New accounts Query Could you remind me whom I contact about opening a new VAX account? Thank you so much. We would like to open account for our new researcher. Response Please contact Pat Bean, extension 7008. By the way, this information is listed in the LBL Computing Newsletter, under "Names and Numbers to Know", under User Resources. Macintosh/PC Macintosh/Versaterm/FOCUS on CSA Query Last week I used "focus" and "tkit" to order supplies via my Mac running VersaTerm emulation. The problem is that for the last few days each time I log onto CSA1 and type "focus" the screen continuously spits out " VAX/FOCUS: " line, after line in rapid fire. The only escape that I have found is to quit VersaTerm, which breaks the connection with CSA1. Is there a problem with > "focus", or maybe with my login.com file or the Response Called this person this AM. Checked as his login.com as he agreed. He uses "set term/page=255" -- so he gets the maximum page possible before programs like MAIL and Software Tools Mail ask him to press the space-bar. Why does he care? So that he can cut and paste from long mail messages without editing out the the "Press space-bar to continue" messages. He doesn't want to save the messages to a file via the functions available in these mail programs and then copy from there. When I telnet into CSA1 from a Macintosh and run focus with normal page settings -- the size of the screen -- it runs fine. When I try "set term/page=255", focus gives me endless "VAX/FOCUS" messages, as he reports. Told him this. He understood; will "set term/page=24" before running FOCUS. Kermit/CSA location Question Where on the CSA cluster disks is kermit.exe located? Response MICROTOOLS:kermit.exe X11/setting the display properly on the remote workstation Query I have a PC-Xview program from NCD. If I want to run Xapplications on ux5 and display them on my PC, what is the sequence of actions I should take on ux5 once I login? Response suppose your IP address is 128.3.xxx.xxx. Then do setenv DISPLAY 128.3.xxx.xxx:0.0 To check, try something like xdpyinfo UNIX/Harmonizing Sun and VMS keyboard mappings with X11 Query I'm trying to rlogin on csa from my SUN/IPX and use EDhT. I have the new type keyboard from SUN and I would like to know what kind of settings I need in order to emulate a vt100 keyboard. At the moment I can only use the arrows and none of the function keys like the select cut paste delete character etc. work. if I show terminal this is what I get: Csa2>sho term Terminal: _VTA5120: Device_Type: VT200_Series Owner: SOMEUSER Physical terminal: _NTY25: Username: SOMEUSER Remote Port Info: myws.lbl.gov [...] Input: 9600 LFfill: 0 Width: 80 Parity: None Output: 9600 Scope Remote Eightbit [...] Response Suggested that she try running these commands on myws, her Sun IPX. # Makes vt100 terminal emulation work much better: xmodmap /usr/local/lib/keymap.vt100.x11 & [xmodmap is a utility for modifying keymaps in X.] Worked fine: she had to hunt around a bit for the "gold" key, which turned out to be a function key, rather than a keypad key. Once she found that, she was quite happy with the setup for EDT. Helped her put the xmodmap command show above into her .xinitrc, so that it will be executed every time she starts up X11. She may still buy a X11-based package Darrell Davis told her about that emulates Vt200 and Vt300 similar terminals on Suns. Costs about $300. PostScript/VMS/ghostview does not preview a certain document properly Query If I'm not mistaken, the file manypages.ps on CSA is 164 copies of the SOMEPACKAGE manual title page. Something is clearly wrong, or ghostview doesn't work on it. try it out. Response: ghostview'ed (gv'ed) manypages.ps No problems. Definitely did NOT see 164 repetitions of the title page. You may recall ghostview -- but not ghostscript -- was confused when it was used to try to preview a PostScript file that was stored in fixed length, 512 byte records. Printing lpr/UNIX/ap1 is an unknown printer?!? Query: lpr on myws.lbl.gov reports ap1: unknown printer when I try to print to it. Response: This may be because the Sun 'lpr' was in your path before the LBL 'lpr'. I made both the same. [lpr now works OK on myws.lbl.gov] Printing/file format for SD1 -- Tektronix Phaser IIsd Query: What is the file format for sd1 printer. It is a color tektronix printer in the IO room. Does it take PostScript? Response: SD1 is a color PostScript printer; it uses "dye-sublimation" technology; we charge $4.25/page for output. Kodak printer/problems -- background and repairs Question: What does this mean? [Error message e-mailed to user after he submitted a job to the Kodak -- and it did not print.] /usr/local/kodak.release/kodak -f dfA040emu.lbl.gov -i sun KODAK ERROR: Unable to open /dev/xl7700. Response: This indicates a system level error that you, the user, can do absolutely nothing about. This message occurs when the driver is unable to access the printer. This situation typically occurs when the driver exits abnormally from a previous job. Due to the lack of cooperation from the company from whom