12 September 1996. Installed Debian Linux 1.1 base system, and created custom boot floppy. Created user odonnell. Booted successfully from hard disk and custom boot diskette. Installed packages: ppp, netbase. netstd failed for lack of cpp. 13 September 1996. Succeeded in connecting by ppp to EarthLink in Iowa City. Created the following files: /etc/ppp/chatscript.avalon, /etc/ppp/chatscript.earthlink.ic, /etc/ppp/options.cua0, /etc/resolv.conf. These files are done crudely. Previous versions are backed up in a Save directory. They need to be reconsidered carefully, and updated with RCS. Also need to set up useful ppp-on, ppp-off scripts. Changed mode of /usr/sbin/pppd to suid, so I can invoke it without being root. Tried to install standard packages using dselect acessing by ftp to sunsite.unc.edu. Made a hash of it. Apparently purged dpkg-ftp, and cannot get dselect to use ftp access any more. 14 September 1996. Updated dpkg to 1.2.11elf. No visible effect. Copied in emacs package by ftp, and used dpkg --install. Seemed to be the right procedure, but installation did not complete due to a dependence on elf-x11r6lib, which I can't find. Tried ftp.debian.org instead of sunsite.unc.edu, with much better results. Apparently the mirror is far from precise/up-to-date. Copied in dpkg-ftp and several perl packages by ftp from ftp.debian.org. dpkg --install complained, but I was able to use ftp through dselect again. Updated all packages in base section from ftp.debian.org, using dselect (assuming that I got it right). A number of X11 packages came in too, and I don't understand why. I did not configure xserver-vga, because I wasn't sure about the parameters of my video processor. Installed all packages with priority standard. Still haven't configured xserver-vga. There were a number of other configuration errors, which I don't understand in detail. QUERY: what are the new priorities, such as required and important? 15 September 1996. Trying to configure X. According to the README.Config, failure to open the mouse device is an OS configuration error. No clue yet what the error is, nor how to fix it. The test of the server always gives a bizarre display of diagonal lines, so I probably have the video configuration wrong, too. Giving up on the configuration script, and editing the configurations file by hand. 21 September 1996. Installed special kernel number 5 from ftp.debian.org. It includes psaux.o. X works now, with a rather primitive use of the screen. 22 September 1996. Switched to the S3 accelerated X server. Succeeded in displaying 1024x768 resolution. 23 September 1996. Switched pppd configuration to use Avalon instead of EarthLink. Created shell script /usr/local/bin/ppp-on to start up ppp. 29 September 1996. Found that epic, eepic, eepicemu are not available in the TeX installation. I seem to have all of the TeX packages from the ftp.debian.org. So, I copied cs.uchicago.edu:/usr/local/lib/texmf/tex/inputs/latex209/eepic-dir/* to /usr/lib/texmf/tex/inputs/latex/eepic. I also copied cs.uchicago.edu:/usr/local/lib/texmf/tex/inputs/latex209/transfig.sty to /usr/lib/texmf/tex/inputs/latex. I need to copy and systematize all of the TeX materials at cs.uchicago.edu. 1 October 1996. Installed non-Debian software for manipulating BIBTeX materials from ftp.tex.ac.uk:/tex-archive/bibliography/bibtex/utils/bibextract, bibclean, bibsort bibextract wanted deroff and checksum, but appears to have provided citetags (which I need for Overton's bct), anyway. 4-5 October 1996. Installed non-Debian ssh (secure shell) from ftp.net.ohio-state.edu/pub/security/ssh/ssh-1.2.14.tar.gz.sig. During the compile by make, discovered missing X libraries needed by ssh-askpass.c. Found much of the missing material in the Debian packages xdevel and xslib, but still cannot open -lX11 to compile ssh-askpass. Installed all but ssh-askpass by make -i install. 5 October 1996. Added -L/usr/X11R6/lib to gcc options for ssh-askpass.c in the Makefile. This leads to successful making of ssh-askpass, but I'd like to understand how to get it to configure right in the normal way, creating the Makefile with ./configure. 16 October 1996. Installed amaya, an alpha-test WWW browser/editor. It doesn't work. 16 October 1996. Installed GNUscape w3, a WWW browser implemented in elisp to work under emacs. It works, but it's not clear that it will be very useful. It recommends xemacs instead of emacs v. 19 for full power. But, xemacs is large, and depends on either Motif or Athena. I haven't figured out how it fits in, so I've left it alone for now. 16 October 1996. Downloaded ctwm, but the make failed for lack of xpm.h. The xpm package appears to be installed. Decided to use fvwm instead. SEE 21 October 1996. 18 October 1996. Installed TkNet, a GUI front end for PPP. Doesn't work, apparently because of reliance on a missing Adobe font. 21 October 1996. Deleted source for ctwm. I'm pretty sure nothing was installed. 21 October 1996. Fixed font specifications in TkNet, but it still doesn't work. Windows come up, but it's incredibly flakey, and the help pages don't appear. 21 October 1996. Tried to install icl calendar system. Failed to find tcl.h, which is not included in the tcl74 package. What's up? 22 October 1996. Succeeded in installing icl calendar system, after finding tcl.h, tk.h in the tcl74-dv, tk40-dev packages. It sort of works, shows a calendar, gives alarms, but a lot of features crash, apparently due to looking for Adobe fonts. 