feat: document what settles each category

Each missing platform entry now carries its evidence: the profile entries that read it with their source reference, the cores of its system examined without finding a reader, or why the question stays open. verify --verbose prints it, --json carries it and the platform pages show it.

scripts/exporter/candidates.py

@@ -241,50 +241,89 @@ class EntryCategories:
     archive, and the platform declares the cores it runs. Anything else,
     and every entry of a platform with no core to examine (BizHawk, MiSTer),
     stays firmware.
+
+    `explain` returns the category with its evidence: the profile entries
+    that read the file, or the cores of its system that were examined.
     """
 
     def __init__(self, config: dict, profiles: dict[str, dict],
                  data_registry: dict | None = None) -> None:
         examined = resolve_platform_cores(config, profiles)
+        self._unresolved = sorted(unresolved_platform_cores(config, profiles))
         # Without a declared core list, the examined cores are those sharing
         # a system id: a guess about what runs, which cannot prove a negative.
-        self._open = (
-            not examined or not config.get("cores")
-            or bool(unresolved_platform_cores(config, profiles))
-        )
-        self._romsets = any(_loads_romsets(profiles[key]) for key in examined)
-        self._firmware = ReadEvidence()
-        self._game = ReadEvidence()
+        self._guessed = not config.get("cores")
+        self._romsets = sorted(key for key in examined if _loads_romsets(profiles[key]))
+        self._profiles = profiles
+        # One index per core and per kind, so a proof names its reader.
+        self._readers: list[tuple[str, str, ReadEvidence]] = []
         for key in sorted(examined):
             profile = profiles[key]
             files = [f for f in profile.get("files") or [] if isinstance(f, dict)]
-            self._firmware.add_profile(
-                {**profile, "files": [f for f in files if f.get("category") != GAME_DATA]},
-                data_registry or {},
-            )
-            self._game.add_profile(
-                {"files": [f for f in files if f.get("category") == GAME_DATA]},
-                data_registry or {},
-            )
-        self._firmware.add_launchers(config.get("launchers") or [])
-        self._systems = {
-            _norm_system_id(str(sid))
-            for key in examined
-            for sid in profiles[key].get("systems") or []
+            for kind, keep in ((BIOS, lambda f: f.get("category") != GAME_DATA),
+                               (GAME_DATA, lambda f: f.get("category") == GAME_DATA)):
+                evidence = ReadEvidence()
+                evidence.add_profile(
+                    {"files": [f for f in files if keep(f)],
+                     "data_directories": profile.get("data_directories") if kind == BIOS else []},
+                    data_registry or {},
+                )
+                self._readers.append((kind, key, evidence))
+        self._launchers = ReadEvidence()
+        self._launchers.add_launchers(config.get("launchers") or [])
+        self._by_system: dict[str, list[str]] = {}
+        for key in sorted(examined):
+            for sid in profiles[key].get("systems") or []:
+                self._by_system.setdefault(_norm_system_id(str(sid)), []).append(key)
+
+    def _matching(self, key: str, kind: str, entry: dict) -> list[dict]:
+        """The profile entries of `key` that name the platform entry."""
+        names = {
+            str(n).casefold()
+            for n in (entry.get("name"), *(entry.get("aliases") or []))
+            if n
         }
+        found = []
+        for f in self._profiles[key].get("files") or []:
+            if not isinstance(f, dict) or (f.get("category") == GAME_DATA) != (kind == GAME_DATA):
+                continue
+            own = {str(n).casefold() for n in (f.get("name"), *(f.get("aliases") or [])) if n}
+            if own & names:
+                found.append({k: f[k] for k in ("name", "source_ref") if f.get(k)})
+        return found
+
+    def explain(self, entry: dict, destination: str, system: str) -> dict:
+        """The category of an entry and what establishes it."""
+        readers: dict[str, list[dict]] = {BIOS: [], GAME_DATA: []}
+        for kind, key, evidence in self._readers:
+            proof = evidence.proof(entry, destination)
+            if proof is not None:
+                readers[kind].append({
+                    "emulator": key, "proof": proof,
+                    "entries": self._matching(key, kind, entry),
+                })
+        if self._launchers.proof(entry, destination) is not None:
+            readers[BIOS].append({"emulator": "launcher", "proof": "launcher", "entries": []})
+        system_cores = self._by_system.get(_norm_system_id(system), [])
+        if readers[BIOS]:
+            return {"category": BIOS, "read_by": readers[BIOS]}
+        if readers[GAME_DATA]:
+            return {"category": GAME_DATA, "read_by": readers[GAME_DATA]}
+        examined = {"cores_examined": system_cores}
+        if self._guessed:
+            return {"category": BIOS, "unsettled": "the platform declares no core list", **examined}
+        if self._unresolved:
+            return {"category": BIOS,
+                    "unsettled": "cores the platform names have no profile: "
+                                 + ", ".join(self._unresolved), **examined}
+        if not system_cores:
+            return {"category": BIOS,
+                    "unsettled": "no examined core emulates this system", **examined}
+        if self._romsets and str(entry.get("name", "")).casefold().endswith(_ARCHIVES):
+            return {"category": BIOS,
+                    "unsettled": "a ROM set, opened by name by " + ", ".join(self._romsets),
+                    **examined}
+        return {"category": UNREAD, **examined}
 
