Datasets:

Qwen
/

RecreationBench

#!/usr/bin/env python3
"""RB TestGen Kit — depth/jump audit, verdict helpers, and portability lint.
LOCATION ON THE VM: deployed read-only to the testgen output dir.
(NEVER edit; `import rb_testgen_kit` or `from rb_uia import rb_showing, rb_reach, ...`)
Tests use pytest format with @pytest.mark.rb() markers. The conftest.py handles
result collection automatically. This kit provides:
  - Verdict helpers: rb_showing, rb_reach, rb_flip, rb_value
  - Depth tracking: rb_enter, rb_back (used by the `rb` fixture)
  - Self-audit: rb_audit, rb_lint, rb_lint_sources, rb_audit_depth, rb_audit_tier
PYTEST TEST FORMAT:
    @pytest.mark.rb(jump_kind="reach", expect="Preferences dialog visible")
    def test_prefs_opens(app, rb):
        rb.enter("prefs")
        before = rb_showing(app)
        # ... trigger ...
        after = rb_showing(app)
        assert rb_reach(before, after, target="Preferences", source="toolbar")
VERDICT HELPERS (prove the jump kind — a bare `assert True` is unbacked):
    rb_reach(before, after, target, source) — tier2 reach (surface jump)
    rb_flip(before, after)                  — tier2 transition (state flip)
    rb_value(observed, expected)            — tier3 value (specific content)
SELF-AUDIT (report-only, NOT a hard gate):
    python -c "import rb_testgen_kit as k; k.rb_audit('<results_dir>')"
    python -c "import rb_testgen_kit as k; k.rb_audit_depth('<testgen_dir>','<results_dir>')"
    python -c "import rb_testgen_kit as k; k.rb_audit_tier('<testgen_dir>','<results_dir>')"
"""
from __future__ import annotations

import ast
import glob
import json
import os
import re

JUMP_KINDS = {"none": 0, "liveness": 1, "reach": 2, "transition": 2, "value": 3}

# Double-axis quota TARGETS (fractions) — goals rb_audit REPORTS, NOT a hard gate. Tune here only.
QUOTAS = {
    "existence_max": 0.30,    # jump_kind=="none" share of ALL tests           (structure gate)
    "interactive_min": 0.60,  # reach|transition|value share of ALL tests
    "depth_ge1_min": 0.40,    # depth>=1 share of INTERACTIVE tests            (real navigation)
    "depth_ge2_min": 0.10,    # depth>=2 share of INTERACTIVE tests            (deep chains)
    "value_min": 0.25,        # jump_kind=="value" share of INTERACTIVE tests  (tier3 strength)
}

# Depth family (naming convention) — see rb_audit_depth. A test "crosses a surface" iff it
# calls a member of the depth family; rb_back* pops one. Semantic aliases (rb_enter_dialog,
# rb_reach_prefs) are matched BY PREFIX so the kit stays small and testcases stay diverse.
DEPTH_FWD_RE = re.compile(r"^rb_(enter|reach)[a-z0-9_]*$")   # forward crossing / proven jump
DEPTH_BACK_RE = re.compile(r"^rb_back[a-z0-9_]*$")           # pop a surface
DEPTH_RESET_RE = re.compile(r"^rb_reset_depth$")
# Verdict family (tier audit): a tier2/3 claim must be backed by its verdict function.
# rb_reach* is counted by rb_audit_depth; rb_flip* and rb_value* are counted here.
VERDICT_RE = re.compile(r"^rb_(reach|flip|value)[a-z0-9_]*$")
_JK_FAMILY = {"reach": "reach", "transition": "flip", "value": "value"}
# normalize plural (v73prep) / singular (kltg74) module names so a JSON module ("button")
# and a file stem ("buttons") join.
_MODULE_ALIAS = {"menus": "menu", "buttons": "button", "states": "state",
                 "structures": "structure", "texts": "text",
                 "computations": "computation", "workflows": "workflow"}

