fangerer has quit [Ping timeout: 264 seconds]
fangerer has joined #hpy
Guest58 has joined #hpy
Guest58 has quit [Client Quit]
dalley has quit [Quit: Leaving]
<antocuni> ronan: do you have a failing test?
<ronan> I'm trying to write one now
<ronan> antocuni: note that I was confused yesterday, it happens in all modes
* antocuni looks
<ronan> the reason is obvious: we unconditionally clobber tp_dealloc with hpytype_dealloc
<antocuni> yes, I imagined something like that but I didn't remember the code by heart
<antocuni> I wonder whether the proper fix is to disallow the legacy tp_dealloc if we have HPyFields
* antocuni looks at top4000 packages
<ronan> that would make a gradual port of something like numpy harder
<antocuni> what do they do in tp_dealloc that can't be done in e.g. hpy_tp_destroy?
<antocuni> ah, I answer myself. E.g. you need it if you want to mix PyObject* and HPyField
<ronan> yes
<antocuni> it's not obvious how to fix it though, especially on CPython. We need to install our own tp_dealloc which also calls the legacy one, but how to we know which function to call?
<antocuni> ah no, we have HPyType_Extra nowadays
<antocuni> we can put it there
<fangerer> I wrote a short blog post about the 0.0.3 release. Please review: https://github.com/hpyproject/hpyproject.org/pull/22
cfbolz has quit [*.net *.split]
qunaibit has quit [*.net *.split]
qunaibit has joined #hpy
cfbolz has joined #hpy