Sun-Spots-Request@RICE.EDU (William LeFebvre) (06/29/88)
SUN-SPOTS DIGEST Tuesday, 28 June 1988 Volume 6 : Issue 123
Today's Topics:
Administrivia: archive server news
Re: Need advice on distributed networks and database systems
Re: Sun losing track of console
Monitor-problem Verified
gethostbyname in 4.0
Problems Installing 4.0
SunOS 3.5 Blues
Error messages not appearing in console window on client
Sun NeWS - command needed to free space allocated by the package
Disk configuration question (Xylogics with No. Telecom on Sun 3)
ndbm routines hang in lseek() or read() -- Any ideas??
Big monitors, and monitors in parallel?
rlogin/vi problems?
how do we use a 3/260 CXP?
Trailblazer modems and Sun386?
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, 28 Jun 88 12:57:10 CDT
From: William LeFebvre <phil@Rice.edu>
Subject: Administrivia: archive server news
I have installed a new version of the archive server software and it seems
to be running rahter smoothly. Among the other things that this version
fixes is the infamous "`system' in the address's comment field" problem.
The server now strips off any comment in the "From" address before
checking for known userids such as "daemon" and "system". Some of you
were having problems because your From address included a string like
"(systems division)". This should now be fixed. I have also added
mechanisms to make it easier to track down the real address for messages
that are returned to me with bad paths. This should make it easier for me
to track down problems some of you have had in getting the server to
respond. Enjoy!
William LeFebvre
Sun-Spots moderator
<phil@Rice.edu>
------------------------------
Date: Thu, 16 Jun 88 17:47:45 EDT
From: bob@allosaur.cis.ohio-state.edu (Bob Sutterfield)
Subject: Re: Need advice on distributed networks and database systems
Reference: v6n107
berlin@bu-albert.bu.edu (David Fickes) writes:
>...It seems slightly silly to be shipping all of that data across the
>thin-net but it might make sense if we upgrade that one segment to
>thicknet and if the server is also busy with other tasks...
Ummm, 10Mb/s Ethernet running on a thin, black wire is still 10Mb/s
Ethernet, just like 10Mb/s Ethernet running on a thick, yellow wire. The
bulk of the wire doesn't mean that more data can flow through it, like if
you were buying a bigger garden hose down at True Value. The main
difference between the two is how far you can stretch the cable, and
that's a network level 1 problem - far below your application, and not
affecting it.
The thing to worry about will be memory resources on the 3/50. I haven't
found anybody yet who sells upgrades for them (though I pumped the
Clearpoint rep pretty hard this morning, she wouldn't say whether they
were developing a kit). You may be better off running your large database
stuff on a machine with possibilities for memory expansion. Even if the
server is busy doing other things, you may be hurt less by using it if it
has enough memory. 4Mb is not enough for real work.
Bob Sutterfield, Department of Computer and Information Science
The Ohio State University; 2036 Neil Ave. Columbus OH USA 43210-1277
bob@cis.ohio-state.edu or ...!{att,pyramid,killer}!cis.ohio-state.edu!bob
------------------------------
Date: 17 Jun 88 04:50:00 GMT
From: tekbspa!tss!joe@uunet.uu.net (Joe Angelo)
Subject: Re: Sun losing track of console
tekbspa!tss!joe@uunet.uu.net (Joe Angelo) says:
| I've seen a number of messages regarding "Sun's losing track of console"
| and wish a working solution were posted as well! Here's some more input to
| anyone trying to solve the problem:
Excuse me for my last posting -- I wasn't in the best mind frame.
According to our Boston Sys Manager, removing /dev/kbd, /dev/fd, etc. has
indeed kept the consoles alive! Amazing...
Rick Genter thinks that /etc/init is the problem and since I have an
inquiring mind: anyone with a source license care to check?
