Skip news as it's a experimental feature.

Does this mean we can correctly optimize across _PY_FRAME_GENERAL now instead of bailing?

I think we were already doing that. This just fixes a bug with overlapping frames (it removes it altogether).

Merging this as backport bugfixes don't normally need approval (or at least, I can approve it if it were done by the bot).

I think we were already doing that. This just fixes a bug with overlapping frames (it removes it altogether).

Maybe I'm misunderstanding. Is there anything preventing us from updating _PY_FRAME_GENERAL in the optimizer to push a new frame and continue optimizing, rather than bailing (like it does now)?

IIRC we used to not be able to determine the size of the nee frame to push which is why we bailed. If that's solved I dont see why we cant toptimize through it.

_Py_uop_frame_new still assumes positional arguments only, but it would be simple enough to drop that assumption by initializing the arguments in the caller.