Occasional segfault with Near2Far adjoint and MPI

[smartalecH]

When running some basic near2far adjoint examples with MPI, there is an occasional segfault -- but at different locations. For example, when running the same script 3 different times, the segfault occurs at three different lines:

meep/src/step.cpp

Line 195 in c844fd8

if (chunks[i]->is_mine()) chunks[i]->step_source(ft, including_integrated);

meep/src/boundaries.cpp

Line 260 in e89ad6b

*(zeroes[ft][i]) = 0.0;

meep/src/step.cpp

Line 155 in c844fd8

comm_blocks[ft][pair][n0 + n] = *(chunks[j]->connections[ft][ip][Outgoing][wh[ip]++]);

Unfortunately, the behavior is completely unpredictable.