On Apr 19, 2012, at 9:49 AM, Igor wrote: > The issue was resolved by writing a wrapper script that passed a > variable with other name and redefined it to LD_LIBRARY_PATH later. > But just FYI I give some comments/answers below. Once again, thanks > for support. > >>> Can I force fipy to use python RPATH? >> >> >> Never heard of RPATH. > http://en.wikipedia.org/wiki/Rpath
This sounds like something that needs to be compiled into the binary files, either the Python executable or the Trilinos shared library or both. That's nothing FiPy has any access or control over and frankly sounds like the wrong approach. Is there no mechanism to define environment variables (including LD_LIBRARY_PATH) on the nodes of your cluster? Can you create a .login or .bashrc or any other file to establish the runtime environment? If this was really a security risk, then the solution would be for that environment variable to not be used at all, but clearly it is since you can pass a different variable and redefine it later. Can you direct us to some documentation for what this "security" policy is with regard to LD_LIBRARY_PATH? How does gridengine expect an executable to find shared libraries? Another option is to try building a statically-linked version of PyTrilinos. See "(*) Building static libraries and exectables:" in http://trilinos.sandia.gov/Trilinos10CMakeQuickstart.txt _______________________________________________ fipy mailing list [email protected] http://www.ctcms.nist.gov/fipy [ NIST internal ONLY: https://email.nist.gov/mailman/listinfo/fipy ]
