Over the past two weeks, I set up a small Symbolics LAN, and thought I would share some of the process and findings. I will write a post about each item, but the topics already concluded cover:

• Performance challenges of the Merlin Ethernet under load, and modern 10Base5 to 10BaseT transceivers
• An improvement to the stability of my first XL1201 which I was unstable, some five years ago
• Y2k problems with the QFILE protocol under Chaosnet, and a "hotfix" to Genera's QFILE client to bypass them (this is deprecated, but was fun to try out)
• IP subnets and routing under Symbolics Namespaces
• Setting up the three Symbolics XL computers as a small LAN with a Linux NFS server

Later, I hope to cover some more topics, including:

• Adding MacIvories, NXP and other computers to the LAN
• Bridging this Symbolics LAN with others over WAN

In turn, I hope to explore and share some of the features of the new (to me) XL1200 with Framethrower that was provided by David Schmidt.

I also hope to show my colleagues at work some of the joys and advantages of the Symbolics model of building software, as compared to our current way of building services (e.g., Java/Spring Boot) and clients (e.g., TypeScript/Vue.js/Stencil).

The LAN currently consists of:

• XENON, XL1201, Chaos 1200 & IP 192.168.5.92
• COBALT, XL1201, Chaos 1202 & IP 192.168.5.94
• IRON, XL1200, Chaos 1204 & IP 192.168.5.95
• IGW, Edgerouter Pro, IP 192.168.5.1/24 and 192.168.1.1/22
• BORON, Intel NUC & Ubuntu, IP 192.168.1.97

The 192.168.5.0/24 segment is separate from the rest of my LAN and solely for vintage computers. No DHCP runs there, and all the broadcast traffic modern computers send out doesn't cross.

I've decided to participate in ProcJam 2016 this year.

For a change of pace, I'm building a small roguelike in Clojure with a Unity3D front-end. This is an interesting challenge as Clojure allows for (even strives for) purity, but doesn't have a well-integrated, easy to use state monad and related syntax like Haskell, making it hard to use (for example) random numbers purely.

So, I'm making an almost-pure version, with the impurities of the random numbers "allowed," for now. I may explore writing/refactoring my code to allow for the use of Clojure's threading macros.

The source code and more details are available in my GitHub if you want to follow along.

In preparation of doing some actual software engineering on Genera, I decided to take a look at the CLIM implementation on my MacIvory III Genera 8.3.

There are a few steps to do first:

• Load System CLIM: This will take a while. It loads a lot of files and several subsystems, including:
  • Loading system CLIM version 66
  • Initializing subsystem CLIM Utilities
  • Initializing subsystem CLIM Silica
  • Initializing subsystem CLIM Standalone
  • Initializing system CLIM version 66
  • Loading patches for system CLIM version 66
  • Patches for CLIM (Current version is 66.3):
• If you attempt to use CLIM at this point, however, it won't work because there is no driver (in my parlance only, I'm not sure the official term). The example code will fail: Error: Cannot find port type: :GENERA
• Load System Genera-CLIM will install the remaining parts

This whole process takes probably 15+ minutes of compiling and crunching, so you might want to save the world and restart after this (although I didn't do this yet).

Oddly enough, at this point, I figured I'd read the CLIM documentation online in the Document Examiner, but alas, it doesn't seem to have it there even if I ask it to show candidates.

Anyway, as usual, Bitsavers has the docs so I decided to follow along with the tutorial to run the puzzle-1.lisp example. (Reproduced at the end.)

To back up one step, let me mention how I have my development environment set up. I have the MacIvory III configured in a Mac Quadra 950 running Mac OS 8 with a SuperMac Thunder II GX running on a Dell 2007FP LCD, but unfortunately only running at 1360x1024 (since I don't have a monitor which can do 1600x1200 at 75Hz on VGA). I couldn't get my second Ethernet card to work (due to a bug in the Genera Ethernet control panel under Mac OS 8) so when Genera is running, the Mac is not on the LAN.

I run Genera, and leave it and the Mac running, but I turn off the monitors and lock the Mac's keyboard/mouse. Then I Telnet into Genera, Start X Screen twice to my iMac, and then Halt Remote Terminal (which ends the Telnet session). In the X screen, I use my Tenkeyless Enhanced keyboard layout with my custom WASD keyboard.

