File size: 12,252 Bytes
6debdcc | 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 | """Verify the integrity and provenance of an archived research repository."""
from __future__ import annotations
import argparse
import csv
import hashlib
import json
import logging
from pathlib import Path
from typing import Any
LOGGER = logging.getLogger(__name__)
ARCHIVE_MANIFEST = Path("results/archive_manifest.json")
RAW_ROWS = Path("results/raw_rows.jsonl")
WEIGHT_MANIFEST = Path("weights/manifest.csv")
ARCHIVE_DIRS = ("images", "results", "samples", "weights")
def _sha256(path: Path) -> str:
digest = hashlib.sha256()
with path.open("rb") as handle:
for chunk in iter(lambda: handle.read(1024 * 1024), b""):
digest.update(chunk)
return digest.hexdigest()
def _safe_relative(root: Path, raw_path: str) -> Path:
candidate = Path(raw_path)
if candidate.is_absolute():
raise ValueError(f"path containment violation: {raw_path}")
raw_candidate = root / candidate
if raw_candidate.is_symlink():
raise ValueError(f"path is a symlink: {raw_path}")
resolved = raw_candidate.resolve()
try:
resolved.relative_to(root.resolve())
except ValueError as exc:
raise ValueError(f"path containment violation: {raw_path}") from exc
return resolved
def _relative_path(root: Path, path: Path) -> str:
return path.relative_to(root).as_posix()
def _validate_entry(root: Path, entry: dict[str, Any], *, label: str) -> Path:
raw_path = entry.get("path", entry.get("artifact_path"))
if not isinstance(raw_path, str) or not raw_path:
raise ValueError(f"{label}: missing path")
path = _safe_relative(root, raw_path)
if path.is_symlink():
raise ValueError(f"{label}: symlink is not allowed: {raw_path}")
if not path.is_file():
raise ValueError(f"{label}: listed file is missing: {raw_path}")
expected_bytes = entry.get("bytes", entry.get("size"))
expected_sha = entry.get("sha256", entry.get("sha"))
if not isinstance(expected_bytes, int) or path.stat().st_size != expected_bytes:
raise ValueError(f"{label}: byte count mismatch: {raw_path}")
if not isinstance(expected_sha, str) or _sha256(path) != expected_sha:
raise ValueError(f"{label}: SHA-256 mismatch: {raw_path}")
return path
def _manifest_entries(value: Any) -> list[dict[str, Any]]:
if isinstance(value, list):
entries = value
elif isinstance(value, dict):
entries = value.get("files", value.get("artifacts"))
else:
entries = None
if not isinstance(entries, list) or not all(isinstance(entry, dict) for entry in entries):
raise ValueError("archive manifest must contain a files/artifacts list")
return entries
def _read_weight_manifest(root: Path) -> list[dict[str, Any]]:
path = _safe_relative(root, str(WEIGHT_MANIFEST))
if path.is_symlink() or not path.is_file():
raise ValueError("weights/manifest.csv is missing or is a symlink")
with path.open(newline="", encoding="utf-8") as handle:
rows = list(csv.DictReader(handle))
if not rows:
raise ValueError("weights/manifest.csv contains no entries")
normalized: list[dict[str, Any]] = []
for index, row in enumerate(rows, 1):
try:
byte_count = int(row.get("bytes", row.get("size", "")))
except ValueError as exc:
raise ValueError(f"weights manifest row {index}: invalid bytes") from exc
normalized.append(
{"path": row.get("path", row.get("artifact_path")), "bytes": byte_count, "sha256": row.get("sha256", row.get("sha"))}
)
return normalized
def _load_jsonl_generation_rows(root: Path, conditions: list[str], prompt_count: int) -> list[dict[str, Any]]:
raw_path = _safe_relative(root, str(RAW_ROWS))
if raw_path.is_symlink() or not raw_path.is_file():
raise ValueError("results/raw_rows.jsonl is missing or is a symlink")
rows: list[dict[str, Any]] = []
seen: set[tuple[str, str]] = set()
try:
lines = raw_path.read_text(encoding="utf-8").splitlines()
except (OSError, UnicodeDecodeError) as exc:
raise ValueError("results/raw_rows.jsonl is not valid UTF-8") from exc
for line_number, line in enumerate(lines, 1):
if not line.strip():
continue
try:
row = json.loads(line)
except json.JSONDecodeError as exc:
raise ValueError(f"raw_rows.jsonl line {line_number} is not valid JSON") from exc
if not isinstance(row, dict):
raise TypeError(f"raw_rows.jsonl line {line_number} is not a JSON object")
if row.get("row_type") != "generation":
continue
condition = row.get("condition")
prompt_id = row.get("prompt_id")
if not isinstance(condition, str) or condition not in conditions:
raise ValueError(f"generation row {line_number}: unknown condition")
if not isinstance(prompt_id, str) or not prompt_id:
raise ValueError(f"generation row {line_number}: missing prompt_id")
key = (condition, prompt_id)
if key in seen:
raise ValueError(f"duplicate generation row: {condition}/{prompt_id}")
seen.add(key)
rows.append(row)
expected_rows = len(conditions) * prompt_count
if len(rows) != expected_rows:
raise ValueError(f"generation row count mismatch: expected {expected_rows}, got {len(rows)}")
prompt_ids = {str(row["prompt_id"]) for row in rows}
if len(prompt_ids) != prompt_count:
raise ValueError(f"prompt count mismatch: expected {prompt_count}, got {len(prompt_ids)}")
return rows
def _validate_generation_provenance(root: Path, rows: list[dict[str, Any]], archive_paths: set[str]) -> int:
successful = 0
for row in rows:
condition = str(row["condition"])
prompt_id = str(row["prompt_id"])
label = f"generation {condition}/{prompt_id}"
provenance = (row.get("sample_path"), row.get("sample_hash"), row.get("sample_sha256"), row.get("sample_bytes"))
