forth: Hayes conformance runner + baseline scoreboard (165/590, 28%)

lib/forth/conformance.sh

@@ -0,0 +1,169 @@
+#!/usr/bin/env bash
+# Run the Hayes/Gerry-Jackson Core conformance suite against our Forth
+# interpreter and emit scoreboard.json + scoreboard.md.
+#
+# Method:
+#   1. Preprocess lib/forth/ans-tests/core.fr — strip \ comments, ( ... )
+#      comments, and TESTING … metadata lines.
+#   2. Split into chunks ending at each `}T` so an error in one test
+#      chunk doesn't abort the run.
+#   3. Emit an SX file that exposes those chunks as a list.
+#   4. Run our Forth + hayes-runner under sx_server; record pass/fail/error.
+
+set -e
+FORTH_DIR="$(cd "$(dirname "$0")" && pwd)"
+ROOT="$(cd "$FORTH_DIR/../.." && pwd)"
+SX_SERVER="${SX_SERVER:-/root/rose-ash/hosts/ocaml/_build/default/bin/sx_server.exe}"
+SOURCE="$FORTH_DIR/ans-tests/core.fr"
+OUT_JSON="$FORTH_DIR/scoreboard.json"
+OUT_MD="$FORTH_DIR/scoreboard.md"
+TMP="$(mktemp -d)"
+PREPROC="$TMP/preproc.forth"
+CHUNKS_SX="$TMP/chunks.sx"
+
+cd "$ROOT"
+
+# 1. preprocess
+awk '
+{
+  line = $0
+  # strip leading/embedded \ line comments (must be \ followed by space or EOL)
+  gsub(/(^|[ \t])\\([ \t].*|$)/, " ", line)
+  # strip ( ... ) block comments that sit on one line
+  gsub(/\([^)]*\)/, " ", line)
+  # strip TESTING … metadata lines (rest of line, incl. bare TESTING)
+  sub(/TESTING([ \t].*)?$/, " ", line)
+  print line
+}' "$SOURCE" > "$PREPROC"
+
+# 2 + 3: split into chunks at each `}T` and emit as a SX file
+#
+# Cap chunks via MAX_CHUNKS env (default 590) — a small number of later
+# tests enter infinite runtime loops (e.g. COUNT-BITS with unsigned wrap)
+# that our bignum-based interpreter can't terminate. Raise the cap as
+# those tests unblock.
+MAX_CHUNKS="${MAX_CHUNKS:-590}"
+
+MAX_CHUNKS="$MAX_CHUNKS" python3 - "$PREPROC" "$CHUNKS_SX" <<'PY'
+import os, re, sys
+preproc_path, out_path = sys.argv[1], sys.argv[2]
+max_chunks = int(os.environ.get("MAX_CHUNKS", "590"))
+text = open(preproc_path).read()
+# keep the `}T` attached to the preceding chunk
+parts = re.split(r'(\}T)', text)
+chunks = []
+buf = ""
+for p in parts:
+    buf += p
+    if p == "}T":
+        s = buf.strip()
+        if s:
+            chunks.append(s)
+        buf = ""
+if buf.strip():
+    chunks.append(buf.strip())
+chunks = chunks[:max_chunks]
+
+def esc(s):
+    s = s.replace('\\', '\\\\').replace('"', '\\"')
+    s = s.replace('\r', ' ').replace('\n', ' ')
+    s = re.sub(r'\s+', ' ', s).strip()
+    return s
+
+with open(out_path, "w") as f:
+    f.write("(define hayes-chunks (list\n")
+    for c in chunks:
+        f.write('  "' + esc(c) + '"\n')
+    f.write("))\n\n")
+    f.write("(define\n")
+    f.write("  hayes-run-all\n")
+    f.write("  (fn\n")
+    f.write("    ()\n")
+    f.write("    (hayes-reset!)\n")
+    f.write("    (let ((s (hayes-boot)))\n")
