Re: patch for gamma test

[Tatsuro MATSUOKA]

Hello

--- Marco Atzeri  wrote:
> > double
> > xgamma (double x)
> > {
> > #if defined (HAVE_TGAMMA)
> >  return tgamma (x);
> 
> on cywin HAVE_TGAMMA=1
> 
> 
> > #else
> >  double result;
> > 
> >  if (xisnan (x))
> >   result = x;
> >  else if ((x <= 0 && D_NINT (x) == x) ||
> > xisinf (x))
> >   result = octave_Inf;
> >  else
> >   F77_XFCN (xdgamma, XDGAMMA, (x, result));
> > 
> >  return result;
> > #endif
> > }
> 
> If the slatec function is correct than only some 
> external tgamma are wrong.
> In any case it is better to report a failure on that platform
> and eventually report the failure upstream than
> as today where the fail is wrongly reported on the 
> correct platform.
> 
> I doubt it is worth to check gamma(-1) during configuration.

My writing perhaps is not enough.
The original slatec gamma function causes an error message and return nothing 
when x=0 and x=(minus
integer).

Octave itself implements that tgamma(0) and tgamma(mines interger) take 
Infinite using 

else if ((x <= 0 && D_NINT (x) == x) || xisinf (x))

when it uses the slatec routine. 

To be honest, I do not understand which is mathematically correct, 
gamma(-1)=NAN or gamma(-1)=inf.
(Excuse me, the above sentence is not correct English. ) 

http://en.wikipedia.org/wiki/Gamma_function
Γ(z) to a meromorphic function defined for all complex numbers z, except z = −n 
for integers n
&#8805; 0, where the function has simple poles with residue (
−1)n/n!.

I think that octave can implement gamma(minus implement) in the way gamma 
function to be platform
independent.

Regards

Tatsuro





--------------------------------------
Get the new Internet Explorer 8 optimized for Yahoo! JAPAN
http://pr.mail.yahoo.co.jp/ie8/