jake/SplitBit-Emulator
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity
Files
588e02aff50f57da432808ec66eabce33e53ce93
SplitBit-Emulator/Tests/makedisks.sh
T
AnachronautandClaude Opus 5 36a1b07b5b D2: the machine walks a path 
sbfsFind takes a path where it used to take a name: names with '/' between
them, walked from the root, with '.' and '..'. Each name is looked for among
the entries whose parent is where the walk has got to. A bare name is a path of
one name, so everything written before directories existed still works and
still costs one walk of the directory.

sbfsMount takes either version. On a version one disk every entry has zeroes
where a parent goes and the walk starts at zero, so the comparison always
agrees - which is how a flat disk reads correctly here with nothing done to it.

PROGRAMS DID NOT HAVE TO BE TAUGHT ANY OF THIS. Resolution sits inside
sbfsFind, below the services, so every osFile call keeps its signature and a
path is simply a longer name. Type, More, Edit and the assembler gained
subdirectories without a line changing in any of them.

Four things this turned up, none of which was the path walk:

load copied the path into a buffer sized for a NAME, so anything over 22
characters was cut short - and cut short into a path that often still resolved.
"/Apps/Deep/../../Apps/Say.sbx" became "/Apps/Deep/../../Apps/" and reported
that the program was a directory. That is the whole of what looked like a bug
in '..', and it cost most of the time here.

load on a directory SUCCEEDED. A directory has no blocks, so reading it reads
nothing and leaves the staging area holding whatever was staged last - which,
if that was a program, still says SBEX and still has a working entry address.
It handed back the program before it. Refused outright now.

delete and rename on a directory are refused, and save refuses one up front
rather than failing at the rename and leaving a temporary behind. Deleting a
directory frees an entry index, and a parent IS an index, so the next file
created would take it and inherit the children.

create writes the parent rather than leaving it zero by luck. It would be zero
- delete wipes all thirty two bytes and a fresh entry never had any - but that
is a fact about two other routines, and a file appearing inside a directory it
was never put in is not a failure anybody would think to look for.

dir marks directories and counts them apart from files, because at this point
it was calling them files of no bytes.

Two hazards written down in the design note turned out not to be real, and both
were checked rather than argued about:

The lookup cache holding 22 bytes of a longer path cannot hand back the wrong
file - textSame wants both strings to end in the same place, so a cut down
entry misses. It can never HIT either, though, so every path longer than a name
went to the disk every time; it holds a whole path now.

The allocator stepping over directories changes nothing any test can see. A
directory has no start as well as no blocks, so its bounds are nought to nought
and no candidate begins before it ends. The four instructions stay, with a
comment saying they are not load bearing today and why they are there anyway.

makedisks.sh resolves its build path before it cds. Given a relative one it
carried on and quietly built disks missing some of their files, which is how
the tree fixture lost a file and sent me looking for a bug in '..'.

Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01E2JrLzFvuFX9fgi1LDRjrW
2026-08-24 19:08:26 -04:00