Back to CLIM. When I first ran the demo:

(setq fp1 (make-application-frame 'fifteen-puzzle-1
        :left 200 :right 400 :top 150 :bottom 350))
(run-frame-top-level fp1)

By selecting this and hitting c-sh-E in Zmacs, nothing happened, although Zmacs kept telling me it was executing the CLIM program. I couldn't figure it out. Furthermore, I couldn't get my nicely mapped ABORT button to do anything, nor c-ABORT. (It took me a while to remember to do c-m-ABORT which worked.) Turns out I had to tell System Preferences in the Mac (OS X.11) to unmap ^F8 from "Move focus to the status menus."

Anyway, I abort and restart the program a few times, both from Zmacs and a Lisp Listener ((clim:run-frame-top-level clim-user::fp1)). Finally I power on the MacIvory screen running Genera and, voila, the puzzle window is running on the console.

Well, needless to say, I may have to not use CLIM to do my Genera development, because I need (well, want) everything to work over an X screen.

Documentation [Updated]

I searched and found that the CLIM documentation was indeed on my filesystem, at ARGON:>sys>doc>clim>. I tried Read Private Document on the various .sab files, but nothing would show up in Document Examiner.

In there was an attractive looking LISP file, clim-doc.lisp, that had lots of interesting references to documentation. So I Load Filed it, but no luck.

Then I typed Load System and COMPLETE and, voila, CLIM Documentation showed up as an available system to load. Actually, it showed up twice (not sure why). I loaded it, and boom, I had my documentation.

TODOs

• Figure out how to get CLIM to run over CLX, which can be loaded using Load System CLX-CLIM

Puzzle-1 Source

;;; -*- Mode: Lisp; Syntax: ANSI-Common-Lisp; Package: CLIM-USER; Base: 10; Lowercase: Yes -*-
;;;>
;;;> *****************************************************************************************
;;;> ** (c) Copyright 1993 Symbolics, Inc.  All rights reserved.
;;;> ** Portions of font library Copyright (c) 1984 Bitstream, Inc.  All Rights Reserved.
;;;>
;;;>    The software, data, and information contained herein are proprietary 
;;;> to, and comprise valuable trade secrets of, Symbolics, Inc., which intends 
;;;> to keep such software, data, and information confidential and to preserve 
;;;> them as trade secrets.  They are given in confidence by Symbolics pursuant 
;;;> to a written license agreement, and may be used, copied, transmitted, and 
;;;> stored only in accordance with the terms of such license.
;;;> 
;;;> Symbolics, Symbolics 3600, Symbolics 3670 (R), Symbolics 3675 (R), Symbolics
;;;> 3630, Symbolics 3640, Symbolics 3645 (R), Symbolics 3650 (R), Symbolics 3653,
;;;> Symbolics 3620 (R), Symbolics 3610 (R), Symbolics Common Lisp (R),
;;;> Symbolics-Lisp (R), Zetalisp (R), Genera (R), Open Genera, Virtual Lisp Machine,
;;;> VLM, Wheels (R), Dynamic Windows (R), SmartStore (R), Semanticue (R), Frame-Up (R),
;;;> Firewall (R), Document Examiner (R), Delivery Document Examiner, Ivory, MacIvory,
;;;> MacIvory model 1, MacIvory model 2, MacIvory model 3, XL400, XL1200, XL1201,
;;;> Symbolics UX400S, Symbolics UX1200S, NXP1000, Symbolics C, Symbolics Pascal (R),
;;;> Symbolics Prolog, Symbolics Fortran (R), CLOE (R), CLOE Application Generator,
;;;> CLOE Developer, CLOE Runtime, Common Lisp Developer, Symbolics Concordia, Joshua,
;;;> Statice (R), and Minima are trademarks of Symbolics, Inc.
;;;> 
;;;> RESTRICTED RIGHTS LEGEND
;;;>    Use, duplication, and disclosure by the Government are subject to restrictions 
;;;> as set forth in subdivision (c)(1)(ii) of the Rights in Technical Data and Computer 
;;;> Software Clause at DFAR 52.227-7013.
;;;> 
;;;>      Symbolics, Inc.
;;;>      6 Concord Farms
;;;>      555 Virginia Road
;;;>      Concord, Massachusetts 01742-2727
;;;>      United States of America
;;;>      508-287-1000
;;;>
;;;> *****************************************************************************************
;;;>

