On Fri, 20 Sep 2013, Manav Bhatia wrote:
> I tried to dig into this and identified that this is due to the
> destructor being called after MPI_finalize() (libmesh.C: line 625). The
> following code is able to repreduce the problem at my end:
>
> MPI_Init(&argc, const_cast(&argv));
> std::auto_
Hi Roy,
I got this to work at my end. I pretty much moved the MPI_Finalize block
(inside ~LibMeshInit() ) to the end of the destructor, and commented out
the line
if (libMesh::on_command_line ("--redirect-stdout"))
so that the perflog is purged before MPI_Finalize(). (Any reason to keep
thi
Another datapoint, I see a random failure under Linux with gcc using
openmpi-1.7.2 that disappears with 1.6.?, so be careful out there. I'm trying
to whittle it down but haven't been able to make much progress yet.
I wish macports let you subscribe to a 'stable' tree!
On Sep 20, 2013, at 1:3
Hi,
Sometime ago I had sent an email concerning a runtime error I was seeing
upon termination of my code. This was happening inside call to the
destructor of libMesh::perflog, leading to a memory error while calling
localtime() at line 46 of timestamp.C.
I tried to dig into this and identif