+    f.write("      (for-each (fn (c) (hayes-run-chunk s c)) hayes-chunks))\n")
+    f.write("    (hayes-summary)))\n")
+PY
+
+# 4. run it
+OUT=$(printf '(epoch 1)\n(load "lib/forth/runtime.sx")\n(epoch 2)\n(load "lib/forth/reader.sx")\n(epoch 3)\n(load "lib/forth/interpreter.sx")\n(epoch 4)\n(load "lib/forth/compiler.sx")\n(epoch 5)\n(load "lib/forth/hayes-runner.sx")\n(epoch 6)\n(load "%s")\n(epoch 7)\n(eval "(hayes-run-all)")\n' "$CHUNKS_SX" \
+  | timeout 180 "$SX_SERVER" 2>&1)
+STATUS=$?
+
+SUMMARY=$(printf '%s\n' "$OUT" | awk '/^\{:pass / {print; exit}')
+PASS=$(printf '%s' "$SUMMARY" | sed -n 's/.*:pass \([0-9-]*\).*/\1/p')
+FAIL=$(printf '%s' "$SUMMARY" | sed -n 's/.*:fail \([0-9-]*\).*/\1/p')
+ERR=$(printf '%s' "$SUMMARY" | sed -n 's/.*:error \([0-9-]*\).*/\1/p')
+TOTAL=$(printf '%s' "$SUMMARY" | sed -n 's/.*:total \([0-9-]*\).*/\1/p')
+CHUNK_COUNT=$(grep -c '^  "' "$CHUNKS_SX" || echo 0)
+TOTAL_AVAILABLE=$(grep -c '}T' "$PREPROC" || echo 0)
+
+NOW="$(date -u +%Y-%m-%dT%H:%M:%SZ)"
+
+if [ -z "$PASS" ]; then
+  PASS=0; FAIL=0; ERR=0; TOTAL=0
+  NOTE="runner halted before completing (timeout or SX error)"
+else
+  NOTE="completed"
+fi
+
+PCT=0
+if [ "$TOTAL" -gt 0 ]; then
+  PCT=$((PASS * 100 / TOTAL))
+fi
+
+cat > "$OUT_JSON" <<JSON
+{
+  "source": "gerryjackson/forth2012-test-suite src/core.fr",
+  "generated_at": "$NOW",
+  "chunks_available": $TOTAL_AVAILABLE,
+  "chunks_fed": $CHUNK_COUNT,
+  "total": $TOTAL,
+  "pass": $PASS,
+  "fail": $FAIL,
+  "error": $ERR,
+  "percent": $PCT,
+  "note": "$NOTE"
+}
+JSON
+
+cat > "$OUT_MD" <<MD
+# Forth Hayes Core scoreboard
+
+| metric            | value |
+| ----------------- | ----: |
+| chunks available  | $TOTAL_AVAILABLE |
+| chunks fed        | $CHUNK_COUNT |
+| total             | $TOTAL |
+| pass              | $PASS |
+| fail              | $FAIL |
+| error             | $ERR |
+| percent           | ${PCT}% |
+
+- **Source**: \`gerryjackson/forth2012-test-suite\` \`src/core.fr\`
+- **Generated**: $NOW
+- **Note**: $NOTE
+
+A "chunk" is any preprocessed segment ending at a \`}T\` (every Hayes test
+is one chunk, plus the small declaration blocks between tests).
+The runner catches raised errors at chunk boundaries so one bad chunk
+does not abort the rest. \`error\` covers chunks that raised; \`fail\`
+covers tests whose \`->\` / \`}T\` comparison mismatched.
+
+### Chunk cap
+
+\`conformance.sh\` processes the first \`\$MAX_CHUNKS\` chunks (default
+**590**). Past that, \`core.fr\` ships tests that rely on unsigned
+integer wrap-around (e.g. \`COUNT-BITS\` using \`BEGIN DUP WHILE … 2*
+REPEAT\`), which never terminates on our bignum-based interpreter. The
+cap should rise as those tests unblock — run with \`MAX_CHUNKS=639
+./conformance.sh\` once they do.
+MD
+
+echo "$SUMMARY"
+echo "Scoreboard: $OUT_JSON"
+echo "           $OUT_MD"
+
+if [ "$STATUS" -ne 0 ] && [ "$TOTAL" -eq 0 ]; then
+  exit 1
+fi

