pnorbert

This simple example exhibits the valgrind problem with long doubles. ``` #include #include #include #include #include #include #include #include int main (int argc, char ** argv) { char filename[] =...

This is a valgrind issue. Filling the buffers with some value (255) before the memcpy does not help. ``` #include #include #include #include #include #include #include int main (int argc,...

NULL selections on transformed variables were never supported. I agree it's inconsistent with the fact that non-transformed variables can be read with NULL selections. This remains a TODO for now.

In this example, the second process is selecting the second writeblock of each array. However, there is only one block per variable per step. It should result in error, not...

Even the reorganize - gsplot.py pipeline works fine although this is very very similar to the one not working properly: `~/Software/ADIOS2-Examples/build/install/share/adios2-examples/gray-scott$ mpirun -n 4 /opt/adios2/bin/adios2_reorganize_mpi gs.bp reorg.bp BP5 "" SST...

Note that the gray-scott -- pdf-calc -- pdfplot.py also have the same issue, so the problem is not tied to the gsplot.py script itself

``` ~/Software/ADIOS2-Examples/build/install/share/adios2-examples/gray-scott$ mpirun -n 4 ../../../../bin/adios2-gray-scott settings-files.json : -n 1 python3 ./gsplot.py -i gs.bp MPI init without MPI_THREAD_MULTIPLE (Externally initialized:true) EVPath Version 4.2.4 -- Mon Dec 4 16:21:33 EST 2017...

The warning is there in the good run (gray-scott -- pdf-calc) as well, but the reader picks up the record ``` CM - Writing record of type ReaderRegister CM -...

SstVerbose=5: why is the reader in python selecting mpi dataplane? ``` Reader 0 (0x55f243779450): Sst set to use sockets as a Control Transport DP Reader 0 (0x55f243779450): Considering DataPlane "evpath"...

Changing gray-scott to use MPI_THREAD_MULTIPLE makes SST pick MPI as default transport and then everything works. So it's that the python mpi4py is probably threaded by default