feat(access-control-prover): 12.3 instrument and hard runtime gates

.kiro/specs/access-control-prover/tasks.md

@@ -270,7 +270,7 @@ its final resolution, dominance, fact-acquisition and measurement semantics with
   - _Requirements: 5.1, 5.3, 5.4, 7.1, 7.6, 9.7, 11.8_
   - _Boundary: Development Runner_
   - _Depends: 10.6, 12.1_
-- [ ] 12.3 Enforce instrument and hard runtime gates
+- [x] 12.3 Enforce instrument and hard runtime gates
   - Preserve zero-byte and stale-anchor refusal and add stage-aware deadline records.
   - Fail the run if any measured change exceeds 120 seconds; never summarize the overrun as a passing result.
   - Done when zero-byte, stale-anchor, deadline and 121-second fixtures all refuse the summary.

benchmarks/valid/development_runner.py

@@ -1,10 +1,12 @@
 """Development runner: per-record provenance, stage funnel and judgment transitions.
 
-Task 12.2 of the access-control-prover spec (Reqs 5.1, 5.3, 5.4, 7.1, 7.6, 9.7, 11.8).
-Every development record carries the bundle digests (analyzer revision, fact set,
-acquisition record, configuration) plus the population and per-change input digests,
-a stage funnel from entry recognition through resolution to the three-way judgment
-counts, and the transition label between compared judgments.
+Tasks 12.2 and 12.3 of the access-control-prover spec (Reqs 5.1, 5.3, 5.4, 5.6,
+6.1, 6.3, 7.1, 7.6, 9.7, 11.8). Every development record carries the bundle
+digests (analyzer revision, fact set, acquisition record, configuration) plus the
+population and per-change input digests, a stage funnel from entry recognition
+through resolution to the three-way judgment counts, and the transition label
+between compared judgments. The runner gates refuse to summarise an instrument
+failure (zero-byte read, stale anchor) or a change over the hard runtime budget.
 """
 
 from __future__ import annotations
@@ -116,4 +118,78 @@ def run_development(
     return tuple(records)
 
 
-__all__ = ["DevelopmentRecord", "judgment_transition", "run_development", "stage_funnel"]
+class GateFailure(RuntimeError):
+    """A measured change exceeded the hard runtime budget (Req 6.1)."""
+
+
+class InstrumentFailure(RuntimeError):
+    """An instrument failure (zero-byte read, stale anchor) refuses the summary (Req 5.6)."""
+
+
+_INSTRUMENT_FAILURE_MARKERS = ("zero-byte", "stale_anchor")
+
+
+def enforce_gates(records: tuple[DevelopmentRecord, ...], *, budget_seconds: float = 120.0) -> None:
+    """Fail when any measured change exceeds the budget; never let an overrun pass (Req 6.1)."""
+    for record in records:
+        if record.seconds > budget_seconds:
+            raise GateFailure(f"change {record.change_id} took {record.seconds:.1f}s over the {budget_seconds:.1f}s budget")
+
+
+def deadline_record(record: DevelopmentRecord, *, budget_seconds: float = 120.0) -> dict[str, object]:
+    """Stamp the stage-aware deadline view of one record (Req 6.3)."""
+    return {
+        "deadline_seconds": budget_seconds,
+        "exceeded": record.seconds > budget_seconds,
+    }
+
+
+def _instrument_failure(record: DevelopmentRecord) -> str | None:
+    for value in _funnel_strings(record.stage_funnel):
+        for marker in _INSTRUMENT_FAILURE_MARKERS:
+            if marker in value:
+                return value
+    return None
+
+
+def _funnel_strings(funnel: dict[str, dict[str, Any]]) -> list[str]:
+    strings: list[str] = []
+    for stage in funnel.values():
+        for value in stage.values():
+            if isinstance(value, str):
+                strings.append(value)
+    return strings
+
+
+def summarize_development(records: tuple[DevelopmentRecord, ...], *, budget_seconds: float = 120.0) -> dict[str, object]:
+    """Summarise a development run; an overrun or instrument failure refuses the summary."""
+    enforce_gates(records, budget_seconds=budget_seconds)
+    for record in records:
+        failure = _instrument_failure(record)
+        if failure is not None:
+            raise InstrumentFailure(f"change {record.change_id} instrument failure: {failure}")
+    judgment_counts = {"unguarded": 0, "discharged": 0, "indeterminate": 0}
+    for record in records:
+        judgment = record.stage_funnel.get("judgment", {})
+        for kind in judgment_counts:
+            judgment_counts[kind] += int(judgment.get(kind, 0))
+    seconds = [record.seconds for record in records]
+    return {
+        "total_changes": len(records),
+        "judgment_counts": judgment_counts,
+        "total_seconds": sum(seconds),
+        "max_seconds": max(seconds, default=0.0),
+    }
+
+
+__all__ = [
+    "DevelopmentRecord",
+    "GateFailure",
+    "InstrumentFailure",
+    "deadline_record",
+    "enforce_gates",
+    "judgment_transition",
+    "run_development",
+    "stage_funnel",
+    "summarize_development",
+]

