chore: stage PR126 v4 final parser hardening

Author: TheHiddenObserver

dev/pr126_v4_final_parser_hardening.py

@@ -0,0 +1,135 @@
+#!/usr/bin/env python3
+from __future__ import annotations
+
+from pathlib import Path
+
+
+def replace_once(path: Path, old: str, new: str, label: str) -> None:
+    text = path.read_text()
+    count = text.count(old)
+    if count != 1:
+        raise RuntimeError(f"{label}: expected one target, found {count}")
+    path.write_text(text.replace(old, new, 1))
+
+
+def harden_parser() -> None:
+    path = Path("dev/benchmarks/frontend_data/parsers/panel_stage_c_identifiability.py")
+    replace_once(
+        path,
+        "historical v1/v2 sources without overwriting or colliding with them.",
+        "historical v1/v2/v3 sources without overwriting or colliding with them.",
+        "parser history docstring",
+    )
+    replace_once(
+        path,
+        '''        repeats = int(row.get("repeats", 0))
+        samples = row.get("samples_seconds")
+        if repeats <= 0 or not isinstance(samples, list) or len(samples) != repeats:
+            raise ValueError("PR126 identifiability Stage-C timing samples/repeats contract failed")''',
+        '''        repeats = int(row.get("repeats", 0))
+        samples = row.get("samples_seconds")
+        if repeats != 3 or not isinstance(samples, list) or len(samples) != 3:
+            raise ValueError(
+                "PR126 identifiability Stage-C timing requires exactly three raw samples"
+            )''',
+        "exact three-sample contract",
+    )
+    replace_once(
+        path,
+        '''        expected_median = float(statistics.median(numeric_samples))
+        if not math.isclose(median, expected_median, rel_tol=1e-12, abs_tol=1e-15):
+            raise ValueError("PR126 identifiability Stage-C reported median does not match raw samples")''',
+        '''        expected_median = float(statistics.median(numeric_samples))
+        if median != expected_median:
+            raise ValueError(
+                "PR126 identifiability Stage-C reported median must exactly match raw samples"
+            )''',
+        "exact median contract",
+    )
+    replace_once(
+        path,
+        '''                            {"metric": "synchronized_timing", "status": "pass"},
+                            {"metric": "raw_samples_finite_positive", "status": "pass"},
+                            {"metric": "median_matches_raw_samples", "status": "pass"},''',
+        '''                            {"metric": "synchronized_timing", "status": "pass"},
+                            {"metric": "exactly_three_raw_samples", "status": "pass"},
+                            {"metric": "raw_samples_finite_positive", "status": "pass"},
+                            {"metric": "median_exactly_matches_raw_samples", "status": "pass"},''',
+        "performance validation checks",
+    )
+
+
+def harden_tests() -> None:
+    path = Path("dev/tests/test_panel_stage_c_identifiability_frontend_source.py")
+    text = path.read_text()
+    if "import math\n" not in text:
+        text = text.replace("import json\n", "import json\nimport math\n", 1)
+    old = '''def test_v4_performance_parser_rejects_matrix_or_median_drift(tmp_path):
+    payload = json.loads(PERFORMANCE.read_text())
+    broken = copy.deepcopy(payload)
+    broken["rows"] = broken["rows"][:-1]
+    with pytest.raises(ValueError, match="60 rows"):
+        parse_panel_stage_c_identifiability_performance(
+            _write(tmp_path, "bad-matrix.json", broken), ENV
+        )
+
+    broken = copy.deepcopy(payload)
+    broken["rows"][0]["median_seconds"] *= 2.0
+    with pytest.raises(ValueError, match="median"):
+        parse_panel_stage_c_identifiability_performance(
+            _write(tmp_path, "bad-median.json", broken), ENV
+        )
+'''
+    new = '''def test_v4_performance_parser_rejects_matrix_repeat_or_median_drift(tmp_path):
+    payload = json.loads(PERFORMANCE.read_text())
+    broken = copy.deepcopy(payload)
+    broken["rows"] = broken["rows"][:-1]
+    with pytest.raises(ValueError, match="60 rows"):
+        parse_panel_stage_c_identifiability_performance(
+            _write(tmp_path, "bad-matrix.json", broken), ENV
+        )
+
+    broken = copy.deepcopy(payload)
+    broken["rows"][0]["repeats"] = 2
+    broken["rows"][0]["samples_seconds"] = broken["rows"][0]["samples_seconds"][:2]
+    broken["rows"][0]["median_seconds"] = sum(broken["rows"][0]["samples_seconds"]) / 2.0
+    with pytest.raises(ValueError, match="exactly three raw samples"):
+        parse_panel_stage_c_identifiability_performance(
+            _write(tmp_path, "bad-repeats.json", broken), ENV
+        )
+
+    broken = copy.deepcopy(payload)
+    current = float(broken["rows"][0]["median_seconds"])
+    broken["rows"][0]["median_seconds"] = math.nextafter(current, math.inf)
+    with pytest.raises(ValueError, match="exactly match"):
+        parse_panel_stage_c_identifiability_performance(
+            _write(tmp_path, "bad-median.json", broken), ENV
+        )
+'''
+    if text.count(old) != 1:
+        raise RuntimeError(f"v4 performance test anchor drifted: {text.count(old)}")
+    path.write_text(text.replace(old, new, 1))
+
+
+def update_review_record() -> None:
+    path = Path("dev/reviews/pr126_round4_autofix_review_2026-08-12.md")
+    text = path.read_text()
+    text = text.replace(
+        "`PHYSICAL_GPU_ACCEPTED / CANONICAL_PROMOTION_PENDING / NOT MERGE-READY`",
+        "`PHYSICAL_GPU_ACCEPTED / CANONICAL_PROMOTED / HOSTED_FINAL_PENDING / NOT MERGE-READY`",
+        1,
+    )
+    if "[MEDIUM][PARSER][fixed]" not in text:
+        text += '''\n\n## Post-promotion independent parser review\n\n[MEDIUM][PARSER][fixed] The first v4 performance parser accepted any positive `repeats` count and used a tolerance-based median comparison, while the immutable source contract requires exactly three raw timing samples and an exactly persisted median. The parser now requires `repeats == 3`, exactly three samples, and exact equality with `statistics.median(samples)`. Corruption tests cover a two-sample row and a one-ULP median drift.\n\n[MEDIUM][ARTIFACT][fixed] The checkpoint header still said `CANONICAL_PROMOTION_PENDING` after canonical promotion commit `72bc21d3d0a1afd23467ecb1ff176d42df709cb4` had passed the dedicated v4 promotion gate. The record now reflects `CANONICAL_PROMOTED / HOSTED_FINAL_PENDING`.\n\nThe post-promotion parser hardening changes only parser/test/review artifacts. It does not touch `statgpu/panel/**`, the correctness runner, the performance runner, or either immutable raw JSON file, so the exact-clean `a99726e1...` P100 evidence remains applicable.\n'''
+    path.write_text(text)
+
+
+def main() -> None:
+    harden_parser()
+    harden_tests()
+    update_review_record()
+    print("PR126 v4 final parser hardening applied")
+
+
+if __name__ == "__main__":
+    main()