skip: sound whole-tile proofs for legacy FI/TN/VS fields (L74); FI -30%, TN -22%

The compiled-analytic stratum interval now certifies a legacy tile uniform only when it stays
beyond the detail envelope (and smooth-min slack for solid) with no intersecting primitive,
edit, native field, composer candidate, passage or active transition; everything else fails
closed. The stream layer uses the proof for whole-tile skips; legacy sub-tile skips stay off
(two block proofs failed the falsifier in TN).

Falsifier: full-grid reference vs skip-enabled run, identical mesh/payload/topology hashes on
every tile of all four owner views, Clang and GCC, W1/W8/W12; 20-camera sweep per strate,
0 violations; a too-tight control changed 3/184/233/238 tiles and was caught.
Whole view (8 workers): SW 2.61->2.48, FI 7.18->5.05, TN 5.97->4.63, VS 2.70->2.40 s.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_016dhxM43dnWvUu2B5Ph1iG1

Standalone/Research/WorldAssemblerM3/Linux74/RunHashSweepL74.py

@@ -0,0 +1,167 @@
+#!/usr/bin/env python3
+"""Exact full-grid versus whole-tile skip sweep at 20 camera positions per stratum."""
+from __future__ import annotations
+
+import json
+import re
+import statistics
+import subprocess
+import time
+from pathlib import Path
+
+
+ROOT = Path(__file__).resolve().parents[4]
+OUT = ROOT / "Saved/Linux/l74/logs"
+DATA = ROOT / "Standalone/data/owner_world.json"
+BIN = ROOT / "Saved/Linux/l74/bin/VoxelForgeCoreHeadless-l74-clang"
+CAMERAS = {
+    0: (8.0, 16.0, -372.2626953125),
+    1: (25.8369140625, 3.19921875, -737.0),
+    2: (19.166015625, -14.2001953125, -753.3740234375),
+    3: (-20.0625, -5.068359375, -900.353515625),
+}
+OFFSETS = [(x, y) for y in (0.1, 4.1, 8.1, 12.1)
+           for x in (0.1, 4.1, 8.1, 12.1, 16.1)]
+HASH = re.compile(
+    r"\[PipelineTileHash\] lod=(\d+) x=(-?\d+) y=(-?\d+) z=(-?\d+) "
+    r"density=(0x[0-9a-fA-F]+) mesh=(0x[0-9a-fA-F]+) "
+    r"payload=(0x[0-9a-fA-F]+) topology=(0x[0-9a-fA-F]+)"
+)
+GENERATED = re.compile(
+    r"\[UniformGridCeiling\] strate=(-?\d+) lod=(\d+) key="
+    r"\((-?\d+),(-?\d+),(-?\d+)\)"
+)
+PIPELINE = re.compile(r"\[StreamPipeline\] (.*)")
+TILE_PIPELINE = re.compile(r"\[WorldAssemblerTilePipeline\] (.*)")
+PAIR = re.compile(r"([a-zA-Z0-9_]+)=([^ ]+)")
+
+
+def read_hashes(path: Path) -> tuple[dict, set[tuple[int, int, int, int]]]:
+    hashes = {}
+    generated = set()
+    for line in path.read_text(encoding="utf-8", errors="replace").splitlines():
+        match = HASH.search(line)
+        if match:
+            lod, x, y, z = (int(match.group(index)) for index in range(1, 5))
+            hashes[(lod, x, y, z)] = {
+                "density": match.group(5).lower(), "mesh": match.group(6).lower(),
+                "payload": match.group(7).lower(), "topology": match.group(8).lower(),
+            }
+        match = GENERATED.search(line)
+        if match:
+            generated.add(tuple(int(match.group(index)) for index in range(2, 6)))
+    return hashes, generated
+
+
+def parse_summary(path: Path) -> dict:
+    content = path.read_text(encoding="utf-8", errors="replace")
+    pipeline = PIPELINE.search(content)
+    tile_pipeline = TILE_PIPELINE.search(content)
+    if pipeline is None or tile_pipeline is None:
+        raise SystemExit(f"pipeline summaries missing; see {path}")
+    values = dict(PAIR.findall(pipeline.group(1)))
+    tile_values = dict(PAIR.findall(tile_pipeline.group(1)))
+    tiles = values.get("whole_tiles", "0/0").split("/")
+    return {
+        "tiles": int(tiles[1]) if len(tiles) == 2 else 0,
+        "whole_tile_skips": int(tile_values.get("whole_tile_skips", "0")),
+        "classification_worker_ms": float(tile_values.get("classification_worker_ms", "0")),
+    }
+
+
+def run(mode: str, view: int, position: int, offset: tuple[float, float]) -> dict:
+    base = CAMERAS[view]
+    xyz = (base[0] + offset[0], base[1] + offset[1], base[2])
+    camera = ",".join(f"{value:.12f}" for value in xyz) + ",0,0"
+    log = OUT / f"sweep-{mode}-s{view}-p{position:02d}.log"
+    hash_flag = "--stream-hash-tiles" if mode == "reference" \
+        else "--stream-hash-tiles-allow-skips"
+    command = [
+        "taskset", "-c", "0-11", "prlimit", "--as=6442450944", "--", "timeout",
+        "1740s", str(BIN), str(DATA), "--world-assembler-prototype", "--stream-benchmark",
