Five conflicts resolved (CURRENT-STATE, ResolvedLayoutInternal.h, StreamBenchmarkHeadless.cpp, StreamBenchmarkPipeline.h, WorldAssemblyPrototype.h) keeping liminal + cave planner; L156 band fit only; evaluator v239 with v239 baselines (18/18 Clang/GCC W1/W8/W12). Fast tier FAILS (FI upper arrival: 40 unsupported chamber samples, 10 chamber-floor shell violations); TN and VS composed sweeps 0/16 (passage, seam, shell gates). Living plane loads at PostConfigInit (UE smoke passes). Source/: VoxelForgeCoreAdapter.cpp; VoxelForge.uplugin. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_016dhxM43dnWvUu2B5Ph1iG1
182 lines
9.1 KiB
Python
182 lines
9.1 KiB
Python
#!/usr/bin/env python3
|
|
"""Run the composed Tunnel Network or Vertical Shafts gates for seeds 0..15."""
|
|
from __future__ import annotations
|
|
|
|
import argparse
|
|
import json
|
|
import re
|
|
import subprocess
|
|
import sys
|
|
import time
|
|
from pathlib import Path
|
|
|
|
|
|
ROOT = Path(__file__).resolve().parents[4]
|
|
OUT = ROOT / "Saved/Linux/l173/cave-sweeps/seeds"
|
|
BIN = ROOT / "Saved/Linux/l173/fast/bin/VoxelForgeCoreHeadless-fast-clang"
|
|
DATA = ROOT / "Standalone/data/owner_world.json"
|
|
STAMP_TOOL = ROOT / "Standalone/Tools/verify/source_stamp.py"
|
|
STAMP_RE = re.compile(r"^\[VoxelForgeSourceStamp\] sha256=([0-9a-f]+)$", re.M)
|
|
DISPATCH = {
|
|
"tunnel-network": re.compile(
|
|
r"^\[DensityDispatch\] strate=2 generator=TunnelNetwork "
|
|
r"recipe=Core\.Caves\.PrimitiveField version=1 "
|
|
r"bound_path=Caves-native adapter=FCavesFieldAdapter$", re.M),
|
|
"vertical-shafts": re.compile(
|
|
r"^\[DensityDispatch\] strate=3 generator=VerticalShafts "
|
|
r"recipe=Core\.Caves\.PrimitiveField version=1 "
|
|
r"bound_path=Caves-native adapter=FCavesFieldAdapter$", re.M),
|
|
}
|
|
ROUTES = {
|
|
"tunnel-network": re.compile(
|
|
r"^\[ResolvedStrateRouteAudit\] strate=2 .*unsupported=0 "
|
|
r"grade_failures=0 .*status=PASS$", re.M),
|
|
"vertical-shafts": re.compile(
|
|
r"^\[ResolvedStrateRouteAudit\] strate=3 .*unsupported=0 "
|
|
r"grade_failures=0 .*status=PASS$", re.M),
|
|
}
|
|
CONTROLS = {
|
|
"tunnel-network": re.compile(
|
|
r"^\[TunnelNetworkBlockedRouteControl\].*expected=FAIL status=PASS$", re.M),
|
|
"vertical-shafts": re.compile(
|
|
r"^\[VerticalShaftsBlockedRouteControl\].*expected=FAIL status=PASS$", re.M),
|
|
}
|
|
REQUIRED = {
|
|
"passages": re.compile(r"^\[ResolvedPassageAuditSummary\].*status=PASS$", re.M),
|
|
"seams": re.compile(r"^\[PassageTileSeamAudit\].*status=PASS$", re.M),
|
|
"shell": re.compile(r"^\[PassageShellAuditSummary\].*status=PASS$", re.M),
|
|
"naturalness": re.compile(r"^\[PassageNaturalnessSummary\].*status=PASS$", re.M),
|
|
"mouth_profile": re.compile(r"^\[PassageMouthProfileSummary\].*status=PASS$", re.M),
|
|
"ripple": re.compile(r"^\[PassageRippleAuditSummary\].*status=PASS$", re.M),
|
|
"passage_interior": re.compile(
|
|
r"^\[PassageInteriorCapsuleAuditSummary\].*status=PASS$", re.M),
|
|
}
|
|
|
|
|
|
def evaluate_seed(kind: str, content: str, stamp: str, layout: Path) -> tuple[dict, str | None]:
|
|
stamp_match = STAMP_RE.search(content)
|
|