22 October 1996. Installed dotfile, a GUI interface to configuration files. It looks handy on the whole, BUT: a lot of features fail, apparently due to looking for Adobe fonts. Also, I will probably have to re-engineer my existing configurations to get started. Perhaps worth it if the font problem is solved. 22 October 1996. Installed tkgoodstuff. It wants tk version 4.1. I intend to load that, as well as tc version 7.5, which are in Debian packages. 25 October 1996. Loaded a Debian package for ical, and deleted the hand-loaded version from /usr/local/src. Key parts still fail due to font problems. TO DO: End use of /dev/cua0 by ppp. There is some correspondence in the Debian newsgroup regarding why, and I hope how, to do this. TO DO: Get the Sound Blaster working. This probably requires compiling a custom kernel. 25-26 October 1996. Compiled a custom version of the 2.0.6 kernel, in /boot/vmlinuz-2.0.6.custom.1, including Sound Blaster support. When I try to boot it, I get an apparently infinite loop with weird diagonal lines on the monitor, periodically blanking out to black. Using the lilo boot image selector, I was able to restart with /boot/vmlinuz-2.0.5. 2.0.0 (which I could swear was running just before, but there is an automatically created /vmlinuz.old pointing to /boot/vmlinuz-2.0.5, so I guess I was running that before) and 2.0.6 seem to have the same problem. Could it be the new startup modules in /etc/modules? Under the newly booted 2.0.5, lsmod reports Module: #pages: Used by: slip 2 1 (autoclean) slhc 2 [slip] 1 (autoclean) psaux 1 1 serial 7 0 cdrom 1 0 This seems strange, since /etc/modules is # /etc/modules: kernel modules to load at boot time. # # This file should contain the names of kernel modules that are # to be loaded at boot time, one per line. Comments begin with # a `#', and everything on the line after them are ignored. # An entry named `auto' will cause the system to start kerneld immediately. # Kerneld loads modules on demand. #auto cdrom serial psaux sound Maybe I compiled in the module startup somehow? For more reference, here is /etc/lilo.conf: boot=/dev/hda2 root=/dev/hda2 compact install=/boot/boot.b map=/boot/map vga=normal delay=20 image=/vmlinuz label=Linux image=/boot/vmlinuz-2.0.0 image=/boot/vmlinuz-2.0.5 image=/boot/vmlinuz-2.0.6 image=/boot/vmlinuz-2.0.6.custom.1 label=2.0.6.custom.1 read-only 27 October 1996. I relinked /vmlinuz to kernel 2.0.5, and arranged for /vmlinuz.old and /vmlinuz.future to be available from lilo. 27 October 1996. Somewhere, I think I started up diald, which starts slip, which captures something that messes up pppd. This is probably tied to the problems of /dev/cua0 somehow. I need to look through the files installed by the diald module to figure out how to deconfigure it until I'm ready for it. 27 October 1996. My method of setting different /etc/resolv.conf files for different Internet access providers doesn't work. Even when I link the file in a script owned by root and marked suid, it refuses permission. I'm fetching the "super" package to see if it have a better approach. Later, I succeeded in running the script properly using super. 28 October 1996. Installed a few new packages, and updated a *lot* (~40 MB) from "unstable". The "diald" installation gave the following advice: -------------------------- (0) IMPORTANT Users have reported that using the diald/pppd combination on /dev/cua? devices causes problems. Use a /dev/ttyS? for diald. Let me repeat this: the man page shows several examples with /dev/cua1. THESE WILL NOT WORK (pppd will complain about the device being busy). Use /dev/ttyS? devices instead. Read /usr/doc/diald/debian.device for a possible explanation. (1) Check the script /etc/init.d/diald. This is configured for a modem on /dev/ttyS0 and it should be changed to match your set up. (2) Copy /usr/doc/diald/diald.options to /etc/diald.options. Edit its content to match your set up. Read the diald(8) man page and the FAQ in /usr/doc/diald for the complete list of available options. (3) After, and only after, you have completed steps (1) and (2) will diald have a chance to work correctly. Now either reboot, or start diald by hand by typing `/etc/init.d/diald stop' followed by `/etc/init.d/diald start'. (4) If anything goes wrong keep in mind that if you remove /etc/diald.options and reboot, diald will not be activated. This should allow you to go back to your original (presumably working ;) configuration and start again. (5) It would be VERY nice if there was a proper configuration script. The mantainer only knows about PPP links with fixed IP, and cannot be trusted write a general tool. The next best thing would be for the users to send him details of their succesful installations (what type of connection, fixed/dynamic IP, diald.options, scripts, tips) so that a collection of sample configurations could be distributed. (6) (0) through (6) can be found in /usr/doc/diald/debian.diald. -------------------------- 29 October 1996. Although I installed the fileutils package, and the xterm-color package, ls --color gives me boldface, but no color. 29 October 1996. I am doing a comparison test of Internet Navigator, Earthlink, and Avalon, using my testnet script based on a suggestion by Chris Barnard. The script and results are in /System_administration/network_test. 30 October 1996. Moved /etc/diald to /etc/Save/diald, and unlinked with the command "update-rc.d diald remove". This should prevent diald from starting up at boot time, but should be easy to reverse after I've switched from /dev/cua0 to /dev/*tty*, and am ready to try diald. 