I should have tried this before I replied. I had a further thought (after I replied, of course) -- I was wondering if one of our components had a reference to ompi_comm_world (and not your application) and that caused the problem. If you installed 1.1 over 1.0.2 and didn't uninstall first, an old component (e.g., pml_teg) could have something like this.
But then I just tried this and I was unable to replicate the problem. Specifically:
- make 1.0
- install 1.0
- compile sample MPI app
- mpirun sample MPI app to ensure it works
- make 1.1 with the same prefix as 1.0
- install 1.1 over 1.0 (i.e., without uninstalling 1.0 first)
- mpirun sample MPI app -- expect it to fail like yours
But mine succeeded with no linking errors.
Can you send a recipe to reproduce this error?
> -----Original Message-----
> From: users-bounces_at_[hidden]
> [mailto:users-bounces_at_[hidden]] On Behalf Of Jeff Squyres
> Sent: Thursday, June 29, 2006 4:25 PM
> To: Open MPI Users
> Subject: Re: [OMPI users] OpenMPI 1.1 backward compatible?
> I think you may have caught us in an unintentional breakage.
> If your Open MPI was compiled as shared libraries and dynamic
> shared objects (the default), this error should not have
> happened since we did not change mpi.h. So there must be a
> second-order effect going on here (somehow the size of a
> back-end data structure caused a problem. Hrm.).
> We'll look into this, because for where all of OMPI's
> functionality is in shared libraries and components, the
> user's application should be isolated from internal changes
> like this (i.e., and we can provide forward compatibility).
> I suspect that something deeper is going on, so let us go
> investigate and come back with a more definitive statement.
> > -----Original Message-----
> > From: users-bounces_at_[hidden]
> > [mailto:users-bounces_at_[hidden]] On Behalf Of Daniël Mantione
> > Sent: Tuesday, June 27, 2006 11:09 AM
> > To: Open MPI Users
> > Subject: Re: [OMPI users] OpenMPI 1.1 backward compatible?
> > On Tue, 27 Jun 2006, PeterKjellström wrote:
> > > On Monday 26 June 2006 16:55, Daniël Mantione wrote:
> > > > Hi!
> > > >
> > > > Just tried out OpenMPI 1.1. First impression is that it
> > doesn't seem to
> > > > be able to run OpenMPI 1.0.2 executables. The result of
> > such an attempt
> > > > can be seen below.
> > > >
> > > > Is it right that OpenMPI 1.1 cannot run 1.0.2
> executables? If yes,
> > > > shouldn't the major version of the library have been increased?
> > >
> > > If I'm not entierly mistaken OpenMPI (like LAM before it)
> > requires you to
> > > rebuild your applications when switching version. This goes
> > for _ANY_ version
> > > even minor ones.
> > Well, no, the FAQ promises that applications will continue to work:
> > > 10. We just upgraded our version of Open MPI; do I need to
> > recompile all my MPI apps?
> > >
> > > It is unlikely. Most MPI applications solely interact with
> > Open MPI through the standardized MPI
> > > API and the constant values it publishes in mpi.h. The
> > MPI-2 API will not change until the MPI
> > > Forum changes it.
> > >
> > > We will try hard to make Open MPI's mpi.h stable such that
> > the values will not change from
> > > release-to-release. While we cannot guarantee that they
> > will stay the same forever, we'll try hard
> > > to make it so.
> > So, it it a bug that things break? Or was it intentional?
> > Daniël
> users mailing list