rif@RICE.EDU (Vicky Riffle) (07/14/87)
SUN-SPOTS DIGEST Thursday, 2 July 1987 Volume 5 : Issue 21 Today's Topics: Re: Generating Sun-2 Binaries on a Sun-3 Re: Sun rasterfiles in a TeX/Postscript document NEC 800 Mb disk + Ciprico controller + Rimfire driver = lousy one console for multiple servers? Device driver for S? ttytype of pseudo-ttys - Sun 3.2? nfs? Sun-3 VME to IBM Channels? batch & accounting on unix Problem with C language loader? franzlisp and 68881's Power supply in Sun-3/160 shuts down without warning? SunIPC board & Interphase 3200 SMD Controllers? Mail? Problems with SunView? Sun rasterfiles in a TeX/Postscript document? Upgraded Sun 1s? ------------------------------------------------------------------------ Date: Mon, 22 Jun 87 20:07:26 From: richard%aiva.edinburgh.ac.uk@cs.ucl.ac.uk (Richard Tobin) Subject: Re: Generating Sun-2 Binaries on a Sun-3 The claim that the files /lib/crt0.o and /lib/Fcrt1.o affect the linking of sun 2 binaries on sun 3s is in fact wrong, since there's no difference between the sun 2 and sun 3 versions (at least under sun os 3.2). Both are 68010 binaries with magic number 200407. Richard Tobin, JANET: R.Tobin@uk.ac.ed AI Applications Institute, ARPA: R.Tobin%uk.ac.ed@cs.ucl.ac.uk Edinburgh University. UUCP: ...!ukc!ed.ac.uk!R.Tobin ------------------------------ Date: Fri, 26 Jun 87 10:40:34 edt From: trevor@grasp.cis.upenn.edu (Trevor Darrell) Subject: Re: Sun rasterfiles in a TeX/Postscript document > Including a Postscript version of a Sun screendump in a TeX file always > puts the image on the bottom, left side of the page. I understand that > this is the default location for graphics on a Postscript printer. > I would like placement of the image to be *relative* to the placement > of the ``\special'' command in the TeX source. I have a package that does what you want and more, called ``psfig/tex'', only problem being at this moment it only works with dvips from ArborText. Dvi2ps support should be in the next release (of dvi2ps); all that is needed is adding a literal postscript pass through to dvi2ps. I have just recieved a beta copy of the new dvips that has psfig support (hasn't even been untar'ed yet...), and I'm going to take a quick look at retrofitting the \special code into the exisiting dvi2ps to tide people over for the short term. In any case you can ftp it from linc.cis.upenn.edu; first get psfigtex.README then read that and decide what files you want. (the source itself is quite small, but the entire documentation bundle has some huge figures) It was also send over mod.sources (aka. comp.sources.unix) several months ago and should be in the various archive sites. > -- Mike Jipping > jipping@cs.uiowa.edu --trevor trevor@linc.cis.upenn.edu ps. (the troff version of this (actually the original) will be availble for ftp shortly, after some distribution issues are resolved with adobe systems). ------------------------------ Date: 23 Jun 87 07:44:30 GMT From: mcvax!inria.inria.fr!shapiro@seismo.css.gov (Marc Shapiro) Subject: NEC 800 Mb disk + Ciprico controller + Rimfire driver = lousy A while ago, I posted an article recommending the (cheap) combination of a NEC 800 Mb disk (23xx series) + Ciprico controller + Rimfire driver sold by Digital Design, commenting "there used to be bugs but we/they ironed them out". That was a bit hasty. The bugs are definitely still there. We experienced a number of crashes which would totally fuck up 1 or 2 partitions. Also, at times of relatively high activity, all of a sudden the system would just sit there, hung, apparently deadlocked. We had to disconnect our NEC drive and revert to the good old Eagle. Digital Design is supposed to come today and install the "final, debugged" release. I don't trust them a bit. If anybody is interested in finding out what happens next, send me mail. -- Marc Shapiro; INRIA Bat. 11; B.P. 105; 78153 Le Chesnay Cedex; France tel. +33 (1) 39-63-53-25 uucp: seismo!mcvax!inria!shapiro; or: shapiro@inria.inria.fr ------------------------------ Date: Tue, 23 Jun 87 14:31:11 From: wwtz%cgcha.uucp%cernax.bitnet@berkeley.edu (Wolfgang Wetz) Subject: one console for multiple servers? We are expanding our SUN