30 October 1996. Recompiled the custom 2.0.6 kernel, using the instructions in debian.README instead of those in README. This creates a kernel_image package, which installs the kernel image. although the instructions insisted on setting a custom version string in debian.rules, the package still installs to 2.0.6, overwriting any older version. That's not good. Perhaps I need to chance the version setting in the Makefile. Also, the kernel doesn't work---same symptoms as before. Perhaps the video card is incorrectly configured, or perhaps the mouse driver is broken. 30 October 1996. When I rebooted, I lost all of the nice *.xpm pixmaps that I was using for fvwm2. I suspect that they were reorganized from the fvwm2 package to fvwm-common, which I will fetch and see. Later: after installing fvwm-common, I have the *.xpm pixmaps again. 30 October 1996. From looking at /var/log/daemon.log, it appears that the custom kernel 2.0.6 is indeed failing due to psaux.o. The complaint is Oct 30 14:26:58 Powdermilk modprobe: No dependancy information for module /lib/modules/2.0.6/misc/psaux.o The place where the dependency information belongs is apparently /lib/modules/2.06/modules.dep. I ran "depmod -a 2.0.6" and it created a new (and much shorter) modules.dep, which contains a mention of psaux.o, with no dependency (the old version had no mention). Alas, I let the old version of modules.dep get away when depmod updated it. 30 October 1996. "ls /dev" produces weird, unreadable, results. Probably a bug in the color option, or configuration. 30 October 1996. I can now run my custom-compiled 2.0.6 kernel. I believe that configuring the dependency for the psaux module in /lib/modules/2.0.6 was the key. But, sound still does not work. lsmod reports Module: #pages: Used by: psaux 1 1 serial 7 0 auinfo reports auinfo: unable to connect to audio server so, apparently the audio server is not running. Hmm, I thought that the sound module was up in the lsmod report earlier. Have to solve a ppp problem first. OK, interleaving work on the sound problem, while waiting for Internet performance tests to run. I was able to start the sound module with insmod. When I try to start the nas server with au, I get the message: Fatal server error: could not create audio connection block info There doesn't appear to be a world-writable audio device: crw-rw---- 1 root audio 14, 4 Oct 21 22:47 /dev/audio crw-rw---- 1 root audio 14, 20 Oct 21 22:47 /dev/audio1 Even as root, I cannot write to /dev/audio: root@Powdermilk:/usr/games/xgalaga/sounds# cat shield.raw | /dev/audio bash: /dev/audio: Permission denied Broken pipe Here's the console message when I attempt au as root: Oct 30 22:12:13 Powdermilk modprobe: Can't locate module char-major-13 Immediately after system startup, a similar problem occurs with 14: Oct 30 18:37:43 Powdermilk modprobe: Can't locate module char-major-14 These dependencies are not in /lib/modules/2.0.6/modules.dep. Here's a diagnostic recommended by the Readme: root@Powdermilk:/usr/src/kernel-source-2.0.6/drivers/sound# cat /dev/sndstat Sound Driver:3.5.4-960630 (Wed Oct 30 13:24:32 CST 1996 root, Linux Powdermilk 2.0.5 #1 Thu Jul 11 21:17:05 PDT 1996 i586) Kernel: Linux Powdermilk 2.0.6 #2 Wed Oct 30 13:27:11 CST 1996 i586 Config options: 0 Installed drivers: Type 1: OPL-2/OPL-3 FM Type 2: Sound Blaster Type 7: SB MPU-401 Card config: (Sound Blaster at 0x220 irq 5 drq 1,5) (SB MPU-401 at 0x330 irq 1 drq 0) (OPL-2/OPL-3 FM at 0x388 drq 0) Audio devices: Synth devices: Midi devices: Timers: 0: System clock Mixers: What does it mean? 30 October 1996. I had /usr/local/bin/ppp-on all screwed up. It was calling EarthLink in all cases, even when I thought I had selected Avalon or Internet Navigator. The problem was /etc/ppp/options.cua0 executing after the command line options, and cancelling them. I fixed it, by removing the setting of provider in options.cua0. The whole structure needs rethinking sometime. I now discover that I can't connect to Avalon: probably never have. The console reports: Oct 30 21:10:38 Powdermilk pppd[918]: Serial connection established. Oct 30 21:10:39 Powdermilk pppd[918]: Using interface ppp0 Oct 30 21:10:39 Powdermilk pppd[918]: Connect: ppp0 <--> /dev/cua0 Oct 30 21:11:09 Powdermilk pppd[918]: LCP: timeout sending Config-Requests Oct 30 21:11:09 Powdermilk pppd[918]: Connection terminated. Oct 30 21:11:09 Powdermilk pppd[918]: Receive serial link is not 8-bit clean: Oct 30 21:11:09 Powdermilk pppd[918]: Problem: all had bit 7 set to 0 Oct 30 21:11:09 Powdermilk pppd[918]: Exit. Oct 30 21:12:27 Powdermilk kernel: PPP: ppp line discipline successfully unregistered Or, perhaps it was just a bad connection. Further attempts through Avalon led to busy signals. 30 October 1996. dselect is broken. Perhaps I installed a broken package function. Here's the comment when I try to UPDATE: Can't locate Net/FTP.pm in @INC at /usr/lib/dpkg/methods/ftp/update line 7. BEGIN failed--compilation aborted at /usr/lib/dpkg/methods/ftp/update line 7. update available list script returned error exit status 2. 31 October 1996. I recompiled kernel 2.0.6 overnight, and installed it as vmlinuz-2.06.custom.2. I recorded a lot of worries in Debian_advice/30, as comments on the kernel compiling instructions. The new kernel failed again, although /lib/modules/2.0.6/modules.dep appears to cover psaux.o appropriately. I screwed up the file system by toggling power, and had to do a brutal rearrangement under fsck, which I didn't understand at all. I *must* learn the gentler way to interrupt before doing another experimental boot. 31 October 1996. I have 3 immediate tasks: 1) find the right way to interrupt a bad system; 2) fix dselect (try rebooting first, then reload relevant packages by hand); 3) compile a kernel to drive the Sound Blaster. 