258 lines
15 KiB
Bash
Executable File
Raw Blame History

#!/usr/bin/env bash
# Builds the disk images the tests read from.
#
# These are made with SplitDisk, which is the other implementation of the same format.
# That is the point of them: a SplitBit program reading one of these is being checked
# against something written by different code from a written specification, rather than
# against itself.
#
# Written by Anachronaut
set -eu
# Made absolute before anything else. This script cds into its own working directory part
# of the way down, so a relative build path would be measured from the wrong place from
# there on - and the way that failed was not an error but a disk quietly missing some of
# the files it was supposed to have, which is a much worse thing to debug.
mkdir -p "$1"
BUILD="$(cd "$1" && pwd)"
ROOT="$(cd "$(dirname "$0")/.." && pwd)"
TOOL="$ROOT/SplitDisk"
DISKS="$BUILD/disks"
WORK="$BUILD/.diskwork"
[ -x "$TOOL" ] || { echo "SplitDisk is not built."; exit 1; }
mkdir -p "$DISKS" "$WORK"
# Two directory blocks, so that a file can be put beyond the first one and the walk from
# block to block gets exercised rather than assumed.
"$TOOL" format "$DISKS/sbfs.img" 64 2 >/dev/null
cd "$WORK"
printf 'hello from a file' > greeting.txt
# Eight files fill the first directory block exactly, so everything after this lands in
# the second one.
for i in 1 2 3 4 5 6 7 8; do printf 'filler %d' "$i" > "filler$i.txt"; done
# Longer than a block, so reading it has to cross from one to the next. The pattern
# repeats every twenty six bytes, which makes a misplaced block obvious to read.
awk 'BEGIN { for (i = 0; i < 700; i++) printf "%c", 65 + (i % 26) }' > across.txt
: > empty.txt
printf 'exactly twenty two!!!!' > longname.txt
"$TOOL" put "$DISKS/sbfs.img" greeting.txt >/dev/null
for i in 1 2 3 4 5 6 7 8; do "$TOOL" put "$DISKS/sbfs.img" "filler$i.txt" >/dev/null; done
"$TOOL" put "$DISKS/sbfs.img" across.txt >/dev/null
"$TOOL" put "$DISKS/sbfs.img" empty.txt >/dev/null
"$TOOL" put "$DISKS/sbfs.img" longname.txt aName22CharactersLong! >/dev/null
# A disk with a loadable program on it. The program is assembled here rather than kept as
# bytes, so that what gets loaded is always built from the source beside it.
"$TOOL" format "$DISKS/load.img" 64 2 >/dev/null
# The assembler writes a loadable program itself, because the source says where it goes.
"$ROOT/Assembler" "$ROOT/Programs/Loader/loadable.asm" -o "$WORK/hello.sbx" >/dev/null
"$TOOL" put "$DISKS/load.img" "$WORK/hello.sbx" >/dev/null
# A disk for CosmOS. greet.sbx asks the system for everything it does rather than talking
# to the hardware itself, so loading and running it exercises the whole path: the loader,
# the vector table, the service handlers, and giving the machine back at the end.
#
# hello.sbx is the opposite case, and that is why it is here: it is the original
# hello.asm, written before any of this existed, and it still writes straight to port
# 0x00 rather than calling osPrintString. A program is allowed to reach past the system
# to the hardware, so something has to check that one still gives the machine back.
#
# notes.txt is there so that loading something that is not a program can be tried too.
# 512K, and four directory blocks for 32 files. It was 64K and 16, which was ample for a
# disk that held programs and nothing else - but the native assembler reads SOURCE from
# here, and the CosmOS sources alone are 104,142 bytes against the 65,536 a 256 block disk
# holds. A machine that is going to assemble itself has to be able to hold its own source.
"$TOOL" format "$DISKS/cosmos.img" 2048 4 >/dev/null
"$ROOT/Assembler" -I "$ROOT/Programs/CosmOS/Source" \
"$ROOT/Programs/CosmOS/Apps/greet.asm" -o "$WORK/greet.sbx" >/dev/null
"$TOOL" put "$DISKS/cosmos.img" "$WORK/greet.sbx" >/dev/null