_DEPTH = 0  # module-global surface counter for rb_enter/rb_back


def tier_of(jump_kind: str) -> int:
    return JUMP_KINDS.get(jump_kind, 2)


def rb_reset_depth() -> None:
    global _DEPTH
    _DEPTH = 0


def rb_enter(_surface: str = "") -> int:
    """Call right after your test navigates into a new surface (dialog/page/submenu)."""
    global _DEPTH
    _DEPTH += 1
    return _DEPTH


def rb_back() -> int:
    global _DEPTH
    _DEPTH = max(0, _DEPTH - 1)
    return _DEPTH


def rb_showing(root) -> set:
    """Snapshot visible UIA elements under root as {"control_type|name", ...}.
    Safe no-op ({}) if pywinauto is unavailable or root is None."""
    out = set()
    if root is None:
        return out
    try:
        seen = 0
        for el in root.descendants():
            seen += 1
            if seen > 6000:
                break
            try:
                if not el.is_visible():
                    continue
                ct = el.element_info.control_type or ""
                nm = el.element_info.name or ""
                out.add(f"{ct}|{nm}")
            except Exception:
                pass
    except Exception:
        pass
    return out


def rb_reach(before, after, target: str, source: str | None = None) -> bool:
    """True-jump verdict: the TARGET surface's discriminating element newly APPEARED
    AND (if given) the SOURCE surface's discriminating element DISAPPEARED. This is
    what separates a real page jump from reading an element that was always in the tree."""
    before = set(before or [])
    after = set(after or [])
    t = target.lower()
    appeared = any(t in k.lower() for k in (after - before))
    if source is None:
        return appeared
    s = source.lower()
    src_before = any(s in k.lower() for k in before)
    src_after = any(s in k.lower() for k in after)
    return appeared and src_before and not src_after


def rb_flip(before, after, expect: str = "") -> bool:
    """In-page STATE-FLIP verdict (tier2 `transition`): the observed state actually CHANGED
    before->after (a CHECKED/ENABLED/mode/selection value read the two times). Pass its return
    as record()'s `passed`. Does NOT touch depth — a flip is in-page (declare depth=0 unless you
    navigated a surface to reach the control). You MAY write your own `rb_flip_<what>()` wrapper."""
    try:
        return before != after
    except Exception:
        return False


def rb_value(observed, expected) -> bool:
    """VALUE verdict (tier3 `value`): a specific expected value/content matched what you OBSERVED
    on the reference. Pass its return as record()'s `passed`. Tolerant: exact match, else the
    stringified `expected` appears in the stringified `observed` (substring/normalized), mirroring
    rb_reach's tolerant match. Does NOT touch depth. You MAY write your own `rb_value_<what>()`."""
    if observed is None:
        return False
    if observed == expected:
        return True
    try:
        return bool(str(expected)) and str(expected) in str(observed)
    except Exception:
        return False


# NOTE: rb_connect() and RBResults have been removed — replaced by:
#   - rb_uia.get_app() for app acquisition (used by conftest.py's `app` fixture)
#   - conftest.py's pytest hooks for automatic result collection
# Tests now use pytest format: `def test_*(app, rb):` with `@pytest.mark.rb(...)`


def _load_all(results_dir: str) -> list[dict]:
    recs: list[dict] = []
    seen_files = set()
    for pat in (os.path.join(results_dir, "*_results.json"),
                os.path.join(results_dir, "results", "*_results.json")):
        for path in glob.glob(pat):
            base = os.path.basename(path)
            if base in seen_files:
                continue
            seen_files.add(base)
            try:
                data = json.load(open(path))
            except Exception:
                continue
            for t in data.get("tests", []):
                if isinstance(t, dict) and "name" in t:
                    recs.append(t)
    return recs