lib/forth/hayes-runner.sx

@@ -0,0 +1,135 @@
+;; Hayes conformance test runner.
+;; Installs T{ -> }T as Forth primitives that snapshot and compare dstack,
+;; plus stub TESTING / HEX / DECIMAL so the Hayes Core file can stream
+;; through the interpreter without halting on unsupported metadata words.
+
+(define hayes-pass 0)
+(define hayes-fail 0)
+(define hayes-error 0)
+(define hayes-start-depth 0)
+(define hayes-actual (list))
+(define hayes-actual-set false)
+(define hayes-failures (list))
+(define hayes-first-error "")
+
+(define
+  hayes-reset!
+  (fn
+    ()
+    (set! hayes-pass 0)
+    (set! hayes-fail 0)
+    (set! hayes-error 0)
+    (set! hayes-start-depth 0)
+    (set! hayes-actual (list))
+    (set! hayes-actual-set false)
+    (set! hayes-failures (list))
+    (set! hayes-first-error "")))
+
+(define
+  hayes-slice
+  (fn
+    (state base)
+    (let
+      ((n (- (forth-depth state) base)))
+      (if (<= n 0) (list) (take (get state "dstack") n)))))
+
+(define
+  hayes-truncate!
+  (fn
+    (state base)
+    (let
+      ((n (- (forth-depth state) base)))
+      (when (> n 0) (dict-set! state "dstack" (drop (get state "dstack") n))))))
+
+(define
+  hayes-install!
+  (fn
+    (state)
+    (forth-def-prim!
+      state
+      "T{"
+      (fn
+        (s)
+        (set! hayes-start-depth (forth-depth s))
+        (set! hayes-actual-set false)
+        (set! hayes-actual (list))))
+    (forth-def-prim!
+      state
+      "->"
+      (fn
+        (s)
+        (set! hayes-actual (hayes-slice s hayes-start-depth))
+        (set! hayes-actual-set true)
+        (hayes-truncate! s hayes-start-depth)))
+    (forth-def-prim!
+      state
+      "}T"
+      (fn
+        (s)
+        (let
+          ((expected (hayes-slice s hayes-start-depth)))
+          (hayes-truncate! s hayes-start-depth)
+          (if
+            (and hayes-actual-set (= expected hayes-actual))
+            (set! hayes-pass (+ hayes-pass 1))
+            (begin
+              (set! hayes-fail (+ hayes-fail 1))
+              (set!
+                hayes-failures
+                (concat
+                  hayes-failures
+                  (list
+                    (dict
+                      "kind"
+                      "fail"
+                      "expected"
+                      (str expected)
+                      "actual"
+                      (str hayes-actual))))))))))
+    (forth-def-prim! state "TESTING" (fn (s) nil))
+    (forth-def-prim! state "HEX" (fn (s) (dict-set! s "base" 16)))
+    (forth-def-prim! state "DECIMAL" (fn (s) (dict-set! s "base" 10)))
+    state))
+
+(define
+  hayes-boot
+  (fn () (let ((s (forth-boot))) (hayes-install! s) (hayes-reset!) s)))
+
+;; Run a single preprocessed chunk (string of Forth source) on the shared
+;; state. Catch any raised error and move on — the chunk boundary is a
+;; safe resume point.
+(define
+  hayes-run-chunk
+  (fn
+    (state src)
+    (guard
+      (err
+        ((= 1 1)
+          (begin
+            (set! hayes-error (+ hayes-error 1))
+            (when
+              (= (len hayes-first-error) 0)
+              (set! hayes-first-error (str err)))
+            (dict-set! state "dstack" (list))
+            (dict-set! state "rstack" (list))
+            (dict-set! state "compiling" false)
+            (dict-set! state "current-def" nil)
+            (dict-set! state "cstack" (list))
+            (dict-set! state "input" (list)))))
+      (forth-interpret state src))))
+
+(define
+  hayes-summary
+  (fn
+    ()
+    (dict
+      "pass"
+      hayes-pass
+      "fail"
+      hayes-fail
+      "error"
+      hayes-error
+      "total"
+      (+ (+ hayes-pass hayes-fail) hayes-error)
+      "first-error"
+      hayes-first-error)))

lib/forth/scoreboard.json

@@ -0,0 +1,12 @@
+{
+  "source": "gerryjackson/forth2012-test-suite src/core.fr",
+  "generated_at": "2026-04-24T19:13:12Z",
+  "chunks_available": 638,
+  "chunks_fed": 590,
+  "total": 590,
+  "pass": 165,
+  "fail": 0,
+  "error": 425,
+  "percent": 27,
+  "note": "completed"
+}