network. We will then have somewhat around six servers (2*3/180 and 4*3/280), where all these servers will be located in one room. Now, I do not like the imagination of having six consoles (wasting a lot of space). I am looking for a solution, where one console will serve all servers. In special I am thinking on some sort of a PC running some fancy software doing this job. Is somebody out there who can give me some hints? If there is a significant response on this topic, I will post a summary to the net. Many thanks in advance. Wolfgang Wetz (...seismo!mcvax!cernvax!cgcha!wwtz) c/o CIBA-GEIGY Ltd., CH-4002 Basel, P.O. Box, Switzerland ------------------------------ Date: Wed, 24 Jun 87 13:06:49 edt From: booker@nrl-aic.arpa (Lashon Booker) Subject: Device driver for S? Does anyone have a SUN device driver for S that works under release 3.2? Thanks. Lashon Booker booker@nrl-aic.arpa ------------------------------ Date: 21 Jun 87 02:45:37 GMT From: mkhaw@teknowledge-vaxc.arpa (Michael Khaw) Subject: ttytype of pseudo-ttys - Sun 3.2? My sun's /etc/ttytype doesn't have any entries for /dev/tty[pqrs]? If I test for $term in my .login BEFORE doing any "set term"s or "tset"s, I get $term == "su". Why is this? Is it a bug (or a feature)? Is it random garbage? Or is what I should always expect when the pseudo-ttys are not listed in /etc/ttytype? Getting $term = "su" when logging in on a pseudo-tty happens with remote logins. I haven't tested it with shelltools opened under suntools. Thanks, Mike Khaw -- internet: mkhaw@teknowledge-vaxc.arpa usenet: {hplabs|sun|ucbvax|decwrl|sri-unix}!mkhaw%teknowledge-vaxc.arpa USnail: Teknowledge Inc, 1850 Embarcadero Rd, POB 10119, Palo Alto, CA 94303 ------------------------------ Date: Mon, 22 Jun 87 19:56:19 From: richard%aiva.edinburgh.ac.uk@cs.ucl.ac.uk (Richard Tobin) Subject: nfs? We have a network of about 20 mixed disked and diskless sun 2s and 3s. They all mount each other's file systems, and we'd like to arrange it so that a piece of software installed on one machine is available on them all. So most people have paths that include directories from a large number of machines. This works well unless one machine is down, and then not only does the software on that machine (of course) become unavailable, but also when people log in the 'set path' command in their .login files hangs, as the shell tries to read directories on the dead machine. We could mount the file systems soft, but that implies read-only (aside: is this really necessary? What kind of damage could occur if they were read-write? And how about a mode that waited to do reads and writes, but timed out for opens - would that be safe?). Alternatively, .login files could check that the relevant machines are up before adding them to the path. A similar problem is that paths should be different on different machines (sun 2s should get 68010 binaries; machines with local disk should get files from that in preference to remote ones). This can be done with a complicated system of symbolic links, or with conditional commands. I'd be interested to hear what solutions other people use for these problems. Richard Tobin, JANET: R.Tobin@uk.ac.ed AI Applications Institute, ARPA: R.Tobin%uk.ac.ed@cs.ucl.ac.uk Edinburgh University. UUCP: ...!ukc!ed.ac.uk!R.Tobin ------------------------------ Date: Thu, 18 Jun 87 16:32:24 EDT From: mike@brl.arpa (Mike Muuss) Subject: Sun-3 VME to IBM Channels? We at BRL need to attach some IBM peripherals to a Sun-3 machine. There are several commercial boards to attach IBM channels to the VME bus. If any of you can provide us with leads, or reports on your experiences, or offer drivers :-), that would be most appreciated. Please reply to <ACST@BRL.ARPA>. In particular, we are looking at going this route to make some of our IBM peripherals network-wide assets. For starters, we have one string (1x8) of IBM 3480 cartridge tapes, two strings (2x1) of IBM 3420 reel-to-reel tapes, and a Xerox 6040 laser printer. Assuming that all of that works, we will then also attach a MASSTORE device using the same interface. Once the Sun(s) have direct access to the IBM peripherals, we will layer on top of that a network-transparant