def rb_audit(results_dir: str, quotas: dict | None = None) -> dict:
    """Aggregate declared depth/jump_kind across a suite and check the double-axis
    quotas. Prints PASS/FAIL per quota with the concrete gap (how many tests to add),
    so the orchestrator can target its Improve round. Call BEFORE finalizing."""
    q = quotas or QUOTAS
    recs = _load_all(results_dir)
    n = len(recs)
    if not n:
        print(f"[rb_audit] no records under {results_dir}")
        return {"total": 0}
    kinds = [r.get("jump_kind", "none") for r in recs]
    none_n = sum(1 for k in kinds if k == "none")
    live_n = sum(1 for k in kinds if k == "liveness")
    inter = [r for r in recs if r.get("jump_kind") in ("reach", "transition", "value")]
    ni = len(inter) or 1
    d1 = sum(1 for r in inter if int(r.get("depth", 0)) >= 1)
    d2 = sum(1 for r in inter if int(r.get("depth", 0)) >= 2)
    val = sum(1 for r in inter if r.get("jump_kind") == "value")

    def line(label, got, need, kind):
        ok = got >= need if kind == "min" else got <= need
        return (ok, f"  [{'ok ' if ok else 'GAP'}] {label}: {got:.0%} "
                    f"({'≥' if kind=='min' else '≤'} {need:.0%})")

    checks = [
        line("existence(none)/all", none_n / n, q["existence_max"], "max"),
        line("interactive/all", len(inter) / n, q["interactive_min"], "min"),
        line("depth≥1 /interactive", d1 / ni, q["depth_ge1_min"], "min"),
        line("depth≥2 /interactive", d2 / ni, q["depth_ge2_min"], "min"),
        line("value(tier3)/interactive", val / ni, q["value_min"], "min"),
    ]
    print(f"[rb_audit] {results_dir}  total={n}  interactive={len(inter)}  "
          f"none={none_n} liveness={live_n} depth≥1={d1} depth≥2={d2} value={val}")
    for _ok, msg in checks:
        print(msg)
    passed = all(ok for ok, _ in checks)
    print(f"[rb_audit] {'ALL TARGETS MET' if passed else 'some targets below — OPTIONALLY add a deeper reach/value test, or finalize if you have already tried enough angles (these are goals, NOT a hard gate)'}")
    return {"total": n, "interactive": len(inter), "none": none_n, "liveness": live_n,
            "depth_ge1": d1, "depth_ge2": d2, "value": val,
            "passed": passed, "checks": [m for _o, m in checks]}


def rb_lint(results_dir: str) -> list:
    """Flag weak/inconsistent declarations: missing expect on a value/transition/reach
    test, unknown jump_kind, or negative depth. Prints and returns the issues."""
    recs = _load_all(results_dir)
    issues = []
    for r in recs:
        jk = r.get("jump_kind", "none")
        nm = r.get("name", "?")
        if jk not in JUMP_KINDS:
            issues.append(f"{nm}: unknown jump_kind {jk!r}")
        if jk in ("reach", "transition", "value") and not str(r.get("expect", "")).strip():
            issues.append(f"{nm}: jump_kind={jk} but empty expect (declare the specific observed target)")
        if jk == "liveness":
            issues.append(f"{nm}: liveness is tier1 — replace with the app-authored effect you can observe")
        if int(r.get("depth", 0)) < 0:
            issues.append(f"{nm}: negative depth")
    if issues:
        print(f"[rb_lint] {len(issues)} issue(s):")
        for i in issues:
            print("  - " + i)
    else:
        print("[rb_lint] no issues")
    return issues


# Non-portable source patterns: pass on the reference (testgen, app launched by runner)
# but break on the recreation (eval, app elsewhere). Each is a hard portability defect.
_BAD_SOURCE_PATTERNS = [
    (r"C:\\\\rb_pipeline|C:/rb_pipeline",
     "hardcodes the pipeline path (varies per task) — use the `app` fixture from conftest (backed by rb_uia.get_app())"),
    (r"\\\\workspace\\\\install|/workspace/install",
     "names a Linux-style install path — not applicable on Windows; use the `app` fixture from conftest"),
    (r"subprocess\.(?:Popen|run|call)\([^)]*[\"'][A-Z]:\\\\",
     "spawns the app by an absolute path — use the `app` fixture for portability"),
    (r"Application\([^)]*\)\.(?:start|connect)\([^)]*path\s*=",
     "uses pywinauto .start() or .connect(path=...) — use the `app` fixture for portability"),
]


