René's URL Explorer Experiment

{"@context":"https://schema.org","@type":"DiscussionForumPosting","headline":"_PyObject_CAST breaks code which assumed a static/dynamic casting","articleBody":"pywin32 has, since the 90s, used c++ to \"inherit\" from a PyObject in some cases. In short, it ends up something like:\r\n\r\n```\r\nclass MyObject : public PyObject {\r\n  virtual void some_helper();\r\n}\r\n```\r\n\r\nWith a layout like this, a `reinterpret_cast` between a `MyObject *` and a `PyObject *` will create invalid pointers.\r\n\r\nConsider something like the following:\r\n\r\n```\r\nvoid MyObject::some_helper() {\r\n  Py_DECREF(this);\r\n}\r\n```\r\nThis ends up being a macro which uses `_PyObject_CAST` on the param passed to `Py_DECREF`. [In 3.10, this uses a traditional \"c\" cast](https://github.com/python/cpython/blob/30015de7235e5b4033298b85a164fcd8d96046b3/Include/object.h#L112) , but [in 3.11 it turned into a `reinterpret_cast`](https://github.com/python/cpython/blob/f3212b1ec7a70b69f44acd763221463a69beee4f/Include/pyport.h#L67). In the above example, this causes `Py_DECREF` to be called with an invalid `PyObject *` - which typically doesn't crash at the time, but instead corrupts memory causing a difficult to diagnose crash some time later.\r\n\r\nChanging the code to use explicit `static_cast\u003c\u003e`, or the old-school `(PyObject *)` cast makes things work, but it would be error prone as it would be easy to miss places where it's used via non-obvious nested macros. It also shouldn't be necessary to make this kind of change - c++ code which has worked for this long should continue to work.","author":{"url":"https://github.com/mhammond","@type":"Person","name":"mhammond"},"datePublished":"2022-07-11T05:29:45.000Z","interactionStatistic":{"@type":"InteractionCounter","interactionType":"https://schema.org/CommentAction","userInteractionCount":19},"url":"https://github.com/94731/cpython/issues/94731"}