BUG: Title sometimes is bytes and not str.

pypdf/generic/_utils.py

@@ -184,6 +184,14 @@ def create_string_object(
                 # version if possible... and the only way to check if that's
                 # possible is to try.
                 # Some strings are strings, some are just byte arrays.
+                try:
+                    text = string.decode("utf-8")
+                    retval = TextStringObject(text)
+                    retval._original_bytes = string
+                    return retval
+                except UnicodeDecodeError:
+                    pass
+
                 retval = TextStringObject(decode_pdfdocencoding(string))
                 retval._original_bytes = string
                 retval.autodetect_pdfdocencoding = True

tests/test_reader.py

@@ -108,9 +108,18 @@ def test_read_metadata(pdf_path, expected):
         docinfo.modification_date
         docinfo.modification_date_raw
         if "/Title" in metadict:
+            assert isinstance(docinfo.title, str)
             assert metadict["/Title"] == docinfo.title
 
 
+def test_read_metadata_title_is_bytes():
+    with open(RESOURCE_ROOT / "bytes.pdf", "rb") as inputfile:
+        reader = PdfReader(inputfile)
+        title = reader.metadata.title
+        # Should be a str.
+        assert title == "Microsoft Word - トランスバース社買収電話会議英語Final.docx"
+
+
 def test_iss1943():
     with PdfReader(RESOURCE_ROOT / "crazyones.pdf") as reader:
         docinfo = reader.metadata