fix: rank archives by name on a shared crc32

segadimm.zip holds fpr23905.ic36 and fpr23905c.ic36 under the same crc32, and segasp.zip four fpr-*.ic72 images. Ranked by crc32 alone, a copy missing some of them tied with the full set and the index order chose it. A crc32 that two declared members share now also requires the member name.

install/recalbox.json

@@ -280,9 +280,9 @@
     }
   ],
   "total_files": 3743,
-  "total_size": 4682168657,
+  "total_size": 4685153679,
   "pack_files": 3745,
-  "pack_size": 4682168657,
+  "pack_size": 4685153679,
   "total_omitted": 62,
   "omitted_files": [
     {
@@ -42218,14 +42218,13 @@
     },
     {
       "dest": "segasp.zip",
-      "sha1": "de78a0b8df78d5ffdb1a274f80d877a96ef204aa",
-      "sha256": "3bca348667d6ff08fd5efe38eec35100ccbe0de3cd1e0b5dd72ae5c6cc201ee9",
-      "size": 1281424,
-      "repo_path": "bios/Arcade/Arcade/segasp.zip",
+      "sha1": "b4c7e46fb510096acf8fb19fb99c8c5a8ab1d9e1",
+      "sha256": "bda96939992d42410034caf3f5296d8fc63eced1693b5b522536c6ac717ded73",
+      "size": 4266446,
+      "repo_path": "bios/Arcade/MAME/segasp.zip",
       "cores": [
         "mame0258",
-        "mame0278",
-        "mame2016"
+        "mame0278"
       ],
       "systems": [
         "3do",
@@ -42238,6 +42237,17 @@
         "scv",
         "stv",
         "ti994a"
+      ],
+      "alternatives": [
+        {
+          "sha1": "de78a0b8df78d5ffdb1a274f80d877a96ef204aa",
+          "sha256": "3bca348667d6ff08fd5efe38eec35100ccbe0de3cd1e0b5dd72ae5c6cc201ee9",
+          "size": 1281424,
+          "repo_path": "bios/Arcade/Arcade/segasp.zip",
+          "cores": [
+            "mame2016"
+          ]
+        }
       ]
     },
     {

scripts/common.py

@@ -1149,19 +1149,30 @@ def contents_members(file_entry: dict) -> list[dict]:
     A member is its crc32 where the profile declares one, its name where the
     code opens it by name alone: Flycast reads epr-21329.ic27 in
     hod2bios.zip by filename, and a copy without it has no Japan BIOS.
+    A crc32 two members share also keeps the name: fpr23905.ic36 and
+    fpr23905c.ic36 both have crc32 ffffffff in segadimm.zip, so the crc
+    alone cannot tell a copy holding one from a copy holding both.
     """
-    return [
-        {"crc32": str(m["crc32"]).lower(), "required": True}
-        if m.get("crc32") else {"name": str(m["name"]), "required": True}
-        for m in file_entry.get("contents") or []
-        if m.get("crc32") or m.get("name")
-    ]
+    listed = [m for m in file_entry.get("contents") or [] if m.get("crc32") or m.get("name")]
+    seen = Counter(str(m["crc32"]).lower() for m in listed if m.get("crc32"))
+    members = []
+    for m in listed:
+        if not m.get("crc32"):
+            members.append({"name": str(m["name"]), "required": True})
+            continue
+        crc = str(m["crc32"]).lower()
+        member = {"crc32": crc, "required": True}
+        if seen[crc] > 1 and m.get("name"):
+            member["name"] = str(m["name"])
+        members.append(member)
+    return members
 
 
 def _member_held(member: dict, crcs: frozenset[str], names: frozenset[str]) -> bool:
+    named = member.get("name", "").rsplit("/", 1)[-1].casefold() in names
     if member.get("crc32"):
-        return member["crc32"] in crcs
-    return member["name"].rsplit("/", 1)[-1].casefold() in names
+        return member["crc32"] in crcs and ("name" not in member or named)
+    return named
 
 
 def _holding_members(paths: list[str], members: list[dict]) -> list[str]:

tests/test_archive_choice.py

@@ -390,6 +390,22 @@ class AnArchiveRankedByWhatItHolds(unittest.TestCase):
             for name in ("epr-21330.ic27", "epr-21332.ic27")]}
         self.assertEqual(self._resolve(entry, db), "bios/MAME/hod2bios.zip")
 
+    def test_a_shared_crc_keeps_the_member_name(self):
+        """segadimm.zip holds fpr23905.ic36 and fpr23905c.ic36, both crc32
+        ffffffff. Ranked by crc alone, a copy without fpr23905c.ic36 tied with
+        the full set and came first in the index."""
+        self.roms["fpr23905.ic36"] = self.roms["fpr23905c.ic36"] = b"\xff" * 16
+        db = self._db([("bios/Arcade/hod2bios.zip", ["epr-21330.ic27", "fpr23905.ic36"]),
+                       ("bios/MAME/hod2bios.zip",
+                        ["epr-21330.ic27", "fpr23905.ic36", "fpr23905c.ic36"])])
+        crc = format(zlib.crc32(b"\xff" * 16), "08x")
+        first = format(zlib.crc32(self.roms["epr-21330.ic27"]), "08x")
+        entry = {"name": "hod2bios.zip", "contents": [
+            {"name": "epr-21330.ic27", "crc32": first},
+            {"name": "fpr23905.ic36", "crc32": crc},
+            {"name": "fpr23905c.ic36", "crc32": crc}]}
+        self.assertEqual(self._resolve(entry, db), "bios/MAME/hod2bios.zip")
+
     def test_a_copy_holding_no_declared_member_is_not_the_archive(self):
         """galgames.zip held only an assembler listing and none of the six
         ROMs MAME declares: the name step served it as the BIOS set."""