chore: drop files whose bytes are not what they claim

scripts/identity.py

@@ -0,0 +1,158 @@
+"""Files whose bytes contradict the identity they are stored under.
+
+The collection is indexed by content, so one sha1 keeps one path. A file
+copied into a second system's directory under a second name therefore becomes
+invisible: the index answers with whichever path was recorded, and every
+consumer of that name is served bytes belonging to another machine. That is
+how the Amiga Kickstart came to be served as a Saturn cartridge firmware, and
+nothing in the pipeline could see it, because on paper both names resolved.
+
+This walks the tree rather than the index, groups by content, and reports a
+group whose paths sit under different manufacturers. A dump legitimately
+shared by two systems of the same maker is common; one shared across makers
+almost never is.
+"""
+
+from __future__ import annotations
+
+import hashlib
+from dataclasses import dataclass
+from pathlib import Path
+
+# A manufacturer directory holding files that genuinely belong to several
+# machines, where a shared dump says nothing about a misfiling.
+SHARED_TREES = frozenset({"Arcade", "Other"})
+
+# One dump, two machines that really did share it. Each entry names the reason,
+# so a crossing that is not on this list is a finding rather than noise.
+KNOWN_SHARED = {
+    "5ea7c2b824672e914525d1d5c419d71b84a426a2": "ZX Spectrum 48K ROM, which the "
+    "Enterprise and the FBNeo Spectrum driver run as well",
+    "16375d42ea109b47edded7a16028de7fdb3013a1": "ZX Spectrum 128K ROM, under both "
+    "Sinclair spellings and the Enterprise Spectrum mode",
+    "45bedc4cbdeac66c7df59e9e599195c778d86a92": "ColecoVision BIOS, carried by MSX "
+    "emulators for their ColecoVision machine",
+    "b2e1955d957a475de2411770452eff4ea19f4cee": "Odyssey 2 and Videopac are one "
+    "machine sold under two names",
+    "0d2ef6e67322f48f4b7e08d8bbe68827e2074561": "Oric Microdisc ROM, the Oric being "
+    "a Tangerine machine",
+    "0334b35f164089df7b2a82d46fa0ec6e43fafa90": "ZX81 overlay, the TS1500 being the "
+    "same machine for Timex",
+    "286b2e1fb21cc79851da01666db6c0b0e88f25e3": "one soundfont shipped by two "
+    "interpreters",
+    "a439fbb390da38cada95a7cbb1d6ca199cd66ef8": "Roland CM-32L control ROM, kept a "
+    "second time in the DOSBox mt32-roms tree",
+    "b083518fffb7f66b03c23b7eb4f868e62dc5a987": "Roland MT-32 control ROM, same",
+    "289cc298ad532b702461bfc738009d9ebe8025ea": "Roland CM-32L PCM ROM, same",
+    "f6b1eebc4b2d200ec6d3d21d51325d5b48c60252": "Roland MT-32 PCM ROM, same",
+}
+
+CHUNK = 1 << 20
+
+
+@dataclass
+class Duplicate:
+    """One content, stored under paths that disagree about what it is."""
+
+    sha1: str
+    size: int
+    paths: list[str]
+    manufacturers: list[str]
+
+    @property
+    def names(self) -> list[str]:
+        return sorted({Path(p).name for p in self.paths})
+
+
+def _manufacturer(path: Path, root: Path) -> str:
+    """The vendor directory a path sits under, relative to the scanned root."""
+    parts = path.relative_to(root).parts
+    return parts[0] if len(parts) > 1 else ""
+
+
+def _sha1(path: Path) -> str:
+    digest = hashlib.sha1()
+    with path.open("rb") as handle:
+        for chunk in iter(lambda: handle.read(CHUNK), b""):
+            digest.update(chunk)
+    return digest.hexdigest()
+
+
+def scan(bios_dir: str = "bios") -> list[Duplicate]:
+    """Group the tree by content and keep the groups that cross makers."""
+    by_content: dict[str, list[Path]] = {}
+    root = Path(bios_dir)
+    for path in sorted(root.rglob("*")):
+        if not path.is_file():
+            continue
+        by_content.setdefault(_sha1(path), []).append(path)
+
+    found: list[Duplicate] = []
+    for sha1, paths in by_content.items():
+        if len(paths) < 2:
+            continue
+        if sha1 in KNOWN_SHARED:
+            continue
+        makers = {_manufacturer(p, root) for p in paths} - SHARED_TREES
+        names = {p.name for p in paths}
+        if len(makers) > 1 and len(names) > 1:
+            found.append(
+                Duplicate(
+                    sha1=sha1,
+                    size=paths[0].stat().st_size,
+                    paths=[str(p) for p in paths],
+                    manufacturers=sorted(makers),
+                )
+            )
+    return sorted(found, key=lambda d: d.paths[0])
+
+
+def format_duplicate(duplicate: Duplicate) -> str:
+    """One line naming the content and every identity it answers to."""
+    return (
+        f"{duplicate.sha1[:12]} ({duplicate.size} bytes) is stored as "
+        + " and as ".join(duplicate.paths)
+    )
+
+
+def main() -> int:
+    import argparse
+
+    parser = argparse.ArgumentParser(
+        description="Report one content stored under identities that disagree.",
+    )
+    parser.add_argument("--bios-dir", default="bios")
+    parser.add_argument("--json", action="store_true", help="JSON output")
+    parser.add_argument(
+        "--strict", action="store_true", help="exit non-zero when any is found"
+    )
+    args = parser.parse_args()
+
+    duplicates = scan(args.bios_dir)
+    if args.json:
+        import json
+
+        print(
+            json.dumps(
+                [
+                    {
+                        "sha1": d.sha1,
+                        "size": d.size,
+                        "paths": d.paths,
+                        "manufacturers": d.manufacturers,
+                    }
+                    for d in duplicates
+                ],
+                indent=2,
+            )
+        )
+    else:
+        for duplicate in duplicates:
+            print(f"  {format_duplicate(duplicate)}")
+        print(f"\n{len(duplicates)} contents stored under contradicting identities.")
+
+    return 1 if (duplicates and args.strict) else 0
+
+
+if __name__ == "__main__":
+    raise SystemExit(main())