Files
SplitBit-Emulator/Tests/run.sh
T
Anachronaut c3188ed657 Seventy becomes seventy one: a machine that can wait
HALT is terminal - stepCPU returns at once when the Halt Flag is up, so a
halted machine does not execute, service devices, or take an interrupt -
and that has to stay true, because every test ends with a halt and "halted"
is how a program says it has finished. The consequence was that SplitBit
had no way to wait at all. Every wait was a spin, and a spin is bus
traffic: 11.5% of Type over a 14K file on a disk of ten thousand cycles,
after read-ahead had already hidden three quarters of the latency.

WAIT is 0xFE, one byte, no operands, sitting under HALT where the
instruction that almost stops the machine belongs. Three decisions in it:

- A line already standing means there is nothing to wait for, so WAIT does
  nothing. That is what makes test-then-wait race-free.
- Any line ends the wait, masked or not, so a program can sleep on a device
  it has no handler for and read its status afterwards. Masking says who
  answers a request, not whether it happened.
- A line that wakes the CPU without being dispatched is taken down by the
  WAIT. Left standing it would be found by the next WAIT, which would
  return at once - the program would spin exactly as before while looking
  as though it slept.

Waiting is NOT a Status bit, and that is the trap avoided rather than a
gap: Status rides into the interrupt frame and comes back out, so a machine
interrupted mid-wait would return from its handler still waiting, and wait
again for what it had already been given. An internal field instead.

Idle cycles are counted apart from bus cycles and the halt line says so
when there are any, which is what makes the difference observable at all -
with the line-clearing removed the total moves by ONE cycle, 20,100 against
20,099, and only the idle half changes, halving to 9,976. A test on
totals could never have seen it. Tests/terminal.sh asks that question,
being the file for things a recorded output cannot see, and fails with the
clear removed while "both reads finished" still passes.

Three collisions, all found by building it:

- 0xFE was the assembler's "not an instruction" sentinel. getOpcode now
  answers a negative NOT_AN_OPCODE, which is outside the range of every
  possible answer instead of inside the unused part of it.
- 0xFE was also what faultTest and faultResumeTest executed to provoke a
  fault. They now use 0xFD and say why, because they did not fail when it
  became an instruction - they HUNG, having started sleeping instead.
- Keys.asm has had a label called "wait" for a year, and mnemonics are
  matched uppercased. What that reported was "Branch without label" at the
  BRQ thirty lines away. The assembler now refuses a label that is already
  an instruction, at the label, by name; every instruction added takes a
  word out of the space of label names, so this will happen again.
2026-08-26 11:11:25 -04:00

262 lines
9.4 KiB
Bash
Executable File

#!/usr/bin/env bash
# run.sh
# Test runner for the SplitBit Emulator and Assembler.
# Written by Anachronaut
#
# Assembles and runs every program listed in Tests/manifest and compares the
# output against the recorded results in Tests/expected.
#
# ./Tests/run.sh Run the suite.
# ./Tests/run.sh --bless Record current output as the expected results.
# ./Tests/run.sh <name>... Run only the named tests.
#
# Programs are built inside Tests/build so that running the suite never touches
# the binaries in Programs/.
set -u
ROOT="$(cd "$(dirname "${BASH_SOURCE[0]}")/.." && pwd)"
TESTS="$ROOT/Tests"
BUILD="$TESTS/build"
EXPECTED="$TESTS/expected"
INPUT="$TESTS/input"
MANIFEST="$TESTS/manifest"
ASSEMBLER="$ROOT/Assembler"
EMULATOR="$ROOT/SplitBit"
RUN_TIMEOUT=10
BLESS=0
ONLY=()
for arg in "$@"; do
case "$arg" in
--bless) BLESS=1 ;;
-h|--help)
sed -n '3,14p' "${BASH_SOURCE[0]}" | sed 's/^# \{0,1\}//'
exit 0
;;
-*) echo "run.sh: unknown option $arg" >&2; exit 2 ;;
*) ONLY+=("$arg") ;;
esac
done
for tool in "$ASSEMBLER" "$EMULATOR"; do
if [ ! -x "$tool" ]; then
echo "run.sh: $tool is missing. Run 'make' first." >&2
exit 2
fi
done
PROGRAMS="$ROOT/Programs"
rm -rf "$BUILD"
mkdir -p "$BUILD" "$EXPECTED"
# Disk images the tests read from are built here, with the host tool, before anything
# runs. The build directory is thrown away above, so they are always freshly made.
if [ -x "$TESTS/makedisks.sh" ]; then
"$TESTS/makedisks.sh" "$BUILD" || { echo "Couldn't build the test disks."; exit 1; }
fi
PASS=0
FAIL=0
BLESSED=0
FAILED_NAMES=()
wanted() {
[ ${#ONLY[@]} -eq 0 ] && return 0
local n
for n in "${ONLY[@]}"; do [ "$n" = "$1" ] && return 0; done
return 1
}
report() {
# report <status> <name> <detail>
printf ' [%-4s] %-20s %s\n' "$1" "$2" "$3"
}
trim() {
local v="$1"
v="${v#"${v%%[![:space:]]*}"}"
v="${v%"${v##*[![:space:]]}"}"
printf '%s' "$v"
}
# Takes the cycle count out of the emulator's last line, in place.
#
# HOW MANY CYCLES A PROGRAM TOOK IS NOT WHAT ANY OF THESE TESTS ARE ABOUT, and having it in
# every recorded result made every one of them fragile in the same way: two instructions
# added to CosmOS moved the count in six unrelated files at once, so a real difference
# would have arrived in a crowd of meaningless ones and had to be picked out by hand.
#
# WHETHER a program stopped on its own or ran into its limit is kept, because that is
# behaviour and several tests exist to check it. Only the number goes.
#
# Anything that genuinely wants to measure cycles should say so out loud in a test of its
# own rather than every test carrying the measurement and nothing asserting anything about
# it.
settle() {
# NOT ANCHORED TO THE START OF A LINE. A program whose last output has no newline on it
# leaves the cursor mid line, and the halt message is printed there - so the count this
# exists to remove was sitting inside a line rather than at the head of one, and
# survived. replCalculator is the one that does that, and it was the only test to churn
# when the machine started charging for memory instead of counting instructions.
sed -i -E 's/Execution halted after [0-9]+ cycles, [0-9]+ of them waiting\./Execution halted./;
s/Execution halted after [0-9]+ cycles\./Execution halted./;
s/Execution stopped after [0-9]+ cycles\. \(cycle limit reached\)/Execution stopped. (cycle limit reached)/' "$1"
}
check() {
# check <name> <actual-file>
local name="$1" actual="$2" golden="$EXPECTED/$1.out"
if [ "$BLESS" -eq 1 ]; then
cp "$actual" "$golden"
BLESSED=$((BLESSED + 1))
report "rec" "$name" "$(wc -c < "$golden" | tr -d ' ') bytes recorded"
return 0
fi
if [ ! -f "$golden" ]; then
FAIL=$((FAIL + 1)); FAILED_NAMES+=("$name")
report "FAIL" "$name" "no recorded output; run with --bless"
return 1
fi
if cmp -s "$actual" "$golden"; then
PASS=$((PASS + 1))
report "ok" "$name" ""
return 0
fi
FAIL=$((FAIL + 1)); FAILED_NAMES+=("$name")
report "FAIL" "$name" "output differs"
diff -u "$golden" "$actual" 2>/dev/null | head -20 | sed 's/^/ /'
return 1
}
assemble() {