Joe Angelo -- Senior Systems Engineer/Systems Manager
at Teknekron Software Systems, Palo Alto 415-325-1025
uunet!tekbspa!joe -OR- tekbspa!joe@uunet.uu.net
------------------------------
Date: Thu, 16 Jun 88 16:57:47 CDT
From: gene@uicsl.csl.uiuc.edu (Gene Gardner)
Subject: Monitor-problem Verified
We just had another "flyback" failure and I would like to COOBERATE the
experience previously mentioned in a Sun-Spot recently by Dean Krafft at
Cornell University. The H.V. lead was touching the metal case, and altho
the arcing was so subtle that the wire still looked O.K., there was a
definite "etching" of the metal surface where the wire had laid.
So I would AGREE that it is probably very good preventive maintenance to
inspect all of your Philips monitors and be sure the lead is clear; I use
a loose "loop" with a nylon-tie to be sure the wire doesn't "drift back"
after a time. By the way, the last new one I installed had an improper
bridge of solder fused between 3 of its terminals, so new ones are
apparently not tested at the factory. It's easier to change the fly- back
if you temporarily remove the power-supply unit ...also easier to position
H.V. lead and a nylon-tie.
------------------------------
Date: Sat, 11 Jun 88 10:33:42 PDT
From: franz!frisky!jkf@ucbarpa.berkeley.edu (John Foderaro)
Subject: gethostbyname in 4.0
Regarding the problems with gethostbyname mentioned in v6n108:
Beware, the 'struct hostent' (netdb.h) structure has changed form in 4.0.
The last field is now points to a vector of addresses, not just one
address. Thus C code using netdb.h compiled on 3.x and linked with 4.0
libraries will lose (and vice-versa).
Another gotcha is that 'struct direct' (sys/dir.h) changed in 4.0. Any
files using dir.h must be compiled with the 4.0 include files if you want
to use the 4.0 libraries.
john foderaro
franz inc.
[[ The changes to the "struct hostent" were made to be compatible with the
name resolver software. It was a necessary change made by the people who
wrote the resolver (in other words: NOT by Sun). As for recompiling
software: it is always a good idea to recompile as much as you can after
installing a major system upgrade (such as 3.x to 4.0). What other
include files might have changed? --wnl ]]
------------------------------
Date: Thu, 16 Jun 88 14:37:55 EDT
From: Graham Campbell <gc@ewok.amd.bnl.gov>
Subject: Problems Installing 4.0
We are considering installing SunOS 4.0 and being seized with an unusual
fit of rationality, decided to install it first on a standalone system.
This system does not have a tape drive. The only tape drive available on
a Sun is on our big server system. Guess what? Sun does not support 4.0
install on a system using a remote tape drive, unless the tapehost is a
4.0 system already!!!!!!
Since 4.0 is such a major change, we are not about to install it first on
the most important system. Has anyone else faced this? Any tips about
problems encountered would be welcome.
Graham
P.S. Is some of the AT&T ineptness rubbing off on Sun now that they are
working so closely together?
------------------------------
Date: 16 Jun 88 19:28:41 GMT
From: ulowell!applix!mark@eddie.mit.edu (Mark Fox)
Subject: SunOS 3.5 Blues
We were told by Sun to upgrade to SunOS 3.5 because they want SunAlis
rebuilt with the new compiler and libraries, etc. At the same time we
upgraded our first server, a 3/180, we added a Hitachi 1GB disk. We had
been running 3.4. Since the upgrade, 3 weeks ago, our server has been up
and down like a yo-yo displaying the following problems:
1) nfs daemons swap out and never come back until the server is rebooted.
This is a documented 3.5 (and 3.4) bug and we have tried several patches
-- including installing 3.5.1 and some .o's from 3.5.2 -- without
success.
2) fsck, when asked to salvage a "partially truncated inode" will not
respond and poses the same prompt repeatedly. we have reverted to the 3.4
fsck...
3) the server panics with a "dup biodone" message and sometimes reports
"xy0a: write failed (zero sector count) blk #0" before the panic message.
we replaced the xylogics 451 controller without success.
Anyone got any suggestions? We are ready to revert to 3.4 or make the jump
to 4.0 -- except we were told that 4.0 froze when 3.5.1 did. Could it be
the disk drive? We previously had a Fujitsu Eagle. Is 3.5 really as bad as
it seems?
