[Bug-gsl] bug in tdistinv.c and patch

[Brian Gough]

Jason Stover writes:
 > Have you decided what convention to use for returning the NAN?  I ask
 > because I'm writing routines to invert the beta and F distributions.
 > I also thought about writing some discrete cdf's. So if you know
 > what this convention is, I can follow it before sending the code along.
 > Have you decided what convention to use for returning the NAN?

Probably similar to the macro DOMAIN_ERROR or DOMAIN_ERROR_MSG in
specfun/error.h.

 > Out of curiosity: What's wrong with GSL_ERROR("blah blah",GSL_EINVAL)
 > or GSL_ERROR("blah blah),GSL_EDOM)? That seems to be what the coding
 > standards suggest.

Those only work if the return value of the function is a status code
(int).

-- 
Brian Gough