def rb_lint_sources(testgen_dir: str) -> list:
    """Scan authored test/helper .py for NON-PORTABLE patterns (hardcoded install paths,
    self-spawn by absolute path) and for helper/test API DRIFT
    (a test calls `H.<attr>()` that the imported helper module does not define — the
    hotspot connect_running/spawn_with_file class of bug). Returns + prints the issues.
    Run BEFORE finalize; a clean result means the suite will drive the recreation, not
    just the reference."""
    import glob
    import re
    import ast
    py = [p for p in (glob.glob(os.path.join(testgen_dir, "*.py"))
                      + glob.glob(os.path.join(testgen_dir, "tests", "*.py")))
          if os.path.basename(p) not in ("rb_testgen_kit.py", "run_uia_tests.py")]
    defined: dict[str, set] = {}
    for p in py:
        mod = os.path.splitext(os.path.basename(p))[0]
        names: set = set()
        try:
            for node in ast.parse(open(p).read()).body:
                if isinstance(node, (ast.FunctionDef, ast.AsyncFunctionDef, ast.ClassDef)):
                    names.add(node.name)
                elif isinstance(node, ast.Assign):
                    for t in node.targets:
                        if isinstance(t, ast.Name):
                            names.add(t.id)
        except Exception:
            pass
        defined.setdefault(mod, set()).update(names)
    issues = []
    for p in py:
        base = os.path.basename(p)
        try:
            src = open(p).read()
        except Exception:
            continue
        for pat, why in _BAD_SOURCE_PATTERNS:
            if re.search(pat, src):
                issues.append(f"{base}: {why}")
        for m in re.finditer(r"import\s+(\w+)\s+as\s+(\w+)", src):
            mod, alias = m.group(1), m.group(2)
            if mod not in defined:
                continue
            for am in re.finditer(rf"\b{alias}\.(\w+)\s*\(", src):
                attr = am.group(1)
                if attr not in defined[mod]:
                    issues.append(
                        f"{base}: calls {alias}.{attr}() but {mod}.py never defines {attr} "
                        f"(helper/test drift — the eval copy will AttributeError)")
                    break
    if issues:
        print(f"[rb_lint_sources] {len(issues)} portability issue(s):")
        for i in issues:
            print("  - " + i)
    else:
        print("[rb_lint_sources] no portability issues")
    return issues


# ── MEASURED DEPTH + TIER cross-check (report-only) ──────────────────────────
def _ast_call_name(node) -> str:
    f = node.func
    if isinstance(f, ast.Attribute):
        return f.attr
    if isinstance(f, ast.Name):
        return f.id
    return ""


def _verdict_family(cn: str) -> str:
    if cn.startswith("rb_reach"):
        return "reach"
    if cn.startswith("rb_flip"):
        return "flip"
    if cn.startswith("rb_value"):
        return "value"
    return ""


def _scan(fn_node, local_funcs, budget: int = 2, seen=None):
    """Walk fn_node (+ up to `budget` levels of file-local functions it calls; cycle-guarded)
    and return (n_fwd, n_back, n_reach, verdicts, guarded):
      DEPTH axis — n_fwd/n_back = rb_enter*|rb_reach* / rb_back* counts;
                   n_reach = the rb_reach* subset.
      TIER axis  — verdicts = verdict families present among {reach, flip, value};
                   guarded = an assert/try gates the code."""
    seen = seen if seen is not None else set()
    n_fwd = n_back = n_reach = 0
    verdicts = set()
    guarded = False
    for node in ast.walk(fn_node):
        if isinstance(node, (ast.Assert, ast.Try)):
            guarded = True
        if not isinstance(node, ast.Call):
            continue
        cn = _ast_call_name(node)
        is_fwd = bool(DEPTH_FWD_RE.match(cn))
        is_back = bool(DEPTH_BACK_RE.match(cn))
        is_verdict = bool(VERDICT_RE.match(cn))
        if is_fwd:
            n_fwd += 1
            if cn.startswith("rb_reach"):
                n_reach += 1
        elif is_back:
            n_back += 1
        if is_verdict:
            verdicts.add(_verdict_family(cn))
        if not (is_fwd or is_back or is_verdict) and budget > 0 and cn in local_funcs and cn not in seen:
            seen.add(cn)
            f, b, r, v, g = _scan(local_funcs[cn], local_funcs, budget - 1, seen)
            n_fwd += f
            n_back += b
            n_reach += r
            verdicts |= v
            guarded = guarded or g
    return n_fwd, n_back, n_reach, verdicts, guarded