Mark Fox
Applix Inc., 112 Turnpike Road, Westboro, MA 01581, (617) 870-0300
uucp: {ames,rutgers}!harvard!m2c!applix!mark
mark@applix.m2c.org
------------------------------
Date: Thu, 16 Jun 88 18:18:26 EDT
From: info@scarecrow.csee.lehigh.edu (Info Directory-x4508)
Subject: Error messages not appearing in console window on client
We have a Sun 3/60 as a client serving off of a sun 3/160. Under suntools
on the 3/60 client, error messages (network, NFS type) and other messages
(like those from FrameMaker....Loading fonts.....etc..) which should
appear in the console window simply do not. Instead they appear
splattered across the console as if no console window has been opened.
Any ideas what might be wrong here?
-Keith Werkman, CSEE Dept., Lehigh University, Bethlehem, PA, 18015
keithw%scarecrow.csee.lehigh.edu@vax1.cc.lehigh.edu: Internet
ihnp4!lehi3b15!scarecrow!keithw: UUCP
------------------------------
Date: 15 Jun 88 13:08:31 GMT
From: mordor!lll-tis!pacbell!ski!robbin@rutgers.edu (Robbin Huls)
Subject: Sun NeWS - command needed to free space allocated by the package
Help! I'm using the "readcanvas" and "imagecanvas" commands of the Sun
NeWS package (post-script based) to display a bit-mapped image on a
"canvas". Each time I do this, lots of space (up to 2 MB) gets allocated.
The space is allocated to save a copy of the image for refresh, but it
does not get ("automagically") freed until the program exits. I have been
told by people at Sun that there exists a program command to free this
space but no one can tell me what it is, nor can I find it in the manual.
Can anyone help??
Thanks, Robbin.
P.S. "makecanvasnonretained" sort of solves the problem by not
allocating the space to save a copy of the image, but if
another window gets placed over the image, the original can't
be refreshed and thus is lost.
e-mail may be addressed to ..!rochester!kodak!fedsys!ski!robbin or I'll be
watching this space for an answer. Thanks again for any help you can
offer. R.
------------------------------
Date: 16 Jun 88 15:29:12 GMT
From: lloyd!wse%hscfvax@husc6.harvard.edu (Bill Edwards)
Subject: Disk configuration question (Xylogics with No. Telecom on Sun 3)
I'm configuring a Northern Telecom disk (model 8312-50) with a Xylogics
451 controller on a Sun 3/180 file server (Sun OS 3.2). What is not clear
to me is what sector size (in bytes) I should give Sun's diag program. Is
anybody using this hardware/sofware combination? Please respond by email.
Thanks. -- Bill Edwards
Bill Edwards wse%lloyd@husc6.harvard.edu (ARPA)
Camex;75 Kneeland St. Boston MA 02111 wse@lloyd.uucp; husc6!lloyd!wse (UUCP)
------------------------------
Date: Thu, 16 Jun 88 14:46:34 EDT
From: sundar@wheaties.ai.mit.edu (Sundar Narasimhan)
Subject: ndbm routines hang in lseek() or read() -- Any ideas??
I have a database written with ndbm routines. It is fairly large, and has
about 500 records (~600 bytes each).
Here's the size of it:
-rw-rw-rw- 1 sundar 196608 Jun 12 05:57 WREQ.dir
-rw-rw-rw- 1 sundar 1579556864 Jun 12 05:57 WREQ.pag
a. Why is the database so large? I thought the size was supposed to be
only 4 times the actual size. (stat() reports a filesize (blocksize *
allocated blocks) to be fairly huge as well).
