[Bug-gsl] nan bug in bessel_Jnu ?

[Benjamin von Ardenne]

Hello everyone,

I have this particular problem where I need

gsl_sf_bessel_Jnu(0.5, 3/2 * pi)

and I tested it with the following code:


#include <stdio.h>
#include <math.h>
#include <gsl/gsl_sf_bessel.h>

int
main (void)
{
  double y1 = gsl_sf_bessel_Jnu(0.5, 3.0/2.0 * M_PI+1e-15);
  double y2 = gsl_sf_bessel_Jnu(0.5, 3.0/2.0 * M_PI);

  printf ("%e vs. %e\n", y1, y2);
  return 0;
}

Output:

-3.675526e-01 vs. nan

MATLAB e.g. does return the correct value for 3/2 * pi. Is this a bug in
the implementation of the function? It seems so and I have not found
documentation about irregularities at 3/2 PI.

Cheers,
Ben