Re: strange behaviour with gsl_cdf_fdist_Q()

[Rajarshi Guha <rajarshi at presidency dot com>]

On Wed, 2003-09-03 at 13:04, Rajarshi Guha wrote:
> Hi,
>   more problems with gsl_cdf_fdist_Q() and very strange ones at that!

I've been investigating this problem a little more. 

I took the code from fdist.c and beta_inc.c and basically renamed the
functions and compiled the code directly into my binary. Now running my
program gives different values of P inside the function and absurd
values of the return value after it has returned from the function!

With the original gsl_* function from the GSL libs:

x = 3.145724 nu1 = 7.666667 nu2 = 3.000000 P = 0.044555
return value = 32.0000

When I call the renamed funtion in my binary I get

x = 3.145724 nu1 = 7.666667 nu2 = 3.000000 P = 0.000981
return value = 521874777.0000

Assuming that there is a problem somewhere I would have expected that
whether I called the code from the library or my own program, the
answers would be consistent.

Is there some detail of the GSL internals that might lead to this
difference? I'm still assuming the problem lies in  my code - some form
of memory corruption somewhere, b ut as soon as I get rid of calls to
gsl_cdf_fdist_Q(), the program runs fine and produces the expected
results.

Thanks,

-------------------------------------------------------------------
Rajarshi Guha <rajarshi@presidency.com> <http://jijo.cjb.net>
GPG Fingerprint: 0CCA 8EE2 2EEB 25E2 AB04 06F7 1BB9 E634 9B87 56EE
-------------------------------------------------------------------
A committee is a group that keeps the minutes and loses hours.
-- Milton Berle