def _record_name_passed(call):
    """(name, passed_node) for a record() call: name = string-literal name arg (else None);
    passed_node = AST node of the `passed` argument."""
    a = call.args
    if a and isinstance(a[0], ast.Constant) and isinstance(a[0].value, str):
        return a[0].value, (a[1] if len(a) >= 2 else None)
    if len(a) >= 2 and isinstance(a[1], ast.Constant) and isinstance(a[1].value, str):
        return a[1].value, (a[2] if len(a) >= 3 else None)
    return None, None


def _passed_form(passed_node, guarded: bool) -> str:
    """Classify record()'s `passed`: 'bare_literal', 'guarded_literal', 'derived', or 'missing'."""
    if passed_node is None:
        return "missing"
    if isinstance(passed_node, ast.Constant) and isinstance(passed_node.value, (bool, int)):
        return "guarded_literal" if guarded else "bare_literal"
    return "derived"


def _enclosing_funcs(tree):
    spans = []
    for n in ast.walk(tree):
        if isinstance(n, (ast.FunctionDef, ast.AsyncFunctionDef)):
            end = max((getattr(c, "lineno", n.lineno) for c in ast.walk(n)), default=n.lineno)
            spans.append((n.lineno, end, n))
    return spans


def _scope_of(line, spans):
    best = None
    for s, e, node in spans:
        if s <= line <= e and (best is None or s > best[0]):
            best = (s, e, node)
    return best[2] if best else None


def _has_depth_family(tree) -> bool:
    return any(isinstance(n, ast.Call) and DEPTH_FWD_RE.match(_ast_call_name(n))
               for n in ast.walk(tree))


def _verdict_families_in(tree) -> set:
    return {_verdict_family(_ast_call_name(n)) for n in ast.walk(tree)
            if isinstance(n, ast.Call) and VERDICT_RE.match(_ast_call_name(n))}


