Re: [Bug-gsl] bug report

[Rhys Ulerich]

> sl_sf_bessel_Knu(), I got some problems: for example
> gsl_sf_bessel_Knu(pi,0.5) gives "domain error"

I'm afraid I'm not very familiar with this code.

If you by hand walk through gsl/specfunc/bessel_Knu.c does the logic
follow the path you'd expect?

For example, gsl_sf_bessel_Knu calls gsl_sf_bessel_Knu_e which calls
both gsl_sf_bessel_Knu_scaled_e and gsl_sf_exp_mult_err_e.  The return
value is a domain error if either of these calls return a domain
error.

> gsl_sf_bessel_Kn_scaled(3.5, 1) and gsl_sf_bessel_Kn_scaled(3.1, 1) give the 
> same results (of
> gsl_sf_bessel_Kn_scaled(3, 1)), it seems it does not work for fractional order

There are calls to gsl_sf_bessel_Knu_e within specfunc/test_bessel.c
that look as if fractional orders are correctly computed.  Could you
suggest a few specific test cases that we might add to capture your
problems?

- Rhys