Fix text-file snapshot positions for Unicode and translated newlines

cloudpickle/cloudpickle.py

@@ -992,17 +992,25 @@ def _file_reduce(obj):
     retval = io.StringIO()
 
     try:
-        # Read the whole file
         curloc = obj.tell()
-        obj.seek(0)
-        contents = obj.read()
-        obj.seek(curloc)
+        try:
+            # TextIOWrapper positions are opaque cookies, not StringIO character
+            # offsets. Count the unread characters to locate the snapshot cursor.
+            remaining_chars = len(obj.read())
+            tail_end = obj.buffer.tell()
+            obj.seek(0)
+            contents = obj.read()
+            file_end = obj.buffer.tell()
+        finally:
+            obj.seek(curloc)
     except OSError as e:
         raise pickle.PicklingError(
             "Cannot pickle file %s as it cannot be read" % name
         ) from e
     retval.write(contents)
-    retval.seek(curloc)
+    # Preserve the existing behavior for files explicitly positioned past EOF.
+    # The public byte-stream positions distinguish this from a large text cookie.
+    retval.seek(curloc if tail_end > file_end else len(contents) - remaining_chars)
 
     retval.name = name
     return _file_reconstructor, (retval,)

tests/test_file_text_positions.py

@@ -0,0 +1,88 @@
+import io
+import pickle
+
+import pytest
+
+import cloudpickle
+
+
+@pytest.mark.parametrize(
+    "encoding, text, count, newline",
+    [
+        ("utf-8", "\u00e9abc", 1, None),
+        ("utf-8", "a\u2603bc", 2, None),
+        ("utf-16", "abcd", 1, None),
+        ("utf-32", "abcd", 1, None),
+        ("utf-8-sig", "abcd", 1, None),
+        ("utf-8", "first\r\nlast", 6, None),
+        ("utf-8", "first\rlast", 6, None),
+        ("utf-8", "first\r\nlast", 7, ""),
+        ("utf-8", "first\r\nlast", 6, ""),
+        ("utf-8", "\u00e9\r\nlast", 3, ""),
+        ("utf-16", "first\r\nlast", 7, ""),
+        ("latin-1", "\u00e9abc", 1, None),
+        ("ascii", "abcd", 1, None),
+        ("utf-16", "abcd", 0, None),
+        ("utf-8", "", 0, None),
+        ("utf-8", "\u00e9abc", 4, None),
+        ("utf-16", "abcd", 4, None),
+        ("utf-8", "first\r", 6, None),
+    ],
+)
+def test_text_file_roundtrip_preserves_character_position(
+    tmp_path, encoding, text, count, newline
+):
+    path = tmp_path / "text.txt"
+    path.write_bytes(text.encode(encoding))
+    with path.open(encoding=encoding, newline=newline) as source:
+        prefix = source.read(count)
+        cookie = source.tell()
+        restored = pickle.loads(cloudpickle.dumps(source))
+        assert source.tell() == cookie
+        assert restored.tell() == len(prefix)
+        assert restored.read() == source.read()
+        source.seek(0)
+        restored.seek(0)
+        assert restored.read() == source.read()
+
+
+@pytest.mark.parametrize("encoding", ["ascii", "utf-8", "utf-16", "utf-32"])
+def test_text_file_roundtrip_preserves_existing_position_beyond_eof(tmp_path, encoding):
+    path = tmp_path / "text.txt"
+    path.write_bytes("abcd".encode(encoding))
+    with path.open(encoding=encoding) as source:
+        source.seek(100)
+        restored = pickle.loads(cloudpickle.dumps(source))
+        assert source.tell() == 100
+        assert restored.tell() == 100
+        assert restored.read() == source.read() == ""
+
+
+def test_stringio_roundtrip_preserves_character_position():
+    source = io.StringIO("\u00e9abc")
+    source.read(1)
+    restored = pickle.loads(cloudpickle.dumps(source))
+    assert source.tell() == restored.tell() == 1
+    assert restored.read() == source.read() == "abc"
+
+
+def test_text_file_position_is_restored_on_decode_error(tmp_path):
+    path = tmp_path / "invalid.txt"
+    path.write_bytes(b"prefix\xff")
+    with path.open(encoding="utf-8") as source:
+        source.seek(2)
+        cookie = source.tell()
+        with pytest.raises(UnicodeDecodeError):
+            cloudpickle.dumps(source)
+        assert source.tell() == cookie
+
+
+def test_text_file_position_is_restored_when_full_read_fails(tmp_path):
+    path = tmp_path / "invalid-prefix.txt"
+    path.write_bytes(b"\xffprefixTAIL")
+    with path.open(encoding="utf-8") as source:
+        source.seek(7)
+        cookie = source.tell()
+        with pytest.raises(UnicodeDecodeError):
+            cloudpickle.dumps(source)
+        assert source.tell() == cookie