we bought the GPIB interface board, certain error recovery procedures at the driver level were never implemented. The appropriate action for you is to forward the error message to trouble. The appropriate response from the ICSD staff is to kill any lingering kodak processes, then do an "ibclear" command on ux6 to reset the interface board. If the "ibclear" doesn't work, there is a printed procedure for resetting the interface board that is a bit more involved. I'll see if I can find the procedure (on a piece of paper somewhere in my office) and pass it around. The queues are currently all clear, and a number of jobs just came out, so things seem to be ok now. UNIX/Security Question: The following message just turned up on my console window, I never heard of the person, and attempts to finger him failed. user@myws:/jobs/ 16% Mar 17 13:20:26 myws login: LOGIN FAILURE ON ttyp7 FROM abc.def.xyz, somebody Response: This message is a result of increased logging I have turned on in the login program. It is most likely a mistake by the user and should be disregarded. You should only get alarmed if you get many of these errors in a row. And especially if the attempt is to log in as users like "guest", or "root" from a non-lbl node. VMS/reminder program Query I have a reminder program installed in my main directory, called REMINDER.EXE. It allows me to set reminder messages for future dates, and it warns me upon logging in. The problem is that, as I was using it today, it crashed when I issued the command "F" inside REMINDER ("F" is supposed to display all future reminder messages in effect). Upon crashing, it gave the following error message: Horrible error messages deleted] This never happened before. Could you look into it? Response As you say, this is your own program and not administered nor supported by the Computer Center. May we please recommend that you try the REMIND program that comes with the MULTINET suite? Although we have not used it much, we should be able to get support from the vendor (TGV). How to start it: (1) Comment out the "set command remind" you have in your login.com (2) Then, login in again and type REMIND. It has internal help, which uses the file MULTINET:remind.help You may well want to copy that file and study it a little. My tests of it worked fine; it sends mail or a message to the terminal or both. RN/UNIX/error on startup Query myws rn unix (Revising soft pointers--be patient.) Unread news in alt.bbs.unixbbs 5 articles Unread news in clari.nb.unix 7 articles Unread news in ca.unix 1 article Unread news in comp.std.unix 11 articles Unread news in comp.unix.admin 169 articles Too many lines in .newsrc [rn quit] [By the way, I don't have a .newsrc file in my home directory!!!] Response It is actually referring to your .Newsrc file. There is a hardwired upper limit for the number of newsgroups that rn can handle at once. The installed version should handle 4000. We are only at 3500 newsgroups now. Please check to see that there are less than 4000 lines in your .Newsrc file. MAIL/UNIX/aliases in the .mailrc Query Mail sent to an alias from the Sun UNIX workstation myws.lbl.gov bounces; I got mail like this From: MAILER-DAEMON (Mail Delivery Subsystem) Subject: Returned mail: Host unknown that contained numerous lines of the form 421 Host abc.def.ghi, not found for mailer tcpld. 550 ... Host unknown [returned message deleted] Response Believe the crux of the problem is that your mailing list was a comma-separated list and UNIX mail aliases in the .mailrc should be space-separated lists. Please note the following from "man mail", subsection BUGS: The format for the alias command is a space-separated list of recipients, while the format for an alias in either the .forward file or the /etc/aliases file is a comma-separated list. Therefore, try aliases like these in your .mailrc file: alias mygroup1 user1 user2 \ user3 \ user5 alias mygroup2 usera userb userc \ userd \ usere As you can see, you do NOT need to start the additional lines in column one. The '\' at the end of the lines that continue the alias must be followed IMMEDIATELY by a carriage return. The "alias" command by itself on the line should display the aliases you have setup. [User replaced ',' in the alias with a space, tried the new alias, and it worked.] INTERNET/rsh command from VMS to UNIX Query I am sending the command $ rsh/password=XXXXXX ux1 -t -me -F Helvetica abc > xyz from my VAXstation myvax.lbl.gov to ux1.lbl.gov to generate a PostScript file from a Troff file. I then use RCP to transfer the PostScript back to my workstation. The problem is that the rsh command is converted to lowercase and psroff does not recognize "helvetica" as a font. If I leave the font designation off or execute the command while on ux1, everything works fine. How can I prevent the case conversion by rsh? Response #1 Put the command in quotes: $ rsh ux1 "psroff -t -me -F Helvetica abc > xyz" Response #2 May I further suggest the following: Make a file, called .rhosts, in your ux1 home directory. Put this line in it: myvax.lbl.gov You