checks = {name: bool(pattern.search(content)) for name, pattern in REQUIRED.items()}
|
|
checks["source_stamp"] = stamp_match is not None and stamp_match.group(1) == stamp
|
|
checks["dispatch"] = bool(DISPATCH[kind].search(content))
|
|
checks["route"] = bool(ROUTES[kind].search(content))
|
|
checks["blocked_route_control"] = bool(CONTROLS[kind].search(content))
|
|
checks["three_passage_interiors"] = len(re.findall(
|
|
r"^\[ResolvedPassageAudit\].*interior_air=PASS", content, re.M)) == 3
|
|
checks["three_interior_negative_controls"] = len(re.findall(
|
|
r"^\[ResolvedPassageInteriorNegativeControl\].*detected=true status=PASS$",
|
|
content, re.M)) == 3
|
|
checks["three_ripple_gap_controls"] = len(re.findall(
|
|
r"^\[PassageRippleSamplingControl\].*status=PASS$", content, re.M)) == 3
|
|
checks["three_blocked_passage_controls"] = len(re.findall(
|
|
r"^\[ResolvedPassageBlockedControl\].*detected=true$", content, re.M)) == 3
|
|
checks["passage_count"] = len(re.findall(
|
|
r"^\[ResolvedPassageAudit\].*walkability=PASS$", content, re.M)) == 3
|
|
layout_identity = None
|
|
try:
|
|
checks["layout_output"] = layout.is_file()
|
|
if layout.is_file():
|
|
layout_identity = json.loads(layout.read_text(encoding="utf-8")).get(
|
|
"stable_identity")
|
|
except (OSError, json.JSONDecodeError):
|
|
checks["layout_output"] = False
|
|
return checks, layout_identity
|
|
|
|
|
|
def main() -> int:
|
|
parser = argparse.ArgumentParser(description=__doc__)
|
|
parser.add_argument("--kind", choices=tuple(DISPATCH), required=True)
|
|
parser.add_argument("--grade-only", action="store_true",
|
|
help="regrade completed logs without rerunning the headless executable")
|
|
args = parser.parse_args()
|
|
if not BIN.is_file() or not DATA.is_file():
|
|
raise SystemExit("missing current Clang binary or composed owner fixture")
|
|
stamp = subprocess.check_output(
|
|
[sys.executable, str(STAMP_TOOL), "stamp", "--root", str(ROOT)], text=True).strip()
|
|
mode_out = OUT / args.kind
|
|
mode_out.mkdir(parents=True, exist_ok=True)
|
|
if args.grade_only:
|
|
partial = mode_out / "runs-partial.json"
|
|
previous = json.loads(partial.read_text(encoding="utf-8")) if partial.is_file() else []
|
|
by_seed = {int(row["seed"]): row for row in previous}
|
|
results = []
|
|
for seed in range(16):
|
|
log = mode_out / f"seed{seed}.log"
|
|
layout = mode_out / f"seed{seed}.json"
|
|
content = log.read_text(encoding="utf-8", errors="replace") \
|
|
if log.is_file() else ""
|
|
checks, identity = evaluate_seed(args.kind, content, stamp, layout)
|
|
code = by_seed.get(seed, {}).get("exit")
|
|
passed = code == 0 and all(checks.values())
|
|
failed_checks = [name for name, ok in checks.items() if not ok]
|
|
row = {"kind": args.kind, "seed": seed,
|
|
"verdict": "PASS" if passed else "FAIL",
|
|
"reason": ", ".join(failed_checks) if failed_checks else
|
|
"all composed gates pass" if code == 0 else f"process exit {code}",
|
|
"exit": code, "checks": checks, "source_stamp": stamp,
|
|
"layout_identity": identity,
|
|
"log": str(log.relative_to(ROOT)), "layout": str(layout.relative_to(ROOT))}
|
|
results.append(row)
|
|
passed_count = sum(row["verdict"] == "PASS" for row in results)
