me hidden parameters used in dimensions specifications
    need to be determined based on the input array size. This is a
    harder problem because f2py has to solve the inverse problem: find
    a parameter `p` such that `upper(p) - lower(p) + 1` equals to the
    size of input array. In the case when this equation cannot be
    solved (e.g. because the input array size is wrong), raise an
    error before calling the Fortran function (that otherwise would
    likely crash Python process when the size of input arrays is
    wrong). f2py currently supports this case only when the equation
    is linear with respect to unknown parameter.

    z