Sun-Spots-Request@RICE.EDU (Vicky Riffle) (01/13/87)
SUN-SPOTS DIGEST Tuesday, 13 January 1987 Volume 5 : Issue 2 Today's Topics: Common Lisp compiler-gc bug to be fixed Object oriented programming in sh Need beta-testers for a multi-player spacewar game Catatonia found and fixed Synch over Sun serial ports ND doesn't use buffer cache effectively Two 3.2 bugs to be wary of SUN-3/ZIP 3232(+) access time questions? Bug in SUN Kernel RPC? Appeal for help communicating with a xerox 1186? Want to save 80,000 disk accesses a day? Help in setting up spice3a7 on a sun3 with FPA? 260's delivered? problem with Sun serial lines? color display? High line rate image hardcopy ? 150U halt problem? Help needed on swap death bug? Running S on Sun workstations? screen saver? RCS on SUN-3's? ------------------------------------------------------------------------ Date: Wed, 17 Dec 86 22:21:26 -0800 From: mk@aero2.ARPA Subject: Common Lisp compiler-gc bug to be fixed Several weeks ago, we reported that the compiler in Version 1.2 of Sun Common Lisp will cause endless garbage collection on 3/160s with FPAs. We have since been told that Sun and Lucid will have this bug fixed i Version 2.0, due for release in late January or early February. ------------------------------ Date: Wed, 17 Dec 86 14:09:45 est From: ramsdell%faron@mitre-bedford.ARPA (John D. Ramsdell) Subject: Object oriented programming in sh Many programming languages have been augmented to include object-oriented programming constructs. Common Lisp has LOOPS or Flavors, for example. I would like for you to focus on the Unix utility make, used to maintain related files as a system. Consider the implications of viewing make as the object-oriented extension for sh. Object-oriented programming consists of objects that respond to messages. Let a directory with that contains a makefile be sh's notion of an object. The methods that an object can handle correspond to target translations given by the makefile. The analog of instance variables, or state information associated with an object, is stored in files in the directory. Usually, all objects that share the same methods are said to be in one class. This corresponds to many directories sharing identical copies of makefiles. The above is summarized in the table below. ----------------------------------------------- object directory plus makefile instance variables files handler or class makefile method makefile rule inheritance included makefiles. ----------------------------------------------- The problem I have found with Sun's version of make comes when implementing the equivalent of inheritance. I would like to specialize classes of objects via the include command. For example, make using the following makefiles :::::::::::::: makefile :::::::::::::: all: a include ../makefile a: a.sh :::::::::::::: ../makefile :::::::::::::: a: a.o -------------- demands that a.sh is present and there is some way to make a.o. I created a.c, which it compiled. However, the make command will not make anything called a. It will not give you any error message. Multiply defined targets totally confuse make. So much for inheritance. John ------------------------------ Date: Thu, 18 Dec 86 13:44:26 PST From: jonab@SDC-CAMARILLO.ARPA (Jonathan P. Biggar) Subject: Need beta-testers for a multi-player spacewar game SPACEWAR Version 2.0 Spacewar is a multi-player spaceship combat game inspired by the original version written for a PDP-1 and by the arcade game of the same name. Version 2.0 runs on Sun-3 workstations with a 68881 math co-processor and a 1152x900 monochrome monitor. Each ship is controlled by the mouse and keyboard of one workstation. Each workstation running the game communicates over the ethernet with a central server machine which keeps track of the state of the game. The game has been up and running at my site for three months, and playing it with my associates consumes much of my free time. I am looking for beta-testers with at least the following hardware: 3 Sun-3 workstations with 68881 chips connected via ethernet. At lease two of the workstations must have monitors. If you are interested, and have the time to test the game and mail comments to me, send me email. I will choose several responders and send them a spacewar kit. NOTE: I do not plan to release the source to the game for two reasons: 1) I wish to maintain control of the source in order to make bug-fixes, improvements, or write new games. 