Files
AnachronautandClaude Opus 5 4b1c3d8e3f A missing file is an error, not a core dump
Naming a pad file that is not there printed the error and then said the
machine had STARTED. MACHINE_OK is nought and the code returned nought, so
the front end ran a machine whose clock had never been set up and divided
by it: a typo in a path came out as a floating point exception and a core
dump.

The trap is two functions in one file with opposite conventions -
machineStart returns MACHINE_OK for worked, machineRestart thirty lines up
returns 1 for worked - and this copied the nearer one. Both of the returns
I added last week had it.

Checked now for all three files the replay suite is about, because the
same mistake fits all of them, and re-broken to be sure: the check comes
back exit 136, which is a signal 8, which is the crash.

Found by somebody typing a path that was not there, which is the fourth
thing this week that no test would have reached.

Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01E2JrLzFvuFX9fgi1LDRjrW
2026-09-03 15:50:40 -04:00

148 lines
6.6 KiB
Bash
Executable File

#!/usr/bin/env bash
# Checks that what the machine records is what it plays back.
#
# --record-pad writes what a controller held, one byte a frame, in exactly the format --pad
# reads. That symmetry is the whole feature and it is the whole of what is checked here: a
# recording is only worth having if playing it back does what was recorded.
#
# ---- Why it exists ----
#
# Some inputs cannot sensibly be written by hand. Flying a lander from one base to another is
# a few hundred frames of steering that has to arrive somewhere eight cells wide, and several
# attempts at hand-authoring one got within two columns and no closer. That is a piloting
# exercise rather than a test. Playing it once and keeping what happened is the answer, and
# this is what says the keeping works.
#
# Written by Anachronaut
set -u
ROOT="$(cd "$(dirname "$0")/.." && pwd)"
BUILD="$ROOT/Tests/build/replay"
ASM="$ROOT/Assembler"
EMU="$ROOT/SplitBit"
for tool in "$ASM" "$EMU"; do
[ -x "$tool" ] || { echo "$(basename "$tool") is not built."; exit 1; }
done
rm -rf "$BUILD"; mkdir -p "$BUILD"
PASS=0
FAIL=0
FAILED_NAMES=()
GREEN=$'\033[32m'; RED=$'\033[31m'; RESET=$'\033[0m'
[ -t 1 ] || { GREEN=""; RED=""; RESET=""; }
result() {
if [ "$1" = "ok" ]; then
PASS=$((PASS + 1)); printf " [%sok %s] %-40s %s\n" "$GREEN" "$RESET" "$2" "$3"
else
FAIL=$((FAIL + 1)); FAILED_NAMES+=("$2")
printf " [%sFAIL%s] %-40s %s\n" "$RED" "$RESET" "$2" "$3"
fi
}
echo "Checking that a recording plays back as itself."
# A program that reads a pad every frame and stops. It is the one in testPrograms rather than
# a new one, because what is being checked is the recording and not the reading.
"$ASM" "$ROOT/Programs/testPrograms/padTest.asm" -o "$BUILD/padTest.bin" > "$BUILD/asm.log" 2>&1 \
|| { echo "padTest did not assemble."; exit 1; }
# ---- Played and recorded at the same time ----
#
# The strongest form of the claim: the recording is made OF a playback, so if the two formats
# ever drift apart the bytes coming out stop matching the bytes going in.
"$EMU" --fast --pad "$ROOT/Tests/input/padTest.pad" \
--record-pad "$BUILD/roundtrip.pad" "$BUILD/padTest.bin" > "$BUILD/roundtrip.out" 2>&1
SAME="$(python3 -c "
first = open('$ROOT/Tests/input/padTest.pad', 'rb').read()
again = open('$BUILD/roundtrip.pad', 'rb').read()
print('yes' if again[:len(first)] == first else 'no')
" 2>/dev/null || echo error)"
[ "$SAME" = "yes" ] \
&& result ok "a recording of a playback is the playback" "byte for byte, in and out" \
|| result no "a recording of a playback is the playback" "the bytes differ"
