[ajnonaka]

Fixes #113.

Problem

Adding sigma, epsilon, or mu (or the LLG mag_Ms_*, mag_alpha_*, mag_exchange_*, mag_anisotropy_* fields) to fields_to_plot segfaults when algo.em_solver_medium is not macroscopic. The crash happens in FullDiagnostics::InitializeFieldFunctors, where MacroscopicProperties returns null MultiFab pointers because the material properties were never initialized.

Diagnostics::BaseReadParameters already has asserts meant to catch this with a clear message. They never fire because they check m_varnames, which is still empty at that point. It only gets filled from m_varnames_fields later in the same function.

Fix

The seven checks now look at m_varnames_fields instead of m_varnames, matching the existing check for G just above them.

Testing

• With algo.em_solver_medium = vacuum and sigma in fields_to_plot: before, a segfault; after, the run stops with sigma in plotfiles only works with algo.em_solver_medium=macroscopic.
• With algo.em_solver_medium = macroscopic: sigma, epsilon, and mu are still written correctly, with and without LLG and when restarting from a checkpoint.