if row.get("success"):
sample_raw, sample_hash, sample_sha256, sample_bytes = provenance
if not isinstance(sample_raw, str) or not sample_raw:
raise ValueError(f"{label}: successful row has no sample path")
sample_path = _safe_relative(root, sample_raw)
if sample_path.is_symlink() or not sample_path.is_file():
raise ValueError(f"{label}: successful sample is missing or is a symlink")
actual_bytes = sample_path.stat().st_size
actual_hash = _sha256(sample_path)
if not isinstance(sample_bytes, int) or actual_bytes != sample_bytes:
raise ValueError(f"{label}: sample byte count mismatch")
hashes = [value for value in (sample_hash, sample_sha256) if value is not None]
if not hashes or any(not isinstance(value, str) or value != actual_hash for value in hashes):
raise ValueError(f"{label}: sample SHA-256 mismatch")
relative = _relative_path(root, sample_path)
if relative not in archive_paths:
raise ValueError(f"{label}: sample is not listed in archive manifest: {relative}")
successful += 1
elif any(value is not None for value in provenance):
raise ValueError(f"{label}: failed row contains sample provenance")
return successful
def _reject_unlisted_tree_files(root: Path, archive_paths: set[str]) -> None:
for directory in ARCHIVE_DIRS:
for path in (root / directory).rglob("*"):
if path.is_symlink():
raise ValueError(f"symlink found in archived tree: {_relative_path(root, path)}")
if path.is_file():
relative = _relative_path(root, path)
if relative == ARCHIVE_MANIFEST.as_posix():
continue
if relative not in archive_paths:
raise ValueError(f"unlisted file in archived tree: {relative}")
def verify_archive(root: str | Path) -> dict[str, int]:
"""Verify archive files, raw generation provenance, containment, and symlink safety."""
archive_root = Path(root).resolve()
if not archive_root.is_dir():
raise ValueError(f"archive root is not a directory: {archive_root}")
for directory in ARCHIVE_DIRS:
path = _safe_relative(archive_root, directory)
if path.is_symlink() or not path.is_dir():
raise ValueError(f"archive directory missing or symlinked: {directory}")
manifest_path = _safe_relative(archive_root, str(ARCHIVE_MANIFEST))
if manifest_path.is_symlink() or not manifest_path.is_file():
raise ValueError("results/archive_manifest.json is missing or is a symlink")
try:
manifest = json.loads(manifest_path.read_text(encoding="utf-8"))
except (OSError, UnicodeDecodeError, json.JSONDecodeError) as exc:
raise ValueError("archive manifest is not valid UTF-8 JSON") from exc
if not isinstance(manifest, dict):
raise TypeError("archive manifest must be a JSON object")
profile = manifest.get("profile")
prompt_count = manifest.get("prompt_count")
conditions = manifest.get("conditions")
if not isinstance(profile, str) or not profile:
raise ValueError("archive manifest profile is missing")
if not isinstance(prompt_count, int) or prompt_count <= 0:
raise ValueError("archive manifest prompt_count is invalid")
if not isinstance(conditions, list) or not conditions or not all(isinstance(item, str) for item in conditions):
raise ValueError("archive manifest conditions are invalid")
if len(set(conditions)) != len(conditions):
raise ValueError("archive manifest conditions contain duplicates")
entries = _manifest_entries(manifest)
archive_paths: set[str] = set()
checked = 0
for index, entry in enumerate(entries, 1):
path = _validate_entry(archive_root, entry, label=f"archive manifest row {index}")
relative = _relative_path(archive_root, path)
if relative in archive_paths:
raise ValueError(f"duplicate archive path: {relative}")
archive_paths.add(relative)
checked += 1
for required in ("images", "results", "samples", "weights"):
if not any(path == required or path.startswith(required + "/") for path in archive_paths):
raise ValueError(f"archive manifest must list a file under {required}")
for required in (RAW_ROWS.as_posix(), WEIGHT_MANIFEST.as_posix()):
if required not in archive_paths:
raise ValueError(f"archive manifest must list {required}")
weight_entries = _read_weight_manifest(archive_root)
weight_paths: set[str] = set()
for index, entry in enumerate(weight_entries, 1):
path = _validate_entry(archive_root, entry, label=f"weights manifest row {index}")
relative = _relative_path(archive_root, path)
if not relative.startswith("weights/"):
raise ValueError(f"weights manifest path is outside weights/: {relative}")
if relative in weight_paths:
raise ValueError(f"duplicate weight path: {relative}")
weight_paths.add(relative)
generation_rows = _load_jsonl_generation_rows(archive_root, conditions, prompt_count)
successful_samples = _validate_generation_provenance(archive_root, generation_rows, archive_paths)
_reject_unlisted_tree_files(archive_root, archive_paths)
result = {
"checked_files": checked + len(weight_entries),
"checked_weight_entries": len(weight_entries),
"generation_rows": len(generation_rows),
"successful_samples": successful_samples,
}
LOGGER.info("archive verified: %s", result)
return result
def main() -> int:
parser = argparse.ArgumentParser(description=__doc__)
parser.add_argument("--root", type=Path, default=Path("."), help="archive repository root")
args = parser.parse_args()
logging.basicConfig(level=logging.INFO, format="%(levelname)s %(message)s")
report = verify_archive(args.root)
print(json.dumps(report, sort_keys=True))
return 0
if __name__ == "__main__":
raise SystemExit(main())
|