[Help-gsl] GSL test programs

[RK]

I have completed compiling the GSL 1.7 source files into static libraries as
well as linking the accompanying test programs under WinXP using MSVC 2003.
I used Jerry St. Clair's GSL 1.6 port (particularly his config.h) as
guidance for my efforts with version 1.7. All the tests run very well except
for a few instances where the tolerances need to be adjusted very slightly
in terms of absolute magnitude (e.g. 1.1e-14 instead of 1e-14).

My question: why is there a difference in the floating point math between
Win and Unix machines? These tests must run just fine on Unix machines with
the provided tolerances. Are there some machine dependent constants in GSL?

Thanks to Brian Gough and all his colleagues for this amazing effort!

Thanks,
RK