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

Mario Juric mjuric at lsst.org
Wed Feb 19 08:46:13 PST 2014


On 2/19/14, 8:22 , Michael Wood-Vasey wrote:
> Hi Chris,
> 
> OK, that makes sense, but I then don't understand why one has to
> install Anaconda Python at all?  The instructions in Mario's email
> stated
> 
> """
> # If you don't have python >= 2.7 with numpy >= 1.5.1 and
> # matplotlib >=1.2.0, use Anaconda python distribution by installing
> # it manually, or use the LSST-packaged one.
> eups distrib install anaconda
> setup anaconda
> """
> 

It doesn't have to be Anaconda, it could be any distribution (or
self-built packages) that satisfy those versioning constraints. Fink is
just fine.

We offer Anaconda above just as an easy way to get everything that's
needed (both Python and that list of dependencies).

> But then "eups distrib install lsst" installs matplotlib, numpy,
> pyfits and other things anyway so why does the above matter?
> 

We're lying a bit (OK, we're lying, period). matplotlib and numpy
packages mentioned in the install log are just stubs that check the
appropriate versions are present on the system. They don't actually
build anything (they did use to!).

I'll get them renamed so they don't cause confusion.

> The log file claims that it's building matplotlib and numpy.
> 
>   [  7/71 ]  matplotlib 1.2.0 (from source)                             done.
> ...
>   [ 18/71 ]  numpy 1.5.1+61f754d329 (from source)                       done.
> 
> Specifically how does the LSST stack choose which Python when
> building?  Does it just look for 'python' in the path?  Is it not
> actually building these and just finding them in system?
> 

Before, we used to build our own python (or you had to hack the infamous
"manifest.remap" file to tell EUPS not to do it and use the system python).

Now, we just use whatever is on the (python)path and the various stubs
just check that what is on the path is compatible.

Hope this helps,
-- 
Mario Juric,
Data Mgmt. Project Scientist, Large Synoptic Survey Telescope
Web : http://research.majuric.org     Phone : +1 617 744 9003



More information about the dm-users mailing list