On Nov 7, 2012, at 5:41 PM, Jed Brown <jedbrown at mcs.anl.gov> wrote:
> On Wed, Nov 7, 2012 at 5:33 PM, Barry Smith <bsmith at mcs.anl.gov> wrote: > Right, but since we are using MPI we live in the world where each compiler > has a different ABI which pkg-config doesn't live in. That is why I say it is > not a great model for us. > > I've complained about pkg-config a lot, mostly because of this issue. I think > it's still useful for us because it finds static library dependencies in a > reliable way and avoids lots of failed links trying to probe these things. I have no objection to supporting it as Matt had stared (even to trying generate one for PETSc) but with or without it is a pretty minor thing; it doesn't solve any hard problems. Plus doesn't any HPC open source package generate a correct one for itself that we can use? Barry