(define-application-frame fifteen-puzzle-1 ()
    ((pieces :initform (make-array '(4 4) :initial-contents '((1 2 3 4)
                                  (5 6 7 8)
                                  (9 10 11 12)
                                  (13 14 15 0)))))
  (:menu-bar nil)
  (:panes
    (display :application
         :text-style '(:fix :bold :very-large)
         :display-function 'draw-the-display
         :scroll-bars nil)
    (menu :command-menu))
  (:layouts
    (main 
      (vertically () display menu))))

;;; this draws the entire display

(defmethod draw-the-display ((application fifteen-puzzle-1) stream)
  (with-slots (pieces) application
    (dotimes (y 4)
      (dotimes (x 4)
    (let ((piece (aref pieces y x)))
      (if (zerop piece)
          (format stream "   ")
          (format stream "~2D " piece))))
      (terpri stream))))

;;; useful macrology - the body will be run with x and y bound to
;;; the coordinates of the empty cell

(defmacro find-empty-piece-and-do ((y x) &body body)
  `(block find-empty-piece
     (dotimes (,y 4)
       (dotimes (,x 4)
     (when (zerop (aref pieces ,y ,x))
       ,@body
       (return-from find-empty-piece))))))

(define-fifteen-puzzle-1-command (down :menu t) ()
  (with-slots (pieces) *application-frame*
    (find-empty-piece-and-do (y x)
      (if (not (zerop y))
      (rotatef (aref pieces y x) (aref pieces (- y 1) x))))))

(define-fifteen-puzzle-1-command (up :menu t) ()
  (with-slots (pieces) *application-frame*
    (find-empty-piece-and-do (y x)
      (if (not (= y 3))
      (rotatef (aref pieces y x) (aref pieces (+ y 1) x))))))

(define-fifteen-puzzle-1-command (left :menu t) ()
  (with-slots (pieces) *application-frame*
    (find-empty-piece-and-do (y x)
      (if (not (= x 3))
      (rotatef (aref pieces y x) (aref pieces y (+ x 1)))))))

(define-fifteen-puzzle-1-command (right :menu t) ()
  (with-slots (pieces) *application-frame*
    (find-empty-piece-and-do (y x)
      (if (not (zerop x))
      (rotatef (aref pieces y x) (aref pieces y (- x 1)))))))

#||
()
(setq fp1 (make-application-frame 'fifteen-puzzle-1
        :left 200 :right 400 :top 150 :bottom 350))
(run-frame-top-level fp1)
||#

One of the theories about the instability of my Symbolics XL1200 is that the CPU has no heatsink and so it is suffering from thermal problems after a while. (15-30-ish minutes)

David K. Schmidt was kind enough to send me an old heatsink. It's 38x38x5 mm and has 11 fins on it, as you'll see on the pictures; slightly smaller than the Ivory CPU itself.

First I cleaned the old epoxy from the heatsink using acetone, which was surprisingly easy to do.

After removing the Merlin II card, I cleaned the top of the Ivory CPU and the bottom of the heatsink with a small 70% Isopropol Alcohol medical swab.

I removed the Ivory CPU and cleaned the socket pins with CRC Contact cleaner and brushed on a small amount of Deoxit D5 on the pins of the Ivory CPU.

Using a minimal amount of Arctic Silver Thermal Adhesive on just the golden "chip" part of the top of the Ivory CPU, I then stuck the chip and heatsink together. I then put the top of the Deoxit D5 bottle on top of the heatsink, and then a 3.5" SCSI hard drive on top of it to apply pressure. After an hour, I removed the hard drive and Deoxit cap, and left the whole setup to cure overnight.

In the morning, I cranked up the air conditioner and reinstalled the Merlin II.

Unfortunately, this does not seem to have solved the problem. Using the system for about 23 minutes has now produced some weird errors, including:

--> The scheduler's data structures are locked, using the cold load stream <--

Error: The generic function PROCESS-ACTIVE-P was applied to the argument #<DPT-CALL-INDIRECT-PREFETCH 22005600001>.  
       The argument was not of the correct type.
       The function is implemented by the flavor PROCESS:PROCESS.

