Re: GSL on OS/2

[Brian Gough <bjg at network-theory dot co dot uk>]

Henry Sobotka writes:
 > In complex, I got the attached failures at all optimization levels I
 > tried; they only went away when I reverted to linkage with fdlibm for
 > acosh and atanh. OTOH, in specfunc I got the attached new failures with
 > the fdlibm versions, and they went away when I used the ones in gslsys;
 > just the three legendre failures remain.

I think this could be the result of differences between your default
system math library and fdlibm because the complex functions which
fail (e.g. gsl_complex_sin) don't use any calls to the inverse
hyperbolic functions.  They do use calls to the standard functions
sin, cos, tan, cosh, sinh and tanh though.  The problems in specfunc
may be connected with this too.