00:16 <Hodgestar> mattip: Welcome back.

00:17 <Hodgestar> mattip: We missed you at the sprint. I think both calls went really well.

00:17 <Hodgestar> mattip: Maybe docutil needs pinning? Or at least that's the only package other than sphinx that I see in the traceback.

04:02 <mattip> I think documentation is updating now. Sorry for all the commits to get it right.

12:45 <mattip> on the hpy-0.0.4 branch, the last commit synced an upstream test that calling pt = mod.FakeArray() will create the c-struct

12:46 <mattip> so __new__ for FakeArray needs to go through interp_type._create_instance

12:46 <mattip> but I am not sure how to override the default descr__new__ for W_HPyTypeObject

12:48 <mattip> one way would be to pre-load an app-level __new__ into the class __dict__, which could be overloaded if a tp_new slot is filled

12:48 <mattip> but how can I wrap a rpython-level function and put it in the __dict__?

12:50 <mattip> interp2app() needs to be done in a TypeDef, and since W_HPyTypeObject inherits from W_TypeObject setting a TypeDef is tricky

12:54 <mattip> this problem is causing a segfault in test_slots.py since the result of mod.FakeArray() is not a W_HPyOjbect, so HPy_AsStruct fails and returns None (NULL), which crashes