fill mask in index with nan when not calling maybe_upcast_putmask

Author: arw2019

pandas/core/frame.py

@@ -4879,8 +4879,12 @@ def _maybe_casted_values(index, labels=None):
                     if issubclass(values_type, DatetimeLikeArray):
                         values = values._data  # TODO: can we de-kludge yet?
 
-                    if mask.any() and isinstance(values, np.ndarray):
-                        values, _ = maybe_upcast_putmask(values, mask, np.nan)
+                    if mask.any():
+                        if isinstance(values, np.ndarray):
+                            values, _ = maybe_upcast_putmask(values, mask, np.nan)
+                        else:
+                            # GH24206
+                            values[mask] = np.nan
 
                     if issubclass(values_type, DatetimeLikeArray):
                         values = values_type(values, dtype=values_dtype)