[[ What is the actual size? "ls -ls" will tell you the actual number of
disk blocks that each file takes up. Remember that ".pag" files contain
holes. --wnl ]]
b. I have a routine that attempts to iterate over all the records;
for(key = dbm_firstkey(db); key.dptr != NULL; key = dbm_nextkey(db)) {
data = dbm_fetch(db, key);
if(data.dptr == NULL) {
check_set_error("Malformed data item - key %s\n", key.dptr);
continue;
} else {
etc etc.. (right out of the man page)
This routine always gets right to the end of the database (i.e. in the
above example with 500 records, it will go upto the 498'th or 499'th
record) and then simply hang. I've caused the program to dump core and
looked at the trace -- it is always inside either lseek() or read() within
the dbm_nextkey() routine.
Why does this happen? How can it be fixed? Thanks for any help you can
give.
-Sundar
[[ How large is each record? There is a restriction in dbm: "the sum of
the sizes of a key/content pair must not exceed the internal block size
(currently 1024 bytes)." Could this be part of the problem? Have you
checked the return value of "dbm_store"? --wnl ]]
------------------------------
Date: Thu, 16 Jun 88 9:28:59 PDT
From: wrs!jerry@sun.com (Jerry Fiddler)
Subject: Big monitors, and monitors in parallel?
I'm looking for a Sun-compatible monochrome large-screen monitor or
projection video system to use at trade shows for demonstrations. The
biggest monitor I've found is a 25" available from Moniterm. Does anyone
know of anything larger, or of a reasonably priced video projector?
Also, will I have any video degradation problem connecting the new monitor
in parallel with the existing monitor? I want them showing exactly the
same image.
------------------------------
Date: Thu, 16 Jun 88 14:38:06 EDT
From: oravax!cammie@cu-arpa.cs.cornell.edu (Cammie Howard)
Subject: rlogin/vi problems?
While logged onto a Sun 3/60, and using Suntools, I rlogged-in to a Vax
(11/750) to edit a file there. I experienced the following behavior:
(This was using vi, but similar events occur using emacs)
The first screenful of text looked fine. Then I did a ^D to go down half a
screenful (or thereabouts), and the new lines appearing had a blank line
between each pair of new text lines. These blank lines disappeared after a
^L. I then began to edit, and discovered that typing a "k" in command mode
didn't move the cursor. After a ^L, the cursor appeared on the line above
theline it had been on when I typed "k". Finally, at least under some
circumstances, when I go into input mode by typing an "i" in command mode,
the characters I type get written over the characters in front of the
cursor, as if I had typed an "s" instead of an "i". After getting out of
input mode and typing ^L, the overwriten characters appear to the right of
the newly added characters, just as they are supposed to.
Thank you for any help you can give me.
Cammie E. Howard, System Administrator
Odyssey Research Assoc.
cammie%oravax.uucp@cu-arpa.cs.cornell.edu
------------------------------
Date: Thu, 16 Jun 88 15:34:01 EDT
From: Ned Danieley <ndd@sunbar.mc.duke.edu>
Subject: how do we use a 3/260 CXP?
We have purchased a couple of 3/260 CXPs, but really don't know how to use
them. We've heard that it is possible to get the graphics speedup without
explicitly writing code as described in the GPSI manual, but can't figure
out how to do so. So we have a few questions:
1) should the demos in /usr/demo/GP1 work on the CXP? I've had
some luck getting flight to run (if I start up suntools, and
a gfxtool), but the screen frequently gets screwed up, and I
get a lot of messages about the graphics processor restarting.
2) can a program written for standard sunview/pixrect use the GP2?
if so, how?
3) does anyone have any simple demos we could run? especially demos
that will run with and without the GP2, so we can get an idea of
how fast it is.
Any info about this beast would be greatly appreciated.
------------------------------
Date: Thu, 16 Jun 88 12:39:58 PDT
From: acad!robert@decwrl.dec.com (Robert Wenig ext 609)
Subject: Trailblazer modems and Sun386?
Has anyone gotten a Telebit modem to work well with a Sun386i and uucp?
I hooked mine up, following the instructions from Mike Ballard at Telebit,
however uucp barely works, with many bad header packets, etc. On the same
machine, with a Hayes 2400 baud modem, all is well.
Robert Wenig
Autodesk
{sun,decwrl}!acad!robert
------------------------------
End of SUN-Spots Digest
***********************