fix: compute correct Content-Length for StringIO with multi-byte chars

StringIO.tell() returns the character position, not the byte offset,
so super_len() returned the wrong value for StringIO objects containing
multi-byte UTF-8 characters (e.g. emoji).  This caused an incorrect
Content-Length header that violates RFC 9110 section 8.6.

Read the remaining text and encode it to UTF-8 to measure the true
byte length, consistent with how plain str bodies are already handled.

Closes #6917

Author: veeceey

src/requests/utils.py

@@ -186,13 +186,30 @@ def super_len(o):
             if hasattr(o, "seek") and total_length is None:
                 # StringIO and BytesIO have seek but no usable fileno
                 try:
-                    # seek to end of file
-                    o.seek(0, 2)
-                    total_length = o.tell()
-
-                    # seek back to current position to support
-                    # partially read file-like objects
-                    o.seek(current_position or 0)
+                    if isinstance(o, io.StringIO):
+                        # StringIO.tell() returns the character
+                        # position, not the byte offset.  Read the
+                        # remaining text and encode it to measure the
+                        # true byte length for Content-Length.
+                        start = current_position or 0
+                        o.seek(start)
+                        total_length = len(o.read().encode("utf-8"))
+
+                        # Reset current_position so the returned value
+                        # is just total_length (the remaining bytes).
+                        current_position = 0
+
+                        # seek back to original position to support
+                        # partially read file-like objects
+                        o.seek(start)
+                    else:
+                        # seek to end of file
+                        o.seek(0, 2)
+                        total_length = o.tell()
+
+                        # seek back to current position to support
+                        # partially read file-like objects
+                        o.seek(current_position or 0)
                 except OSError:
                     total_length = 0
 

tests/test_utils.py

@@ -142,6 +142,30 @@ def test_super_len_with_tell(self):
         foo.read(2)
         assert super_len(foo) == 3
 
+    def test_super_len_stringio_multibyte(self):
+        """Ensure StringIO with multi-byte characters returns the UTF-8
+        byte length rather than the character count.  See #6917."""
+        # Single emoji: 1 character, 4 bytes in UTF-8
+        foo = StringIO.StringIO("\U0001F4A9")
+        assert super_len(foo) == 4
+
+        # Mixed ASCII and multi-byte
+        foo = StringIO.StringIO("hello \U0001F4A9 world")
+        assert super_len(foo) == len("hello \U0001F4A9 world".encode("utf-8"))
+
+        # Partially read StringIO with multi-byte characters
+        foo = StringIO.StringIO("hello \U0001F4A9 world")
+        foo.read(6)  # read "hello "
+        remaining_bytes = len("\U0001F4A9 world".encode("utf-8"))
+        assert super_len(foo) == remaining_bytes
+
+        # Position should be preserved after super_len call
+        foo = StringIO.StringIO("hello \U0001F4A9 world")
+        foo.read(3)
+        pos_before = foo.tell()
+        super_len(foo)
+        assert foo.tell() == pos_before
+
     def test_super_len_with_fileno(self):
         with open(__file__, "rb") as f:
             length = super_len(f)