 'music' (i.e. notes, rests, clefs) screen font
> for a Sun? Any pointer will be greatly appreciated.
The 1987 Sun Users Group tape contains a program to convert Macintosh
fonts to Sun fonts (look in sunview/suntroff/LaserWriterFonts on the
tape). You might have better luck finding a music font for the Mac, then
converting it.
- Glenn Roberts, MITRE Corp.
gfr%wolfgang@gateway.mitre.org
------------------------------
Date: Wed, 21 Sep 88 05:17 EDT
From: "SDRRTR::PSI%PRSRTR::PSI%SCRVX2::BLUE::IN%\"'m_mailnow::m_sdr::davis'@scr-gateway\""@sdr.slb.com
Subject: a slow Sun network
I'm looking for clues or comments on the generally slow performance of our
Sun network.
We currently have 15 Sun 3's (50's, 110's, 140, 160, 180, 260) sitting on
a broadband Ethernet along with two VAXen and a number of microVAXen.
There is currently a single file server (a 3/180) though we're upgrading
to a 4/260 second server this week (along with another 7 seats).
Performance is, bluntly, atrocious and subjectively appears to have always
been that way (at least for the past 9 months or so). We recently added 4
new 3/50's, and that may have had some impact on the network performance,
but netstat-ing does not show an overly high collision rate - in fact, to
my untrained eye, the ethernet looks fins. The sort of features we're
seeing are phenomenally slow SunView operations (my 3/50 takes more than a
second to focus input on a window) and slow file access (Emacs [GNU, of
course] takes a *long* time (2-3 seconds) to display a newly visited
file).
We are considering adding an ethernet bridge to help things along once the
second file server is in, so that we then have `two' ethernets with one
server and about 11 seats each, but its not clear if this kind of problem
is truly a function of network loading at all. Someone here hooked up his
home Sun last weekend, with a local disk, and found the speed *phenomenal*
in comparison to ours - thats our only lead. Its worth noting that X
(10.4) is considerably faster than SunView (but until we get X11R2, we
have the problem of vanishing windows).
Any comments or advice would be appreciated.
Paul Davis
Systems Engineer (yuk!)
------------------------------
Date: Tue, 20 Sep 88 18:57:38 CDT
From: strong%stetsun@mcc.com (Michael Strong)
Subject: Problems with strtod in 4.0
strtod fails every time I call it if I compile my program normally;
however, if I compile it using the -Bstatic option, it seems to work.
This was a big disappointment since this function has been broken for
several releases now. Its failure has been intermittent but demonstrable
at least as far back as SunOS 3.2, I believe.
Sun thought they had this fixed in 4.0 but... In all fairness it does
appear to be a linker problem rather than a strtod problem, but the net
affect is the same.
Over the years I have had to fall back on sprintf often. That has some
disadvantages and strtod is supposedly going to be an ANSI standard
routine. It would be nice if it worked.
I know -- the answer is, all bets are off until 4.1 for dynamic linking
wierdnesses, right?
MJS
P.S. Have your Sun 4/110's been getting spurious Watchdog resets?
Peace.
------------------------------
Date: Wed, 21 Sep 88 11:30:30 EST
From: trinkle@purdue.edu
Subject: NGROUPS problems on SunOS 4.0
Sun decided to increase NGROUPS from 8 to 16 with SunOS 4.0. Locally,
being in more than 8 groups works fine. However, NFS access to an NFS
server that is not running SunOS 4.0 fails with an Authentication error.
I could understand Sun saying to hell with other vendors (they usually
do), but the problem is this also breaks NFS with other Suns running SunOS
3.x.
Daniel Trinkle trinkle@cs.purdue.edu ARPA
Department of Computer Sciences trinkle%purdue.edu@relay.cs.net CSNET
Purdue University {ucbvax,decvax}!purdue!trinkle UUCP
West Lafayette, IN 47907 (317) 494-7844 PHONE
------------------------------
Date: Wed, 21 Sep 88 11:16:44 EST
From: trinkle@purdue.edu
Subject: csh '*' expansion is not consistent (SunOS 4.0 on Sun3)
The following typescript shows what I consider to be a bug. The files
referenced are really symbolic links to (in the context of the server
where the script was run) a non-existent file. The echo command should
show the link path names in all cases and ls -l should list these files as
being symbolic links regardless of the value of the link.
Note that everything works as expected when there is a final '*' on the
file name argument. Everything works as expected with the Bourne shell,
so this is only a csh problem.
