On Sat, Aug 13, 2011 at 6:18 PM, Jed Brown <jedbrown at mcs.anl.gov> wrote:
> Parmetis-4.0 has a much nicer build process and eliminated the historical > conflict between libmetis and libparmetis. This is good, however SuiteSparse > (contains Umfpack and Cholmod) used a private (I guess) API (the closest > thing to NodeComputeSeparator is ComputeVertexSeparator, but it has one less > argument; there may be other differences) that has since changed, so the > latest Parmetis cannot be used with them. On the first pass, it looks like > MUMPS is in a similar situation. Do we wait on updating our Parmetis, until > these packages update, update PETSc --download-parmetis now and watch the > breakage, or try fixing the dependent packages? I think we definitely update petsc-dev (not the release), and then guilt people into fixing their code. This means then that the release clone should be made on Monday, so that work can begin in dev on the new ParMetis. Matt -- What most experimenters take for granted before they begin their experiments is infinitely more interesting than any results to which their experiments lead. -- Norbert Wiener -------------- next part -------------- An HTML attachment was scrubbed... URL: <http://lists.mcs.anl.gov/pipermail/petsc-dev/attachments/20110813/7d762651/attachment.html>