(FLAVOR:METHOD TV:SCREEN-GET-WHO-LINE-PROCESS TV:GENERIC-WHO-LINE-SCREEN-MIXIN)
  Arg 0 (SYS:SELF-MAPPING-TABLE): #<Map to flavor TV:GENERIC-WHO-LINE-SCREEN-MIXIN 20012145454>
  Arg1 (SELF): #<WHO-LINE-SCREEN Who Line Screen 20006200000 exposed>
...

Then, eventually within a few minutes it just gave random errors, like:

Lisp stopped itself  
Error trap 10000006612 not in emulator mode ...  

So, it appears thermal stress on the Ivory CPU isn't the problem, or that the thermal stress is too great for that small heatsink to keep up with.

Sigh.

Next step is to clean and reseat all the chips on the I/O board, in a few weeks.

UPDATE: I clipped and removed the broken capacitor. Replacing it will be the task of another day.

I spent two days removing, cleaning and reseating each DIP chip on the Symbolics XL1201 Merlin II Main (CPU) board. Earlier I closely examined the solder side of the board and found a few minor discrepancies. This time I also encountered three discrepancies:

• Bent pin on chip BB27 - unbent
• Broken bypass cap next to chip GG7 - I clipped this one off entirely and will replace another day
• One long bodge wire running along the entire top has come loose. This could get snagged on something and fall off.

Additionally, I removed and resocketed the Weitek Floating Point co-processor without doing any cleaning. It looked sparkling clean.

While I was at it, I removed the metal gunk in the spare hole of BB12.

Finally, I examined the component-side of the bad-looking solder joint between CC25-26/DD25-26. It seems this was likely caused by the soldering of a bodge wire to this connection, on a resistor pack/network.

Note that I did not remove the two other PGA chips on the board. One of them is almost certainly the Ivory. They are:

• VTC VIC068PGF4 9050 985 VIC ES: Probably the VMEBus controller chip.
• 110730 USA 1XC9-0401 9051: Probably the Ivory CPU, right in the center of the board.

Discrepancies Detailed

I found a broken capacitor between chip FF7 and GG7. Top label says "A5E104Z". Octopart shows a few places to get it, two of which seem to think it's a 0.1 uF 50V cap. One says "A5E104Z ARCO .1MFD-50VDC" and the other says "CAP CER DIP .1UF 50V 80-20% 7.5LS BULK". It's about 6.25mm wide, 2.5mm tall and the top is about 4mm off the circuit board. GG7 is a 74F00N which is a Quad 2-input NAND chip. It likely is a bypass capacitor for this chip, as almost every single chip has one immediately to its left.

After moving the capacitor around, it was clearly broken all the way across.

Additionally, I found one bent pin on chip BB27, which also is the only chip with an unusual label, "985-DS/N 582." I believe that this is the chip that contains the serial number, as the board has a sticker saying "S/N 582" as well as "582" hand-scratched into the board. The corresponding socket hole was also slightly marred. I very carefully unbent it and reinstalled it.

After-repair:

Finally, there are two places with no chip installed into a socket. One iS JJ30 silkscreened "WRPOST" and the other is KK30, silkscreened "AUX-TIMER". There are numerous jumper posts nearby which may enable these, though for what purpose I don't know. There are a total of about eight jumper blocks scattered around the board.

Procedure

• Photograph board with sufficient detail to ensure all chips can be replaced correctly
• Remove each DIP chip, using Wiha Chip Lifter
• Write down the chip location, the silkscreen written under the chip, and the label on the chip (if any)
• Spray some Deoxit D5 onto a Q-tip and wipe the chip's pins clean and set aside to dry
• Spray some contact cleaner into the chip socket, and agitate a bit with a toothbrush
• Air-can dry the socket and, if necessary, the chip
• Reinsert chip
• Verify chip reinserted properly
  • Correct location
  • Correct orientation
  • All pins in their holes correctly

I replaced the Q-tip every 3-6 chips as it would become gray after a short time.