Script started on Wed Sep 21 10:39:30 1988
roland 1: echo /export/root/*/etc/termcap
echo: No match.
roland 2: echo /export/root/*/etc/termcap*
/export/root/aristotle/etc/termcap /export/root/athena/etc/termcap /export/root/homer/etc/termcap /export/root/howell/etc/termcap /export/root/icarus/etc/termcap /export/root/logos/etc/termcap /export/root/mycroft/etc/termcap /export/root/odysseus/etc/ter
mcap /export/root/pan/etc/termcap /export/root/vision/etc/termcap
roland 3: ls -l /export/root/*/etc/termcap
No match.
roland 4: ls -l /export/root/*/etc/termcap*
lrwxrwxrwx 1 root 24 Jul 29 16:30 /export/root/aristotle/etc/termcap -> ../usr/share/lib/termcap
lrwxrwxrwx 1 root 24 Jul 29 16:41 /export/root/athena/etc/termcap -> ../usr/share/lib/termcap
lrwxrwxrwx 1 root 24 Aug 4 16:41 /export/root/homer/etc/termcap -> ../usr/share/lib/termcap
lrwxrwxrwx 1 root 24 Jul 29 16:25 /export/root/howell/etc/termcap -> ../usr/share/lib/termcap
lrwxrwxrwx 1 root 24 Jul 29 16:27 /export/root/icarus/etc/termcap -> ../usr/share/lib/termcap
lrwxrwxrwx 1 root 24 Aug 9 07:29 /export/root/logos/etc/termcap -> ../usr/share/lib/termcap
lrwxrwxrwx 1 root 24 Aug 9 07:26 /export/root/mycroft/etc/termcap -> ../usr/share/lib/termcap
lrwxrwxrwx 1 root 24 Jul 29 16:44 /export/root/odysseus/etc/termcap -> ../usr/share/lib/termcap
lrwxrwxrwx 1 root 24 Jul 29 16:38 /export/root/pan/etc/termcap -> ../usr/share/lib/termcap
lrwxrwxrwx 1 root 24 Jul 29 16:33 /export/root/vision/etc/termcap -> ../usr/share/lib/termcap
roland 5: ^D
script done on Wed Sep 21 10:40:21 1988
Daniel Trinkle trinkle@cs.purdue.edu ARPA
Department of Computer Sciences trinkle%purdue.edu@relay.cs.net CSNET
Purdue University {ucbvax,decvax}!purdue!trinkle UUCP
West Lafayette, IN 47907 (317) 494-7844 PHONE
------------------------------
Date: Wed, 21 Sep 88 11:41:54 +0930
From: Kevin J. Maciunas <munnari!cs.flinders.oz.au!kevin@uunet.uu.net>
Subject: NFS cache bug - reported to Sun
There is a really nasty bug in SunOS4.0's NFS cacheing code.
The situation:
(1) Local disk
(2) NFS mount a source file system. No exported root access (ie map to -2)
Copying a file from (2) to (1) with cp will complain "permission denied"
ON THE OUTPUT FILE (!!) and, even worse, will create an 8192byte file of
zeros (1 FS block). Furthermore, access by the owner of the file in (2)
can't get at the file properly for a while or until it is updated.
This would seem to be a very serious bug!
Here's a sample script:
[my home directory "~kevin" is NFS mounted]
Script started on Wed Sep 21 11:23:09 1988
bremerer 24> ls -ldg ./awm . ..
drwxrwxr-x 2 kevin staff 1536 Sep 21 11:20 .
drwxr-xr-x 20 kevin staff 2048 Sep 21 11:15 ..
-rwx------ 1 kevin staff 262144 Sep 21 11:20 ./awm
bremerer 25> file ./awm
./awm: mc68020 demand paged dynamically linked executable not stripped
bremerer 26> whoami
kevin
bremerer 27> su
Password:
bremerer 1# cp ./awm /tmp/foo
cp: /tmp/foo: Permission denied
bremerer 2# ls -l /tmp/foo
-rwx------ 1 root 8192 Sep 21 11:23 /tmp/foo
bremerer 3# file /tmp/foo
/tmp/foo: commands text
bremerer 4# file ./awm
./awm: commands text
bremerer 5# ^D
bremerer 28> file ./awm
./awm: commands text
bremerer 29> less ./awm
./awm: Read Error
bremerer 30> file ./awm
./awm: commands text
bremerer 31> touch ./awm
bremerer 32> file ./awm
./awm: mc68020 demand paged dynamically linked executable not stripped
bremerer 33> ^D
script done on Wed Sep 21 11:24:32 1988
...Hope this saves someone a disaster - I discovered it while writing a
script to make my local disk cache my NFS directory - surprise surprise
when your 30+MB directory fits on a 41MB partition with 30+MB free!!
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: Tue, 20 Sep 88 13:34:12 +0100
From: Scott Williamson <mcvax!memex.co.uk!scott@uunet.uu.net>
Subject: Problems with VME address spaces
We have a device which looks like a piece of memory in the vme24d16
address space of the VME-bus. It is usually accessed by mmap'ing it into a
process's virtual memory space.