# ---- And it does not matter which controller it was ----
#
# The first recording ever made with this came back 1,766 frames of nothing. It recorded pad
# NOUGHT and the controller was somewhere else - which pad one lands on is an accident of the
# host, and the flight had to be flown again for nothing.
#
# So every pad is or-ed into the byte. Here the same recording is played on pad ONE, with
# nought holding nothing, and what comes out has to be what went in.
python3 -c "open('$BUILD/idle.pad','wb').write(b'\x00' * 64)"
"$EMU" --fast --pad "$BUILD/idle.pad" --pad "$ROOT/Tests/input/padTest.pad" \
--record-pad "$BUILD/padone.pad" "$BUILD/padTest.bin" > "$BUILD/padone.out" 2>&1
ELSEWHERE="$(python3 -c "
first = open('$ROOT/Tests/input/padTest.pad', 'rb').read()
again = open('$BUILD/padone.pad', 'rb').read()
print('yes' if again[:len(first)] == first else 'no')
" 2>/dev/null || echo error)"
[ "$ELSEWHERE" = "yes" ] \
&& result ok "a controller on any pad is recorded" "flown on pad one, written all the same" \
|| result no "a controller on any pad is recorded" "the bytes differ"
# ---- One byte a frame, and the frame is the machine's ----
#
# A recording is a TIMELINE. If it were a byte a read, a program that polled twice in one
# frame would record twice as fast as it flew; if it skipped frames nobody looked at, it would
# play back faster than it was flown. So the count is the number of frames the machine ran,
# which at 16,667 cycles a frame is arithmetic rather than a guess.
"$EMU" --fast --cycles 500000 --record-pad "$BUILD/timeline.pad" \
"$BUILD/padTest.bin" > "$BUILD/timeline.out" 2>&1
WROTE="$(wc -c < "$BUILD/timeline.pad" | tr -d ' ')"
# The machine halts of its own accord well before the limit, so what is checked is that the
# count is frames-of-something rather than bytes-of-nothing: more than none, and fewer than
# the limit could possibly hold.
[ "$WROTE" -gt 0 ] && [ "$WROTE" -le 30 ] \
&& result ok "a byte for every frame and no more" "$WROTE frames of it" \
|| result no "a byte for every frame and no more" "$WROTE bytes, which is not a frame count"
# ---- And a machine with no controller records that honestly ----
#
# Nothing held is nought, which is what a pad nobody is touching reports and what a machine
# with no pad at all reports. A recorder that wrote something else would make a fixture that
# pressed buttons nobody pressed.
QUIET="$(python3 -c "
data = open('$BUILD/timeline.pad', 'rb').read()
print('yes' if set(data) <= {0} else 'no')
" 2>/dev/null || echo error)"
[ "$QUIET" = "yes" ] \
&& result ok "and nothing held is written as nothing" "no buttons nobody pressed" \
|| result no "and nothing held is written as nothing" "the recording holds something"
# ---- A file that is not there is an error, not a crash ----
#
# Naming a pad file that does not exist used to print the error and then say the machine had
# STARTED: MACHINE_OK is nought, and the code returned nought. The front end then ran a
# machine whose clock had never been set up and divided by it, so a typo in a path came out as
# a floating point exception and a core dump.
#
# The trap was two functions in one file with opposite conventions - machineRestart returns 1
# for worked - and the wrong neighbour being copied. Checked here for all three files this
# suite is about, because the same mistake fits all of them.
for missing in "--pad" "--keyboard" "--record-pad"; do
case "$missing" in
--record-pad) where="/nowhere/at/all.pad" ;;
*) where="$BUILD/there-is-no-such-file" ;;
esac
"$EMU" $missing "$where" "$BUILD/padTest.bin" > "$BUILD/missing.out" 2>&1
status=$?
if [ "$status" = "1" ] && grep -q "Error:" "$BUILD/missing.out"; then
result ok "$missing with no file stops cleanly" "an error and a status, not a signal"
else
result no "$missing with no file stops cleanly" "exit $status"
fi
done
echo
if [ "$FAIL" -eq 0 ]; then
echo "All $PASS replay checks passed."
exit 0
fi
echo "$PASS passed, $FAIL failed: ${FAILED_NAMES[*]}"
exit 1