Hi Pablo,

On 16/09/16 12:15, Pablo Escobar Lopez wrote:
I remember we have commented in some hackathon (while having some beers I think) about the possibility of distributing easybuild as a self-contained app. Using something like mini-conda (http://conda.pydata.org/miniconda.html) with easybuild already installed on top of hit could do the trick. With this approach installing easybuild would be as simple as uncompressing a tarball, modifying PATH and use it (supposing a modules tool is already available). Has this been discussed again or it got lost with the beers? ;)

Maybe easybuild 3.x could be a good moment to do this change. IMHO, it would simplify the development of easybuild as only one python environment needs to be supported and tested, it would make easybuild more portable and would give the main developers more freedom to choose the python version to use. The provided mini-conda could also provide all the required dependencies so the --dep-graph option or the git integration works out of the box (right now you need to add extra dependencies to get those options working). Maybe we could also include a working Lmod installation inside the mini-conda tarball so users can choose to use Lmod from the tarball or not. Maybe it would also help for installations reproducibility.

Probably I am missing some drawbacks about this approach but I don't see them now.

That certainly sounds like a promising avenue, but in my view it would just be another way of installing EasyBuild. I don't think we should force something like this onto everyone; people like to have options...

Of course, if this turns out to work well, it may quickly become the standard way of installing EasyBuild. Especially the possibility to include optional dependencies sounds promising!

Let me know if you need any help with this, or if you have something that can be tested by others.




2016-09-16 11:33 GMT+02:00 Jens Timmerman <jens.timmer...@ugent.be <mailto:jens.timmer...@ugent.be>>:

    On 09/16/2016 10:34 AM, Riccardo Murri wrote:
    > Hi all,
    >> If you would be so kind to fill this in it would give me a
    sense of what is going on in the
    >> python 2.x/3.x support landscape.
    > It might be worth to note that `pip` is dropping support for
    Python 2.6
    > in release 9.0, scheduled beginning of next year: see
    on that note, here is a list of scientific python applications
    that have pledged to drop python 2.X support in 2020:

    EB can install these and provide a python3 module, so I expect no
    issues on using these on HPC systems,
    but it is a statement alright.

    > Ciao,
    > R
    > --


Pablo Escobar López
HPC systems engineer
sciCORE, University of Basel
SIB Swiss Institute of Bioinformatics

Reply via email to