chore: drop copies only an install provides

A file that arrives with the emulator is already on the user's machine
and the packs leave it out. 4558 collected copies (1.4 GB) were reached
only by such entries, with no platform, other profile entry or install
manifest needing them. They are removed, and two release-asset copies
leave the ignore list. bundled_only.py now lists every copy of a listed
file, since the database keeps one path per content, and pipeline step
4d fails while one remains.

scripts/bundled_only.py

@@ -19,6 +19,7 @@ import json
 from pathlib import Path
 
 from common import (
+    compute_hashes,
     expand_directory_entries,
     list_registered_platforms,
     load_database,
@@ -70,12 +71,39 @@ def bundled_only(db: dict, profiles: dict, platforms_dir: str, install_dir: Path
             else:
                 needed.add(sha1)
     files = db.get("files", {})
-    return [
+    rows = [
         {"sha1": sha1, "path": files[sha1]["path"], "size": files[sha1].get("size", 0),
          "emulators": sorted(emulators)}
         for sha1, emulators in sorted(bundled.items(), key=lambda kv: files[kv[0]]["path"])
         if sha1 not in needed and sha1 in files
     ]
+    return rows + other_copies(rows, Path(db.get("bios_dir") or "bios"))
+
+
+def other_copies(rows: list[dict], bios_dir: Path) -> list[dict]:
+    """The same bytes under other paths of the collection.
+
+    The database keeps one path per sha1, and the trees kept whole
+    (NODEDUP: splitwolf's sprite directories, c64-emu's machine folders)
+    hold the same file several times. Listing only the indexed path took
+    five passes to empty splitwolf. A file is hashed only when its size is
+    one a listed file has.
+    """
+    by_size: dict[int, dict[str, dict]] = {}
+    for row in rows:
+        by_size.setdefault(int(row["size"]), {})[row["sha1"]] = row
+    listed = {row["path"] for row in rows}
+    found = []
+    for path in sorted(bios_dir.rglob("*")):
+        if not path.is_file() or path.is_symlink() or path.as_posix() in listed:
+            continue
+        same_size = by_size.get(path.stat().st_size)
+        if not same_size:
+            continue
+        sha1 = compute_hashes(path, frozenset({"sha1"}))["sha1"]
+        if sha1 in same_size:
+            found.append({**same_size[sha1], "path": path.as_posix()})
+    return found
 
 
 def main() -> int:

scripts/identity.py

@@ -51,10 +51,6 @@ KNOWN_SHARED: dict[str, tuple[frozenset[str], str]] = {
         frozenset({"Oric", "Tangerine"}),
         "Oric Microdisc ROM, the Oric being a Tangerine machine",
     ),
-    "0334b35f164089df7b2a82d46fa0ec6e43fafa90": (
-        frozenset({"Sinclair", "Timex"}),
-        "ZX81 overlay, the TS1500 being the same machine for Timex",
-    ),
     "286b2e1fb21cc79851da01666db6c0b0e88f25e3": (
         frozenset({"Infocom", "ZC"}),
         "one soundfont shipped by two interpreters",

scripts/pipeline.py

@@ -20,6 +20,7 @@ Steps, numbered as the run prints them:
   4    generate_pack.py --all       ZIP packs
   4b   generate_pack.py --manifest  install manifests
   4c   --manifest-targets           per-target core lists
+  4d   bundled_only.py --strict     copies only an emulator install provides
   5    consistency check            verify counts == pack counts
   6    --verify-packs               pack integrity, extraction and hashes
   7    generate_readme.py           README.md and CONTRIBUTING.md
@@ -728,6 +729,17 @@ def main() -> None:
         print("\n--- 4c/8 generate target manifests: SKIPPED (--skip-packs) ---")
         results["generate_target_manifests"] = SKIPPED
 
+    # Step 4d: A collected copy that only an emulator's own install provides.
+    # Packs leave such files out, so a copy nothing else asks for is weight
+    # the collection carries for nobody. Read after the manifests, which say
+    # what an installation ships.
+    ok, _ = run(
+        [sys.executable, "scripts/bundled_only.py", "--strict"],
+        "4d/8 copies only an emulator install provides",
+    )
+    results["bundled_only"] = ok
+    all_ok &= ok
+
     # Step 5: Consistency check. verify describes the full pack, and a run that
     # built only a reduced one (--source platform or truth) has nothing to
     # compare it with, and says so instead of failing on an empty parse.