Re: How is mpi-default-dev supposed to work?
On 23/03/10 at 19:01 +0100, Manuel Prinz wrote:
> Am Dienstag, den 23.03.2010, 13:16 +0100 schrieb Lucas Nussbaum:
> > I'm not the mpi-defaults maintainer :-) Manuel? Adam?
>
> Sorry for replying late!
>
> mpi-defaults was supposed to be an interim solution and in an ideal
> world gone already. I know it's not the best solution out there, but the
> best we have at the moment.
>
> To comment on the discussions you had: The problem Thibaut mentions only
> exists on the platforms where both implementations exist and mpich2 is
> installed later into a buildd chroot. Lucas, you're right with what you
> said about the buildds but I'm not sure if we can (or should, for that
> matter) rely on that. On platforms where Open MPI is not available,
> there is no problem, as Open MPI never built on those anyway.
>
> The easiest (and cleanest) solution seems to be to add mpicc.default and
> friends (or whatever we call it) to mpi-defaults-dev, being symlinks to
> the default mpicc on that platform. Those can be used by the packages
> using only mpi-defaults-dev. The only downside I see right now it the
> fact that we should fix all packages not using it also which has a
> negative effect on the whole transition. On the other hand, we do not
> need to; it won't break anything if the assumptions made are valid. (But
> as said, we can't really rely to it.)
>
> I could also increase the update-alternatives priority of Open MPI to
> something higher than MPICH2. This should have no negative side effects
> but does of course not fix the current problem. What do you think about
> that? (Planning to upload some minor fixes to Open MPI anyway.)
>
> Last, I recently changed the arch list in mpi-defaults to match that of
> Open MPI on all platforms. There are some more platforms supported now,
> and I hope this goes on. Just wanted to mention that, it does have
> nothing to do with this really.
I think that the best way to solve this problem is to explicitely
conflict with the other implementations in mpi-defaults. I don't see any
other solution that really fixes the problem we are having and doesn't
require an upload of all packages using mpi-defaults.
--
| Lucas Nussbaum
| lucas@lucas-nussbaum.net http://www.lucas-nussbaum.net/ |
| jabber: lucas@nussbaum.fr GPG: 1024D/023B3F4F |
Reply to: