
The Euler gamma function code is somewhat limited in accuracy,
especially for negative values that are close to negative integers.
For values within +- 0.1 of a negative integer, the approximation can
only yield 5-6 figure accuracy.  I'm working on how to fix this.