[lsst-dm-stack-users] A test development version of Winter'14 stack

Chris Walter chris.walter at duke.edu
Wed Feb 19 07:42:06 PST 2014


Hi Michael,

I think what Mario is saying is that the configure etc is only relevant for building of EUPS itself.  For that you don't care about matplot lib etc.  He is recommending that eups is built with the system python so it doesn't depend on anything else and doesn't break if alternative pythons get wiped out.

Building the actual DM stack is where anaconda or fink is relevant.

Is that right Mario?

-Chris



On Feb 19, 2014, at 10:30 AM, Michael Wood-Vasey <wmwv at pitt.edu> wrote:

> On Feb 19, 2014, at 0:02, Mario Juric <majuric at gmail.com> wrote:
> 
>> A better choice may have been --with-python=/usr/bin/python. We should
>> change that warning to make it clear that using /usr/bin/python is
>> strongly preferred.
> 
> Hmmm...  I understand this touches on the long-running discussion of different pythons, etc.
> 
> To explain more clearly I did this because I wanted to use the numpy and matplotlib already installed with Fink.  Otherwise I have to install Anaconda (easy enough, I hear; but I wanted to keep this simple).  The system Python numpy is 1.6 and matplotlib is 1.1 which are older than the requirements.  
> 
> Aside from the downsides of blowing away /sw and breaking eups, are there any other concerns with using the /sw Python?
> 
> Using Anaconda to provide numpy and matplotlib seem to also bear the same risk that if I (re)move the Anaconda installation things break as well.  suppose it will be just the LSST stack using python rather than eups.  But for many users who are just using eups to install the LSST stack this is somewhat the same.
> 
> - Michael
> 
> 
> _______________________________________________
> lsst-dm-stack-users mailing list
> lsst-dm-stack-users at lsstcorp.org
> http://listserv.lsstcorp.org/mailman/listinfo/lsst-dm-stack-users





More information about the dm-users mailing list