remote tape interface, so that all the machines on the network will be able to use the IBM devices. Once that is done, we plan to implement a transparant file migration/ file archiving system, using network access to the MASSTORE to support automatic file reloads, much like "older" operating systems have always provided. BRL will, of course, make this software available to the community when it is operational. Please don't shower us with requests for it, yet. At this point, what we are looking for is advice on the Sun-3 VME bus to IBM Channel interface, which is the building block for this project. (We intend to make it possible for the software to also drive the IBM peripherals via a Hyperchannel A515, but do not at present have a Hyperchannel A515 to work with, so...). Best, -Mike ------------------------------ Date: Tue, 23 Jun 87 11:22 EDT From: SYSRUTH%UTORPHYS.BITNET@icsa.rice.edu (Ruth Milner) Subject: batch & accounting on unix (My apologies if these are naive questions; I am new to unix) Does anybody know of a batch queuing system available for unix, or a reasonable facsimile? I would like to be able to set up something like what we have on our VMS machines, i.e. to limit the number of jobs at various nicenesses and distribute the CPU accordingly. Is there some reasonably straightforward way of setting up and maintaining one using the tools available in SUN unix? (BTW, we have an AT&T source licence). Similarly, does the accounting keep track of a process' niceness? We would like to be able to charge people less for running nicer jobs, but I haven't noticed this being recorded anywhere. Does anyone have a package (or know of one, preferably PD) for reading the accounting information and outputting a nicely-formatted report including such information as niceness, CPU time, elapsed time, username, group name, I/O operations, memory use, etc? Can anyone direct me to some documentation on the existing accounting (what I have seen so far is pretty skimpy)? Does anyone have a fix for the bug that causes "last" to think that processes logging in through dna (DECnet/sunlink) have never logged out? Please reply directly to me if possible as I am not a regular reader of this list. I will post the results, if any. Thanks in advance. Ruth Milner Systems Manager University of Toronto Physics SYSRUTH@UTORPHYS.BITNET (sorry, I'm not sure of the UUCP/Usenet path in) ------------------------------ Date: Tue, 23 Jun 87 11:03:19 CST From: 79343382%TRINITY.BITNET@wiscvm.wisc.edu (Aaron Konstam) Subject: Problem with C language loader? I do have a problem I would like to ask about. We are running version 3.2 of the Sun operating system and are having the following problem with the loader when programming in C. If we are using, for example, the floating point routines by including math.h and using the -lm flag when calling cc the following problem occurs. Instead of only the library routines that we are using in the C program being included in the a.out file all the math routines are included in the executable file. Why is this happening and/or what are we doing wrong. Aaron Konstam (512)-736-7484 or 79343382@TRINITY on Bitnet ------------------------------ Date: Tue, 23 Jun 87 13:31:44 PDT From: guyton%condor@rand-unix.arpa (Jim Guyton) Subject: franzlisp and 68881's Yes, if Franzlisp is built on a machine with the 68881 chip, and then run on one without it, then the first time franz tries to do a garbage collect, it does a floating pt instruction to calc some percentages and dies with an illegal instruction trap. The problem is that the build of franz includes an "unexec" or "sysout" of existing memory to an a.out file. This unexec happens AFTER the run-time initialization of Sun's that checks to see what type of floating pt hardware you have, and if it changes Franz has problems ... -- Jim p.s. I'm talking about the PD franz that came with 4.2/4.3BSD. I have no idea if the above is true about the supported version of franzlisp from Franz Inc. ------------------------------ Date: Wed, 24 Jun 87 09:04:24 PDT From: hoptoad.UUCP!gnu@cgl.ucsf.edu (John Gilmore) Subject: Power supply in Sun-3/160 shuts down without warning? Hoptoad has