# Assembled to a different working name so it cannot tread on load.img's hello.sbx above,
# then put under the name the shell asks for.
"$ROOT/Assembler" -I "$ROOT/Programs/CosmOS/Source" \
"$ROOT/Programs/CosmOS/Apps/hello.asm" -o "$WORK/appHello.sbx" >/dev/null
"$TOOL" put "$DISKS/cosmos.img" "$WORK/appHello.sbx" hello.sbx >/dev/null
# Life.sbx is the one that had to be taught to stop. It runs to a still life and returns
# on its own, so the test needs no cycle limit: whether it ends is the thing being checked
# and a limit would hide the answer by supplying one.
"$ROOT/Assembler" -I "$ROOT/Programs/Libraries" -I "$ROOT/Programs/CosmOS/Source" \
"$ROOT/Programs/CosmOS/Apps/Life.asm" -o "$WORK/Life.sbx" >/dev/null
"$TOOL" put "$DISKS/cosmos.img" "$WORK/Life.sbx" >/dev/null
# Snake.sbx is the one that is played rather than watched. It reads the console a key at a
# time without ever waiting for one, so a script of moves drives it a move to a frame, and
# what is recorded is a whole game: turning, eating, growing, and running into a wall.
"$ROOT/Assembler" -I "$ROOT/Programs/Libraries" -I "$ROOT/Programs/CosmOS/Source" \
"$ROOT/Programs/CosmOS/Apps/Snake.asm" -o "$WORK/Snake.sbx" >/dev/null
"$TOOL" put "$DISKS/cosmos.img" "$WORK/Snake.sbx" >/dev/null
# Keys.sbx brings a vector of its own, which is what the version two format exists for. It
# is the only program here the system has to install anything for, so it is what says the
# whole path works: written into the file, installed at run, taken back out at exit.
"$ROOT/Assembler" -I "$ROOT/Programs/Libraries" -I "$ROOT/Programs/CosmOS/Source" \
"$ROOT/Programs/CosmOS/Apps/Keys.asm" -o "$WORK/Keys.sbx" >/dev/null
"$TOOL" put "$DISKS/cosmos.img" "$WORK/Keys.sbx" >/dev/null
# Say.sbx is the first program that can be told anything. Everything before it did the same
# thing however it was started.
"$ROOT/Assembler" -I "$ROOT/Programs/CosmOS/Source" \
"$ROOT/Programs/CosmOS/Apps/Say.asm" -o "$WORK/Say.sbx" >/dev/null
"$TOOL" put "$DISKS/cosmos.img" "$WORK/Say.sbx" >/dev/null
# Break.sbx stops itself twice and shows what the registers were each time. It needs no disk
# of its own: it only prints.
"$ROOT/Assembler" -I "$ROOT/Programs/CosmOS/Source" \
"$ROOT/Programs/CosmOS/Apps/Break.asm" -o "$WORK/Break.sbx" >/dev/null
"$TOOL" put "$DISKS/cosmos.img" "$WORK/Break.sbx" >/dev/null
printf 'this is not a program' > notes.txt
"$TOOL" put "$DISKS/cosmos.img" notes.txt >/dev/null
# A disk of its own for the writing test, with one file already on it so that what it
# writes has to be placed somewhere that does not tread on what is there.
"$TOOL" format "$DISKS/write.img" 32 1 >/dev/null
printf 'already here' > here.txt
"$TOOL" put "$DISKS/write.img" here.txt >/dev/null
# A disk of its own for the editing test, which deletes and renames things and would
# otherwise leave the writing test's disk looking nothing like the writing test expects.
# It starts with one file on it so that a document written here has to be placed around
# something, and so that the disk can be compared afterwards against one that never had
# any of it: the metadata should come back exactly as it was.
"$TOOL" format "$DISKS/edit.img" 32 1 >/dev/null
"$TOOL" put "$DISKS/edit.img" here.txt >/dev/null
# A disk for the shell's delete and rename, which is its own because those change what is
# on it. A fixture whose name has a directory in it is used as it stands rather than being
# made fresh per test, so a test that writes to a shared one would quietly change what
# every test after it sees.
"$TOOL" format "$DISKS/files.img" 32 1 >/dev/null
printf 'the first one' > one.txt
printf 'the second one' > two.txt
"$TOOL" put "$DISKS/files.img" one.txt >/dev/null
"$TOOL" put "$DISKS/files.img" two.txt >/dev/null
