For me things work fine if I install Anaconda with Pyhton 2.x (do not use
3.x !!) https://www.continuum.io/downloads#_windows
Then, all he has to do is:
1. launch cmd
2. type: conda install libpython
3. reboot
4. ensure ENV["PYTHON"] evals to the python.exe in the anaconda dir (i.e.
add a user environment variable).
then Pkg.build("PyCall") works just fine.
Am Montag, 12. Oktober 2015 16:59:27 UTC+2 schrieb Alex Dowling:
>
> Thank you for the suggestions. I'm actually helping a colleague debug
> this, so I don't know the exact sequence of commands that were tried. I'm
> going to recommend just nuking the Julia v0.4 user directory, upgrading to
> the Julia v0.4 stable release and trying to install PyPlot again. If this
> doesn't work, I'll post again here.
>
> On Thursday, October 8, 2015 at 10:39:06 AM UTC-5, Steven G. Johnson wrote:
>>
>>
>>
>> On Thursday, October 8, 2015 at 11:24:28 AM UTC-4, Steven G. Johnson
>> wrote:
>>>
>>>
>>>
>>> On Thursday, October 8, 2015 at 10:42:38 AM UTC-4, Luthaf wrote:
>>>>
>>>> I guess that is a bug in PyCall, so you should do your bug report there
>>>> : https://github.com/stevengj/PyCall.jl/issues/new. Please ping me
>>>> (@Luthaf) on the issue, because this might be related with Conda.jl
>>>>
>>>
>>> Conda.PYTHONDIR was defined in recent versions of Conda, so I'm guessing
>>> you have an out-of-date Conda. That, combined with the fact that your
>>> PyCall package was not already built the first time you tried it, leads me
>>> to suspect that you did a Pkg.checkout("PyCall") at some point to get the
>>> latest (master) version. By doing this, you turned off automatic
>>> dependency management, which is why you were able to update PyCall without
>>> updating Conda.
>>>
>>
>> (Either that, or you're hitting
>> https://github.com/JuliaLang/julia/issues/13458 ... just start a new
>> Julia session and do Pkg.update() followed by Pkg.build("PyCall") ...)
>>
>