|
|
summary = {"kind": args.kind, "source_stamp": stamp,
|
|
"grading_mode": "existing-captured-logs", "seeds": list(range(16)),
|
|
"passed": passed_count, "failed": 16 - passed_count, "results": results}
|
|
(mode_out / "runs-regraded.json").write_text(
|
|
json.dumps(results, indent=2) + "\n", encoding="utf-8")
|
|
(mode_out / "summary.json").write_text(
|
|
json.dumps(summary, indent=2) + "\n", encoding="utf-8")
|
|
print(f"[L173CaveSeedSweepSummary] kind={args.kind} "
|
|
f"passed={passed_count}/16 failed={16-passed_count} "
|
|
f"status={'PASS' if passed_count == 16 else 'FAIL'} grading=existing-logs",
|
|
flush=True)
|
|
return 0 if passed_count == 16 else 1
|
|
results = []
|
|
partial = mode_out / "runs-partial.json"
|
|
for seed in range(16):
|
|
log = mode_out / f"seed{seed}.log"
|
|
layout = mode_out / f"seed{seed}.json"
|
|
command = [
|
|
"taskset", "-c", "0-11", "prlimit", "--as=6442450944", "--",
|
|
"timeout", "1740s", str(BIN), str(DATA), "--world-assembler-prototype",
|
|
"--stream-benchmark", "--stream-pipeline", "--workers=8", "--clip-radius=3",
|
|
"--max-clip-level=8", "--max-level=0", "--force-generate", f"--seed={seed}",
|
|
"--recipe=surface-world", "--density-dispatch-report",
|
|
"--audit-resolved-layout-passages", "--audit-passage-tile-seams",
|
|
f"--resolved-layout-output={layout}",
|
|
]
|
|
started = time.monotonic()
|
|
try:
|
|
with log.open("w", encoding="utf-8") as output:
|
|
proc = subprocess.run(command, cwd=ROOT, stdout=output,
|
|
stderr=subprocess.STDOUT, check=False, timeout=1740)
|
|
code = proc.returncode
|
|
except subprocess.TimeoutExpired:
|
|
code = "OUTER_TIMEOUT"
|
|
content = log.read_text(encoding="utf-8", errors="replace") if log.exists() else ""
|
|
checks, layout_identity = evaluate_seed(args.kind, content, stamp, layout)
|
|
passed = code == 0 and all(checks.values())
|
|
failed_checks = [name for name, ok in checks.items() if not ok]
|
|
row = {
|
|
"kind": args.kind, "seed": seed, "verdict": "PASS" if passed else "FAIL",
|
|
"reason": ", ".join(failed_checks) if failed_checks else "all composed gates pass",
|
|
"exit": code, "checks": checks,
|
|
"source_stamp": stamp,
|
|
"layout_identity": layout_identity,
|
|
"elapsed_s": round(time.monotonic() - started, 3),
|
|
"log": str(log.relative_to(ROOT)), "layout": str(layout.relative_to(ROOT)),
|
|
}
|
|
results.append(row)
|
|
print("[L173CaveSeedSweep] " + json.dumps(row, sort_keys=True), flush=True)
|
|
partial.write_text(json.dumps(results, indent=2) + "\n", encoding="utf-8")
|
|
passed_count = sum(row["verdict"] == "PASS" for row in results)
|
|
summary = {"kind": args.kind, "source_stamp": stamp, "seeds": list(range(16)),
|
|
"passed": passed_count, "failed": 16 - passed_count, "results": results}
|
|
(mode_out / "summary.json").write_text(
|
|
json.dumps(summary, indent=2) + "\n", encoding="utf-8")
|
|
print(f"[L173CaveSeedSweepSummary] kind={args.kind} passed={passed_count}/16 "
|
|
f"failed={16-passed_count} status={'PASS' if passed_count == 16 else 'FAIL'}", flush=True)
|
|
return 0 if passed_count == 16 else 1
|
|
|
|
|
|
if __name__ == "__main__":
|
|
raise SystemExit(main())
|