31 October 1996. Regarding task 1 above, it seems to be determined by the following line in /ec/inittab: # What to do when CTRL-ALT-DEL is pressed. ca:12345:ctrlaltdel:/sbin/shutdown -t1 -r now Well, I tried CTRL-ALT-DEL in the login window, and it had no impact. OK, it appears that CTRL-ALT-F1 takes me to single user mode, even from X. CTRL-ALT-DEL is trapped by X, but causes a reboot during single-user mode, and probably other times as well. 1 November 1996. Regarding the breaking of dselect, somehow, I deinstalled dpkg-ftp. I am reinstalling it. That was the problem: dselect is now working again. 2 November 1996. dselect appears to work, except that the INSTALL step hung up installing gpm (presumably because it tried to start gpm, and the mouse was already taken by X). I killed the process, and then it hung again at the end of INSTALL, and again trying to CONFIGURE gpm. This is probably a fault of the gpm package, and not of dselect. I probably should deinstall gpm entirely. 2 November 1996. I am recompiling kernel 2.0.6 with sound support, following the instructions in ~odonnell/System_administration/Debian_advice/howtow_recompile rather closely. There were steps that appear to be crucial that I missed before. But, I am perturbed that depmod -a reports errors: root@Powdermilk:/lib/modules/2.0.6# depmod -a *** Unresolved symbols in module /lib/modules/2.0.6/fs/fat.o *** Unresolved symbols in module /lib/modules/2.0.6/fs/isofs.o *** Unresolved symbols in module /lib/modules/2.0.6/fs/minix.o *** Unresolved symbols in module /lib/modules/2.0.6/fs/msdos.o *** Unresolved symbols in module /lib/modules/2.0.6/fs/nfs.o *** Unresolved symbols in module /lib/modules/2.0.6/fs/vfat.o *** Unresolved symbols in module /lib/modules/2.0.6/misc/lp.o *** Unresolved symbols in module /lib/modules/2.0.6/misc/serial.o *** Unresolved symbols in module /lib/modules/2.0.6/misc/sound.o *** Unresolved symbols in module /lib/modules/2.0.6/net/dummy.o *** Unresolved symbols in module /lib/modules/2.0.6/net/slhc.o *** Unresolved symbols in module /lib/modules/2.0.6/net/slip.o Also, /lib/modules/2.0.6/modules.dep still fails to mention psaux.o. I will try the system without any fix, but I'm sure that it will fail. Well, I didn't even get that far. lilo reports: root@Powdermilk:/etc# lilo Added Current_Linux * Added Previous_Linux Kernel /vmlinuz.test is too big 2 November 1996. OK, I am again compiling a kernel that runs X, but not sound. The "too big" problem was stupid: I had installed the executable vmlinux instead of the zImage. Stangely, /lib/modules/2.0.6/modules.dep now knows psaux.o, but considers it dependent on misc.o. A clue regarding sound: root@Powdermilk:/home/odonnell# insmod sound sound_unload_lowlevel_drivers undefined sound_init_lowlevel_drivers undefined Loading failed! The module symbols (from linux-2.0.6) don't match your linux-2.0.6 So, I will try configuring in the low level drivers. Well, they were configured in, so I will try leaving them out. 2 November 1996. OK, I'm running a kernel (custom.5) with no low-level sound drivers. insmod and modprobe install sound.o, but I can't get sound to the speakers. /dev/audio and /dev/audio1 are unwriteable, which is a bad sign. When I try to start the audio server, I get root@Powdermilk:/usr/games/xgalaga/sounds# au Fatal server error: could not create audio connection block info /usr/src/kernel-source-2.0.6/drivers/sound/Readme has some advice: root@Powdermilk:/usr/games/xgalaga/sounds# cat /dev/sndstat Sound Driver:3.5.4-960630 (Sat Nov 2 20:35:37 CST 1996 root, Linux Powdermilk 2.0.6 #3 Sat Nov 2 19:14:11 CST 1996 i586) Kernel: Linux Powdermilk 2.0.6 #4 Sat Nov 2 20:38:33 CST 1996 i586 Config options: 0 Installed drivers: Type 1: OPL-2/OPL-3 FM Type 2: Sound Blaster Type 7: SB MPU-401 Card config: (Sound Blaster at 0x220 irq 5 drq 1,5) (SB MPU-401 irq 1 drq 0) (OPL-2/OPL-3 FM at 0x388 drq 0) Audio devices: Synth devices: Midi devices: Timers: 0: System clock Mixers: /dev/dsp is somehow relevant: root@Powdermilk:/usr/games/xgalaga/sounds# /bin/ls -l /dev/dsp crw-rw---- 1 root audio 14, 3 Oct 21 22:47 /dev/dsp I tried "make config" instead of "make xconfig", to see if there were any options that failed to get into the X interface. It appears that there are not. I set the MPU-401 interface to 330, which is what the SB16 manual seems to call for. It had been 0. The other obvious thing to try is to set the I/O base to 200 (which the manual gives as the "Game Port"), instead of 220 ("Audio Interface" in the manual). I got /dev/audio etc. to be writable by running a script at the end of kernel-source-*/drivers/sound/Reame.linux. Still can't cat to the devices, nor start tha audio server au. 3 November 1996. I am pretty sure that the problem with sound is in the installation/configuration of the SB card. It's pretty clear that the card is not being detected at all by the system. Some configuration software came with the card. I'm trying to run it under dosemu/freedos, with no success. Perhaps I should try booting freedos directly from the floppy. Will I risk data on the hard drive? DOS should be able to run entirely on the floppy. 