[Numpy-discussion] How to upload to pypi

David Cournapeau cournape at gmail.com
Thu Feb 14 06:56:46 EST 2013


On Wed, Feb 13, 2013 at 5:35 AM, Ondřej Čertík <ondrej.certik at gmail.com> wrote:
> David,
>
> On Tue, Feb 12, 2013 at 6:46 AM, David Cournapeau <cournape at gmail.com> wrote:
>> On Tue, Feb 12, 2013 at 5:49 AM, Ondřej Čertík <ondrej.certik at gmail.com> wrote:
>>> Hi,
>>>
>>> I have uploaded the NumPy 1.7.0 source distribution to pypi:
>>>
>>> http://pypi.python.org/pypi/numpy/1.7.0
>>>
>>> I did it by uploading the file PKG-INFO from numpy-1.7.0.tar.gz. It
>>> said "Error processing form. Form Failure; reset form submission"
>>> about 3x times,
>>> but on the 4th try it went through. I reported the issue here:
>>>
>>> https://sourceforge.net/tracker/?func=detail&aid=3604194&group_id=66150&atid=513504
>>>
>>> I then attached the numpy-1.7.0.tar.gz and numpy-1.7.0.zip source files.
>>>
>>> Now I am having trouble attaching the windows installers, just like
>>> they are here:
>>>
>>> http://pypi.python.org/pypi/numpy/1.6.2
>>
>> Those installers are ones built through bdist_wininst. You should
>> *not* upload superpack installers there, as most python tools will not
>> know what to do with it. For example, easy_install will not work with
>> those, even though it does with simple installers from bdist_wininst.
>>
>> So ideally, one should build simple (== bdist_wininst-generated)
>> installers using the lowest common denominator for architecture (i.e.
>> pure blas/lapack, not atlas), and the superpack installer on
>> sourceforge. Incidentally, that's why the super pack installer uses a
>> different filename, to avoid confusion.
>
> I see. I looked into my scripts and it turns out that actually I do
> build the bdist_wininst versions as well, I just didn't know what they
> are for, so I ignored them. Now I can see that those will get uploaded
> to pypi, so I did that now and it works!
>
> http://pypi.python.org/pypi/numpy/1.7.0
>
> I learned something new today, thanks for the explanation. So pypi
> should be done.

Great, thanks a lot for doing all this !

David



More information about the NumPy-Discussion mailing list