[petsc-dev] moab nightlybuild failure

[Satish Balay]

On Fri, 28 Jun 2013, Jed Brown wrote:

> Satish Balay <balay at mcs.anl.gov> writes:
> 
> > Yes most complexity is manageable at [at some cost and tradeoffs]. Its
> > a matter of whats required.
> >
> > For ex: If tracking moab-dev is required - one simplification is to
> > use the nightly tarballs [as before] - not the git repo.
> 
> The challenge is knowing when an update is required.  If the URL doesn't
> change, then we don't know that an update is needed.  My intent was that
> we would eventually use the gitcommit field to know when an update is
> needed, but note that that is quite different from tracking whatever is
> in 'master'.

With such extra metadata - one could potentially do such tarcking of
updates with tarballs aswell.

But for what Barry wants - we need subrepo type support where we keep track of
snapshot relations at each step of progress of petsc-dev  & moab-dev.

As it stands now - if you checkout petsc-dev from a few weeks back [or
working on a branch for that long] - you get latest moab that could
potentially have conflicts.

Satish