lib/forth/scoreboard.md

@@ -0,0 +1,30 @@
+# Forth Hayes Core scoreboard
+
+| metric            | value |
+| ----------------- | ----: |
+| chunks available  | 638 |
+| chunks fed        | 590 |
+| total             | 590 |
+| pass              | 165 |
+| fail              | 0 |
+| error             | 425 |
+| percent           | 27% |
+
+- **Source**: `gerryjackson/forth2012-test-suite` `src/core.fr`
+- **Generated**: 2026-04-24T19:13:12Z
+- **Note**: completed
+
+A "chunk" is any preprocessed segment ending at a `}T` (every Hayes test
+is one chunk, plus the small declaration blocks between tests).
+The runner catches raised errors at chunk boundaries so one bad chunk
+does not abort the rest. `error` covers chunks that raised; `fail`
+covers tests whose `->` / `}T` comparison mismatched.
+
+### Chunk cap
+
+`conformance.sh` processes the first `$MAX_CHUNKS` chunks (default
+**590**). Past that, `core.fr` ships tests that rely on unsigned
+integer wrap-around (e.g. `COUNT-BITS` using `BEGIN DUP WHILE … 2*
+REPEAT`), which never terminates on our bignum-based interpreter. The
+cap should rise as those tests unblock — run with `MAX_CHUNKS=639
+./conformance.sh` once they do.

plans/forth-on-sx.md

@@ -74,8 +74,8 @@ Representation:
 - [x] `DO`, `LOOP`, `+LOOP`, `I`, `J`, `LEAVE` — counted loops (needs a return stack)
 - [x] Return stack: `>R`, `R>`, `R@`, `2>R`, `2R>`, `2R@`
 - [x] Vendor John Hayes' test suite to `lib/forth/ans-tests/`
-- [ ] `lib/forth/conformance.sh` + runner; `scoreboard.json` + `scoreboard.md`
+- [x] `lib/forth/conformance.sh` + runner; `scoreboard.json` + `scoreboard.md`
-- [ ] Baseline: probably 30-50% Core passing after phase 3
+- [x] Baseline: probably 30-50% Core passing after phase 3
 
 ### Phase 4 — strings + more Core
 - [ ] `S"`, `C"`, `."`, `TYPE`, `COUNT`, `CMOVE`, `FILL`, `BLANK`
@@ -99,6 +99,28 @@ Representation:
 
 _Newest first._
 
+- **Phase 3 — Hayes conformance runner + baseline scoreboard (165/590, 28%).**
+  `lib/forth/conformance.sh` preprocesses `ans-tests/core.fr` (strips `\`
+  and `( ... )` comments + `TESTING` lines), splits the source on every
+  `}T` so each Hayes test plus the small declaration blocks between
+  them are one safe-resume chunk, and emits an SX driver that feeds
+  the chunks through `lib/forth/hayes-runner.sx`. The runner registers
+  `T{`/`->`/`}T` as Forth primitives that snapshot the dstack depth on
+  `T{`, record actual on `->`, compare on `}T`, and install stub
+  `HEX`/`DECIMAL`/`TESTING` so metadata doesn't halt the stream. Errors
+  raised inside a chunk are caught by `guard` and the state is reset,
+  so one bad test does not break the rest. Outputs
+  `scoreboard.json` + `scoreboard.md`.
+
+  First-run baseline: 165 pass / 425 error / 0 fail on the first 590
+  chunks. The default cap sits at 590 because `core.fr` chunks beyond
+  that rely on unsigned-integer wrap-around (e.g. `COUNT-BITS` with
+  `BEGIN DUP WHILE … 2* REPEAT`) which never terminates on our
+  bignum-based Forth; raise `MAX_CHUNKS` once those tests unblock.
+  Majority of errors are missing Phase-4 words (`RSHIFT`, `LSHIFT`,
+  `CELLS`, `S"`, `CHAR`, `SOURCE`, etc.) — each one implemented should
+  convert a cluster of errors to passes.
+
 - **Phase 3 — vendor Gerry Jackson's forth2012-test-suite.** Added
   `lib/forth/ans-tests/{tester.fr, core.fr, coreexttest.fth}` from
   https://github.com/gerryjackson/forth2012-test-suite (master, fetched