def _depth_backing_map(testgen_dir: str):
    """Parse the suite's test_uia_*.py (+ imported LOCAL helpers) and return
    (per_record, per_module) carrying BOTH axes' source evidence:
      per_record[(module,name)] = {measured_depth, measured_backing, n_reach,   # DEPTH axis
                                   verdicts:set, passed_form:str}               # TIER axis
      per_module[module]        = {module_backing:bool, verdict_families:set, unmatched:int}."""
    all_py = [p for p in (glob.glob(os.path.join(testgen_dir, "*.py"))
                          + glob.glob(os.path.join(testgen_dir, "tests", "*.py")))
              if os.path.basename(p) not in ("rb_testgen_kit.py", "run_uia_tests.py")]
    trees = {}
    helper_depth = {}       # stem -> bool  (imported-helper DEPTH backing)
    helper_verdicts = {}    # stem -> set   (imported-helper TIER verdicts)
    for p in all_py:
        try:
            t = ast.parse(open(p, errors="replace").read())
        except Exception:
            continue
        trees[p] = t
        stem = os.path.splitext(os.path.basename(p))[0]
        helper_depth[stem] = _has_depth_family(t)
        helper_verdicts[stem] = _verdict_families_in(t)
    per_record, per_module = {}, {}
    for p in [q for q in all_py if os.path.basename(q).startswith("test_uia_")]:
        t = trees.get(p)
        if t is None:
            continue
        stem = os.path.basename(p)[len("test_uia_"):-len(".py")]
        mod = _MODULE_ALIAS.get(stem, stem)
        local_funcs = {n.name: n for n in ast.walk(t)
                       if isinstance(n, (ast.FunctionDef, ast.AsyncFunctionDef))}
        spans = _enclosing_funcs(t)
        imported = set()
        for n in ast.walk(t):
            if isinstance(n, ast.Import):
                for al in n.names:
                    imported.add(al.name.split(".")[0])
            elif isinstance(n, ast.ImportFrom) and n.module:
                imported.add(n.module.split(".")[0])
        module_backing = _has_depth_family(t) or any(helper_depth.get(h) for h in imported)
        verdict_families = set(_verdict_families_in(t))
        for h in imported:
            verdict_families |= helper_verdicts.get(h, set())
        unmatched = 0
        for n in ast.walk(t):
            if not (isinstance(n, ast.Call) and _ast_call_name(n) == "record"):
                continue
            nm, passed = _record_name_passed(n)
            fn = _scope_of(n.lineno, spans) if nm is not None else None
            if nm is None or fn is None:
                unmatched += 1
                continue
            n_fwd, n_back, n_reach, verdicts, guarded = _scan(fn, local_funcs)
            measured_depth = max(0, max((n_fwd - n_reach) - n_back, 1 if n_reach >= 1 else 0))
            per_record[(mod, nm)] = {"measured_depth": measured_depth,
                                     "measured_backing": n_fwd >= 1, "n_reach": n_reach,
                                     "verdicts": verdicts,
                                     "passed_form": _passed_form(passed, guarded)}
        per_module[mod] = {"module_backing": module_backing,
                           "verdict_families": verdict_families, "unmatched": unmatched}
    return per_record, per_module


def _load_declared_by_module(results_dir: str):
    """(module, name, depth:int, jump_kind) declared records — KEEPS the module (unlike
    _load_all, whose per-test dicts drop it)."""
    out, seen = [], set()
    for pat in (os.path.join(results_dir, "*_results.json"),
                os.path.join(results_dir, "results", "*_results.json")):
        for path in glob.glob(pat):
            base = os.path.basename(path)
            if base in seen or base == "baseline_results.json":
                continue
            seen.add(base)
            try:
                data = json.load(open(path))
            except Exception:
                continue
            m = data.get("module", "") if isinstance(data, dict) else ""
            mod = _MODULE_ALIAS.get(m, m) or base[: -len("_results.json")]
            for t in (data.get("tests", []) if isinstance(data, dict) else []):
                if not (isinstance(t, dict) and "name" in t):
                    continue
                try:
                    d = int(t.get("depth", 0) or 0)
                except Exception:
                    d = 0
                out.append((mod, t["name"], d, t.get("jump_kind", "none")))
    return out


def rb_audit_depth(testgen_dir: str, results_dir: str | None = None) -> list:
    """MEASURE navigation depth from the test SOURCE (AST) and cross-check the DECLARED
    depth/jump_kind in <module>_results.json — so a test can't just TYPE depth=2. REPORT-ONLY
    (NOT a finalize gate): flags a declared depth>=1 or jump_kind="reach" that has NO backing
    call from the depth family rb_enter*/rb_reach* (directly, in a local wrapper it calls, or in
    an imported local helper). Variable-named records fall back to module-level backing so the
    wrapper/closure idiom is not false-flagged. depth-only; jump verdicts are a later phase."""
    results_dir = results_dir or testgen_dir
    per_record, per_module = _depth_backing_map(testgen_dir)
    issues = []
    for mod, name, depth, jk in _load_declared_by_module(results_dir):
        strong = depth >= 1 or jk == "reach"
        rec = per_record.get((mod, name))
        if rec is not None:
            if strong and not rec["measured_backing"]:
                issues.append(f"{mod}:{name}: declared depth={depth}/jump_kind={jk} but NO "
                              f"backing rb_enter*/rb_reach* call (fake depth)")
            elif jk == "reach" and rec["n_reach"] == 0:
                issues.append(f"{mod}:{name}: jump_kind=reach but no rb_reach* call in its code")
        elif strong and not per_module.get(mod, {}).get("module_backing", True):
            issues.append(f"{mod}:{name}: declared depth={depth}/jump_kind={jk} but the module "
                          f"has NO rb_enter*/rb_reach* call anywhere (fake depth)")
    if issues:
        print(f"[rb_audit_depth] {len(issues)} fake-depth issue(s) (REPORT-ONLY — fix by adding "
              f"the rb_reach/rb_enter you navigated with, or lower the declared depth):")
        for i in issues:
            print("  - " + i)
    else:
        print("[rb_audit_depth] no fake-depth issues (declared depth/reach backed by nav calls)")
    unm = sum(v.get("unmatched", 0) for v in per_module.values())
    if unm:
        print(f"[rb_audit_depth] note: {unm} record(s) use non-literal names "
              f"(joined at module level, not per-test)")
    return issues