Chips Reseated

• Position: Silkscreen: Chip Label
• AA11: BC-BNK-BUS: 115932-A
• AA12: BC-RFL-ADD: 115914-A
• AA13: BC-OPER-GEN: 115915-A
• AA16: BC-LCLK-G_N: 115958-A
• AA18: (forgot to write down)
• AA22: ID IMBU CT(R?)L: 115921-A
• AA27: ID-DE-GATING: 115962-A
• AA28: ID-STR-CTRL: 115963-A
• BB11: BC-ADDR-PIPE: 115942-A (engraved)
• BB17: BC-QUE-0: 115931-A
• BB18: BC-QUE-1: 115930-A
• BB19: BC-QUE-2: 115929-A
• BB20: BC-QUE-CONT: 115928-A
• BB21: BC-QUE-RFL: 115927-A
• BB22: ID-INTERRUPT: 115920-A
• BB24: CY7C291: 115982-D
  • BB24-28 are a different type of chip than all the rest, in a different package. Looks sort of like a KitKat, and harder to get back into the socket. See pictures.
• BB25: CY7C291: 115983-D
• BB26: (forgot to write down)
• BB27: CY7C291: 985-DS/N 582
  • This chip had its bottom right-most pin bent in and hence was probably not making solid contact.
  • I unbent the pin and reinserted into the socket. See pictures.
• BB28: CY7C291: 115986-D
• BB30: ID-RESET: 115956-A
• CC11: TC 5328 (no label)
• CC12: TC 55328 (no label)
• CC13: BD-RD-PIPE: 115935-A (engraved)
• CC14: BC-DATA-SCH: 115941-A (engraved)
• CC15: BC-WR-PIPE: 115957-A
• CC16: BC-DATA-MON: 115937-A (engraved)
• CC31: ID-ALATCH: 115964-A
• DD7: MEM-WR-CTL: 115971-A
• DD8: BANK-STATE: 115975-A
• DD9: MEM-WR-CTL: 115971-A
• DD10: MEM-RD-CTL: 115972-A
• DD12: CACHE-STBS: 115951-A
• DD13: TC55328 (no label)
• DD14: TC55328 (no label)
• DD16: BC-REQ-DEC: 115965-A (engraved)
• DD18: BC-ADDR-CONT: 115944-A (engraved)
• DD19: BC-QUE01-RDY: 115949-A (engraved)
• DD20: BC-RD-ACK: 115961-C (engraved)
• DD21: BC-2-RFL-RDY: 115948-A (engraved)
• DD22: BC-ACK-GEN: 115947-A (engraved)
• DD29: IM-DRIVE: 115955-B
  • Three separate bodge wires go to various pins of this chip's socket.
• EE16: CACHE-REFILL: 115946-A (engraved)
• EE18: CACHE-CNTRL: 115945-A (engraved)
• EE19: CACHE-REQ: 115926-A
• EE22: MII-TIMEOUT: 115968-A
• EE24: SHUFFLE: 115939-B
• EE25: SHUFFLE: 115987-A
• EE26: SHUFFLE: 115987-A
• EE27: SHUFFLE: 115939-B
• EE28: REQ-DECODE: 115966-A
• EE29: V-REQ-ACK: 115992-A
• FF1: MEM-STB: 115976-A (vertical)
• FF4: MEM-STB: 115976-A (vertical)
• FF7: MEM-ACK-BUF: 115973-B
• FF8: MEM-REQ-CNT: 115970-A
• FF9: MEM-RD-CTL: 115972-A
• GG10: MEM-REQ-STATE: 115980-A
• GG11: MEM-REQ-CTL: 115979-A
• FF13: FPA-CLOCK: 115950-A (engraved)
• FF22: IB-REQ-ACKED: 115953-A (engraved)
• FF23: IB-CNTRL: 115960-A
• FF24: POLLED-CNTRL: 115991-A
• FF25: PLD-DECODE: 115943-A
• FF27: V-ARBITOR: 115933-D (very faint)
• FF28: ILB-CNTRL: 115934-C
• FF29: SHUFFLE: 115987-A
• FF30: SHUFFLE: 115939-B
• GG15: IDT6178S: (no label)
• GG16: IDT6178S: (no label)
• GG17: IDT6178S: (no label)
• GG18: IDT6178S: (no label)
• GG19: IB-DECODE: 115952-A (engraved)
• GG21: MII-REFRESH: 115969-A
• GG23: REQ-DECODE: 115954-A
• GG31: SHUFFLE: 115939-B
• HH15: DT6178S: (no label)
• HH16: DT6178S: (no label)
• HH17: DT6178S: (no label)
• HH18: DT6178S: (no label)
• HH25: MEM-LTRL: 115950-A (engraved)
• HH31: SHUFFLE: 115987-A
• JJ15: MII-FP0-WR: 115916-B
• JJ23: TRIGGER: 115924-B
• JJ24: PARITY: 115925-A
• JJ25: QUINT-MUX: 115981-A
• JJ26: QUINT-MUX: 115981-A
• JJ28: VME-DECODE: 115923-D
• JJ30: WRPOST (no chip installed)
• KK11: MII-FP0-RD: 115917-B
• KK12: MII-TAG-CNTR: 115919-B
• KK13: MII-FP1: 115918-B
• KK14: MII-FPA-MISC: 115938-A (engraved)
• KK31: AUX-TIMER (no chip installed)
• Weitek FPA: Removed, inspected & reinserted only (not cleaned)

