07:25 <fangerer> antocuni: btw. I've addressed all your comments for https://github.com/hpyproject/hpy/pull/331 . Please have another look. Thanks!

12:44 <fangerer> antocuni: You wrote that `tp_vectorcall`can be used with heap types. You are maybe right but CPython doesn't allow it in the type spec: https://docs.python.org/3/c-api/type.html?highlight=pytype_spec#c.PyType_Spec

20:20 <antocuni> fangerer: uhm, you are right. But at the same time, this piece of the docs seems to hint that vectorcalls can somehow work with heap types: https://docs.python.org/3/c-api/typeobj.html#c.PyTypeObject.tp_vectorcall_offset

20:22 <antocuni> we should investigate more; it might be possible to support it in HPy, e.g. by providing HPySlot_vectorcall and implement it by manually patch the TypeObject after calling PyType_FromSpec

20:29 <Hodgestar> fangerer, antocuni: Should we ask vstinner if we should open an issue on CPython to allow some of those unsupported slots in CPython?

20:33 <ronan> tp_vectorcall_offset is supported by defining a __vectorcalloffset__ member: https://docs.python.org/3/c-api/structures.html#c.PyMemberDef

20:34 <antocuni> fangerer: PR 331 merged, thank you!

20:35 <antocuni> Hodgestar: yes, asking the CPython devs sounds like a good idea. I'd also know whether the "no vectorcall on heap types" rule is a fundamental limitation or just a missing feature