Just a comment: During the tests of the matrix constructions for the anisotropic materials, there are still some constraints on the different quantities appearing in those equations:
The vector n is a unit vector and therefore is always != 0
The epsilon tensor is eye(3) in vacuum and has only slight (maybe complex) deviation from the identity matrix. Therefore it may also not vanish.
The violation of these constraints leads to problems:
with construction of the polynomial in calcXiNormZeros and therefore in test_anisotropic_xi_calculation_polynomial_zeros, test_anisotropic_xi_eigenvalues
with construction of the GLEVP from the M, C, K (since they may be zero) and therefore in test_anisotropic_xi_determinants and also test_anisotropic_xi_eigenvectors
I would suggest to let rnd_data1, rnd_data2, rnd_data3 be chosen in the specified ranges (i.e. [0.1, 1)). But rnd_data4 and rnd_data5 can be chosen from [0,1). If the last choice leads to problems, these maybe real bugs.
Just a comment: During the tests of the matrix constructions for the anisotropic materials, there are still some constraints on the different quantities appearing in those equations:
The violation of these constraints leads to problems:
I would suggest to let rnd_data1, rnd_data2, rnd_data3 be chosen in the specified ranges (i.e. [0.1, 1)). But rnd_data4 and rnd_data5 can be chosen from [0,1). If the last choice leads to problems, these maybe real bugs.