Additional Pictures

To see if I can isolate the cause of the XL1201 instability, I am going to physically inspect the two main cards with a loupe and flashlight. If I see anything suspect, I'll take a picture with a (cheap) USB microscope.

I am doing a detailed inspection of the solder side of the Merlin II CPU card first.

Overview picture:

Discrepancies

There are only two that are concerning to me, but several others were found. The two of most concern are:

• Seemingly bad solder joint or problematic solder joint
• Material in a empty socket hole - could it make shorts between the layers? Irrelevant?

Material in a hole of empty socket BB12

Bad-looking solder joint between CC25-26, DD-25-26

Black blemish near CD23

Poorly cut lead near EE25 but seems fine

FF11 - Slightly discolored joint

FF30 - Black stuff near solder joint, seems to be just a bit of fabric?

Well, after living with the default keyboard mapping for Genera under X-Windows, I decided to make an optimized one for an 87-key ANSI layout. The design goals were:

• Utilize all the 16 top row keys efficiently, with most common functions in logical places
• Utilize the 6-key quadrant for other common functions
• Leave the arrow keys for cursor motion (because even though I have been using Emacs for at least 26 years I like my arrow keys)

Some of the important things that these led me to were:

• The esc key is now Genera SELECT, as that is one of the most commonly used key somewhat analoguous to Command-TAB in macOS.
• The line key is nice to have available for use in Zmacs to start a new line and auto-indent, so put that somewhere reasonably close. In fact, I thought about putting it where the backslash
• Genera Backspace is not very useful (equivalent to c-B) so it gets demoted to a middle area and could even be entirely removed

Some of the challenges were:

• Modifier keys in the bottom and second to bottom row are too few. I put the most common ones on the left side (CONTROL, SUPER, META) and the less common ones on the right side along with the two more common ones (META, CONTROL, SYMBOL, HYPER). Just in case someone doesn't have a Menu/Application key, I also mapped SYMBOL to F5, but didn't put a keycap on it. I had an "extra key" and wasn't sure what to put there, and this seemed the most useful thing.
  • The SYMBOL key (Menu/Application key) probably needs to be marked as a modifier in Xquartz. Right now it autorepeats, which is very annoying. I haven't tried fixing the problem though, since the key is so rarely used. (Mostly for ≣ and ⇄ (or is it ⇆).)
• SUSPEND, RESUME, ABORT seem like they might be more safely put in F13, F14, and F15 to prevent accidental keypresses. On the other hand, they're used so often during software development that they're easier to reach in the middle. So, they went into the middle and the shape keys into the upper right.
• CAPS LOCK and MODE LOCK don't seem to do much useful at all. Indeed, on my Mac I generally map CAPS LOCK to "noop" wasting a completely useful key. The Symbolics 3600 keyboard puts a backspace there. I tried that, but the big problem is that Xquartz or macOS always captures the CAPS LOCK to toggle caps lock, regardless of how you map the key in Xquartz or Genera.
  • One solution would be to remap the CAPS LOCK key to something else like F16 using something like Karabiner or System Preferences. The former works great but has to be installed everywhere. The latter only lets you map it to pre-existing modifiers. If it were mapped to CONTROL (or whatever), then we could put a HYPER on the left. But, HYPER just isn't used much, so... Whatever.
• LOCAL seems to serve no purpose on MacIvory or on X11, so it's omitted. Same with REPEAT, although it is an interesting key for people without auto-repeat.
• COMPLETE is used frequently, but so is CLEAR INPUT (at least by me), so I wasn't sure which one belonged in the easily reached F12.
• HELP is also used frequently. It was a challenge to decide if it should stay in the F12 spot or in the (previously) Mac standard spot.
• GitHub doesn't work well with the character encoding of Genera. It considers those .lisp files with Genera characters (like ≣ and ⇄) to be binary and won't show them in listings.
  • I should really write an iconv plugin that will translate Unicode formats to the Genera format.
  • Another minor detail is that many Mac fonts seem to show the ≣ character with four horizontal lines instead of three. There are also two similar Unicode code points for this sort of a character. (≡ and ≣)

