[petsc-dev] unfreed MPI keyvals

[Jed Brown]

On Fri, 21 May 2010 00:03:08 -0300, Lisandro Dalcin <dalcinl at gmail.com> wrote:
> I'm getting this... In the past I remember getting more, now just
> one... Any of you was working on this?

I fixed some of them in the thread that wraps up here, I think the last
couple can be cleaned up in the same way.

  http://lists.mcs.anl.gov/pipermail/petsc-dev/2009-November/001769.html

Note that they also show up in valgrind (at least with Open MPI, I think
MPICH2 frees that structure by default, but of course tells you where it
happened when built with the debugging options).

Jed