gh-94401: Remove unneeded overlapped module state #94403
Conversation
erlend-aasland
commented
Jun 29, 2022
•
- Issue: gh-94401
| @@ -2045,14 +2011,15 @@ overlapped_exec(PyObject *module) | |||
| return -1; | |||
| } | |||
|
|
|||
| OverlappedState *st = overlapped_get_state(module); | |||
| st->overlapped_type = (PyTypeObject *)PyType_FromModuleAndSpec( | |||
| PyTypeObject *overlapped_type = (PyTypeObject *)PyType_FromModuleAndSpec( | |||
There was a problem hiding this comment.
Is PyType_FromModuleAndSpec the correct API to retrieve a type object from a module by spec? It calls PyType_FromMetaclass, which does a lot of extra work.
There was a problem hiding this comment.
All the PyType_From* APIs call PyType_FromMetaclass. The recent changes to the PyType_From* APIs should not result in a lot of extra work. If that is the case, PyType_FromMetaclass should be optimised, if possible.
There was a problem hiding this comment.
I was under the assumption that the PyType_From* APIs are used to create a new type object from a spec, not fetch an existing type. It may even leak memory . PyType_FromMetaclass calls PyMem_Malloc and eventually assigns it to type->tp_name eventually without a check.
There was a problem hiding this comment.
Each call allocates a whole new type on the heap memory.
There was a problem hiding this comment.
IMO, if there are issues with PEP-489 and the PyType_From* APIs they should be addressed in a separate issue and discussed there. Suggesting to mark this conversation as resolved and continue this discussion in an appropriate issue if needed.
|
Thanks for the review! |