do not need to specify your username -- someone -- because it is the same on both myvax and ux1/ux5. Then, run the command chmod 700 .rhosts so that the file is readable only by you. Then, you can omit the password on your rsh command. Text processing -- psroff, TeX, ... Psroff/troubles on ux5 Report: In the past I have used the psroff command in UNIX to capture the Postscript on to a file using the -t option, e.g. psroff -me -F Helvetica filein -t > fileout This no longer works for me. I get the error message " can't open file /usr/lib/me/eqn.me; line 16, file stack: EQ If I use psroff with the substitution -P hp50, then I get normal looking output. Apparently the equation processor gets lost when -t is specified. This is a serious problem for me. Response: Please note that psroff -me -F Helvetica filein -t > fileout now works on ux5. A directory (/usr/lib/me) was empty; part of the fallout from the disastrous disk crash on ux5 that happened about six weeks ago. Restored that directory and /usr/lib/ms, which was also empty. May I please point out that "man psroff" suggests using psroff -t -me -F Helvetica filein -t > fileout as in its synopsis: SYNOPSIS psroff [ -t ] [ troff options ] [ spool options ] [ files ] psroff -- OK on ux1, problems on ux5 Report: I am having trouble using psroff in ux5 to convert troff to Postscript and save the output to another file. The commands I use are psroff -t -me -F Helvetica filein > fileout This used to work for me six months ago. Now it fails when it encounters eqn definitions. Without the "> fileout" I get a normal output on the laserprinter. I need to do this to insert Postscript figures into text for a large book I am writing. Response You will have to do this on ux1. When we went to Sun4 SunOS 4.1.1 the ditroff (psroff) software broke. UCSF no longer maintains the software since the maintainer, Dr. Rick Rogers M.D. left UCSF and took a at position in Maryland. I communicated with Rick via e-mail and Rick said he would be willing to do one more ditroff release which would run under 4.1.x providing he could get Adobe Systems to agree to lend him the use of two programmers for one month. Last December Adobe agreed to this but have not committed the two programmers to-date due to other priorities within Adobe. I haven't been in touch with Rick since February so I will send Rick some mail and see what the status of a new distribution is. Conclusion: So, if you need to insert PostScript figures into your ditroff output, you will have to use ux1 (Sun3) in order for psroff to work. [Note: A major problem with psroff on ux5 is with specifying the PostScript font family with the -F flag, as in psroff -t -me -F Helvetica filein > fileout psroff on ux5 does much better if user does not try to change font families. TeX/RevTeX macro package: Query #1: Could somebody point me to RevTeX? Is there an online help file or an instruction manual in the user's office? Query #2: Please forget/delete/ignore my request for information about REV\TeX; I found it all on TEX_DISK:[TEX.APS_LATEX.V3_0]. It probably doesn't matter anyway, but the files TEX_DISK:[TEX.APS_LATEX.V3_0]manintro.dvi and TEX_DISK:[TEX.APS_LATEX.V3_0]manend.dvi appear to be empty--at least, 0 pages come out. The others print fine. Response: Improved the HELP article LaTeX (and "man latex") so you can find REVTEX much more easily. Fixed the font problem that caused the printing problems -- the "unprintable" DVI files called for a very unusual, nonstandard, size of a certain font (cmcsc10). FORTRAN/VMS/changes with VAX FORTRAN version 6.0 Report: I have recompiled a fortran program which ran fine on March 17 and got compilation warnings about uninitialized variables. No such warnings were given before. When I ran the newly compiled program several print statements, involving those variables, behaved badly. Apparently, an uninitialized integer is now undefined. The new version of the Fortran compiler is not compatible with the old one. Is this intentional? Response: You should fix the code to initialize the uninitialized integer. Please read Section 1.4.15 of the Fortran Release notes. $ type SYS$HELP:FORT060.RELEASE_NOTES FORTRAN/VMS/Version 6.0 Query I am having trouble compiling fortran code on CSA. Code which worked fine two weeks ago now gives images which I cannot link on my VAXstation. Is this a result of the new version of fortran on CSA? What can I do to make things worK? Response #1 Your VAXstation must not be running Fortran V6. In order to link the object code compiled on CSA on your VAXstation, you have to use the V5 Fortran compiler on CSA. Do the following to use the V5 VMS FORTRAN on CSA: $ define/job FORTRAN$MAIN FORTRAN In order to link an image on CSA that will run on your VAXstation which does have the enhanced RTL images, you need to do the following: $ define mthrtl fortran$mthrtl-vms $ define vmthrtl fortran$vmthrtl-vms Please read the Fortran release notes in SYS$HELP for more details. Response #2 If your VAXstation doesn't have a Fortran license, you will not get these new RTLs until you upgrade