The 2500 bootstrap ROM expects to receive a block loader first, which
will run and recieve the actual payload which is the microcode and
font data divided into blocks.
The assembler will recieve a GC-OVERFLOW interrupt. Other TT2500
files set the GC-OVERFLOW variable to a dummy function, which seems to
appease the interrupt.
Since the TVDIS code was written before backquote was introduced to
Maclisp, it uses comma characters without quoting. To accomodate this
a call to SETSYNTAX overrides the new syntax for comma.
File names where compared against historical backup tapes, and time
stamps where listed in a text file. Before a tape image is made, all
files are updated with timestamps from the listing.
Where no historical timestamp can be determined, the latest git author
date is taken.
The script takes two file names on the command line: an old tape, and
a new tape. Any updated files are written to a new tape diffs.tap.
The intent is that diffs.tap can be extradted with a DUMP command like
LOAD CRDIR LINKS to update an ITS system.
The script tries to avoid including binary files that only differ in
the symbol table or creation time. Some false positives are expected.
Both of these are from MIT's zork-1978-01 release. MADMAN; MADADV SAVE
is from 1978-01-28 (it's madadv.save_3 there), and TAA; ZORK 3 is from
an archive dated 1978-01-27. Unfortunately this isn't the final version
of Zork -- in particular, it doesn't have the endgame.
The launcher will also work with the other 1977/78 Zork images MIT have
released, provided you copy them to MADMAN; MADADV SAVE.
Note that we already have the non-DM fake Zork in SYS3; TS ZORK, but
the real Zork was in SYS2; on DM, so the recommended ZORK^K will find it
first.
These two databases contain pure code routines that Muddle images can
refer to. We don't currently have whatever tool was used to maintain
them, so this is a MIDAS program that creates and populates them.
SAV FILE includes some stubs for functions in the LSRTNS and MUDDLE
libraries -- these are the functions that the existing Muddle 54 Zork
images need.
For the FIELD function in LSRTNS, returning FALSE is equivalent to the
user not being found. I have no idea what C-FCN in MUDDLE is for, but
very early Zork calls it after a command is entered, and seems happy
enough with it doing nothing.
This source was reconstructed to match MUDSAV; TS MUD54 from 1977-07-02,
using a combination of all the surviving Muddle source files. The memory
layout and pure code is the same.
No AGC MUD54 has survived, so the AGC code was adjusted to match the
TOPS-20 agc.mud104 from Chicago that Rich Alderson provided (the only
ITS conditional is the page size). There's a one-instruction difference
in the symbol positions, which I've converted into a patch at the end of
the code to maintain the original layout on ITS.
The INITM code, which doesn't appear in the final executable, is
a best guess but it's probably fairly close, since it generates objects
in the right order and locations, and the symbol locations match the
original.
The 1977 executable has a very large number of patches, which I've
replicated in MUD54 INIT. The code that the patches were replacing --
marked with "XXX patched" in the source -- is also a best guess.
I haven't checked that the TOPS-20 code is correct; it could be adjusted
to match mdl104.exe in the future. It would need STENEX, which could be
linked from MUDSYS;.
Previously the first XFILE printed out some commands to run after STINK
had completed, which is awkward if you're building Muddle repeatedly.
This automates the second half of the build process.
Lars has found a couple of examples of TS MUD54 files, and they're the
initialised version (with references to internal functions filled in,
etc.). TAA's ZORK launcher also expects TS MUDxx to be the initialised
version. So it appears that the ITS version didn't generate TS MDLxx.
(Why does the initialisation process write out TS MUD56, then, if you're
only going to pdump the result over it? Because the last step in
initialisation is to invoke the GC, and getting back from the GC to the
interpreter requires mapping pages back in from the executable.)
Found from comparison with a TS MUD54 binary. The Muddle 56/106 source
came from TOPS-20 originally, and it had been extracted with newline
conversion but not ITS encoding.
Most of these are cosmetic, but there are a couple of VALRET strings
with embedded \rs -- including the one used to exit initialisation.
So successful initialisation now finishes with a *, rather than opening
a random location.
This replaces the existing MUDSYS;STINK 2 that was installed as TS
STINKM. The bad-EXPUNGE patch in there is no longer needed now Muddle is
built with MIDAS 73.
MACDMP MOBY1 has 340 support, but only works with the old PDP-6
microtape device. MACDMP 6U32 has both microtape and TD10 support,
but no 340 code. Both programs can be adjusted for core size.
Older versions of MIDAS expect you to type a command at them, which
makes them hard to automate with :x or XFILEs. This patch backports the
JCLINI code from later versions of MIDAS, which pre-stuffs the command
buffer with the JCL string if available.
This was edited from later versions of the source to match MUDDLE; TS
MIDAS circa 1973, which was built with MOBY==0. The program itself dates
from no later than 1971. Comments, and results when built with other
options, are probably not historically accurate.
MIDAS 73 is useful because 74 and all later versions have completely
rewritten code for relocatable output (and for IO), and they don't
support some of STINK's features correctly (or at all). A ported version
of MIDAS 73 was still being used to build Muddle in the 1980s, going by
<mdl.int>midas.exe.5.
