Allow python-config for Python 2 and Python 3

[lgarrison]

Fixes virtualenv install, which only links python-config instead of python3-config (closes #156)

I was able to reproduce the installation problem reported in #156 and confirm that this fixes it.

[astropy-bot[bot]]

Hi there @lgarrison :wave: - thanks for the pull request! I'm just a friendly :robot: that checks for issues related to the changelog and making sure that this pull request is milestoned and labeled correctly. This is mainly intended for the maintainers, so if you are not a maintainer you can ignore this, and a maintainer will let you know if any action is required on your part :smiley:.

Everything looks good from my point of view! :+1:

If there are any issues with this message, please report them here.

[lgarrison]

This works on Travis for all but Python 2.6, which doesn't have the sysconfig module, so sysconfig.get_path('scripts') fails. distutils.sysconfig provides similar functionality on Python 2.6, but the canonical way to get the bin/ dir via distutils.sysconfig.get_config_var('BINDIR') does not give the correct answer inside a virtualenv (points to the original location).

So we could use the hacky distutils.sysconfig.EXEC_PREFIX + '/bin', or just drop Python 2.6 support...

[manodeep]

Thanks for sorting this out. I think we should completely remove the python-config dependence; all the flags we need are stored under sysconfig

One reason why I am proposing such a change is that I just looked at my python situation on my laptop and looks like /usr/bin/python-config is being picked up, even though I am using the python from anaconda3/bin. I also know of people that use python as an alias...

Separately, I think it is fine to drop python2.6 support but we should make plans for future releases and document them, perhaps on the wiki. Such a drop for python2.6 should probably be done in a minor version update 2.2 (with as-yet un-released 2.1 for DDsmu*)

[lgarrison]

What path do you get if you run python -c "import sysconfig; print(sysconfig.get_path('scripts'));"? That's the solution that's implemented in this PR; the global python-config is never called.

[manodeep]

I get /Users/msinha/anaconda3/bin and there are a bunch of python*-config executables within there.

I would still say, when python3 is being used, we check for python3-config first before checking python-config

[lgarrison]

Makes sense, although hopefully it won't make a difference either way. I've pushed that and it seems to have passed Travis.

I also went ahead and dropped Python 2.6 support and edited the docs to reflect this.

[manodeep]

Since we are potentially introducing breaking changes, should we merge this into a different minor version?

[lgarrison]

Probably! I figured we'd work out release versioning over Skype.

[lgarrison]

We're going to wait until v2.1 is released in master before merging this in.

[manodeep]

New Corrfunc 2.1.0 is now live on PyPI

[manodeep]

@lgarrison Is it okay to delete the venv_fix branch now?

[lgarrison]

I think so? Unless we wanted to ask the original issue poster to verify the fix out of master.

On Fri, Aug 17, 2018, 6:50 PM Manodeep Sinha notifications@github.com wrote:

@lgarrison https://github.com/lgarrison Is it okay to delete the venv_fix branch now?

— You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub https://github.com/manodeep/Corrfunc/pull/159#issuecomment-414006824, or mute the thread https://github.com/notifications/unsubscribe-auth/AFYPy1e867_SIqnIFMSyTYoU8ACou_tNks5uR0i9gaJpZM4U2rTH .

[manodeep]

I think the resolution was suggested by @rainwoodman - so should be okay. Deleting the branch; we can always un-delete if required.

Related, should we add a python_requires into the setup.py file?

[lgarrison]

Yes, I think so. I'll do that in my capture_stderr branch so we can CI it in one go.