2) Part of the release contains a patch to the SunView window library that I can only distribute in binary form pre-linked with my own object files. Jonathan Biggar jonab@SDC-CAMARILLO.ARPA or jonab@sdcrdcf.UUCP or sdcrdcf!jonab ------------------------------ Date: Mon, 22 Dec 86 11:16:36 PST From: krasner.pa@Xerox.COM (Glenn Krasner) Subject: Catatonia found and fixed A while back, I described a problem ("catatonia") where our Suns would lock up often and at unexpected times. It was most highly correlated with running Smalltalk, using NFS, and swapping. Since then, a bug that caused catatonia was discovered in the 3.0 kernel, and has been fixed in 3.2. If you are running 3.0, you can patch this bug if you set the kernel variable ws_set_favor to zero. % adb -w /vmunix ws_set_favor?W 0 ^D % /etc/reboot Many thanks to the support staff at Sun for finding and fixing the bug, and to the Smalltalkers at Univ. of North Carolina for testing it. Glenn Krasner ParcPlace Systems (Krasner.pa@Xerox.com) ------------------------------ Date: Wed, 24 Dec 86 18:26:00 CST From: texsun!hemaneh!tj@sun.UUCP (Cal Thixton) Subject: Synch over Sun serial ports Date: Tue, 25 Nov 86 05:18:02 EST From: mark@markssun.cs.umd.edu (Mark Weiser) Subject: Synch over Sun serial ports? Do the normal pair of sun serial ports brought ought to the back have enough hardware behind them to support externally-clocked synchronous communication (at 56kb)? I presume so, since there are Sun products which talk to them this way. Yes, they do, this is how the RJE BSC Sunlink product works. However, the current zs driver does not support it, and it is not as if you can just add another line dicipline like SLIP does. Where can I find the specifications for programming these ports synchronously? The ZS manual entry refers to a piece of documentation called 800-1052-1 (or at least that is its name, it probably isn't called that). enough of the kernel is provided that you could write your own driver for the zs ports and install it. however, the fastest i've heard the zs cpu ports going to date is about 38.4kb. past this and you are dedicating most of your machine as a byte pusher. I am thinking about using Interoffice LAN from the phone company to connect a couple of semi-distant Suns using SLIP and synchronous communication at 56kb. Anyone with experience to share about any piece of this, I'd like to hear from you. (I currently connect these Suns using SLIP over AJ 4800 baud modems over the switched network, but the AJ's are slowly dying and I am looking into a soon replacement). -mark I have also done this. SLIP can use ordinary serial lines over your average asynchronous modems (like a hayes) to connect two sun nets. there are a few problems and bugs in SLIP; the way SLIP was written, it has the potential to panic your system (though it never did while i was using it). the other thing that really is a drawback is none of the routers are told of this link. it is not broadcasted, so, to let the suns on either nets know about the other side, you'll have to explicitly add routing info at boot time (see route(8)) to each sun that will want to go through that gateway. To do internetwork routing, one really needs a fast link. the expense of TCP is really noticable at anything under 19.2k. Now, a word from our sponsor: The Sunlink INR product does just what SLIP does, but is more robust and the routing code knows how to propagate the link to the net, so explicit routes do not have to be added at boot time. Using the Sunlink