RE: [boost] rational.hpp and Borland C++

From: Moore, Paul (paul.moore_at_[hidden])
Date: 2000-04-27 06:40:57

From: John Maddock [mailto:John_Maddock_at_[hidden]]
> I have some C++ Builder specific fixes for rational.hpp, I've 
> attached the updated code to this mail, the specific fixes are:

Thanks for the fixes, I have just downloaded the free Borland C++ compiler
and was intending to try it and see what the results were. I also now have
the mingw GCC compiler, and if I can ever get it to believe that the std::
namespace exists (problems with the supplied standard library mean that it's
not as simple as -fhonor-std :-() I intend to try with that, too.

I haven't had much time for C++ coding just recently, I'll try to pull
together the various changes which are outstanding and get a new version out
soonish...

> BTW shouldn't gcd either be in its own implementation namespace
> or moved out to utility.hpp and documented - this seems like the
> kind of function that several libraries might want to use.

I'm happy for gcd to go into utility.hpp - does anyone have any problems
with this?

> In the friend function abs:  bcb has problems with friend 
> functions that aren't implemented class-inline, likewise friend
> functions can't access the class template parameters, but can
> access member typedefs.  So either IntType needs to be typedef'ed
> in the class declaration and the implementation moved class inline,
> or abs can be implemented as a non-friend non-member function,
> the latter is the approach I've taken here.

abs() was the big problem for MSVC, too. I don't know if the version there
at the moment could be said to really work in any case. I'll look at your
version and think some more. (The big problem with MSVC was that I could
never work out what MSVC behaviour was buggy - your analysis might help me
understand better...)

Paul.