Peter Donovan

Here is a catalog of things that did not work... Just writing them down so that I don't forget and waste time trying them again. 1. In CI, the error...

I said that our code didn't appear in the stack trace, but I was wrong. It's right there, when we call `PyType_Ready(&py_port_capsule_t)`. So I am optimistic that we can fix...

Thanks Christian for taking a look. I guess the debug print would have to be entered into a fork/local version of the CPython repo in order for it to work,...

Status update from core dump (the problem is that `%rax` is null; note also that the function `PyDict_New` does not take arguments, which makes it difficult to figure out what...

Here is the situation in the parent frame (`PyDict_Ready`): ``` (lldb) x/20i $pc-32 0x100a91abf: 6d insl %dx, %es:(%rdi) 0x100a91ac0: d0 0f rorb (%rdi) 0x100a91ac2: 84 a8 03 00 00 48...

Okay, I guess I just got confused between Intel and AT&T syntax. What Christian said clears that up now. Here is another confusion that I have: ``` 0x108f6f25c: 48 ff...

Following a conversation with @edwardalee, I tried to reproduce the segfault in an SSH session in the GitHub Actions runner and succeeded, but only when using the same debug version...

We just compile the C extension with `-fPIC`, I believe, and `empty_keys_struct` is in the `cpython` repository, with the documentation, "This immutable, empty PyDictKeysObject is used for PyDict_Clear() (which cannot...

If I try to do something completely different than before (acquire the GIL), I still get a segfault from accessing the address `NULL + 0x10`. I think the Python interpreter...

I did as Christian suggested and can confirm that CMake is discovers a different version of Python than what is discovered by `which python`. To be specific, invocations of `gcc`...