dalley has quit [Ping timeout: 272 seconds]
dalley has joined #hpy
<cfbolz> antocuni, Hodgestar: how was the BoF?
<mattip> Any thoughts about this failure on macosx arm64 in rpython.flowspace.test.test_objspace::test_all_opcodes_defined
<mattip> LOAD_REVDB_VAR seems to be missing
<mattip> why is that showing up in rpython, it is defined in pypy/interpreter
<mattip> whoops, wrong channl, sorry
ddavis has joined #hpy
ddavis has quit [Remote host closed the connection]
<antocuni> cfbolz: it went very well! There were ~15 people: some of them were only curious but some seems very interested
<cfbolz> awesome
<antocuni> generally speaking, everyone agrees that the C API is a problem
<cfbolz> heh
<antocuni> and many people seems very interested in universal binaries
<antocuni> more than performance on pypy
<cfbolz> yeah, to be expected :-/
<antocuni> we also showed them some of the nice tweaks that we did to the API, like having HPy_Add and HPy_GetAttr instead of PyNumber_*/PyObject_*
<antocuni> or the fact that when you use HPyDef_METH, you get a static typecheck in the signature of the C function
<antocuni> and it was nice to see many of them nodding enthusiastically
<antocuni> some suggested to experiment with nanobind, which should be much easier than pybind11
<antocuni> and we also had great conversations with Sebastian Berg w.r.t. numpy
<cfbolz> antocuni: cool, what was the outcome of that
<cfbolz> ?
<antocuni> he's generally positive about migrating numpy to hpy
<cfbolz> awesome
<antocuni> and it seems that none of them should be a big concern
<antocuni> the most annoying thing is "PyArrayObject* -> HPy removes type information and type checking", but he seems happy to live with passing both HPy and PyArrayObject* around
<cfbolz> right
<antocuni> he's also moving to Germany, so maybe we might convince him to come to the Duesseldorf sprint
<antocuni> but I don't remember whether I asked him or not 😅
<cfbolz> cool!