# A disk for the editor, holding nothing but the editor. Its own, because the whole point
# of it is that it writes: a document made on a shared fixture would turn up in the file
# listing of every test that came after it.
"$TOOL" format "$DISKS/editor.img" 256 2 >/dev/null
"$ROOT/Assembler" -I "$ROOT/Programs/Libraries" -I "$ROOT/Programs/CosmOS/Source" \
"$ROOT/Programs/CosmOS/Apps/Edit.asm" -o "$WORK/Edit.sbx" >/dev/null
"$TOOL" put "$DISKS/editor.img" "$WORK/Edit.sbx" >/dev/null
# A disk that has directories on it AND gets written to, which nothing else covers: every
# other version two fixture is read only, and every fixture that is written to is flat.
# Making a file runs the allocator over a directory entry, and a directory has no blocks,
# so it must be in nobody's way while a run of free ones is looked for. Its own disk, for
# the same reason the editor has one: what it writes would otherwise turn up in the
# listing of every test that came after it.
"$TOOL" format "$DISKS/treewrite.img" 256 2 >/dev/null
"$TOOL" mkdir "$DISKS/treewrite.img" /Folder >/dev/null
"$TOOL" mkdir "$DISKS/treewrite.img" /Folder/Inner >/dev/null
"$TOOL" put "$DISKS/treewrite.img" "$WORK/Edit.sbx" >/dev/null
# A disk for the file services, holding nothing but the program that exercises them. Its
# own, because that program writes: it tidies up after itself, but a run that stopped part
# way would leave a document behind on a fixture every later test reads.
"$TOOL" format "$DISKS/services.img" 256 2 >/dev/null
"$ROOT/Assembler" -I "$ROOT/Programs/CosmOS/Source" \
"$ROOT/Programs/CosmOS/Apps/Files.asm" -o "$WORK/Files.sbx" >/dev/null
"$TOOL" put "$DISKS/services.img" "$WORK/Files.sbx" >/dev/null
# A disk for streaming, and the only one here that has to be big: the file on it is bigger
# than the machine's Data Memory, which is the entire point of the services it checks.
#
# THE FILE IS GENERATED RATHER THAN TAKEN FROM THE REPOSITORY. Concatenating the CosmOS
# sources would be a truer picture of what streaming is for, and would change the recorded
# checksum every time a line of CosmOS was edited - so a real difference would arrive in a
# crowd of meaningless ones, which is the same trap the cycle counts used to set. This is
# 84000 bytes, which is 328 whole blocks and 32 bytes over, so the short block at the end
# is exercised rather than assumed.
"$TOOL" format "$DISKS/stream.img" 1024 2 >/dev/null
awk 'BEGIN { for (i = 0; i < 4000; i++) printf "streaming line %05d\n", i }' > big.txt
awk 'BEGIN { for (i = 0; i < 50; i++) printf "small %05d\n", i }' > small.txt
"$TOOL" put "$DISKS/stream.img" big.txt >/dev/null
"$TOOL" put "$DISKS/stream.img" small.txt >/dev/null
"$ROOT/Assembler" -I "$ROOT/Programs/CosmOS/Source" \
"$ROOT/Programs/CosmOS/Apps/Stream.asm" -o "$WORK/Stream.sbx" >/dev/null
"$TOOL" put "$DISKS/stream.img" "$WORK/Stream.sbx" >/dev/null
# A disk for Type. The first file crosses several blocks and deliberately has no zero byte
# for the application to mistake for an end marker. The second proves that zero blocks is
# a valid empty file rather than an error.
"$TOOL" format "$DISKS/type.img" 64 2 >/dev/null
printf 'first line\nsecond line\n' > readable.txt
awk 'BEGIN { for (i = 0; i < 30; i++) printf "line %02d: ABCDEFGHIJKLMNOPQRSTUVWXYZ\n", i }' >> readable.txt
: > empty.txt
"$TOOL" put "$DISKS/type.img" readable.txt >/dev/null
"$TOOL" put "$DISKS/type.img" empty.txt >/dev/null
"$ROOT/Assembler" -I "$ROOT/Programs/CosmOS/Source" \
"$ROOT/Programs/CosmOS/Apps/Type.asm" -o "$WORK/Type.sbx" >/dev/null
"$TOOL" put "$DISKS/type.img" "$WORK/Type.sbx" >/dev/null
"$ROOT/Assembler" -I "$ROOT/Programs/CosmOS/Source" \
"$ROOT/Programs/CosmOS/Apps/More.asm" -o "$WORK/More.sbx" >/dev/null
"$TOOL" put "$DISKS/type.img" "$WORK/More.sbx" >/dev/null
