[Bug-gsl] bug in scaled modified spherical Bessel functions of first kind

[Roman Schmied]

GSL version: 1.6
hardware/OS: Mac OS 10.4.1 on a PowerPC G4 (1.2)
compiler: powerpc-apple-darwin8-gcc-4.0.0 (GCC) 4.0.0 20041026 (Apple  
Computer, Inc. build 4061)
compiler options: none other than "-lgsl"

The scaled modified spherical Bessel functions of first kind return  
wrong values for x=0. The correct results are 1 for L=0 and 0 for  
L>=1. In particular:

- gsl_sf_bessel_il_scaled(L, 0) returns zero for any value of L, even  
for L=0 where it should return 1.

- gsl_sf_bessel_il_scaled_array(Lmax, 0, y) returns "nan" values  
instead of {1,0,0,0,0,...}

example code: (compiled with "gcc -lgsl BesselTest.c")

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

int main() {
  double y[11];
  int i;
  printf("should be equal to 1: %g\n", gsl_sf_bessel_il_scaled(0, 0));
  gsl_sf_bessel_il_scaled_array(10, 0, y);
  printf("should be equal to 1: %g\n", y[0]);
  for(i=1; i<=10; i++)
    printf("should be equal to 0: %g\n", y[i]);
  return 0;
}