While the scan of src/cmd/als.s clearly refers to ferror,
that routine is entirely missing from the scans of ald.
There are two identical ferror routines in src/cmd/adm.s and
src/cmd/apr.s printing the file name of a file that cannot be
openend.
ald calls ferror when unable to create a file. For ald to have
a similar ferror routine when creation fails appears sensible.
Therefore a similar routine is added to src/cmd/ald.s with
modifications to the variable containing the file name and
the label ferror jumps back to once done.
Psych was added to the build system. It requires fops.s
for maths. Two notable things need to be fixed to make it
assemble:
* d1 must be defined, presumably to a 1 following the
patterns of other decimal numbers.
* variable xx must be renamed so it doesn't clash with
the xx instruction mnemonic used by the perl cross assembler.
The naive order of the assembly files given to the assembler was to
put the st?.s files first followed by fop.s at the end. This caused
Space Travel to crash, why that happend is explained below.
st1.s through st5.s contain code, while st6. and st7.s contain
constants and data storage respectively. The order of the given
assembler files meant that the code st?.s ended up at 010000 onwards,
while the code in fop.s ended up after the data storage defined in
st7.s. In particular the function fmp ended up immediately after dspl.
At the end of st7.s there is a display list, dspl. At runtime this
list is pointed to by the clistp variable stored at the auto-indexing
memory register at address 017. Auto-indexing means that the pointer
will be incremented after use. However the display list dspl does not
reserve memory for the entire list, instead it assumes that the memory
immediately after dpsl is unused. This conflicts with storing fmp from
fop.s immediately after st7.s!
Despite this conflict the first few frames of space travel could be
rendered because clistp was never referenced. By the time it was
referenced it changed the code in fmp to invalid code. At a later
stage when fmp was called the program crashed.
The solution I chose is to reorder the files at the command-line given
to the assembler; specifically I opted to put fop.s after all the code
in st1.s though st4.s.
This means that Space Travel now runs without crashing!
simplify boot paper tape (do one xfer of 4K, instead of 3K+1K)
Combine cmd/scope.cas cmd/cas.cas into sys/cas.in
Add src/other/chrtbl.s
write chrtbl.out in "ptr" format
source code is now in src/alt. All the files in src/sys and src/cmd
are only annotated versions of the files in scans/.
The build/Makefile has been updated so that it can build the
original PDP-7 Unix or the alternative version.
Phil has fixed the Graphics-2 code, so that there is no need to
have any #ifdef code in the original kernel.
system which corresponds to a later development point, around mid-1971,
where the system had lost the "dd" directory and gained . and ..
entries. This is close to the system as described in
http://www.tuhs.org/Archive/PDP-11/Distributions/research/McIlroy_v0/UnixEditionZero.txt
except there are no pathnames.
To use this version: cd build; make alt; make altrun
Sebastian Rasmussen