communications board (SCP), you can easily go up to 56kb or more without using most of our cycles. Without the SCP, you can go up to 19.2k, perhaps even 38.4kb, just on the cpu zs posts. cal thixton sun consulting dallas sun!tj p.s. INR is synchronous wheras SLIP is asynchronous, btw. ------------------------------ Date: Fri, 19 Dec 86 14:04:57 PST From: decwrl!fluke!jeff@ucbvax.Berkeley.EDU (Jeff Stearns) Subject: ND doesn't use buffer cache effectively Is anyone else bothered by the way their diskless Sun can swamp the Ethernet with unnecessary ND packets? Sun (finally, after more than a year!) tells me that it is due to a ND client buffer cache mismanagement bug, and that it won't be fixed because ND will be going away in the future. Our Ethernet is busy enough, thank you, and I'm not real happy about the extra load and corresponding sloooooooowness of our diskless machines. The simplest demonstration is to run (in sh): while /bin/true; do /bin/date; done It doesn't matter how big I make the system buffer pool -- my diskless client will still miss it most of the time, swamping the fileserver with ND read requests. (What's especially annoying is the fact that, every now and then, the buffering will start to *work* and this demonstration just screams along.) I have seen this bug in every Sun UNIX release since ND was invented. Food for thought: Sun's stated goal is to make NFS as fast as ND. Which implementation of ND? This one? Yech! Jeff Stearns jeff@fluke.COM John Fluke Mfg. Co, Inc. (206) 356-5064 {uw-beaver, decvax!microsoft, ucbvax!lbl-csam, allegra, sun}!fluke!jeff ------------------------------ Date: Mon, 22 Dec 86 20:02:18 PST From: vern@lbl-csam.arpa (Vern Paxson) Subject: Two 3.2 bugs to be wary of Two bugs in Release 3.2 that users should be aware of: 1. "grep -i" can fail to match when it should. For example, "grep -i 'a*b'" won't match "aB". Work-around: use egrep instead of grep, now that egrep supports '-i'. 2. "f77 -O" can generate incorrect code when de-referencing an array. Unfortunately I can't be more specific, since the program I have that demonstrates this problem works fine if it is simplified very much, so I don't know what is causing the problem. If you are interested in looking at the program (it's about 60 lines long), let me know and I'll send you a copy. Work-around: don't compile with -O. Sun has been notified of these problems and is working on fixes, but they may not appear until after the next release. Vern Paxson Real Time Systems Group Lawrence Berkeley Laboratory (415) 486-6411 vern@lbl-csam.arpa ------------------------------ Date: 12 Dec 86 20:16:20 GMT From: tekgvs.tek.com!benw@soma.UUCP (Benjamin Ward) Subject: SUN-3/ZIP 3232(+) access time questions? I am interested in knowing how quickly the Sun-3 to Mercury ZIP 3232+ (or similar array processor) interface runs. For example, suppose you want to multiply two 32 bit real numbers A and B using the ZIP. Assume that the multiply routine is already resident in the ZIP. The overall multiply will require two DMA calls and a routine (multiply) call to the ZIP. How long does this take? (what I am asking is what is the "real" overhead in using this array processor over the VME bus?) If anyone has run some benchmarks I whould love to converse with them by phone/mail. I am Benjamin Ward, Tektronix Inc. 50-370, Beaverton, OR 97077, (503)627-6024, with a net address of tektronix!tekgvs!benw. Please respond directly and thank you to anyone who does respond. ------------------------------ Date: Tue, 16 Dec 86 20:59:15 EST From: dpk@BRL.ARPA (Doug Kingston) Subject: Bug in SUN Kernel RPC? Description: The SUN kernel mode RPC can hang while doing remote RPC that should timeout. An example is NFS when you remote mount a filesystem "soft". Some RPC's to this filesystem will hang. Mount is one such RPC. This problem was found in our Gould kernels which contain code almost identical to the SUN code. I know some other vendors using this code also have