+        "--stream-pipeline", "--force-generate", "--workers=8", "--clip-radius=3",
+        "--max-clip-level=8", "--max-level=8", f"--camera={camera}", hash_flag,
+    ]
+    started = time.monotonic()
+    with log.open("w", encoding="utf-8") as output:
+        result = subprocess.run(command, cwd=ROOT, stdout=output, stderr=subprocess.STDOUT,
+                                check=False, timeout=1740)
+    hashes, _ = read_hashes(log)
+    if result.returncode != 0 or not hashes:
+        raise SystemExit(f"{mode} hash generation failed; see {log}")
+    summary = parse_summary(log)
+    row = {
+        "mode": mode, "strate": view, "position": position, "offset_m": list(offset),
+        "camera_m": list(xyz), "exit_code": result.returncode,
+        "wall_s": round(time.monotonic() - started, 3), "hash_rows": len(hashes),
+        **summary, "log": str(log.relative_to(ROOT)),
+    }
+    print("[L74HashSweepRun] " + json.dumps(row, sort_keys=True), flush=True)
+    return row
+
+
+def compare(reference: Path, candidate: Path) -> dict:
+    expected, _ = read_hashes(reference)
+    actual, candidate_generated = read_hashes(candidate)
+    missing = expected.keys() - actual.keys()
+    extra = actual.keys() - expected.keys()
+    mismatches = {field: 0 for field in ("mesh", "payload", "topology", "density_processed")}
+    for key in expected.keys() & actual.keys():
+        fields = ["mesh", "payload", "topology"]
+        if key in candidate_generated:
+            fields.append("density")
+        for field in fields:
+            if expected[key][field] != actual[key][field]:
+                mismatches["density_processed" if field == "density" else field] += 1
+    return {
+        "reference": str(reference.relative_to(ROOT)),
+        "candidate": str(candidate.relative_to(ROOT)),
+        "reference_tiles": len(expected), "candidate_tiles": len(actual),
+        "missing_keys": len(missing), "extra_keys": len(extra),
+        "mismatches": mismatches,
+        "equal": not missing and not extra and not any(mismatches.values()),
+    }
+
+
+def main() -> int:
+    if not DATA.is_file() or not BIN.is_file():
+        raise SystemExit("missing owner world or L74 Clang headless binary")
+    references, candidates, comparisons = [], [], []
+    for view in range(4):
+        for position, offset in enumerate(OFFSETS):
+            reference = run("reference", view, position, offset)
+            candidate = run("skip", view, position, offset)
+            comparison = compare(ROOT / reference["log"], ROOT / candidate["log"])
+            comparison.update(strate=view, position=position)
+            references.append(reference)
+            candidates.append(candidate)
+            comparisons.append(comparison)
+            print("[L74HashSweepCompare] " + json.dumps(comparison, sort_keys=True), flush=True)
+            if not comparison["equal"]:
+                raise SystemExit(f"whole-tile proof falsified; see {comparison['candidate']}")
+    summaries = []
+    for view in range(4):
+        rows = [row for row in candidates if row["strate"] == view]
+        skips = sum(row["whole_tile_skips"] for row in rows)
+        tiles = sum(row["tiles"] for row in rows)
+        costs = [row["classification_worker_ms"] for row in rows]
+        summaries.append({
+            "strate": view, "camera_positions": len(rows), "hash_comparisons": len(rows),
+            "violations": sum(not row["equal"] for row in comparisons
+                              if row["strate"] == view),
+            "whole_tile_proof_hits": skips, "tiles_visited": tiles,
+            "proof_hit_pct": round(100.0 * skips / max(tiles, 1), 2),
+            "classifier_ms_p50": round(statistics.median(costs), 3),
+            "classifier_ms_p95": round(sorted(costs)[int(0.95 * (len(costs) - 1))], 3),
+        })
+    result = {"runs": {"reference": references, "skip": candidates},
+              "comparisons": comparisons, "summary": summaries}
+    (OUT / "hash-sweep.json").write_text(json.dumps(result, indent=2) + "\n",
+                                         encoding="utf-8")
+    for row in summaries:
+        print("[L74HashSweepSummary] " + json.dumps(row, sort_keys=True), flush=True)
+    return 0
+
+
+if __name__ == "__main__":
+    raise SystemExit(main())