The main outcome I got from my stream today was that @encukou and @vstinner (and presumably the rest of the Core Devs) want to regard CPython C API symbols prefixed with _Py as internal APIs which don't carry any version stability (not even necessarily in the same Python minor version patch series).
I think this means that we should consider removing all such definitions with this _Py prefix from pyo3-ffi, or at the very least making users think very hard before they choose to use these, if there's a good reason for any of these to be exposed in PyO3.
I took some notes on stream about where we use these definitions in PyO3 itself, they're on my desktop PC along with a small patch which removes use of a couple, I'll aim to push both tomorrow.
The main outcome I got from my stream today was that @encukou and @vstinner (and presumably the rest of the Core Devs) want to regard CPython C API symbols prefixed with
_Pyas internal APIs which don't carry any version stability (not even necessarily in the same Python minor version patch series).I think this means that we should consider removing all such definitions with this
_Pyprefix frompyo3-ffi, or at the very least making users think very hard before they choose to use these, if there's a good reason for any of these to be exposed in PyO3.I took some notes on stream about where we use these definitions in PyO3 itself, they're on my desktop PC along with a small patch which removes use of a couple, I'll aim to push both tomorrow.