non-null polarization_state in fields::dump (unsupported)

[radscience]

Probably related to #1767 .

Trying to dump fields in python in any of several ways, invariably get

meep: non-null polarization_state in fields::dump (unsupported)

although dumping structures completes (haven't tried loading yet).

make -j 4 check failed on only the dump_load test, but running that with mpirun -np <all-cores> didn't give any error (though mpirun -np <less-than-all-cores> fails).

This was compiled in an ubuntu-jaunty singularity container, based on the ubuntu-bionic compile instructions on readthedocs with minimal modifications. It does seem to work with python otherwise.

Alternatively, is there anything that can be done to the sim object that would not throw it into this error but preserve enough information that it could be loaded & transformed back? I have a large simulation to do relative to my resources & would really like to be able to step & restart.

[oskooi]

This is a feature and not a bug. The load and dump feature only supports dumping a structure with dispersive materials but it does not support polarization fields $\vec{P}$ which need to be stored and timestepped when dispersive materials are present. What you are probably trying to do which is triggering the error is dumping the fields from a simulation involving dispersive materials. Unfortunately, this feature is not yet supported.

We should probably update the documentation to make this point clearer.