the problem. Specifically, the problem is that the function clntkudp_callit does a sleep on &so->so_rcv.sb_cc. The timeout routine, ckuwakeup(), had an incorrect wakeup value which is corrected below. Repeat By: Edit /etc/fstab to remote mount a filesystem. Shutdown the remote system. Reboot the your system. Watch your system hang when it attemps to mount the NFS filesystems. Fix: Apply the following diff: *** /tmp/,RCSt1016522 Tue Dec 16 20:37:15 1986 --- clnt_kudp.c Mon Dec 8 22:34:48 1986 *************** *** 498,504 **** rpc_debug(4, "cku_timeout\n"); #endif p->cku_flags |= CKU_TIMEDOUT; ! sbwakeup(&p->cku_sock->so_rcv); } /* --- 498,504 ---- rpc_debug(4, "cku_timeout\n"); #endif p->cku_flags |= CKU_TIMEDOUT; ! sbwakeup(&p->cku_sock->so_rcv.sb_cc); } /* ------------------------------ Date: Wed, 17 Dec 86 14:50:37 PST From: wiley!mike@lll-lcc.ARPA (Michael Crocker) Subject: Appeal for help communicating with a xerox 1186? I am working on a project that will try to use an 1186 Xerox as an interface device and a color Sun-3 as a graphics device. I need information : 1. Has this been tried before? 2. What are the problems with this? 3. Is there code available to handle the communications? Any help would be appreciated. Michael Crocker ------------------------------ Date: Thu, 18 Dec 86 22:46:06 EST From: mark@mimsy.umd.edu (Mark Weiser) Subject: Want to save 80,000 disk accesses a day? Scott Carson here at Maryland measured rwhod on our suns doing 80,000 disk accesses a day (per sun), eating up loads of arm motion and keeping the buffer cache full of unimportant stuff. We considered various options, like a ram disk for /usr/spool/rwho, but Scott finally hacked rwhod to take an option, '-s', which causes it to receive no packets nor update the files in /usr/spool/rwho, but merely keep the world up-to-date with its broadcasts. The wonderful result: all our suns immediately 'feel' a little faster (We have about 50, mostly Sun-3's with 4M and local shoeboxes). But how does ruptime get the right values (I hear you cry)? Well, a simple nfs mount of the file server's /usr/spool/rwho partition onto the local /usr/spool/rwho (and making sure that the server runs rwhod *without* '-s') takes care of that. For those of you still running rwhod (in spite of Sun's comment in /etc/rc), I recommend this patch. It is a little long for sun spots, but the context diffs are available for anonymous ftp on host mimsy.umd.edu in file 'rwhoddiff', or I can mail them to uucp sites (or post to mod.sources, should that prove appropriate). -mark ---------------- Spoken: Mark Weiser ARPA: mark@mimsy.umd.edu Phone: +1-301-454-7817 CSNet: mark@mimsy UUCP: seismo!mimsy!mark USPS: Computer Science Dept., University of Maryland, College Park, MD 20742 ------------------------------ Date: Fri, 19 Dec 86 20:48:30 PST From: sllu@venera.isi.edu (Shih-Lien Lu) Subject: help in setting up spice3a7 on a sun3 with FPA? Has anyone successfully setup spice3a7 to run on a SUN 3/160 with FPA? (I have no problem with SUN 3/160 but the PFA is giving me unknow Floating Point exception errors.) I would like to ask couple questions. If enough people show interest, I will post summary. Shih-Lien Lu sllu@vlsif.isi.edu ...!ihnp4!seismo!sllu@vlsif.isi.edu ------------------------------ Date: Sat, 20 Dec 86 14:59:08 EST From: mark@markssun.cs.umd.edu (Mark Weiser) Subject: 260's delivered? I would like help quashing a silly rumor. Have any sun-3/260's been shipped to regular customers yet (not beta)? I assume yes, but an Apollo buff was telling me that 25mhz 68020's were still vaporware. Thanks in advance. -mark ------------------------------ Date: Fri, 2 Jan 87 12:51:47 EST From: cdh@bfly-vax.bbn.com (Carl D. Howe) Subject: problem with Sun serial lines? We've been having a terrible time with our Sun serial lines hanging when exiting tip. What happens is that the process gets hung in the "exiting" state and never unhangs, thereby tying up that tty line until we reboot. We're running