I have been evaluating Laguerre polynomials for large n and alpha = 0 
for largish n on several platforms including i686-cygwin, x86_64-linux, 
powerpc-apple-darwin7.9.0 an it seems that I get bad results on all of them.

I am comparing with Mathematica.  (I know, non-free and otherwise 
annoying.)  They use high-precision arithmetic and I tend to trust their 
results.

I believe the error comes from switching to a hypergeometric expression 
rather than sticking with recursion at alpha=0.
I have found that staying with recursion works just fine at alpha = 0 
for n = 50 or 100.

The special function testsuite is clean on all platforms when I make the 
one-character change.

Change line 288 of laguerre.c
from:
  else if(a > 0.0 || (x > 0.0 && a < -n-1)) {
to:
  else if(a >= 0.0 || (x > 0.0 && a < -n-1)) {


Ed Smith-Rowland