20 November 1996. On 18 November, I copied over a lot of updates from the ftp server, where a new version had just been stabilized. The installation of some part wanted to shut down xdm, so I cancelled that and waited until later when I was willing to sacrifice the fvwm2 state. On 19 November, I went back to dselect, and finished the installation. The shutting down of xdm killed all of my processes, including dselect, and put me into single-user mode (I think), or at least some sort of non-X console mode. I restarted dselect, and finished the installation. dselect seems to be pretty good about keeping its own state, and picking up where it left off. Then, I rebooted, and got stuck in an infinite loop with a login prompt on a tty (non-X) console, but no response to typing of characters, and the whole display flashing. Somewhat similar to the flashing in my earlier X failures, but without the diagonal stripes of color that I saw before. I could get a response only to CTL-ALT-DEL, which sent me around the reboot cycle, ending in the same flashing state. I tried different kernels, including the ancient simple one that I kept on a diskette, with the same effect. So, the problem was evidently in some sort of configuration outside the kernel. I learned that the option "S", with no hyphen, to a boot image at the LILO prompt takes the system to single-user mode. Did that, and hunted through configuration files. Several had changed (is there a way to make package installation use rcs?). I remembered accidentally allowing the installation of an alternative Xserver to make it the default, and suspected that this change was behind the failure. I found that /etc/X11/Xserver was changed on 19 November, and that it contained the initial line /usr/X11R6/bin/XF86_VGA16, which had formerly been /usr/X11R6/bin/XF86_S3. I'm glad that I logged this particular file through RCS. There was also an Xserver.old, but it appears that it gets wiped out by the second change. I restored the old XF86_S3, and now I am Xing again. Mysteriously, color ls, which did not work before, now works. 22 November 1996. Recently (21 November?) I installed LINK from http://dimacs.rutgers.edu/Projects/LINK.html into /usr/local/src/link. It didn't work, for want of STk (Scheme-Tk). I am now installing STk version 3.0 into /usr/local/src/STk-3.0 from ftp://cs.nyu.edu/pub/local/fox/stk. There is a later version availabe (3.1?), but LINK says that it requires version 3.0. 22 November 1996. Someone on the Debian news group recommended a free Quicken replacement, which is supposed to be able to import old Quicken files. It is reported to be available at http://www2.me.umn.edu/~clolsen/cbb/cbb.html, but another user reported that the URL is wrong. 22 November 1996. LINK is a flakey installation. I have edited several of the source scripts to fix path problems. I also created /usr/local/lib/stk/3.0b2/include, and filled it with links to the *.h files in /usr/local/src/STk-3.0/Src/*.h, /usr/local/src/STk-3.0/Mp/*.h, and /usr/local/src/STk-3.0/Tk/*.h to solve another path problem. UNDO THIS LATER. The INSTALL document doesn't quite match the state of the files, but I think I muddled through. It also created /root/STk_inst (yes, this is created by LINK rather than by STk), which should be eliminated later. The right way to fix the first path problem would be to use the same shell that the scripts use, and set $NEWLINKBASE there, instead of editing each file that refers to it. Yet another path problem: I symbolically linked /usr/local/lib/stk/3.0b2/Linux-2.0.5-ix86 to /usr/local/src/STk-3.0/Tcl/*.a, /usr/local/src/STk-3.0/Tk/*.a, /usr/local/src/STk-3.0/Mp/*.a, and /usr/local/src/STk-3.0/Src. I probably could have handled this one by setting Config.mk properly. 24 November 1996. I changed /etc/ppp/chatscript.internet_navigator, eliminating the spaces expected after "ogin:" and "assword:". I think that this avoids some of the failures that I used to get when connecting. 24 November 1996. I solved the missing font problem. There were not enough font paths listed in /etc/X11/XF86Config. 27 November 1996. In order to format the documentation for STk, I need some additional LaTeX styles. I copied the contents of cs.uchicago.edu:/usr/local/lib/texmf/tex/inputs/latex/contrib/other/misc to /usr/lib/texmf/tex/inputs/latex/local_stuff/contrib/other/misc in order to get a4wide.sty. I also copied the contents of /usr/local/lib/texmf/tex/inputs/latex/contrib/other/seminar/inputs/fancybox.sty to /usr/lib/texmf/tex/inputs/latex/local_stuff/contrib/other/seminar/inputs in order to get fancybox.sty. Finally, I copied /usr/local/lib/texmf/tex/inputs/latex209/calc.sty and psfig.sty to /usr/lib/texmf/tex/inputs/local_stuff/latex209 I had to link psfig.sty to psfig.tex, since one of the programs wanted it under that name. 28 November 1996. After a system update with dselect, I found that my ppp startup didn't work. The permission on /usr/sbin/pppd had been set executable only by root. I fixed this, but then running ppp-on from the command line as odonnell, I got: odonnell@Powdermilk:~$ ppp-on I Connect PPP through Internet Navigator, North Liberty IA Sorry - this system lacks PPP kernel support I was able to run it as root, though. I had installed, and then de-installed, diald. This may explain the changed permission, or there may have been an update of ppp itself. Perhaps the problem is the suid. Yup. I changed /usr/sbin/pppd to suid, and was able to fire it up again as odonnell. Highly misleading error message! 