> I don't see a benefit from this patch.

Oh, sorry, I forgot to explain the motivation. Performances of the warnings module are not critical module. The motivation here is to avoid to encoding string to UTF-8 for correctness. For example, _PyUnicode_AsString(filename) fails if the filename contains a surrogate character.

>>> warnings.warn_explicit("text", RuntimeError, "filename", 5)
filename:5: RuntimeError: text
>>> warnings.warn_explicit("text", RuntimeError, "filename\udc80", 5)
Traceback (most recent call last):
  File "<stdin>", line 1, in <module>
UnicodeEncodeError: 'utf-8' codec can't encode character '\udc80' in position 8: surrogates not allowed


Another example where a string to encoded to UTF-8 and decoded from UTF-8 a few instructions later:

PyObject *to_str = PyObject_Str(item);
err_str = _PyUnicode_AsString(to_str);
...
PyErr_Format(PyExc_RuntimeError,  "...%s", err_str);

Using "%R" avoids any encoding conversion.