Support pickling of super object

[serhiy-storchaka]

Feature or enhancement

As was noted in #125714 (comment), the super object is not pickleable. For example:

import pickle
class X: pass
s = super(X, X())
pickle.dumps(s)

Produces a traceback:

Traceback (most recent call last):
  File "<python-input-0>", line 4, in <module>
    pickle.dumps(s)
    ~~~~~~~~~~~~^^^
_pickle.PicklingError: first argument to __newobj__() must be <class 'super'>, not <class '__main__.X'>
when serializing super object

This is because the special methods like __reduce_ex__() are looked up in an instance and translated to a lookup in the underlying object.

>>> super(X, X()).__reduce_ex__(5)
(<function __newobj__ at 0x7fd9e8a0ad50>, (<class '__main__.X'>,), None, None, None)

This cannot be solved by implementing the __reduce_ex__() method in the super class, because the current behavior is expected when super() is used in the __reduce_ex__() implementation of some subclass. The super class likely should be registered in the global dispatch table.

There may be similar issue with shallow and deep copying.

Linked PRs

• gh-125767: Fix pickling and copying of super objects #125781