On Mon, Jun 1, 2015 at 3:43 AM, Ralf Gommers <[email protected]> wrote:
> > > On Fri, May 29, 2015 at 7:28 PM, Benjamin Root <[email protected]> wrote: > >> Speaking from the matplotlib project, our binaries are substantial due to >> our suite of test images. Pypi worked with us on relaxing size constraints. >> Also, I think the new cheese shop/warehouse server they are using scales >> better, so size is not nearly the same concern as before. >> >> Ben Root >> On May 29, 2015 1:43 AM, "Todd" <[email protected]> wrote: >> >>> On May 28, 2015 7:06 PM, "David Cournapeau" <[email protected]> wrote: >>> > On Fri, May 29, 2015 at 2:00 AM, Andrew Collette < >>> [email protected]> wrote: >>> >> >>> >> In any case I've always been surprised that NumPy is distributed >>> >> through SourceForge, which has been sketchy for years now. Could it >>> >> simply be hosted on PyPI? >>> > >>> > >>> > They don't accept arbitrary binaries like SF does, and some of our >>> installer formats can't be uploaded there. >>> > >>> > David >>> >>> Is that something that could be fixed? >>> >> > For the current .exe installers that cannot be fixed, because neither pip > nor easy_install can handle those. We actually have to ensure that we don't > link from pypi directly to the sourceforge folder with the latest release, > because then easy_install will follow the link, download the .exe and fail. > > Dmg's were another non-supported format, but we'll stop using those. So > if/when it's SSE2 .exe installers only (make with bdist_wininst and no > NSIS) then PyPi works. Size constraints are not an issue for Numpy I think. > > Ralf > What about adding some mechanism in pypi to flag that certain files should not by downloaded with pip?
_______________________________________________ NumPy-Discussion mailing list [email protected] http://mail.scipy.org/mailman/listinfo/numpy-discussion
