cfbolz changed the topic of #pypy to: PyPy, the flexible snake (IRC logs: https://quodlibet.duckdns.org/irc/pypy/latest.log.html#irc-end ) | use cffi for calling C | if a pep adds a mere 25-30 [C-API] functions or so, it's a drop in the ocean (cough) - Armin
<tos9>
I suspect the american translation is "shit's hard"
<cfbolz>
haha
<cfbolz>
:-)
dddddd has joined #pypy
<nedbat>
cfbolz: the link is good, thanks :)
<cfbolz>
Dejan, nedbat: a more serious answer to the question is: various people make slow progress on various things. hpy, gnarly windows problems, 3.7
<Dejan>
:D
<cfbolz>
arigato: around for a quick question? I am pondering what to do about sre in 3.7. it uses a different set of bytecodes and I don't quite know what to do about the fact that sre is in rlib
<cfbolz>
Dejan: you doing ok? you're in london, right? stay safe!
<Dejan>
Yea
<Dejan>
working from home these days
<Dejan>
obligatory
<cfbolz>
+1
<Dejan>
if someone told me that I will ever be forced to work from home I would say it is a joke :)
<Dejan>
luckily we can... some guys in different industries cannot
<mjacob>
i transported some equipment from the office to home. a policeman stopped me and asked what i wanted to do with the "TV set"...
alexge50 has joined #pypy
<tos9>
mjacob: he thought you were looting, or what?
<mjacob>
tos9: either that, or he just wanted to show off
YannickJadoul has joined #pypy
<YannickJadoul>
There's still a problem with 3.6 on macOS, it seems? I have access to a mac machine, if useful, but I don't really know where to start debugging
<cfbolz>
YannickJadoul: hey! unrelatedly, I worked on your branch ;-)
<cfbolz>
YannickJadoul: I decided to fix the error message problem by postponing it: on 3.7 positional only args aren't really a full concept yet, so I decided to mirror the error msgs that CPython 3.7 produces
<cfbolz>
and when we get to 3.8, we can change them again
<YannickJadoul>
cfbolz: Hi! Oh, great, that does sound like a nice solution! :-)
<YannickJadoul>
Just out of curiosity: did I start that branch into the right direction, or not?
<YannickJadoul>
Because that really seemed like a magical part of PyPy
<mattip>
YannickJadoul: distutils/sysconfig/setuptools is indeed more mundane. What is the problem?
BPL has quit [Read error: Connection reset by peer]
zmt00 has quit [Read error: Connection reset by peer]
BPL has joined #pypy
zmt00 has joined #pypy
<cfbolz>
YannickJadoul: the branch was basically fine. the main thing I did is that I wrote tests and fix the surfacing problems :-P
<cfbolz>
arguments.py is a pretty hairy part of the machinery, and quite well tested (interpreter/test/test_argument.py)
<YannickJadoul>
mattip: I was trying to look into the build failure on macOS that's also happening on buildbot
<YannickJadoul>
Let me rerun translation and see. It's got something to do with the SSL part, after translation
<YannickJadoul>
cfbolz: Ah, it was really cool to see that part, and mess with it, though :D
<YannickJadoul>
It seems trunk is being packaged correctly, but py3.6 isn't?
<ronan>
pfff... 2daf9ca169e causes a regression on #3096, so I guess I should revert it. But then extra_tests/test_recursion.py is going to fail...
<mattip>
there are tons of test failures in cffi
<mattip>
YannickJadoul: that is what is annoying to see
jvesely has joined #pypy
<YannickJadoul>
mattip: Yes. Anything I could try out, there?
<mattip>
sure. Try one of the tests: pypy3 -mensurepip; pypy3 -mpip install extra_tests/requirements.txt; pypy3 extra_tests/cffi_tests/cffi1/test_recompiler.py
<YannickJadoul>
With just 7.3.0?
<YannickJadoul>
> `$ PYTHONPATH=. ~/pypy3.6-v7.3.0-osx64/bin/pypy3 extra_tests/cffi_tests/cffi1/test_recompiler.py` gives no output. So I guess that means tests pass?
<YannickJadoul>
OK, nvm, I need to run with "-m pytest", but that also succeeds
<YannickJadoul>
mattip: Also, I got a the translation fully working, with export PKG_CONFIG_PATH="/usr/local/opt/libffi/lib/pkgconfig:/usr/local/opt/openssl/lib/pkgconfig" and export CPPFLAGS=-I/usr/X11/include/, so that was just a configuration issue, it seems
<YannickJadoul>
Weirdly enough, the first time after translation, _curses_build.py fails, but rerunning build_cffi_imports.py makes it work somehow
Smigwell has quit [Ping timeout: 240 seconds]
<mattip>
yeah, I think it needs a working _ssl for some reason
<mattip>
hmm. So I wonder why the buildbot fails those tests?
<mattip>
all I can see in the logs is "VerificationError: LinkError: command 'c++' failed with exit status 1"
<mattip>
without the actual error
<mattip>
ahh, no, there is this:
<mattip>
clang: warning: libstdc++ is deprecated; move to libc++ with a minimum deployment target of OS X 10.9 [-Wdeprecated]
<mattip>
ld: library not found for -lstdc++
<mattip>
YannickJadoul: I think the buildbot is using catalina
<mattip>
what does the buildbot need to install for -lstdc++ to work?
<Dejan>
For some reason I thought there is 7.3.1 release ...
<Dejan>
or that is in RC state?
<mattip>
Dejan: I had started it, but want to give 3.7 a chance to get to alpha first
<Dejan>
awesome, thanks mattip
<mattip>
also, conda have begun building pypy3 packages, and some tweaks to header files are needed that should go into 7.3.1
<YannickJadoul>
mattip: I haven't done anything special for OS X 10.9 or MACOSX_DEPLOYMENT_TARGET, as far as I know
<YannickJadoul>
And, $ g++ -v
<YannickJadoul>
Apple clang version 11.0.0 (clang-1100.0.33.12)