14 January 1997. I went through the problem with permissions on pppd again. The fact that every update to ppp screws this up suggests that I am not using ppp in the intended way. I should look into this along with switching away from cu. 14 January 1997. Somehow, the latest update broke xclock. It cannot resolve certain symbols. Presumably, xclock needs to be recompiled to match some change in a library. .xinitrc is not in ~odonnell. Did I remove it for some reason? 26 January 1997. After a routine update with dselect, I decided to run the "Remove unwanted software" step. It started removing lots of wanted stuff. I noticed xfig at first. It was just removing xbase when I hit lots of ^Cs and stopped it. The system was left in a state where I could only bring up single-user mode. I reloaded everything with dselect in single-user mode, and did *not* do the "Remove" section this time. Things seem to work, and xclock is back. I'm not sure what led to the gross amount of removal. A conflict between xlibraries and xlib6 may have been at the bottom of it. Furthermore, since I had to reinstall xbase, dselect restarted xdm, which gave me the X login display, but left dselect running, with locks set. I had to kill the running dselect, remove the locks, and run dselect again. 13 March 1997. Yesterday, performance was extremely sluggish in emacs. I was waiting a second for the mouse to select a bit of text, and most other operations were also slow. So, I rebooted. LILO hung with "LI". I used the "Custom Boot" diskette, which booted, but went into a loop flashing the screen with diagonally distorted noise. I remembered eventually that the "S" boot option gives single-user mode. I got on as root in single-user mode booting from the diskette, and ran "lilo". Then, I was able to boot from the hard disk, but I got to the same noisy flashing loop. I found a note from 20 November describing a similar problem that resulted from rewriting /etc/X11/Xserver. So, I checked /etc/X11/Xserver, but it was OK, and unchanged since 20 November. That's where I'm stuck now. The same failure occurs when I run "startx", so the problem is not entirely with "xdm". I found a reference to /usr/X11R6/bin/X instead of /usr/X11R6/bin/XF86_S3 in /etc/X11/xdm/Xservers, which had been changed on 26 January 1997. All other files in the directory were from 7 May 1996.I changed the reference, with no effect. There may be some problem with modules. I tried to start ppp, and found that there is no kernel support. There is a strange new modules package from the unstable directory, introducing something new called "modutils". To wash out any problems that might have been introduced by stopping at single-user mode, I moved /etc/init.d/xdm into Save to disable, and I rebooted to multi-user mode. I still get the message Sorry --- this system lacks PPP kernel support when I run ppp-on. I think that the X problem may have to do with the psaux mouse module being missing. I am unable to find manual pages for module stuff. "lsmod" isn't found any more. It is mentioned in the note for 25-26 October 1996. I can't find "insmod" or "modprobe" either. 14 March 1997. The problem was indeed the modules package. I brought a copy of the old modules package (modules_2.0.0-15.deb) from the stable distribution. The problematic package was modules_2.1.23-1.deb from unstable. ftp.debian.org still has the bad version as the standard (although they may have fixed the dependency so that it won't install until modutils is there). I couldn't find a modutils package anywhere. Anyway, I reinstalled the old modules package. ppp and xwindows worked even without rebooting. I rebooted after checking things out a bit, and rebooting worked fine. The funny little arrow icons on the upper right corner of the fvwm windows seem to be slightly degraded in visual quality, but perhaps that's an illusion. 14 March 1997. I also couldn't start Netscape. The netscape package is still there. It won't configure, because it doesn't find the netscape distribution file in /tmp. I fetched the distribution file by ftp from ftp.netscape.com, and reconfigured. That worked. Since I have now seen two packages that apparently trashed old material while failing to get the new, I wonder whether some change to dselect or dpkg has caused the problem. 30 March 1997. Updated with dselect. modules_2.1.23-1.deb from unstable wiped out my modules functions again, but I immediately reinstalled modules_2.0.0-15.deb. 1 April 1997. Updated with dselect. modutils is there now, but it did not install successfully. It wanted to deinstall modules, but modules refused to deinstall because it was used by the current kernel. Perhaps the new "dummy" modules package is supposed to solve this. But, while the new modules package recommends modutils, modutils claims to conflict with modules. 1 April 1997. The latest update moved xdvi. The already running shells are unable to find it, but presumably new shells will succeed. The new xdvi is using huge fonts, even at the smallest magnification. 8 August 1997. I tried to update to Debian 1.3 in a time of bad karma last month, and screwed everything up. Using the 1.3 "rescue" disk, I made copies of personal and config files on diskettes. The names got DOSified, and the permissions got screwed up, but the data appear to be readable. I will have to trash local installations. I will append a listing of /usr/local/src below: TeX-MF ssh-1.2.14 Amaya GNUscape bugsx TkNet dotfile-2.0b1 link teamrooms-1.0b3-linux cbb-0.67 TkGoodStuff STk-3.0 STk-3.1.1 Sndan I don't understand why TeX-MF is there. 