Sun 3.0 on Sun 3 systems. Any idea with what causes this? Is it perhaps that we are not correctly asserting or deasserting some modem control line (I believe that none of the modem control lines are tied to anything)? Assuming that the Sun emulates a DTE, perhaps the problem is that it is looking for the deassertion of DCD, and the reassertion later. Any idea if the Sun serial line driver cares about the modem control lines? Please mail me any ideas; I don't normally receive sunspots. Carl ------------------------------ Date: Mon, 15 Dec 86 12:31:31 est From: mcgrew@topaz.rutgers.edu (Charles) Subject: color display? I'm looking for information on a good 1024x1024, 24 bits-per-pixel display for use with a Sun. If this has been discussed here before, could someone point me to an issue number? Thanks, Charles ------------------------------ Date: Fri, 19 Dec 86 12:24:28 CST From: pun%cui.unige.chunet@RELAY.CS.NET (Thierry Pun) Subject: Q: high line rate image hardcopy ? Does anyone has experience with a reliable image hardcopy, monochrome and color, preferably on film (35mm, Polaroid ..), whose interface would allow very high transfer rate ? The source images come from SUN 3's (/50,/160): 1152*900 points, 8 bits per point (mono or color), 66 Hz NON interleaved refresh rate. The easily accessible connectors are R,G,B, sync. We would like to avoid solutions implying halftoning (electrostatic or ink jet). More precisely : * has anyone succeeded in hooking a Matrix 3000 up to a SUN ? It seems that there is no way to transfer the whole image (too much data). * has anyone experience with the MicroColor system from Dunn ? Thanks a lot for any information. Thierry Pun UUCP : seismo!mcvax!cernvax!cui!pun BITNET : pun@cgeuge51 ------------------------------ Date: Fri, 2 Jan 87 16:27:20 CST From: oddjob!jrf@lbl-csam.arpa (Jim Fowler) Subject: 150U halt problem? I have an interesting problem which I wonder if the older Sun guru's can solve. We have a model 150U which halts for no apparent reason. The symptoms are: at apparently random times the system comes to a complete halt. The video screen goes black and all the lights on the cpu board go out. The lights on the keyboard also go out ( we have the old vt100 style with black keys ). We have not been able to correlate this to any other actions occuring on or near the system. Furthermore the system can stay up for two weeks without an occurence, other times it won't stay up for ten minutes. Needless to say this can be frustrating. Our system configuration is: model 150U running Sun Unix release 3.0 2 Sun 1Mb memory boards ( talk to me about slow response!) 1 archive tape drive ( the old original ones) 1 Xylogics 450 controller (rev C proms ) 1 Eagle disk drive 1 Fuji 2312 disk drive ( the old 84 Mb unformatted ) 2 CDC oct-ser interface 1 3Com ethernet board ( installed but unused ) 1 Sun color board and 19in monitor 1 BW video console We have tried pulling all these boards and we still have the problem occur. In addition it does not require Unix to be running as we have seen the system go down while in monitor mode. In addition we have swapped most of these boards into our other machines and we do not see the problem changing machines. Since the problem is occasional we have been trouble shooting this for almost a year now. At first it meant rebooting the system in the morning but now it is almost impossible to do any work. If any one has seen this problem before and found some sort of solution for it I would be gratefull for any help. We have had these machines running continuously for 3 years now ( the machine serial number is C-022 so it is an early one). Many Thanks Jim Fowler ihnp4!oddjob!jrf jrf@oddjob.uchicago.edu ------------------------------ Date: Mon, 5 Jan 87 08:12:22 PST From: gerolima@Ford-wdl1.ARPA (Mark Gerolimatos) Subject: Help needed on swap death bug? Assorted bs: 3/160 with a 3/75 client Running 3.0 (not any more, though) 2 x 75Mb SCSI disks 1 SCSI tape drive, same SCSI bus (correct terminology?) 