|Date: 2016-08-17 20:21|
Sender: sbgrid admin
I've seen the same - install.sh completes with status 0 and then the test fails (not using Anaconda python).
As an aside - it looks like pip by default will install (the different) "openmm" that is rev 0.2, but "OpenMM" is apparently not known to pip by default. And pip apparently cannot distinguish between "openmm" and "OpenMM" once *either* one of them is installed in a given python distribution. Or apparently pip cannot work with/list/install both "openmm" and "OpenMM". It seems there is a kind of "package name space pollution" going on that is incompatible with the standard install tools.
|Date: 2016-06-02 22:46|
Sender: D Greene
I don't think I have multiple copies of OpenMM installed as I had never attempted to install it before last night. I have installed other things like numpy and openeye to run on python 2.7 and those run fine (to verify I just did a check a few moments ago, and I was able to import and run openeye and numpy on an ipython notebook without any issues).
I should mention that I did try to install mdtraj, nglview, parmed, and openmoltools last night before trying to install the main OpenMM package (I actually only attempted to install the main OpenMM after I got an error while running the test script for openmoltools).
Here are the contents of my /usr/local/lib/python2.7/dist-packages folder if it helps:
Thanks for the help!
|Date: 2016-06-02 21:12|
Sender: Peter Eastman
Just a guess, but do you have multiple copies of OpenMM installed on your computer? That looks to me like the Python module is from one release, but it's trying to link to the C++ libraries from a different release. That could happen, for example, if your LD_LIBRARY_PATH points to an older version.