9 August 1997. I reinstalled Debian 1.3.1. I used floppies, copied from the net, to create the kernel, because I couldn't get access to the CDROM for booting. But, I installed packages from an "official" CDROM provided by Linux System Labs. So far, I only have the default packages. Mainly, I made sure that the mouse works under GPM, since the mouse was the biggest source of failures for starting up X. I restored a number of personal files, but I lost the list of their correct names, so I may not have named them all properly (the names were truncated in funny ways by the version of cp that I used for the "rescue" backup). Now, I will try to fire up X. OK, I am running X under xdm, with the VGA16 server. The GUI configuration was flakey, but it worked eventually. I have not nearly enough resolution, and not enough colors. Will tune it up tomorrow. It appears that I can use the GUI configuration program with the SVGA server, as long as I have the VGA16 package loaded. 10 August 1997. Reinstalled my old /etc/X11/XF86Config. Got the old flashing-screen loop when I rebooted. The problem was that I had worked out XF86Config for the S3 server. I fixed /etc/X11/Xserver to specify S3, and installed the xserver-s3 package. Now I'm cooking with a high-resolution display. Need better fonts, a lot of applications that are missing. Having a lot of trouble getting ppp going again. In addition to the ppp problems per se, I can't start up emacs as root. The X server won't accept the connection. I succeeded with ppp by following the new instructions in /usr/doc/ppp/README.debian, after creating /dev/cua0 and linking /dev/modem to it. I now can only dial to one server (Earthlink in Northbrook). I am still puzzled by the old advice *not* to use cua*. Needs more work. 11 August 1997. Installed *lots* of miscellaneous software. .xsession doesn't seem to be executed, so I'm not getting my startup applications. I found that .Xdefaults needs to be known as .Xresources for xdm. Also, .xsession needs to be executable. That solved the basic startup problems for xdm. I tried installing xbanner, but it was buggy, and left a mess on the screen, so I removed it from the xdm startup/setup scripts. 12 August 1997. Installed ssh, and it works to Pity, including X redirection. Did not start sshd. /etc/rc.local doesn't exist. I think that Debian has another appropriate place for starting sshd, but I'm not sure where it is. Installed non-free stuff through ftp. amaya fails with the message: /usr/lib/Thot/LINUX-ELF/bin/amaya: can't load library 'libXm.so.2' I can't find a package with a name like libXm. Installed Netscape 3.01. 15 August 1997. I'm trying again to get the SoundBlaster working. The "isapnp" tools appear to be just what I need to solve the problem of soft configuration on the board itself. "pnpdump" finds the SoundBlaster all right. Now, I need to compile a kernel with sound support. I can't figure out whether I should use the 2.0.29 kernel source or the 2.0.30. /var/log/messages reports that I am running 2.0.29 now, so I'll try that first. On second thought, it appears that 2.0.31 has the same features as 2.0.29, but some bug patches, so I'll use the later version. 16 August 1997. I configured xntpd3 to use gargoyle.cs.uchicago.edu as a server. I created /var/lib/ntp/ and /var/log/ntpstats/ since they are mentioned in the /etc/ntp.conf from the xntpd3 package. I stopped and restarted xntpd, using /etc/init.d/xntp3, to read the new /etc/ntp.conf file. /var/log/xntpd shows activity now, and /var/log/ntpstats has files in it, but /var/lib/ntp is still empty. 17 August 1997. Emacs eliminates the execute permissions on .xsession every time I edit it. 28 August 1997. Downloaded cbb-latest.deb from file://ftp.me.umn.edu/pub/finance/. This is "Check Book Balancer", a freeware system similar to Quicken, and capable of importing Quicken data. cbb-latest.deb is a Debian package, but not in the standard archives. To accommodate it, I created /var/lib/dpkg/methods/manual, and placed cbb-latest.deb in there. Then, I ran dpkg --install cbb-latest.deb. Seemed to work. Changed the file name to cbb_0-73.1_all.deb, which is the more descriptive alias on the ftp site. I need to find the right way to install manually loaded Debian packages. 29 August 1997. I compiled a new kernel with SoundBlaster support. "make modules" complains that SoundBlaster support should be configured with the "CONFIG_AUDIO" option. I edited /usr/src/linux/.config to have "CONFIG_AUDIO=y", and recompiled, but it seems to make no difference. I used "make zlilo", which turns out to be dangerous. It saves /vmlinuz into /vmlinuz.old. But, doing it twice kills the old version. I installed the kernel-image-2.0.29 and kernel-image-2.0.30 packages, and created an /etc/lilo.conf giving access to both of those, as well as the newly compiled "/vmlinuz.test". Miraculously, it seems to work, booting up 2.0.30 by default. Now, I'll try booting the custom version. I need a systematic scheme for naming custom kernels. I succeeded in booting the test kernel, but can't get any sound. "saytime" reports "opening /dev/audio: Permission denied". I also can't start PPP: the console reports "Aug 29 22:04:27 Powdermilk insmod: /lib/modules/2.0.30/net/ppp.o: unresolved". There may be a problem with modules compiled for the wrong kernel. PPP still works with the packaged 2.0.30 kernel. There is also a report of the network audio system coming up. I'm not sure what to make of it. I think I forgot to do "make modules_install" for custom.1. OK, 2.0.30_custom.2 boots and does PPP. I'm pretty sure that it was the omission of "make modules_install" that screwed PPP up before. I found the right button in "make xconfig" to set up