Not sure this will ever actually be an issue in practice, but right now the bessel function interpolator is constructed using a point at x =0 (were we have j_\ell(x)) when \eta = \eta_final.
At least for the polarization source function, there is a problem because j_2(x)/x^2 |x=0 is not zero, but the way the source function integration is done is by calling the bessel interpolator at x=0 and then multiplying it into the source function.
I caught a NaN in the source function for x=0, but the bessel interpolator returning j_2(0)=0 will still be wrong.
This might be solved by the new Bessel integration?
Not sure this will ever actually be an issue in practice, but right now the bessel function interpolator is constructed using a point at x =0 (were we have j_\ell(x)) when \eta = \eta_final.
At least for the polarization source function, there is a problem because j_2(x)/x^2 |x=0 is not zero, but the way the source function integration is done is by calling the bessel interpolator at x=0 and then multiplying it into the source function.
I caught a NaN in the source function for x=0, but the bessel interpolator returning j_2(0)=0 will still be wrong.
This might be solved by the new Bessel integration?