Log-gamma-func
evaluates .x > 0. It is based on Chebyshev expansions.
If x is too large there is a danger of setting overflow so the function evaluation must fail.
For more information please review the s14abc function in the NAG document.