[openib-general] mpi drop in openib tree

[Troy Benjegerdes]

On Thu, Aug 25, 2005 at 12:50:04PM -0400, Dhabaleswar Panda wrote:
> Hi Roland,
> 
> > As for whether MPI should be in the OpenIB subversion tree or not, my
> > personal opinion is that having MPI there is only appropriate if the
> > svn tree is being used as the primary development source tree.  I
> > don't think it's appropriate to use the OpenIB svn server as a release
> > distribution mechanism.
> 
> Yes, we plan to use the svn tree as the primary development source
> tree for mvapich-gen2 not as a release distribution mechanism.
> 
> In fact, during the last two days we have received a couple of
> patches, bug fixes, and suggestions and have checked-in these patches,
> fixes, and enhancements to the SVN tree with credits to the people who
> supplied us the patches. Thanks to those who sent us the patches!!
> 
> We encourage people in the OpenIB community to test the latest version
> of mvapich-gen2 from the svn tree and provide us feedbacks and
> comments so that we can keep on enhancing it.

What is the plan for merging with future MPICH updates? I've dealt with
keeping full copies of the linux kernel in various version control
systems, and merging with upstream always turns into a huge headache.
Of course, if there aren't going to be any more MPICH-1 updates, then
this is really no big deal.

My first suggestion is to remove the java .jar and configure files.
There is also probably a lot of other stuff that really doesn't need to
be there as well. Do we really need to keep windows specific code in the
(currently linux-only) OpenIB project svn tree?