feat: refuse a binary ref without a place in it

A source_ref part that names a compiled image now has to carry an
address, a decompiled function or an IL method. An import directory, a
string table or readelf output only says the binary exists. A name
under a directory is a file of the tree and a .dll.map is not a DLL.
The exception list is empty: every closed profile now cites code.

scripts/validate_schemas.py

@@ -161,34 +161,46 @@ _CODE_SUFFIXES = frozenset(
     "go rs lua pas pp asm s sh bash zsh ps1 bat cmd vbs swift d zig nim qml qc rb pl php "
     "scala ml hs dart vue ino z80 smali".split()
 )
-_CODE_FILE = re.compile(r"[\w+.-]\.([A-Za-z0-9+]+)\b")
-_BINARY_EVIDENCE = re.compile(
-    r"\b0x[0-9a-fA-F]+\b|\bFUN_[0-9a-fA-F]+\b|\.(?:so|dll|exe|dylib|dex|apk|jar|elf|nro|xex)\b",
+_CODE_FILE = re.compile(r"(?<=[\w+-])\.([A-Za-z0-9+]+)\b")
+# A compiled image: a part that names one has to say where in it the file
+# is read, an address, a decompiled function or an IL method.
+# A name under a directory is a file of the tree it sits in, and a further
+# extension (.dll.map) makes it another file.
+_BINARY_FILE = re.compile(
+    r"(?<![\w/+.-])[\w+.-]+\.(?:so|dll|exe|dylib|dex|elf|nro|xex)(?:\.\d+)*(?![\w.-]*\w)",
     re.IGNORECASE,
 )
+_BINARY_EVIDENCE = re.compile(
+    r"\b0x[0-9a-fA-F]+\b"
+    r"|\b(?:FUN|sub|thunk|LAB)_[0-9a-fA-F]+\b"
+    r"|\bIL_[0-9a-fA-F]{4}\b"
+    r"|::|\.\.c?ctor\b"
+    r"|\b[A-Za-z_][\w.$]*\("
+)
 
 # Profiles whose files cite no published code. Each reason says why the
 # check cannot apply yet.
-DATA_ONLY_REF_EXCEPTIONS: dict[str, str] = {
-    "ace-dl": "closed source: the refs describe what the shipped binary opens",
-    "cpcemu": "closed source: the refs describe what the shipped binary opens",
-    "drastic": "closed source: the shader format files are read by the shipped binary",
-    "lexaloffle": "closed source: PICO-8 reads pico8_gpio, the refs are its readme and the KNULLI install script",
-    "mfme": "closed source: the refs are the release notes of the shipped binary",
-}
+DATA_ONLY_REF_EXCEPTIONS: dict[str, str] = {}
 
 
 def _ref_cites_code(ref: str) -> bool:
-    """A source_ref part names a source file, or a binary with an address."""
+    """A source_ref cites a source file or a place in a compiled image.
+
+    A part naming a binary without an address, a decompiled function or an
+    IL method says only that the binary exists: "import directory",
+    "string table" and "readelf -d" are not the code that reads a file.
+    """
     from sourceref import split_source_ref
 
-    if _BINARY_EVIDENCE.search(ref):
-        return True
+    cited = False
     for part in split_source_ref(ref):
-        for suffix in _CODE_FILE.findall(part.raw):
-            if suffix.lower() in _CODE_SUFFIXES:
-                return True
-    return False
+        located = bool(_BINARY_EVIDENCE.search(part.raw)) or any(
+            suffix.lower() in _CODE_SUFFIXES for suffix in _CODE_FILE.findall(part.raw)
+        )
+        if _BINARY_FILE.search(part.raw) and not located:
+            return False
+        cited = cited or located
+    return cited
 
 
 def _data_only_source_refs(path: Path, document: object) -> list[str]:

tests/test_data_only_refs.py

@@ -53,17 +53,39 @@ class DataOnlySourceRefs(unittest.TestCase):
 
     def test_a_binary_with_an_address_passes(self):
         for ref in ("libemucore.so 0x1a2b3c", "Fusion.exe 0x00401000",
-                    "bigpemu FUN_006d8be0 (BootROM read)", "nethersx2 v2.2n classes.dex icons/star-%d.png"):
+                    "bigpemu FUN_006d8be0 (BootROM read)",
+                    "CSpect.exe 3.3.1.0 a.H::e() IL_00bd-IL_00c9",
+                    "GameAssembly.dll DSP1..ctor (name and CRC32)",
+                    "libpizzaboy-jni.so pizza.c:931-937 (path buffer)"):
             with self.subTest(ref=ref):
                 self.assertEqual(self.errors(ref), [])
 
+    def test_a_binary_named_without_a_place_in_it_is_refused(self):
+        for ref in ("ArcadeFlashWeb v1.0.2 pepflashplayer.dll string table",
+                    "Future Pinball.exe 1.9.1.20101231 import directory",
+                    "Main.java:27302, libfpse.so readelf -d",
+                    "nethersx2 v2.2n classes.dex icons/star-%d.png",
+                    "Fusion.exe 0x00401000, PS3Eye.dll .rdata"):
+            with self.subTest(ref=ref):
+                self.assertEqual(len(self.errors(ref)), 1)
+
+    def test_a_binary_file_of_a_source_tree_is_a_path(self):
+        self.assertEqual(self.errors(
+            "src/loader/elfLoader/elfLoader.cpp:365, libs/win32/ll-deps/zlib1.dll"), [])
+        self.assertEqual(self.errors(
+            "src/emu/kernel/src/libmanager.cpp:549-566, src/patch/ecam/group/ecam.dll.map:190-193"), [])
+
     def test_each_build_mode_is_judged(self):
         self.assertEqual(len(self.errors({"standalone": "src/main.cpp:10", "libretro": "data/list.txt:3"})), 1)
 
     def test_an_excepted_profile_is_not_judged(self):
-        name = next(iter(validate_schemas.DATA_ONLY_REF_EXCEPTIONS))
-        path = REPO_ROOT / "emulators" / f"{name}.yml"
-        self.assertEqual(validate_schemas._data_only_source_refs(path, _profile("history.txt:38")), [])
+        path = REPO_ROOT / "emulators" / "synthetic.yml"
+        exceptions = validate_schemas.DATA_ONLY_REF_EXCEPTIONS
+        exceptions["synthetic"] = "binary found nowhere"
+        try:
+            self.assertEqual(validate_schemas._data_only_source_refs(path, _profile("history.txt:38")), [])
+        finally:
+            del exceptions["synthetic"]
 
     def test_every_exception_names_a_profile_and_a_reason(self):
         for name, reason in validate_schemas.DATA_ONLY_REF_EXCEPTIONS.items():