# A disk with directories on it, which is the whole of what version two adds. Built by the
# host tool, because at this rung the machine can read a tree and not yet make one - and
# that split is the point: the two implementations are checked against each other rather
# than each against itself.
#
# ONE NAME IS PUT DOWN TWICE, in two directories, because a name meaning something
# different in each place is what directories are for and what a flat filesystem cannot do.
#
# The two are DIFFERENT PROGRAMS on purpose. Say echoes whatever it is given and hello
# prints one fixed line, so which of them ran is written in the output. Two copies of one
# program would have been the easier fixture and a useless one: resolving to the wrong
# Say.sbx would have printed exactly what resolving to the right one prints, and the test
# would have passed while the path walk was ignoring directories entirely.
"$TOOL" format "$DISKS/tree.img" 256 4 >/dev/null
"$ROOT/Assembler" -I "$ROOT/Programs/CosmOS/Source" \
"$ROOT/Programs/CosmOS/Apps/Say.asm" -o "$WORK/Say.sbx" >/dev/null
"$ROOT/Assembler" -I "$ROOT/Programs/CosmOS/Source" \
"$ROOT/Programs/CosmOS/Apps/hello.asm" -o "$WORK/treeHello.sbx" >/dev/null
"$TOOL" mkdir "$DISKS/tree.img" /Apps >/dev/null
"$TOOL" mkdir "$DISKS/tree.img" /Apps/Deep >/dev/null
"$TOOL" put "$DISKS/tree.img" "$WORK/Say.sbx" /Apps/Say.sbx >/dev/null
"$TOOL" put "$DISKS/tree.img" "$WORK/treeHello.sbx" /Apps/Deep/Say.sbx >/dev/null
printf 'this is not a program' > rooted.txt
"$TOOL" put "$DISKS/tree.img" rooted.txt >/dev/null
# A disk for invoking a program by typing its name. Four files, each there to say one
# thing about how a typed word turns into a file name.
#
# dir.sbx is a working program under the name of a built-in command, which is the only way
# to check that the built-ins really are tried first. If the search ever moved ahead of the
# dispatch chain, this disk would start answering "dir" with a program.
#
# notes.sbx is text under a program's name, so that a file that IS found and IS NOT a
# program can be told apart from a word that names nothing. notes.txt is the same text
# under its own name, which no typed word can reach: "notes.txt" looks for notes.txt.sbx.
"$TOOL" format "$DISKS/invoke.img" 64 2 >/dev/null
"$ROOT/Assembler" -I "$ROOT/Programs/CosmOS/Source" \
"$ROOT/Programs/CosmOS/Apps/Say.asm" -o "$WORK/Say.sbx" >/dev/null
"$TOOL" put "$DISKS/invoke.img" "$WORK/Say.sbx" >/dev/null
"$TOOL" put "$DISKS/invoke.img" "$WORK/Say.sbx" dir.sbx >/dev/null
printf 'this is not a program' > notes.txt
"$TOOL" put "$DISKS/invoke.img" notes.txt >/dev/null
"$TOOL" put "$DISKS/invoke.img" notes.txt notes.sbx >/dev/null
# A disk for the assembler that runs on the machine. It holds a source file and the two
# programs that check the front end - the reader on its own and the tokenizer on its own -
# because a fault in either of those would otherwise turn up much later as a mysterious
# wrong byte in an output file.
#
# hello.asm is here rather than something written for the occasion because it is the
# oldest program in the repository: the first thing this machine ever ran is the first
# thing it assembles for itself.
"$TOOL" format "$DISKS/asm.img" 2048 4 >/dev/null
cp "$ROOT/Programs/Examples/hello.asm" hello.asm
"$TOOL" put "$DISKS/asm.img" hello.asm >/dev/null
"$ROOT/Assembler" -I "$ROOT/Programs/CosmOS/Source" -I "$ROOT/Programs/CosmOS/Assembler" \
"$ROOT/Programs/CosmOS/Assembler/readTest.asm" -o "$WORK/readTest.sbx" >/dev/null
"$TOOL" put "$DISKS/asm.img" "$WORK/readTest.sbx" >/dev/null
"$ROOT/Assembler" -I "$ROOT/Programs/CosmOS/Source" -I "$ROOT/Programs/CosmOS/Assembler" \
"$ROOT/Programs/CosmOS/Assembler/tokenTest.asm" -o "$WORK/tokenTest.sbx" >/dev/null
"$TOOL" put "$DISKS/asm.img" "$WORK/tokenTest.sbx" >/dev/null
Reference in New Issue View Git Blame Copy Permalink