<vstinner> Hodgestar: people don't care about the C API. they use it because it's there and it works. nobody thinks about the consequences of using the C API
<vstinner> Hodgestar: a few people handle incompatible C API changes at each Python version, that's it. the majority is not aware of these incompatible changes
<vstinner> Hodgestar: there is only a short warning somewhere in the C API doc saying "maybe it's a bad idea to use this API. otherwise, here is a giant doc explaining how to use it" :-D
<mattip> the latest untranslated tests on pypy are getting a link error for the pypy.module._hpy_universal.test._vendored.test_00_basic tests
<mattip> LINK : error LNK2001: unresolved external symbol initmytest
<mattip> MSVC is more picky than gcc when it comes to linking: it will try to resolve all functions whether they are called or not
<mattip> also translation fails on win64 with a "arithmetic not suppported on <UINT>" error
<mattip> okay, the translation is fixed
<mattip> I am confused aobut the initmytest function for a module called mytest
<mattip> shouldn't it be HPyInit_mytest ?
<vstinner> Hodgestar: hi. i'm not sure that i give you the answer that you was expecting :-) I'm not sure what do you expect from the SC
<vstinner> Hodgestar: usually, the SC approves PEPs
<vstinner> Hodgestar: for core devs, what do you expect? deprecate the C API? remove the C API? mention HPy in the doc? you're free to propose a PR to mention HPy in the C API doc, as Cython is mentioned there
<vstinner> Hodgestar: IMO it's a bad idea to consume the C API directly, it should be used at least with an abstraction like Cython
<mattip> I can't find where initmytest is even getting used. There is no such function in the source files, does it come form some c macro manipulation?
ammar2 has quit [Ping timeout: 265 seconds]
<mattip> ahh, it is the "export_symbols" argument to distutils.command.build_ext,
<mattip> which is being run with python2.7 untranslated
<mattip> so it picks up the python2-style export_symbol
the_drow has quit [Quit: Bridge terminating on SIGTERM]
jevinskie[m] has quit [Quit: Bridge terminating on SIGTERM]
jboi has quit [Quit: Bridge terminating on SIGTERM]
jboi has joined #hpy
ammar2 has joined #hpy
the_drow has joined #hpy
jevinskie[m] has joined #hpy
<Hodgestar> mattip: Is this a bug in how PyPy is building HPy, or is something wrong with the HPy distutils wrangling? If the latter, I can try fix if I understand the issue a bit more.
<Hodgestar> mattip: Some background (which you may already have discovered) -- "mytest" is the default name of test modules. I'm not sure where the "initmytest" comes from -- I can't see how the HPy code would ever put that in export symbols regardless of how it was being run. Maybe base distutils on Windows puts it there?
<Hodgestar> vstinner: For me a win would be for the SC to say "we agree that these are real problems and we'd like to see them fixed" and maybe even "HPy looks like the right way to do this" and "we can make some small improvements to the C API in the meantime". It's largely a concensus building excercise though. If we can point to a clear concensus, it has the potential to change many other discussions for the better. And maybe it is also a marketing exercise --
<Hodgestar> we could definitely do with more contributors.
<Hodgestar> vstinner: Anyway, I will try find some time to help with the PEP. My main comment is that it is currently quite rambling. I think it needs to be a lot crisper -- otherwise readers will get bored and stop reading. I can try suggest some changes to the structure.
<vstinner> Hodgestar: suggestions are welcomed :)
<Hodgestar> Thanks!
<Hodgestar> I have some other stuff to take care of right now and PyConZA to run next Thursday and Friday, so I'm not 100% sure how much time I will have this coming week, but let's see. And feel free to bug me if updates become more urgent.
<Hodgestar> vstinner: Are there any other core developers we could contact for feedback on the PEP once it's a bit further along? It's very helpful to get negative feedback early so we can improve it rather than waiting to get feedback from everyone at once on the mailing list. I.e. we need some alpha testers for the PEP text. :)
<vstinner> Hodgestar: i didn't know about PyConZA. do you give a talk?
<Hodgestar> vstinner: I am trying not to give a talk because I will have to much behind the scenes wrangling to do. :)
<vstinner> Hodgestar: hehe, ok
<vstinner> Hodgestar: as i told you, basically nobody cares about the C API, so it's hard to get feedback :)
<vstinner> Hodgestar: that's why i spammed 3 mailing lists :)
<vstinner> Hodgestar: recently, i announced that Py_TYPE(obj)=type becomes a compiler error in Python 3.11. in this email thread, i got feedback :-) https://mail.python.org/archives/list/capi-sig@python.org/thread/WGRLTHTHC32DQTACPPX36TPR2GLJAFRB/
<vstinner> Hodgestar: people only cares when incompatible changes are introduced
<vstinner> Hodgestar: a python-dev is vaguely related to the C API, it's more about the new CPython release cycle (1 major release per year): https://mail.python.org/archives/list/python-dev@python.org/thread/7ALUHJBJCZGFGXHK6TTDDR3ST4YVSAWR/
<mattip> Hodgestar: fixed locally in 893cc5b8b0c7 (in the pypy repo)
<mattip> once the buildbot confirms it is fixed, I will submit a PR to upstream the change to hpy
_0az3 has quit [Remote host closed the connection]
_0az3 has joined #hpy
<Hodgestar> mattip: Thanks!
_0az3 has quit [Quit: afk]
_0az3 has joined #hpy