been having an intermittent problem in which all DC power in the 160 chassis suddenly shuts off without warning. There is still AC power to the box, the light in the on/off switch is still lit, and the blowers in the power supply itself are running, but the rest of the machine is dead, including the rest of the fans. The problem seems to correlate with heat, and has occurred several times a day in hot weather, though usually it only happens once a week or less. Has anybody else seen this problem? Know of a fix other than replacing the power supply? John ------------------------------ Date: Mon, 22 Jun 87 17:19:50 From: mcvax!olsen!lance@seismo.css.gov (Lance Berc) Subject: SunIPC board & Interphase 3200 SMD Controllers? I plugged a SunIPC board into an empty (other than CPU) 3/160 and had major problems getting PC-NFS running correctly. Loading files took forever, many retries were required, etc. When I plugged the same board into a 3/180 (also diskless) that had two other IPC boards the problem dissapeared. The Suns use the same kernel and the IPC boards use identical drive C virtual disks (with modifications for the different host). The only symptom I have is that the me0 (memory ethernet) device registers excessive output errors going from the CPU to the IPC board. I tried moving the IPC board to other slots and reseated all of the backplane jumpers. The IPC board passes its diagnostics with no errors. I tried asking SunUK about this, but as is usual with Sun, their expert is on vacation, nobody in the office knows what's going on, and they don't return my calls. They weren't even interested in documentation errors I found during the installation. Sun World Headquarters refuses to talk to me because I'm in Europe, even when I pay the $3/minute and the UK office is closed. Has anyone out there got a clue? ------ I also evaluated a set of Interphase SMD 3200 VME disk controllers. Unfortunatley the controllers failed the evaluation, returning incorrect data blocks (i.e. didn't read what was written) about three times a day when running a set of random seek-write-read thrash tests over disjoint partitions. The errors persisted even after upgrading all of the boards to both the released 150 and in-beta-test 15E prom levels. I say unfortunately because Interphase, unlike SMI, has been an absolute pleasure to deal with at all levels, from my Swiss distributor through their European office to their headquarters in Dallas. Their people have been knowledgeable, competent, honest, and have taken a serious interest in fixing the problem. They've returned calls and have even had helpful suggestions. What a difference people make. I'll post another note when they fix the problem. Until then I really wouldn't trust data that's written with a 3200. This doesn't hold for the 4200 `Cheetah' high-speed, gee-whiz, super-duper board that I'll be evaluating this week. I'd be glad to send my disk/controller test programs to whoever's interested. Lance Berc lance@pescadero.stanford.edu (forwards to:) Olsen & Associates mcvax!olsen!lance@seismo.css.gov Seefeldstrasse 233 8008 Zurich, Switz ------------------------------ Date: Wed, 24 Jun 87 13:57 EST From LYMAN%IASSNS.BITNET@icsa.rice.edu (Lyman Hurd) Subject: Mail? We have a network of about two dozen workstations with two major file servers. At the moment we don't have source, but also we are not averse to super-user access over the net. What is the easiest way to make mail work independently of where one logs in? Currently thanks to yellow pages and NFS, this is the only machine-specific function which ties users to one machine or another. Thanks, Lyman Hurd Institute for Advanced Study ------------------------------ Date: Wed, 24 Jun 87 09:56:55 cdt From: texsun!radian!johnl@sun.com (John Lange) Subject: Problems with SunView? I am writing an application using SunView in which I want to use all 256 colors on the screen at one time. BTW, I'm using a 3/160 with a color monitor. The problem I'm having is this: If I run an application with a 256 entry colormap, the window system automatically makes the parent window's background and foreground entries numbers 254 and 255 in the colormap. Hence, my