tests/test_runner_gates.py

@@ -0,0 +1,95 @@
+"""Task 12.3 -- runner instrument and hard runtime gates (Reqs 5.6, 6.1, 6.3).
+
+Gate fixtures refuse the summary: zero-byte and stale-anchor records raise
+InstrumentFailure, a change over the 120-second budget raises GateFailure, and
+the deadline record stamps budget + exceeded flag per record.
+"""
+
+from __future__ import annotations
+
+import sys
+from pathlib import Path
+
+import pytest
+
+sys.path.insert(0, str(Path(__file__).resolve().parents[1]))
+
+from benchmarks.valid import development_runner  # noqa: E402
+from benchmarks.valid.development_runner import (  # noqa: E402
+    DevelopmentRecord,
+    InstrumentFailure,
+    deadline_record,
+    enforce_gates,
+    summarize_development,
+)
+
+GateFailure = development_runner.GateFailure
+
+
+def _record(
+    *,
+    seconds: float = 1.0,
+    change_id: str = "change-1",
+    instrument_reason: str | None = None,
+) -> DevelopmentRecord:
+    judgment: dict[str, object] = {"unguarded": 0, "discharged": 0, "indeterminate": 1}
+    funnel: dict[str, dict[str, object]] = {"judgment": judgment}
+    if instrument_reason is not None:
+        funnel["instrument"] = {"reason": instrument_reason}
+    return DevelopmentRecord(
+        change_id=change_id,
+        analyzer_revision="0" * 12,
+        fact_set_digest="1" * 12,
+        acquisition_record_digest="2" * 12,
+        configuration_digest="3" * 12,
+        population_digest="4" * 12,
+        input_digest="5" * 12,
+        stage_funnel=funnel,
+        judgment_transition="same",
+        seconds=seconds,
+    )
+
+
+def test_enforce_gates_passes_clean_records() -> None:
+    enforce_gates((_record(seconds=1.0), _record(seconds=119.9, change_id="change-2")))
+
+
+def test_enforce_gates_fails_a_change_over_the_budget() -> None:
+    with pytest.raises(GateFailure) as caught:
+        enforce_gates((_record(seconds=0.5), _record(seconds=121.0, change_id="slow-change")))
+    assert "slow-change" in str(caught.value)
+
+
+def test_budget_is_overridable_for_fixtures() -> None:
+    with pytest.raises(GateFailure):
+        enforce_gates((_record(seconds=6.0),), budget_seconds=5.0)
+
+
+def test_deadline_record_stamps_budget_and_exceeded() -> None:
+    kept = deadline_record(_record(seconds=1.0), budget_seconds=120.0)
+    assert kept == {"deadline_seconds": 120.0, "exceeded": False}
+    blown = deadline_record(_record(seconds=121.0), budget_seconds=120.0)
+    assert blown == {"deadline_seconds": 120.0, "exceeded": True}
+
+
+def test_summary_refuses_an_overrun_never_reports_it_passing() -> None:
+    with pytest.raises(GateFailure):
+        summarize_development((_record(seconds=121.0),))
+
+
+def test_summary_reports_clean_records() -> None:
+    summary = summarize_development((_record(seconds=1.0), _record(seconds=2.0, change_id="change-2")))
+    assert summary["total_changes"] == 2
+    assert summary["max_seconds"] == 2.0
+
+
+def test_zero_byte_record_refuses_the_summary() -> None:
+    with pytest.raises(InstrumentFailure) as caught:
+        summarize_development((_record(instrument_reason="zero-byte read"),))
+    assert "zero-byte" in str(caught.value)
+
+
+def test_stale_anchor_record_refuses_the_summary() -> None:
+    with pytest.raises(InstrumentFailure) as caught:
+        summarize_development((_record(instrument_reason="stale_anchor"),))
+    assert "stale_anchor" in str(caught.value)