     def category(self, entry: dict, destination: str, system: str) -> str:
-        """`system` is the platform's system id the entry is declared under.
-
-        Unread also needs a core of that system among those examined: an
-        entry of a system no profiled core emulates is simply unexamined.
-        """
-        if self._firmware.proof(entry, destination) is not None:
-            return BIOS
-        if self._game.proof(entry, destination) is not None:
-            return GAME_DATA
-        if self._open or _norm_system_id(system) not in self._systems:
-            return BIOS
-        if self._romsets and str(entry.get("name", "")).casefold().endswith(_ARCHIVES):
-            return BIOS
-        return UNREAD
+        return self.explain(entry, destination, system)["category"]

scripts/generate_site.py

@@ -1993,6 +1993,12 @@ def _render_platform_systems(
                 status_display = f"untested: {reason}" if reason else "untested"
             elif status == "missing":
                 status_display = "**missing**"
+                # Not every missing entry is firmware: the platform may list a
+                # game's data or a file no core it runs opens.
+                kind = {"game_data": "game data",
+                        "unread": "no core the platform runs opens it"}.get(f.get("category", ""))
+                if kind:
+                    status_display += f", {kind}"
             else:
                 status_display = status
 
@@ -2027,6 +2033,8 @@ def _render_platform_systems(
                 if sha256:
                     hash_parts.append(f"SHA256: `{sha256}`")
                 lines.append(f"{pad}    {' | '.join(hash_parts)}")
+            if status == "missing" and f.get("category_evidence"):
+                lines.append(f"{pad}    {_category_evidence_md(f['category_evidence'])}")
 
         lines.append("")
 
@@ -2053,6 +2061,27 @@ def _platform_data_links(name: str) -> str:
     return ", ".join(links)
 
 
+def _category_evidence_md(evidence: dict) -> str:
+    """What establishes a missing entry's category, as the page shows it."""
+    if evidence.get("read_by"):
+        readers = []
+        for reader in evidence["read_by"]:
+            refs = [
+                f"`{e['name']}` (`{e['source_ref']}`)" if e.get("source_ref") else f"`{e['name']}`"
+                for e in reader.get("entries", [])
+            ]
+            name = (_emulator_link(reader["emulator"], "../")
+                    if reader["emulator"] != "launcher" else "the platform's launcher")
+            readers.append(f"{name}: {', '.join(refs)}" if refs else f"{name} (by {reader['proof']})")
+        return "Read by " + "; ".join(readers)
+    examined = evidence.get("cores_examined") or []
+    cores = ("cores examined: " + ", ".join(_emulator_link(c, "../") for c in examined)
+             if examined else "no core examined")
+    if evidence.get("unsettled"):
+        return f"No reader found, not settled: {evidence['unsettled']} ({cores})"
+    return f"No reader found ({cores})"
+
+
 def generate_platform_page(
     name: str,
     cov: dict,

scripts/verify.py

@@ -1356,10 +1356,12 @@ def verify_platform(
             prev = file_status.get(dest)
             hle = hle_index.get(file_entry.get("name", ""), False)
             sev = compute_severity(cur, required, mode, hle)
-            category = (
-                categories.category(file_entry, destination, sys_id)
-                if cur == Status.MISSING else BIOS
-            )
+            if cur == Status.MISSING:
+                explained = categories.explain(file_entry, destination, sys_id)
+                category = explained.pop("category")
+                result["category_evidence"] = explained
+            else:
+                category = BIOS
             result["category"] = category
             if file_category.get(dest, category) == BIOS:
                 category = BIOS
@@ -1608,6 +1610,33 @@ _NOT_FIRMWARE = {
 }
 
 
+def category_evidence_text(evidence: dict) -> str:
+    """One line saying what establishes a missing entry's category."""
+    if evidence.get("read_by"):
+        readers = []
+        for reader in evidence["read_by"]:
+            refs = [
+                f"{e['name']} ({e['source_ref']})" if e.get("source_ref") else e["name"]
+                for e in reader.get("entries", [])
+            ]
+            readers.append(
+                f"{reader['emulator']}: {', '.join(refs)}" if refs
+                else f"{reader['emulator']} (by {reader['proof']})"
+            )
+        return "read by " + "; ".join(readers)
+    examined = evidence.get("cores_examined") or []
+    cores = f"cores examined: {', '.join(examined)}" if examined else "no core examined"
+    if evidence.get("unsettled"):
+        return f"no reader found, not settled: {evidence['unsettled']} ({cores})"
+    return f"no reader found ({cores})"
+
+
+def _print_category_evidence(detail: dict) -> None:
+    evidence = detail.get("category_evidence")
+    if evidence:
+        print(f"      {category_evidence_text(evidence)}")
+
+
 def _print_detail_entries(details: list[dict], seen: set[str], verbose: bool) -> None:
     """Print UNTESTED, MISSING, and DISCREPANCY entries from verification details."""
     for d in details:
@@ -1636,11 +1665,14 @@ def _print_detail_entries(details: list[dict], seen: set[str], verbose: bool) ->
                 if verbose:
                     seen.add(key)
                     print(f"  MISSING ({label}): {key}")
+                    _print_category_evidence(d)
                 continue
             seen.add(key)
             req = "required" if d.get("required", True) else "optional"
             hle = ", HLE available" if d.get("hle_fallback") else ""
             print(f"  MISSING ({req}{hle}): {key}")
+            if verbose:
+                _print_category_evidence(d)
             _print_ground_truth(d.get("ground_truth", []), verbose)
     for d in details:
         disc = d.get("discrepancy")