[mpich-discuss] Totalview error with MPICH2

[Pavan Balaji]

This is just to let you know that we are looking into the totalview
issues with MPICH2. This hasn't fallen off the radar.

I've added this discussion into one of our (related) tickets on
totalview (https://trac.mcs.anl.gov/projects/mpich2/ticket/962) and
cc'ed both of you on it.

If others are interested in following this discussion, please feel free
to add yourself to the Cc list of the above ticket.

Thanks,

 -- Pavan

On 12/14/2009 07:05 AM, Ashley Pittman wrote:
> On Sat, 2009-12-12 at 11:18 -0600, Yusong Wang wrote:
>> "The MPI library's definition of MPIR_PROCDESC has changed"
> 
>> Is there any way to fix the problem?
> 
> I don't know but I'm interested in the problem, MPIR_PROCDESC is one of
> the ways that the Runtime can communicate to a external tool the
> location of processes in a job.
> 
> The standard is documented here:
> 
> http://www.mcs.anl.gov/research/projects/mpi/mpi-debug/mpich-attach.txt
> 
> where as quick google shows MPIR_PROCDESC in the MPICH2 src being
> different which would indeed cause a problem.
> 
> https://svn.mcs.anl.gov/repos/mpi/mpich2/trunk/src/pm/mpd/mtv.c
> 
> I don't understand what is going on though, as far as I'm aware mpd
> doesn't support this interface and no build of mpd I've seen has this
> symbol in at all, despite what the source says...
> 
> Ashley,

-- 
Pavan Balaji
http://www.mcs.anl.gov/~balaji