09:17 <fijal> cfbolz: do you have any clue where do I put the following info - ll2ctypes (and ctypes) is broken on newer OS X for variadic-style functions, we skip all tests which require that. Notable example of functions - open, fcntl

09:27 <cfbolz> fijal: file a bug at least?

09:28 <fijal> against?

09:29 <arigato> I would say, put a global flag in ll2ctypes, and then do something like whenever a function call is issued via ll2ctypes, check the flag and the function name against a black list

09:31 <fijal> that might be a good idea, but it seems liek a lot of work (most calls to open() are fine, it's only if you pass extra args or something along those lines)

09:32 <fijal> but maybe we can put fcntl there

09:33 <arigato> Well, add more cases under the 'if flag:' case. It's not like keeping ll2ctypes clean is a goal at all

09:33 <fijal> haha, yes ok

09:39 <fijal> yay rlib/test passes

11:34 <fijal> cfbolz: ok, so I have the first cool crash - callbacks require writable/executable memory

11:34 <fijal> I would imagine it's impossible to get any tests going without callbacks

11:40 <cfbolz> fijal: yay/annoying ;-)

11:41 <fijal> yes

11:42 <arigato> So callbacks are also buggy on cpython ctypes on M1?

11:42 <fijal> yes, of course

11:42 <fijal> remember we are using old cpython

11:42 <fijal> I don't know, maybe one way would be to either patch cpython or compile a jit-less pypy for testing

11:43 <arigato> Or rewrite ll2ctypes to use cffi but well

11:43 <fijal> sounds horrendous

11:44 <cfbolz> Both horrendous and a fantastic cleanup

11:44 <fijal> ... but in this particular case it is a case of C compiled thing

11:44 <fijal> (so I'm not sure what it is)

11:44 <fijal> cfbolz: yes, but I'm not sure I want to put "fantastic cleanup" on a todo for a project that's already 4 months late

11:44 <cfbolz> Sure, sure

11:45 <cfbolz> 'somebody ought to do it'

11:45 <arigato> Or write a general wrapper that gives the ctypes interface with a vff

11:45 <arigato> A cffi implementation

11:45 <arigato> Cough

11:46 <arigato> Maybe indeed get a translated pypy2

11:59 <fijal> ... which leads us to another mess which is why some of our tests don't pass on pypy

11:59 <fijal> ok, but this machine is too broken to do anything more I'm afraid

11:59 <fijal> (I don

11:59 <fijal> 't know why)

12:08 <arigato> You would think that somebody somewhere has already patched cpython 2.7 for his own usage, but who am I to guess

12:09 <arigato> Yes, at some point we need to think again about the whole 2.7-is-going-away mess

12:39 <cfbolz> arigato: yes, new platforms really are annoying for the whole 'we're just using PyPy to bootstrap'

12:58 <fijal> I might just patch cpython in the meantime, I think

13:26 <mgorny> fijal: btw i'm "maintaining" largish cpython patchset for gentoo; the idea was to just backport security fixes but if it's needed to bootstrap pypy, i suppose i can also include more patches

13:26 <mgorny> https://gitweb.gentoo.org/fork/cpython.git/log/?h=gentoo-2.7-vanilla that's the branch

13:26 <mgorny> ("vanilla" because it doesn't include gentoo-specific patches)

13:27 <fijal> oh that sounds good

13:27 <fijal> mgorny: I think what we need here is updating libffi to something non-ancient (so de-vendoring it), does your branch do that?

13:28 <mgorny> fijal: we're building --with-system-ffi so i didn't need to touch that, i think

13:30 <mgorny> that said, it's time for me to take a break from looking at the screen; be back later

13:33 <fijal> oh is that a thing you can do?

13:34 <LarstiQ> the --with-system-ffi or taking breaks from screen? :)

15:01 <mgorny> fijal: well, apparently

15:02 <mgorny> admittedly, that part of python on gentoo predates my time

15:03 <Corbin> Hm. We still have Python 2.7 in NixOS, too. Maybe we should pool resources at some point. PyPI and setuptools and pip are continuing to deteriorate Python 2 support, so we may need to find/build a common repository for tarballs.

16:53 <trace987> can I just take the bytecode instead needing everytime the original source?

19:02 <mgorny> Corbin: to be honest, i'd love to see it gone entirely but the reality is reality

19:02 <mgorny> alternatively, i was thinking of stripping stdlib to remove the most "vulnerable" parts

19:02 <mgorny> i.e. keeping the bare minimum needed to bootstrap pypy

19:03 <mgorny> but these are fun projects for if i have loads of free time ;-)