So, with all that said, let me share the layout: (Keyboard-Layout-Editor link)

Additionally, I made a WASD Keyboards custom layout file (which the original is an Inkscape SVG file, sources located here):

Finally, of course, there is xquartz-87-enh.lisp
in my GitHub Genera X11 Repository. The X11 Repo README.md file has installation instructions.

When I get the custom keyboard from WASD, I'll post pictures here.

I welcome thoughts and feedback on this layout.

Genera View

Here is how this keyboard looks in Genera.

Next Steps

I'm going to try to make a version of this layout for these keyboards that I also regularly use:

• MacBook Air/Pro laptop keyboard
• Microsoft Sculpt keyboard

I also have a USB Sun Type 3 keyboard that I may try to create an X mapping for, and a Sparcstation 20 with a Sun Type 5 keyboard. Finally, I may try to make a similar mapping for Linux.

It has taken some effort, but I finally created a reasonable (and close-to-default) keyboard layout and mapping for Genera for using a modern, 87-key "tenkeyless" keyboard under Xquartz. I became reasonably familiar (at a reading level) with Flavors and Genera's X implementation while reverse engineering all this stuff. I also used the apropos function a lot, of course.

This layout stays very close to the default one provided by Genera while adding a few tweaks such as:

• Improved modifier mappings
• Calling RUBOUT for Backspace/Delete
• Ability to use the Keyboard Control activity to further remap things visually

I also created a patch file that will fix an X-Windows bug in Genera with a modern X server when it tries to set the modifier mappings. This will show up in the Genera herald so you know it's there.

The keyboard layout is called :xquartz-87-tenkeyless and nicely printed as Xquartz 87-key Tenkeyless. A keyboard reference is available.

Xquartz must be configured with these three checkboxes set: (others are at your discretion)

You need to allow remote connections, so be sure to do something like xhost + in your xterm window.

Furthermore, you should put the Genera fonts into your X11 font path as described here or things will look "wrong" and be less enjoyable.

One note: There seems to be only 6-key rollover when using the Key Test feature of Keyboard Control. I'm not sure if this is an Xquartz limitation or Genera limitation, or something else. Regardless, it's sufficient for our needs. (Shift-hyper-meta-super-control-X anyone? Can't add symbol in there... )

Use

To use the keyboard layout in Genera:

• Download the XLib patch file and Keyboard Mapping file from my GitHub repo
  • Warning: these contain non-ASCII/Unicode characters in them (Genera symbol characters)
• Load File xlib-patch.lisp
• Load File xquartz-87.lisp
• Save World ... and update your boot file
• Restart Genera
• Start your Xquartz Server and configure as above
• Start X Screen INTERNET|10.10.10.10 ... on your host IP or hostname (if configured in your Namespace)
• Show X Keyboard Mapping will confirm that this mapping is in use by saying The keyboard layout type is :XQUARTZ-87-TENKEYLESS.
  • If not, see below
• Show Keyboard Layout will show details on the layout
• Select Activity Keyboard Control will let you remap things further to your hearts content, then put on the kill ring the modifications for your lismp-init.lisp file.