# assemble <name> <source>; echoes the built binary path on success.
# Everything builds from Programs/ with Libraries/ and CosmOS/ on the include path,
# and the binary goes to Tests/build, so the source tree is never written to.
# CosmOS is there because it owns the filesystem library and the service names, which
# test programs outside it include.
local name="$1" src="$2"
local bin="$BUILD/$name.bin"
if ( cd "$PROGRAMS" && "$ASSEMBLER" -I Libraries -I CosmOS/Source -o "$bin" "$src" ) >"$BUILD/.assemble.log" 2>&1; then
echo "$bin"
return 0
fi
return 1
}
while IFS='|' read -r name src mode stdin limit disk; do
name="$(trim "$name")"
[ -z "$name" ] && continue
case "$name" in \#*) continue ;; esac
src="$(trim "$src")"
mode="$(trim "$mode")"; stdin="$(trim "$stdin")"
limit="$(trim "$limit")"; disk="$(trim "$disk")"
[ -z "$disk" ] && disk="-"
wanted "$name" || continue
if [ "$mode" = "xfail" ]; then
if assemble "$name" "$src" >/dev/null; then
FAIL=$((FAIL + 1)); FAILED_NAMES+=("$name")
report "FAIL" "$name" "expected assembly to fail, but it succeeded"
else
PASS=$((PASS + 1))
report "ok" "$name" "fails as recorded: $(head -1 "$BUILD/.assemble.log" | tr -d '\033' | sed 's/\[[0-9;]*m//g')"
fi
continue
fi
if ! BIN="$(assemble "$name" "$src")"; then
FAIL=$((FAIL + 1)); FAILED_NAMES+=("$name")
report "FAIL" "$name" "assembly failed"
head -3 "$BUILD/.assemble.log" | tr -d '\033' | sed 's/\[[0-9;]*m//g' | sed 's/^/ /'
continue
fi
if [ "$mode" = "assemble" ]; then
PASS=$((PASS + 1))
report "ok" "$name" "assembles"
continue
fi
if [ "$stdin" != "-" ] && [ ! -f "$INPUT/$stdin" ]; then
FAIL=$((FAIL + 1)); FAILED_NAMES+=("$name")
report "FAIL" "$name" "missing input fixture $stdin"
continue
fi
IN=/dev/null
[ "$stdin" != "-" ] && IN="$INPUT/$stdin"
OUT="$BUILD/.out"
case "$mode" in
run)
# --fast because there is nothing to learn from waiting out the emulated
# clock, and --cycles for programs that never halt on their own, which
# bounds them by cycle count rather than by wall clock.
EMUARGS=(--fast)
[ "$limit" != "-" ] && EMUARGS+=(--cycles "$limit")
# A disk starts fresh for every run, so a test cannot pass because of what a
# previous one left lying on it. The emulator makes the image if it is
# missing, which is what removing it first arranges for.
if [ "$disk" != "-" ]; then
# A trailing :ro attaches the image write protected, so that a test can
# check the device bars writes rather than the filesystem asking nicely.
DISKFILE="${disk%:ro}"
# And a trailing @N gives the disk a latency, so that a test can check the
# filesystem waits for it. Every other test runs with the answer there
# before the next instruction, which is the one condition under which not
# waiting looks like working.
DISKWAIT=""
case "$DISKFILE" in
*@*) DISKWAIT="${DISKFILE##*@}"; DISKFILE="${DISKFILE%@*}" ;;
esac
# A name with a directory in it is one of the images makedisks.sh built,
# and is used as it stands. A bare name is scratch: it is removed first so
# that nothing a test writes can be seen by the next one, and the emulator
# makes a blank image in its place.
case "$DISKFILE" in
*/*) ;;
*) rm -f "$BUILD/$DISKFILE" ;;
esac
EMUARGS+=(--disk "$BUILD/$DISKFILE")
[ -n "$DISKWAIT" ] && EMUARGS+=(--disk-cycles "$DISKWAIT")
case "$disk" in *:ro) EMUARGS+=(--write-protect) ;; esac
fi
timeout "$RUN_TIMEOUT" "$EMULATOR" "${EMUARGS[@]}" "$BIN" <"$IN" >"$OUT" 2>&1
STATUS=$?
if [ "$STATUS" -eq 124 ]; then
FAIL=$((FAIL + 1)); FAILED_NAMES+=("$name")
report "FAIL" "$name" "did not finish within ${RUN_TIMEOUT}s"
continue
fi
# What a program exits with is part of what it does, so it is recorded
# with the output rather than thrown away. A program that faults is
# supposed to exit non zero, and that should be just as pinned down as
# what it printed.
printf '[exit %d]\n' "$STATUS" >> "$OUT"
settle "$OUT"
check "$name" "$OUT"
;;
*)
FAIL=$((FAIL + 1)); FAILED_NAMES+=("$name")
report "FAIL" "$name" "unknown mode '$mode' in manifest"
;;
esac
done < "$MANIFEST"
echo
if [ "$BLESS" -eq 1 ]; then
echo "Recorded $BLESSED expected results into Tests/expected."
exit 0
fi
if [ "$FAIL" -eq 0 ]; then
echo "All $PASS tests passed."
exit 0
fi
echo "$PASS passed, $FAIL failed: ${FAILED_NAMES[*]}"
exit 1