4 Mb memory on server and client No special hardware Seems that when lots of processes are being fired up on the server (ie: bringing up s'tools), and something gets hit with an interrupt (a ^C does the job, fine), the OS stops swapping. Anything whose working set is already in core is unaffected (clock, any other assorted loopy programs), and will keep running, so it seems that the machine is NOT down. We're probably talking some kind of OS death, here. This problem only comes up once a month or so, and then under 3.0. Of course, when 3.2 came along, I converted right away, and so may not see the problem again. However, the guys at Sun are baffled by this apparent bug, and my attempts at getting a core image of the kernel (using prom-monitor command 'g0') kept on failing (yes, I kept on choking). Anyone else out there had the same problem (under any 3.x version)? "For almost a quarter of a century..." Mark Gerolimatos ARPA: gerolima@ford-wdl1.{arpa,com} "Let's take our neighbor for a ride! UUCP: {sun,fortune}!wdl1!gerolima Would you like to go for a ride!?" AT&T: (415) 852-4105 "Uuhhhh...no thanks." USPS: c/o Ford Aerospace "No what!?" "No, I wouldn't like to go." 3939 Fabian Way "Go Where!?" "Uuhh...for a ride..." Palo Alto CA 94303 "GOOD IDEA!!! LET'S GO FOR A RIDE!!!" Mail Stop X20 ------------------------------ Date: 5 Jan 87 06:45:31 GMT From: munnari!natmlab.oz!ronb@seismo.CSS.GOV (Ron Baxter) Subject: Running S on Sun workstations? I am aware that people have S running on Suns, but the ones I have seen only seem to have Tektronix 4014 style graphics capabilities. So my questions are: - does anyone have e better graphics driver for the Sun that would support color-fill, perhaps different fonts, or any other advance on tek14. - has anyone implemented any functions for the Sun which have functionality similar to the "scatmat" and "spin" functions that S has for the "blit" Thanks in advance for any info that is forthcoming. -- Ron Baxter, CSIRO Div Maths & Stats, PO Box 218, Lindfield, NSW, Australia. PHONE: +61 2 467 6059 ACSNET: ronb@natmlab.oz ARPA: ronb%natmlab.oz@seismo.css.gov UUCP: ....{seismo,hplabs,mcvax,ukc,nttlab}!munnari!natmlab.oz!ronb ------------------------------ Date: Tue, 6 Jan 87 12:40:32 EST From: mckay@courageous.ecn.purdue.edu (Dwight D McKay) Subject: screen saver? Has anyone built a screen saver for the SUNs? What I'm thinking of is a program, similar to those available for MacIntoshes, which blanks the screen after a period of inactivity. To bring back the screen a user touches any key or clicks the mouse. Such a tool would be great for saving the SUN monitor from being "burned-in" when an image is left on it for hours at a time. Some people here would like one and I'd prefer not to reinvent the wheel if you've already written such a tool. Please send me a note if you'd like to share a screen saver with us. Thanks! --Dwight Mckay, ECN Text & Workstation Software Support [arpanet: mckay@ee.ecn.purdue.edu, usenet: ...ihnp4!pur-ee!mckay] [Compu-serve: 75776,1521, office: EE 348B, phone: (317) 494-3561] ------------------------------ Date: Wed, 7 Jan 87 13:16:29 CST From: wucs1!wudma!chc@seismo.CSS.GOV (Charles Hampton Curley) Subject: RCS on Sun-3's? I'm trying to get RCS running on a Sun-3 and having a little difficulty. It hangs when it tries to determine the diffs between a new version being checked in. I really don't want to have to figure this out unless I have no other choice. I'm using the version of RCS right off the 4.2 BSD tape with a few changes to make it compile. I had to add stuff to the rdiff source for the sobuf stuff that doesn't seem to be part of the sun's libraries and I had to remove the -d2 option from the makefiles of rdiff and rdiff3 to stop it from getting the text overflow error upon link. Could someone send me the diffs or something like that via e-mail? Thanks, charles ------------------------------ End of SUN-Spots Digest ***********************