On Thu, Aug 19, 2021 at 11:28 AM Duncan Grisby

Thanks. Victor Stinner has now commented on the defect I raised to say that it is actually a bug in the traceback module, in that it should not be using Py_DECREF at all inside a signal handler. The fact that it leads to use of the old trashcan macros is not the cause of the crash that I saw.

I therefore don't think it's necessary for me to check with the updated frameobject.c. Do you agree? I can try to reproduce the situation, but I can't just quickly test it because the infrastructure where the crash was seen is now being used for other things.

Well, in order to verify that we fixed the problem you reported we need to be able to test the fix and see that the problem is gone. Right now we have two theories of what the cause could be, because there are two known bugs in that area. But you may have stumbled upon an unrelated third bug.

The best scenario is that we close a bpo issue because we can reproduce the bug and see that it has been fixed. When that's not possible, there will come a point where we will close it and say "the code has changed since you observed the problem and we fixed things that could have caused it. Please open a new issue if you still see this bug".