In the case where a 2nd/3rd derivative is symbolically not zero but numerically
zero at the evaluation point, the last lines of the g2/g3 matrices (in
KordpDynare::calcDerivativesAtSteady()) where uninitialized (these matrices
store the sparse hessian/3rd-deriv in coordinate list form, i.e. with 3 columns
and as many rows as non-zero elements). When reconstructing the dense
hessian/3rd-deriv matrix out of g2/g3, this would result in invalid memory
accesses.
- in the 3rd derivatives matrix, among symmetric elements, the first one had
the right values but the following ones were set to zero
- moreover, the k-order DLL was trying to add all the symmetric elements in the
folded tensor, instead of only keeping one value among all the symmetric ones
- hopefully, Ondra's tensor library was (silently) refusing to add symmetric
elements after the first (and right) value had been added
- so the final result was correct