On Jun 19, 2007, at 11:35 AM, Alf Wachsmann wrote:
> In line 568 of openmpi-1.2.2/orte/mca/pls/rsh/pls_rsh_module.c
> the call "p = getpwuid(getuid());" returns an invalid shell on our
> nodes. This leads to "pls:rsh: local csh: 0, local sh: 0", i.e. the
> shell is not defined and only the user's ~/.profile gets exectuted
> in lines
> 649ff. This forces users to set their LD_LIBRARY_PATH instead of
> OpenMPI do this for them in lines 981ff.
Wow -- neat! We certainly didn't think of this case. :-)
> Before LSF starts a user job, it sets their complete environment
> the SHELL environment variable. I am wondering whether OpenMPI
> could look
> at that env. variable in lines 567ff in addition to or instead of the
> getpwuid() call.
A good idea; yes, we can do this. I'm sorry that it won't make the
1.2.3 release, though. I'll file a ticket.
FWIW: we're actively working on native LSF support in Open MPI such
that this kind of rsh tomfoolery won't be necessary in the future.
We hope to have it ready for the 1.3 series.