INADA Naoki <songofaca...@gmail.com> added the comment:
I think I can describe what two function does/for.
But I'm not good English writer. Would you write document instead?
---
"Finalizer" is __del__ in Python and tp_finalize in C.
Calling finalizer requires some tricks, including avoiding call twice.
That is what PyObject_CallFinalizer() does. Use it instead of call
`type->tp_finalize(obj)` directly.
But finalizer is called from destructor (tp_dealloc) usually, and calling
finalizer from destructor need more hack (e.g. temporary increment refcount).
This is what PyObject_CallFinalizerFromDealloc does.
Generally speaking, this API is used only when you want to use both of custom
tp_dealloc and tp_finalize. This is very rare because you can write cleanup
code in tp_dealloc and skip using tp_finalize.
When you need really need both of tp_dealloc and tp_finalize, note that:
* Finalizer may resurrect the object.
* When the type is subclassed from Python, tp_finalize is called automatically.
You must not call PyObject_CallFinalizerFromDealloc then.
This is tp_dealloc of Future object in asyncio. This object uses both of
tp_dealloc and tp_finalize for compatibility with Future object implemented in
pure Python. This covers many edge cases of tp_dealloc.
```
static void
FutureObj_dealloc(PyObject *self)
{
FutureObj *fut = (FutureObj *)self;
if (Future_CheckExact(fut)) {
/* When fut is subclass of Future, finalizer is called from
* subtype_dealloc.
*/
if (PyObject_CallFinalizerFromDealloc(self) < 0) {
// resurrected.
return;
}
}
// Weakref callback and finalizer of other objects may start GC.
// So you need to untrack this object from GC before calling
PyObject_ClearWeakRefs()
// or any Py_DECREF()s
// Otherwise, GC will cause segfault because refcount of this object is 0.
PyObject_GC_UnTrack(self);
// Handle weakrefs
if (fut->fut_weakreflist != NULL) {
PyObject_ClearWeakRefs(self);
}
// Reuse tp_clear() for clearing all members.
(void)FutureObj_clear(fut);
// Free this object at last.
Py_TYPE(fut)->tp_free(fut);
}
```
----------
nosy: +inada.naoki
_______________________________________
Python tracker <rep...@bugs.python.org>
<https://bugs.python.org/issue33909>
_______________________________________
_______________________________________________
Python-bugs-list mailing list
Unsubscribe:
https://mail.python.org/mailman/options/python-bugs-list/archive%40mail-archive.com