the SoundBlaster with CONFIG_AUDIO enabled. I did the audio directly in the kernel, not as modules, this time. Still no sound. "au" says "Fatal server error: could not create audio connection block info". Relevant material from /var/log/messages is: Aug 29 22:40:22 Powdermilk kernel: PS/2 auxiliary pointing device detected -- driver installed. Aug 29 22:40:22 Powdermilk kernel: Sound initialization started Aug 29 22:40:22 Powdermilk kernel: Sound initialization complete Aug 29 22:40:22 Powdermilk kernel: ide: i82371 PIIX (Triton) on PCI bus 0 function 57 Aug 29 22:40:22 Powdermilk kernel: ide0: BM-DMA at 0xffa0-0xffa7 Aug 29 22:40:22 Powdermilk kernel: ide1: BM-DMA at 0xffa8-0xffaf Aug 29 22:40:22 Powdermilk kernel: hda: WDC AC21000H, 1033MB w/128kB Cache, LBA, CHS=525/64/63 No clue so far. Rebooting with the stable kernel while thinking about it. 30 August 1997. Now PPP fails with the packages 2.0.30 kernel. I think it's a problem with incompatibility between the PPP modules between 2.0.30 and 2.0.30_custom.1. TO DO: Design a sensible strategy for managing different kernels through LILO. The current state is *very* confusing. 2 September 1997. I am running 2.0.30_custom.2 kernel to get PPP, but I neglected to compile in support for ISO file systems, so I can't use the CD-ROM drive. 2 September 1997. Switched from GNU Emacs to XEmacs. Seems to work, but some annoying differences in details. 5 September 1997. In fact, I have module support for ISO file systems in 2.0.30_custom.2. This is very appropriate, since I make sporadic use of the CDROM drive. Somehow, the module doesn't get loaded. I did "depmod -a" and "modprobe isofs" manually, and now I can use the drive. /usr/src/linux/Documentation/modules.txt has some discussion of the problem, indicating that something needs to go in some startup script. I didn't understand it completely. 10 September 1997. Yesterday I installed an Epson Stylus Color 400 printer. It printed ASCII immediately, using the lpr/lpd spooler. I installed magicfilter, but selected the Espson "9 pin" option on a bad guess, and got gibberish from PostScript files. I switched to apsfilter. The first time through the installation, I asked for more information on filters, and got into a nasty loop, thrashing the disk with a lot of seeks. I restarted with the button, and fortunately there was no serious file system damage. Then, I found printer information in /usr/doc/apsfilter/printer.gz, indicating that the right printer id is "escp2", since the control code for the printer is called "ESC/P". Now, when I try to print PostScript, the printer prints the error "Unknown device: escp2". 11 September 1997. I tried re-installing magicfilter. It appears not to have the right Epson filter (although it *might* be there under a name that I don't recognize). So, I went back to apsfilter. I discovered that the "Unknown device: escp2" message comes from GhostScript, which uses the name "stcolor" rather than "escp2". I found the two places in /usr/lib/apsfilter/bin/apsfilter where the device name is passed to gs, and I changed them from "${PRINTER}" to "stcolor", after which I was able to print ASCII and DVI. I need to find the right way to configure aps for this purpose. I wrote to the package maintainer, and found out that the package is orphaned. I made the patch to apsfilter slightly more robust, and let that stand for now. 13 September 1997. I tried to set up xdm so that different X sessions could run on different virtual consoles. I edited /etc/X11/xdm/xdm-config and Xservers to try to produce a :1 display in addition to the :0. The result was that xdm hung up when I rebooted, with a blank, faintly grey display. I restored the two files, but saved the experimental versions in /etc/X11/xdm/Experiment. There should be a way to make xdm interruptible when it gets screwed up. I will also seek a way to print a reminder about the "S" option for single-user mode at the LILO boot prompt. I can never remember that one on the first try. I added a "Single_user" label to /etc/lilo.conf, using the "append" option to provide the "s" parameter to the "Stable" kernel, which takes one to single-user mode. I also added "message=lilo.bootpromptmessage", and wrote a descriptive message into /etc/lilo.bootpromptmessage, to produce helpful instructions immediately after typing when LILO starts. It would be particularly nice to have LILO remind one to type for the boot prompt, but I can't find a way to do that. I noticed that one must run "lilo" again to get the message loaded, and it takes the path name quite literally, so one must be in the /etc directory. 19 September 1997. I installed bbdb_1.51-2.deb, using dpkg --ignoredepends=emacs --install bbdb_1.51-2.deb. The package is set up to depend on GNU emacs, but the software seems to be specifically designed for X Emacs. I copied some configuration from the info file into .emacs and .xemacs-options (I need to figure out which is really right). I could only find the bbdb commands in emacs after explicitly doing load-file /usr/lib/emacs/site-lisp/bbdb-init.el. Need to find the right way to automate this. Well, it's worse than that. I can't execute any of the bbdb commands, because they claim to be "compiled for Emacs 19.29 or later". 20 September 1997. dselect tries to get rid of xemacs and install GNU emacs, because I have hand-installed bbdb. I couldn't find any marking that straightened this out, so I just de-installed bbdb. I think that I need to fetch bbdb from the main distribution and compile my own version for xemacs.