to OpenVMS V6. So you'll have to continue to use the Version 5 Fortran compiler on CSA. We will not remove the V5 compiler unless we are certain that it is no longer needed. Fortran V6 has been installed on all of the VAXstations that we manage. UNIX/Programming FORTRAN/UNIX/puzzling errors Query I have been having some very weird things happening to a code I have run at least 50 times before. I tried the following piece of code which did NOT print 'success' when run. Am I missing something stupid or what? real x,y x=1. y=1. if(x.eq.y)then write(*,*) 'success' write(*,*) 'success' write(*,*) 'success' write(*,*) 'success' endif end Response Made the code shown above into a ten line program. Compiled it; ran it on ux5, ux6, myws, and csr; it ran fine in all places: printed success success success success Are you using /usr/lang/f77? On what machine are you doing the compilation and using what command line? If these ten lines are part of a much bigger code, please use f77 -C Note that "man f77" says -C Compile code to check that subscripts are within the declared array bounds. This option frequently makes the source of errors much clearer. [In fact, the user recompiled his large code with the -C option, reran it, and got a meaningful run-time error message that pinpointed the error. Of course, the simple test code fragment was just fine: the code was corrupting some part of itself at run-time it when it was built without the -C option.] UNIX/dbx problem on csr Query Since a few months a ago, I started having problems using dbx to locate segmentation fault positions in my code. After even compiling a small program that triggers a segmentation fault, I usually get a 8M byte core dump file, and when I try to run "dbx" with the core file, I get the following message: dbx m Reading symbolic information... Read 142 symbols warning: core file read error: address not in data space warning: core file read error: address not in data space warning: core file read error: address not in data space warning: core file read error: address not in data space attempt to read stack failed - bad frame pointer (dbx) quit Response Believe you are still experiencing the problem with reported some time ago [April 29, 1992] with dbx on the Sun 4...MP machines like csr. Please note the following: Date: Wed, 29 Apr 92 11:46:05 PDT Today, (4-29-92), Sun Languages Support (Darlene Batease), confirmed that there is indeed a bug with dbx on the Sun 4...MP machines like csr (which is a 4/690MP). The bug is NOT with SunOS 4.1.2 per se, because dbx works fine on Sparc 2 workstations with the same OS - such as pi4. BUG: we can NOT read core files generated on Sun 4...MP machines with dbx; no fix yet. dbx execfile works fine on Sun 4...MP machines, such as csr; but dbx execfile core will indeed fail. UNIX/Makefile syntax I am having some small problems with UNIX Makefiles. I would like the system to compile (with 'acc') if EITHER one of the object module changes OR the include file, spline_inc.h, changes. The following file ignores changes in the include file. I also understand there is a more elegant way of specifying the list of files so that either CSOURCE is given or OBJECTS but not both. Can you please indicate how this should be done? Also, is there a decent reference for this sort of thing? The 'man' file for 'make' isn't too helpful. ================================================================================ # Makefile for library: CSOURCE = spline_add.c spline_bas.c spline_cur.c spline_dot.c spline_end.c OBJECTS = spline_add.o spline_bas.o spline_cur.o spline_dot.o spline_end.o CFLAGS = -c spline.a: $(OBJECTS) /bin/rm spline.a ar cr spline.a $(OBJECTS) ranlib spline.a .c.o: $(CSOURCE) spline_inc.h acc $(CFLAGS) $? %%%%%%%%%%%%%%%%% Simple-minded changes I'd suggest: (1) Remove the line "CSOURCE =" altogether (2) After the line CFLAGS = -c add the line CC = acc (3) Replace the last two lines by $(OBJS): spline_inc.h (4) Please also read "man makedepend" and run it. For fuller documentation on "make", see the Sun documentation set, available via the Help Desk, extension 5981. UNIX/Multiple versions of gcc (GNU cc compiler) Query I received the messge below from one of our users concerning "gcc". After poking around I noticed that there is one version, 1.37.1, of "gcc" in "/usr/local/bin" on "csr" (and our workstations) and other version, 2.3.1, of "gcc" in "/usr/local/GNU/bin". Thus, if you don't load the module "gnu" OR if you load the module "local" AFTER "gnu" you get the OLD version. Should this old version be hanging around? It seems too easy to get by mistake (and the man page you get is for version 2.0 or greater). Response Spoke to the appropriate person in our UNIX project about this matter. We are going to remove /usr/local/bin/gcc, so that gcc will be available ONLY after "module load GNU". However, we can NOT do that today, because we need to give adequate notice. Ergo, please make sure to "module load GNU" AFTER loading "local". [There will be a newsletter article, in fact in this issue, explaining this change.]