On Wed, Jun 8, 2016 at 1:55 PM, Jason Swails <jason.swails.gmail.com> wrote:
> On Wed, Jun 8, 2016 at 1:33 PM, David A Case <david.case.rutgers.edu>
> wrote:
>
> > On Wed, Jun 08, 2016, Claudia Ramírez wrote:
> > >
> > > I did a git pull on Monday (June 6), with gnu 4.8 compilers and doing
> > just
> > > ./configure gnu (serial for now) in Ubuntu 14.04 LTS. And it has 4
> tests
> > on
> > > AmberTools that aren't working (this is without my modifications), and
> I
> > > don't know why. I'm attaching the log file to see if anyone could help
> me
> > > make it work.
> >
> > All the problems are python-related problems, related either to pytraj or
> > to the sander api interface.
> >
> > What is the PYTHON variable in your config.h? If you are not already
> > using a miniconda python, try going to $AMBERHOME/AmberTools/src and
> > running
> > configure_python. Then go back to AMBERHOME and re-run configure.
> >
> > [I still am pushing the (so-far, minority) viewpoint that all users
> should
> > be offered the choice to install the miniconda python. Users that have
> > both
> > experience with python, and the desire/need to use a different
> interpreter,
> > are welcome to decline the offer. But configure2 is still prone to
> > deciding
> > that the python in a user's path is OK when it is in fact not OK. So,
> > letting
> > us know which python your Amber installation is using would be a help
> > here.]
> >
>
> I give :). Based on citations, questions to the mailing list, and age of
> the various Python projects, I think the predominant use of Python in Amber
> is MMPBSA.py, which carries with it only what ParmEd needs to run (Python
> 2.7 + numpy these days, which is a lot more than it used to require).
+ mpi4py if using MPI :P
>
> For that, miniconda+everything seems like overkill. But on the flip side,
> deciding on a "standardized" setup *drastically* reduces the effort we need
> to spend to fully test all "common" user configurations.
> My only
> trepidation here is that, on OS X, the libc++/libstdc++ issues still seem
> to be wreaking havoc with pytraj and cpptraj.
FYI: pytraj/cpptraj has only issue with *osx/clang/miniconda* (Dan and I
are working on this). If users are using gnu compiler, there shouldn't be
an issue.
> I hope we can soon find
> something to settle on to make deployment easy, because the added benefit
> from stability, performance, and avoiding effort duplication we get by
> integrating with mature Python packages provided through the Anaconda stack
> is a big win.
>
>
>
Totally agree with you
cheers
Hai
_______________________________________________
AMBER-Developers mailing list
AMBER-Developers.ambermd.org
http://lists.ambermd.org/mailman/listinfo/amber-developers
Received on Wed Jun 08 2016 - 11:30:02 PDT