On sun 3 machines running SunOS 3.4 or 3.5 we have had no problem in
writing to it using bcopy() or dereferencing integer pointers. We
understood that any 32 data bit accesses were being handled by the MMU and
split into 2 * 16 bit accesses.
However, on a Sun 4/260 running SunOS 4.0 a bus error is given when
bcopy() is attempted to the board, even when both addresses are page
aligned and the copy size is a multiple of the page size. Similarly
dereferencing an integer pointer in the mapped memory gives a bus error,
even when on a word boundary. Does anyone have any suggestion as to why it
is not working? Is it related to the cache on the Sun 4 processor board?
Or does the SunOS 4.0 memory management system not handle this situation?
Also, using the read() system call to read from a file or a raw disk
directly into the mmap'ed memory works fine on the Sun 3 but hangs the Sun
4 (the disk controller being SCSI on the Sun 3 and Xylogics SMD on the Sun
4). Presumably the DVMA system on the Sun 3 handles this ok. Is DVMA used
in Sun 4/260 device drivers or is direct DMA performed into the VME
memory?
Any comments on experiences related to the above would be appreciated.
Scott Williamson
scott@memex.co.uk
------------------------------
Date: Wed, 21 Sep 88 11:42:52 EST
From: trinkle@purdue.edu
Subject: Mail is lost if / fills up on SunOS 4.0 clients
If mail is received on diskless clients when the /var/spool file system is
full, the mail is "accepted" (sendmail does not get an error from
/bin/mail), but the user never gets it. With Sun OS 3.x, /usr/spool/mail
was in /private/usr/spool/mail, which was a 4.2 mounted (ND) file system.
/bin/mail recognized a write failure with full 4.2 type file systems. Now
that /var/spool/mail is in an NFS mounted file system, /bin/mail sees no
write errors with a full file system. Only close(2) returns an error on
full NFS file systems.
Unfortunately, now that the root partitions of clients share a common
/export/root file system (at least the way Sun does it by default), any
client can fill up the / file system for all clients. With the current
problem with /bin/mail, this means one runaway process on one client can
cause users to lose mail on all clients that share a common /export/root
file system.
Daniel Trinkle trinkle@cs.purdue.edu ARPA
Department of Computer Sciences trinkle%purdue.edu@relay.cs.net CSNET
Purdue University {ucbvax,decvax}!purdue!trinkle UUCP
West Lafayette, IN 47907 (317) 494-7844 PHONE
------------------------------
Date: Wed, 21 Sep 88 12:23:53 BST
From: Eric Ole Barber <mcvax!nw.stl.stc.co.uk!sizex@uunet.uu.net>
Subject: wiring questions
we have a piece of Japanese equipment supplied with a separate power cable
without a plug at the male end. The power wires are black and white, but
we can't see which is which at the molded plug at the female end. The
power sockets at the female end are labelled w and s. What do w and s
stand for, and are blac and white to be wired up to 'live' and 'neutral'?
Our current answers are weiss and schwartz (well-known English words used
in the export trade), and yes. Any other answers? I don't think 'live'
and 'neutral' would apply in the U.S. since both pins are 55V with
respect to ground, but do they apply in Germany (if it is weiss and
schwartz)? Thanks, Eric
------------------------------
Date: Wed, 21 Sep 88 11:44:43 MDT
From: dbd%benden@lanl.gov (Dan Davison)
Subject: Getting a dead 386i serviced?
There has been some discussion of the 386i in this list lately, and we
have been having such a bad time I thought I would mention it to everyone.
We bought a 386i-250, the faster one, with a 327 MB SCSI drive and a
single floppy. After a few (2 or 3, no more) weeks of use an attempt to
reboot the machine failed with the message that there was no sd device.
This is amusing since it had been booted off that device regularly. So we
go to boot off floppies, right? No floppies. We did get some 1/4" tapes,
which is less than useful since we did not get the expansion chassis and
tape drive.
Calling Sun (local TSEs and sales, and two separate calls to the hotline)
has resulted in a machine dead for the past 8 or 9 weeks. The local
hardware people know nothing about the machine; apparently the Right Coast
Sun people do not talk to the Left Coast Sun people.
When it worked it was a fine machine, but I cannot recommend anyone
getting one...unless you spring for the expansion chassis and tape drive.
I'm also impressed that a large site like ours (last count > 500 Suns are
around here) gets this kind of service. What do small sites do?
BTW, I have left out names because as far as I can tell all the people I
have spoken to locally and at CSD in California are frustrated by this
too.
dan davison/ theoretical biology / t-10 ms k710 / los alamos national lab
dd@lanl.gov / dd@lanl.UUCP / ...cmcl2!lanl!dd / dd%lanl.gov@CUNYVM.BITNET
------------------------------
End of SUN-Spots Digest
***********************