cfbolz changed the topic of #pypy to: PyPy, the flexible snake (IRC logs: https://botbot.me/freenode/pypy/ ) | use cffi for calling C | the secret reason for us trying to get PyPy users: to test the JIT well enough that we're somewhat confident about it
<arigo>
though admittedly, the "core" tests listed there only seem to have a few shallow failures
* cfbolz
waves
<cfbolz>
danchr, arigo: there are a few problems in the os module related to the new path stuff
<danchr>
cfbolz: do you think it's usable, or close, generally speaking?
<cfbolz>
danchr: probably not quite, but if you have something to try it on, that would be interesting
<arigo>
hi
<danchr>
is testing python 3.5 code of interest? my haven't upgraded yet at my work…
<danchr>
s/me/we/
<cfbolz>
danchr: the 3.5 parts shouldn't have changed much
<danchr>
I assumed so
<danchr>
ironically, the most useful feature in Python 3.6 seems to be format literals — and PyPy already supports those in the 3.5 branch
<cfbolz>
:-)
<cfbolz>
There you go
<cfbolz>
Though it seems people use type annotations
<danchr>
yeah — the variable annotations seem useful too
<danchr>
I can immagine some of my colleagues appreciating those :)
<danchr>
(gah, I can't type today)
<danchr>
anyway — what's the plan with 3.6; just making it the new release, or an alpha/beta first?
<arigo>
it won't be the new release until we fix or decide that all remaining failures are acceptable, with an unclear timeline at the moment
<arigo>
but it could be released as "alpha"
<danchr>
it sounds to me like an alpha would be useful — if nothing else to get feedback on whether it's useful…
<danchr>
er, in good state, I mean
<danchr>
heh, I just found a bug in PyPy3: the second argument to isinstance is named differently from CPython 3.5
<cfbolz>
Eh
<cfbolz>
There are a ton of those
<cfbolz>
arigo: no, I think we should fix some of the remaining failures first. Also, I have no idea about windows and OS X
antocuni has joined #pypy
<danchr>
Actually, it seems to not matter; CPython doesn't allow keyword arguments to isinstance…
<danchr>
cfbolz: I'll do a translation of the py3.6 branch on my Mac and see if it works…
<cfbolz>
Cool
<danchr>
(perhaps I should add create pypy-devel port for MacPorts; that might allow abusing their build infrastructure for testing.)
<cfbolz>
I could also kick off a build, but only when I am back from vacation
<cfbolz>
danchr: we have a buildbot for Mac
<danchr>
oh, cool :)
<danchr>
I think it was semiunstable last I checked
<danchr>
does it use the embedded dependencies support I wrote earlier?
<danchr>
Is there any documentation of how to do development in builtin modules without a full translation?
<cfbolz>
danchr: using tests, as usual
<danchr>
OK
<danchr>
one thing I'd like to do to PyPy is to reorganise the directory layout to be more in line with the usual expectations — but that seems hard to test
<danchr>
is it possible to run PyPy in a seminstalled state, nested in another interpreter?
<danchr>
(the installed layout, that is)
<danchr>
I think most packagers actually install in a separate prefix…
<arigo>
ah, you're talking about the directory layout of installation
<arigo>
the problem is to figure out how /usr/bin/pypy would find files from e.g. /usr/lib/pypy/*.py
<danchr>
well, one option is to make something like distutils' sysconfig a builtin module, and essentially have the same logic in the RPython side?
<arigo>
I don't think that involving distutils in this plan is a good idea
<arigo>
and the problem is more, how to initialize "sys.path" so that we can at least start importing the stdlib
<danchr>
rather than @executable_path/../site-packages you use e.g. @executable_path/../lib/pypy3.5/site-packages
<danchr>
for me, distutils is connected to this issue — it seems it should utilise the same logic rather than essentially duplicating it
<danchr>
so if you configured PyPy with a different library path, distutils would call some builtin module and know about it…
<arigo>
ok, but CPython does not do that (using @executable_path), and it duplicates sysconfig.py and distutils/sysconfig.py
<arigo>
or triplicates, rather: once in sysconfig.py, once in distutils/sysconfig.py, and once in the logic setting up sys.path
<arigo>
and I don't even know what setuptools does
<danchr>
ah, but I think it generates part of distutils using configure, doesn't it?
<arigo>
probably
<danchr>
that's the thing — PyPy isn't really configurable in that sense
<arigo>
my position about that is highly conservative: it's a large mess I don't want to understand, and if it works well enough so far, better not touch it. I know that some distribution maintainers are not too happy about /opt. So yes, contributions welcome, as long as they take all use cases in mind
<danchr>
“all uses cases” is a rather lofty goal — I'd like to start with seeing whether I can get it to work, and take it from there :)
<arigo>
yes, but be warned that you'll *have* to consider all use cases before we can merge any of your work
<danchr>
I know :)
<arigo>
e.g. a cffi embedded installation, etc.
<antocuni>
and virtualenv, of course
<danchr>
and venv
lestaty has joined #pypy
lestaty is now known as Guest63409
Guest63409 has quit [Remote host closed the connection]
dddddd has joined #pypy
<kenaan>
arigo cffi/cffi ce24777ce311 /c/misc_thread_common.h: msvc support
PolarizedIons16 has joined #pypy
PolarizedIons16 has quit [Remote host closed the connection]
nimaje has quit [Quit: WeeChat 2.2]
jamesaxl has joined #pypy
nimaje has joined #pypy
<kenaan>
arigo cffi/cffi 1dd7bdcf4993 /testing/cffi0/test_ownlib.py: Issue #375 Port another test from CPython, which also passes already on win64.
fkz has joined #pypy
fkz has quit [Killed (Unit193 (Spam is not permitted on freenode.))]
gbutnaru has left #pypy ["Leaving"]
tsglove7 has joined #pypy
tsglove7 has quit [Killed (Sigyn (Spam is off topic on freenode.))]
montag4511 has joined #pypy
antocuni has quit [Ping timeout: 240 seconds]
montag4511 has quit [Ping timeout: 256 seconds]
nullcone has joined #pypy
nullcone is now known as Guest77800
Guest77800 has quit [Remote host closed the connection]
<kenaan>
arigo default a233de8fbd4b /: update to cffi/1dd7bdcf4993
<kenaan>
arigo default 1a106114b2a3 /pypy/module/test_lib_pypy/cffi_tests/cffi0/test_ownlib.py: update to cffi/1dd7bdcf4993 (really this time)
October has joined #pypy
October has quit [K-Lined]
<danchr>
at least the PyPy 3.6 branch translates on my Mac, although some of the compiler warnings seem suspicious: http://dpaste.com/19SYPFH
Guest72310 has joined #pypy
Guest72310 has quit [K-Lined]
jcea has joined #pypy
antocuni has joined #pypy
forgottenone has quit [Quit: Konversation terminated!]
[Arfrever] has joined #pypy
smidlers23 has joined #pypy
smidlers23 has quit [K-Lined]
nug700 has joined #pypy
Victorsueca23 has joined #pypy
Victorsueca23 has quit [Remote host closed the connection]