Message368852
I consider that all things that could be done have already been done, so I close the issue. Thanks for Hai and Dong-hee who helped ;-)
> * _functools: Py_CLEAR(kwd_mark); is commented in _functools_free()
Sadly, PEP 573 implementation is not complete enough to use it in _functools: see PR 20055.
> * _abc: abc_invalidation_counter
Fixed in bpo-40566 by:
commit 77c614624b6bf2145bef69830d0f499d8b55ec0c
Author: Dong-hee Na <donghee.na92@gmail.com>
Date: Sat May 9 17:31:40 2020 +0900
bpo-40566: Apply PEP 573 to abc module (GH-20005)
> scanner_traverse(), scanner_clear(), encoder_traverse() and encoder_clear()
tp_clear slot cannot get the defining type (PEP 573 may be extended later to allow that).
Using PyType_GetModule(Py_TYPE(self)) to access types stored in the module state and then compare Py_TYPE(self) to these types... looks badly broken :-) If we get the wrong type, PyType_GetModule(Py_TYPE(self)) will return the wrong module and so we cannot get the json state from the wrong module... It's a chicken-and-egg issue :-)
I think that it's not worth it to attempt to add back these assertions. It's very unlikely to get the wrong types in practice. |
|
Date |
User |
Action |
Args |
2020-05-14 17:03:45 | vstinner | set | recipients:
+ vstinner, phsilva, corona10, shihai1991 |
2020-05-14 17:03:45 | vstinner | set | messageid: <1589475825.45.0.304369150178.issue40137@roundup.psfhosted.org> |
2020-05-14 17:03:45 | vstinner | link | issue40137 messages |
2020-05-14 17:03:45 | vstinner | create | |
|