application looks OK when it's in one of its own windows (all its windows have the same colormap), but the background and foreground for the screen are strange unless I write, say, black and white for 254 and 255 of the colormap. When the mouse is then moved to outside the application, the application windows all go monochrome, and so does the parent window's background and foreground. I would like to be able to make the default colormap 256 entries and be able to control which entries function as background and foreground. I would also like to have more control over the colormap sharing between windows. I've tried using pixrects (pr_putcolormap) with '/dev/fb', but that doesn't do what I want. What might work is resetting the cms_mono.h file for 256 entries and reconfiguring the kernal, but I haven't tried this yet. Even if this works, it's not ideal because it's not under software control. Any ideas, anyone?? Also, in using SunView, I'm trying to read events with window_read_event() documented in SUNVIEW documentation, chap. 6, pg. 87. I can get it to work fine if I know which window I'll be reading from in advance. However, I want to do different things depending upon which window I'm in. I think I just missed something in the documentation, as this is something that should not be hard to do. I currently first check the status of a particular window with the WIN_KBD_FOCUS attribute, and if that window is active I call window_read_event. Should I be using the attribute KBD_REQUEST instead? It may be the way I have it set up in the loop, but for some reason, I seem to be MISSING EVERY OTHER EVENT. Has anyone encountered this phenomenon before or found a solution? Thanks in advance, John Lange Radian Corporation ph. 512-454-4797 x. 5747 Arpanet johnl@altair ------------------------------ Date: Thu, 25 Jun 87 14:56:16 CDT From: jipping%swen%cs.uiowa.edu@relay.cs.net (Mike Jipping) Subject: Sun rasterfiles in a TeX/Postscript document? I'm sending this out to several newsgroups to ensure maximal coverage. Apologies to those who get multiple copies... I've got a problem with TeX -> Postscript translation that includes Sun raster dumps. First, some background: We just a new LaserWriter Plus from Apple that's driven off our Sun 3/280 file server. We also just installed the Adobe Transcript soft- ware to drive the printer. Now we've got the tools to (a) dump a bit-image of a Sun monitor's screen, (b) convert that bit-image to a Postscript file. We also have TeX/LaTeX software, complete with "dvi2ps" conversion. Thus, we can (c) run TeX documents on our printer, and (d) use ``\special'' to include Postscript code in a TeX ``dvi'' file. I'm trying to put an image of a Sun screen in a figure in a TeX/LaTeX document. Here's the problem: Including a Postscript version of a Sun screendump in a TeX file always puts the image on the bottom, left side of the page. I understand that this is the default location for graphics on a Postscript printer. I would like placement of the image to be *relative* to the placement of the ``\special'' command in the TeX source. I know that the ``dvi2ps'' software supports horiz and vertical offsets, but I could be adjusting for days to get it right -- and then a change in the TeX source could change the placement of the image. Does anyone have a solution to this problem? I'll readily admit that I don't know Postscript -- so don't flame me for being dumb. I'll get to it, but I'd rather have a solution soon. Thanks in advance. -- Mike Jipping University of Iowa Dept of CS jipping@cs.uiowa.edu ------------------------------ Date: 26 Jun 87 02:46:17 GMT From: umnd-cs!umn-cs!haberman@rutgers.edu (Joe Habermann) Subject: Upgraded Sun 1s? Is there anyone out there who has successfully upgraded their Sun-1's to 4 megabytes or more via a third party vendor such as Helios? If so, are they running "a lot" faster due to the increased cache? Is there anyone else still running Sun-1s? Thanks in advance. Joe Habermann Systems Group Dept. of Computer Science University of Minnesota (612) 625-0876 tundra!mmm!umn-cs!haberman haberman@umn-cs.ARPA ------------------------------ End of SUN-Spots Digest ******** sp sp djob