[Numpy-discussion] What is consensus anyway

Ralf Gommers ralf.gommers@googlemail....
Thu Apr 26 12:19:36 CDT 2012


On Thu, Apr 26, 2012 at 7:02 PM, Chris Barker <chris.barker@noaa.gov> wrote:

> On Mon, Apr 23, 2012 at 11:18 PM, Ralf Gommers
>
> >> Perhaps a more formal "development release" system could help here.
> >> IIUC, numpy pretty much has two things:
>
> > This is a good idea - not for development releases but for master.
> Building
> > nightly/weekly binaries would help more people try out new features.
>
> good start, but I think master may fluctuate too quickly (and how
> often is it broken?) but better than nothing, yes?
>

How often is it broken? A couple of failing tests yes, but hardly ever
seriously broken.


> >> 2) there is the wxversion system
>
> > wxversion was broken for a long time on Ubuntu too (~5 yrs ago). I don't
> > exactly remember it as a good idea.
>
> well, it was a good idea, maybe not a good implementation -- and it
> was vary helpful a few years back when wx was in major flux. What we
> really need is python itself providing a package version selection
> mechanism, but Guido&c  never saw the need (the existence of
> virtualenv proves the need if you ask me....)
>
> agreed


> > virtualenv also doesn't help, because if you can use that you know how
> to build from source anyway.
>
> not true -- lots of folks use easy_install and/or pip with virtualenv.
>

Pip only installs from source, so if you haven't got the right compilers,
development headers etc. it will fail for numpy. easy_install is also a
lottery, and only works for numpy on Windows unless you are set up to build
from source.

Ralf
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://mail.scipy.org/pipermail/numpy-discussion/attachments/20120426/b9a60fec/attachment.html 


More information about the NumPy-Discussion mailing list