#9 is not fixed on PyPy

[skirpichev]

See commit skirpichev/diofant@cfe8a53, which return -n auto and failing CI build: https://travis-ci.org/diofant/diofant/jobs/407725598

CC @greut

[greut]

os.sched_getaffinity seems to be Python3 only and restricted to some platforms..., so it won't fix it for anything Python2-ish, or as it seems PyPy(3).

[skirpichev]

@greut, above build was on

$ python --version
Python 3.5.3 (3f6eaa010fce, Jan 11 2018, 04:44:35)
[PyPy 5.10.1 with GCC 6.2.0 20160901]

It's Python3-compatible version of PyPy.

[greut]

% pypy3
Python 3.5.3 (fdd60ed87e941677e8ea11acf9f1819466521bf2, Jun 22 2018, 02:39:05)
[PyPy 6.0.0 with GCC 8.1.1 20180531] on linux
Type "help", "copyright", "credits" or "license" for more information.
>>>> import os
>>>> os.sched_getaffinity(0)
Traceback (most recent call last):
  File "<stdin>", line 1, in <module>
AttributeError: module 'os' has no attribute 'sched_getaffinity'

% python3 
Python 3.6.6 (default, Jun 27 2018, 13:11:40) 
[GCC 8.1.1 20180531] on linux
Type "help", "copyright", "credits" or "license" for more information.
>>> import os
>>> os.sched_getaffinity(0)
{0, 1, 2, 3}

[skirpichev]

CPython docs says "New in version 3.3", so, IIUIC it must be in PyPy3. Sounds like PyPy3 bug.

psutils seems to be working, as it was suggested in related issue travis-ci/travis-ci#4696 (comment):

In [1]: import psutil

In [2]: psutil.Process().cpu_affinity()
Out[2]: [0, 1]

In [3]: import sys

In [4]: sys.version
Out[4]: '3.5.3 (fdd60ed87e94, Apr 24 2018, 06:10:04)\n[PyPy 6.0.0 with GCC 6.2.0 20160901]'

Not sure if it worth adding additional deps. You may want implement instead a hack for travis-ci, suggested in #9.

[nicoddemus]

I don't mind adding:

if os.environ.get('TRAVIS') == 'true':
    return 2

To our function that discovers the number of cpus when auto is given. It is a hack, but it is just 2 lines of code and will solve a real problem users have right now.

What do you think @RonnyPfannschmidt?

[RonnyPfannschmidt]

sounds okish as a hack

can we factor the cpu number finding into a function and add a few unit/integration tests while we are on it

[nicoddemus]

Sure, sounds good.

Anybody would like to take this on? Seems like a quick contribution. I would be happy to review/merge this.

[skirpichev]

I'll do. But first, I believe, pypy3 bug should be created. Can you release a minor version after merge?

…

On Jul 25, 2018 2:59 PM, "Bruno Oliveira" ***@***.***> wrote: Sure, sounds good. Anybody would like to take this on? Seems like a quick contribution. I would be happy to review/merge this. — You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub <#316 (comment)>, or mute the thread <https://github.com/notifications/unsubscribe-auth/ACDlGOIAWj6xJVdWREHxxnrWGYzEAjLDks5uKF2VgaJpZM4VfhHX> .

[greut]

@skirpichev https://bitbucket.org/pypy/pypy/issues/2375/missing-posix-attributes-in-py3k

[nicoddemus]

Can you release a minor version after merge?

Sure!