Add a decode test with invalid utf8

Author: hukkin

src/mdurl/_decode.py

@@ -82,7 +82,7 @@ def repl_func_with_cache(match: re.Match, cache: Sequence[str]) -> str:
                 continue
 
         if (b1 & 0xF8) == 0xF0 and (i + 9 < l):
-            # 111110xx 10xxxxxx 10xxxxxx 10xxxxxx
+            # 11110xxx 10xxxxxx 10xxxxxx 10xxxxxx
             b2 = int(seq[i + 4 : i + 6], 16)
             b3 = int(seq[i + 7 : i + 9], 16)
             b4 = int(seq[i + 10 : i + 12], 16)

tests/test_decode.py

@@ -3,3 +3,14 @@
 
 def test_decode_multi_byte():
     assert decode("https://host.invalid/%F0%9F%91%A9") == "https://host.invalid/👩"
+
+
+def test_decode_invalid_utf8():
+    assert decode("https://host.invalid/%CF") == "https://host.invalid/\ufffd"
+    assert decode("https://host.invalid/%C0%bf") == "https://host.invalid/\ufffd\ufffd"
+    # This is different from `urllib.parse.unquote`. We add 3 * \ufffd as does
+    # Javascript upstream, urllib only adds 2 * \ufffd.
+    assert (
+        decode("https://host.invalid/%F1%81%d1%45")
+        == "https://host.invalid/\ufffd\ufffd\ufffdE"
+    )