def rb_audit_tier(testgen_dir: str, results_dir: str | None = None) -> list:
    """MEASURE jump RIGOR from the test SOURCE (AST) and cross-check the DECLARED jump_kind.
    REPORT-ONLY (NOT a finalize gate). A tier2/3 claim must be backed by its verdict function —
    transition->rb_flip*, value->rb_value* — AND that verdict's bool result must be the record()'s
    `passed` (a bare-literal True is unbacked). reach's rb_reach* presence is already covered by
    rb_audit_depth, so here reach is only checked for a bare-literal pass."""
    results_dir = results_dir or testgen_dir
    per_record, per_module = _depth_backing_map(testgen_dir)
    issues = []
    for mod, name, _depth, jk in _load_declared_by_module(results_dir):
        fam = _JK_FAMILY.get(jk)
        if fam is None:
            continue
        rec = per_record.get((mod, name))
        if rec is not None:
            has = fam in rec["verdicts"]
            if jk == "reach":
                if has and rec["passed_form"] == "bare_literal":
                    issues.append(f"{mod}:{name}: jump_kind=reach — rb_reach* present but the "
                                  f"pass is a bare literal True (verdict not used as the result)")
            elif not has:
                issues.append(f"{mod}:{name}: declared jump_kind={jk} (tier {tier_of(jk)}) but no "
                              f"rb_{fam}* verdict call — the tier is unbacked")
            elif rec["passed_form"] == "bare_literal":
                issues.append(f"{mod}:{name}: declared jump_kind={jk} — rb_{fam}* present but the "
                              f"pass is a bare literal True (verdict not used as the result)")
        elif jk != "reach" and fam not in per_module.get(mod, {}).get("verdict_families", set()):
            issues.append(f"{mod}:{name}: declared jump_kind={jk} but the module has no "
                          f"rb_{fam}* verdict anywhere")
    if issues:
        print(f"[rb_audit_tier] {len(issues)} unbacked-tier issue(s) (REPORT-ONLY — back a "
              f"transition with rb_flip(), a value with rb_value(), and pass its result as the "
              f"test's `passed`; or drop to a weaker jump_kind you can actually prove):")
        for i in issues:
            print("  - " + i)
    else:
        print("[rb_audit_tier] no unbacked-tier issues (transition/value backed by verdict calls)")
    return issues


if __name__ == "__main__":
    import sys
    _fns = {"audit": rb_audit, "lint": rb_lint, "lint_sources": rb_lint_sources}
    _fns2 = {"audit_depth": rb_audit_depth, "audit_tier": rb_audit_tier}
    if sys.argv[1:2] and sys.argv[1] in _fns2 and len(sys.argv) >= 3:
        _fns2[sys.argv[1]](sys.argv[2], sys.argv[3] if len(sys.argv) >= 4 else None)
    elif len(sys.argv) >= 3 and sys.argv[1] in _fns:
        _fns[sys.argv[1]](sys.argv[2])
    else:
        print("usage: rb_testgen_kit.py {audit|lint|lint_sources <dir> | "
              "{audit_depth|audit_tier} <testgen_dir> [results_dir]}")