If your keyboard isn't working, it's probably because Genera couldn't match it. One thing to try is this:

Command: sys:*consoles*  
(#<X-SCREEN::X-CONSOLE NEON:0.0 0 (Genera on ARGON) 140001042> #<MTB:MAC-RPC-MAIN-CONSOLE 140005535>)
Command: (xlib:display-vendor (x-screen::x-console-display (car sys:*consoles*)))  
"The X.Org Foundation"
11803000  
Command:  

Then, edit the xquartz-87.lisp file to match, in the signature:

(x-screen:define-keyboard-signature :xquartz-87-tenkeyless
  (:keycode-offset 8
   :vendor-name "The X.Org Foundation"
   :vendor-version 11803000)

After that, check the keys against the output of your xmodmap -pk to make sure they match. If not, you'll have to edit almost everything else. (xmodmap will start with ...KeyCodes range from 8 to... - that's where the :keycode-offset comes from.)

Forcing the Keyboard Mapping to Be Used

UPDATE: I cobbled together the Set X Keyboard Mapping for Genera 8.3 from a few pieces I found on the Internet. Load set-x-kb-map.lisp into your world (save it if you want the command permanently) and then you can say Set X Keyboard Mapping [screen] Xquartz-87-Tenkeyless and it will load this mapping.

Note about F14 and F15

F14 and F15 are usually set to lower and raise brightness on the Mac. Disable this in System Preferences -> Keyboard -> Shortcuts -> Display

Next Steps

I'm going to make an improved layout that better utilizes the keys and puts the most frequently accessed ones closer to the home key. Having SELECT on the esc key is also nice (and familiar from using the MacIvory on the console). I'll share that here later.

Pictures

Here are some screenshots of the keyboard in use on my iMac from the MacIvory III.

In an effort to rule out the power supply causing the XL1201's instability, I sent it out for overhaul by Power Clinic. After reinstalling it, I booted the XL1201 and just let it sit in the Document Examiner.

Unfortunately, the computer decided to reboot itself after about 15-16 minutes. So, the power supply would seem not to be the problem.

I will have to disassemble the XL1201 and give it a thorough examination to see if there are any obvious physical defects. If none are found, I will try to remove, clean, and reseat all socketed chips, and see if that helps.

I still have a feeling that the capacitors on the backplane and/or the two boards could be the problem, but at this point I just don't know. It worked fine for several weeks (including one or two days of overnight runs) before becoming unstable.

Power supply before picture:

Power supply after overhaul picture:

Space is a premium in my little vintage computing paradise, so I had the bright idea of transplanting my MacIvory III from a desktop Quadra 650 to a tower Quadra 950 and clear some desktop space. David K. Schmidt said this was a simple matter, and he was entirely correct. I effectively put the Quadra 650 hard drive and the MacIvory card into the 950, rebooted, and blam, it "just worked." (I actually put a clone of the Quadra 650 drive from 3 years ago in to test first.)

The Quadra 950 takes a long time to start up due to 256MB of RAM, and its fan is louder than the 650. However, it also has two Apple Ethernet devices on it (one with AAUI, one with Twisted Pair). This should be a good thing, but it turns out it isn't.

At boot time, you get a warning to select the Ethernet device for Genera to use. Unfortunately, as it turns out, there's a bug with the "MacIvory Control Panel" app and Mac OS 8. Apparently one work around for this is to boot into Mac OS 7, configure Ethernet through that control panel, and then boot back into Mac OS 8 and the MacIvory software will remember the setting. I don't have a way to do that that I know of, but I believe Apple distributes Mac OS 7 on their site for free.

I wonder where MacIvory software saves this setting. Maybe I can just manually edit that file.

I used the opportunity to take a picture of the MacIvory though. Enjoy!

So, I learned something about the Todd MAX-354-1252 power supply in the Symbolics XL1201 today that I thought I'd share. Not that I understand it, mind you.

The power supply has six trim pots. Four are obviously labeled for adjusting the four output voltages:

• 5V at 50A
• 12V at 8A, 12A peak
• -12V at 2A
• -5.2V at 10A (for ECL circuits)

However, according to the expert engineers at Power Clinic "The two [remaining] trim pots in question are related to the front-end boost circuits. One is for operation frequency and the other is for boost fold. NEITHER should ever be adjusted."

As I understand it, switching power supplies work by switching the output voltage on and off extremely quickly (on the order of 100kHz) from the high input voltage, so obviously one of these tweaks that frequency. No idea what the "boost fold" is - but fortunately according to the engineers I never have to touch it.