Re: [boost] Interval library, bjam and Borland CC

From: fernando_cacciola_at_[hidden]
Date: 2002-10-17 15:10:30

Hi Guillaume,

I just now managed to be able to look at the interval library.

I am unable to use bjam with my BCB+Boost installation... I've given up...
but I manually built test_execution_monitor.lib and created a BCB project
for each interval test (all taken from a fresh cvs update from boost and
checkout from boost-sandbox, late yesterday)

So...

In arith2.hpp, pow<> function, I moved the using declaration (using
interval_lib::detail::pow_aux),
to line 72, right before the first 'if':

template<class T, class Policies> inline
interval<T, Policies> pow(const interval<T, Policies>& x, int pwr)
{
  using interval_lib::detail::pow_aux;  // <<<<<<<<<<<<<<<<<<<<<<<<<

  if (interval_lib::detail::test_input(x))
    return interval<T, Policies>::empty();
...

Apparently, since this first 'if' is true at compile-time in some cases, the
appearance of the using declaration in the unreachable section of the
function's body jokes this compiler.

I also found a mismatch between the functions 'equal' and 'overlap' as
forward declared in 'interval.hpp' and defined in 'utility.hpp'. I've fixed
that:

  // template<class T, class Traits>
  //  bool equal(const interval<T,Traits>& x, const interval<T,Traits>& y);

  template<class T, class Policies1, class Policies2>
    bool equal(const interval<T, Policies1>& x, const interval<T,
Policies2>& y) ;

  // template<class T, class Traits>
  //   bool overlap(const interval<T,Traits>& x, const interval<T,Traits>&
y);

  template<class T, class Policies1, class Policies2>
    bool overlap(const interval<T, Policies1>& x,
                 const interval<T, Policies2>& y) ;

Lastly, the function widen<>, in 'utility.hpp':

template<class T, class Policies> inline
interval<T, Policies> widen(const interval<T, Policies>& x, const T& v) ;

is not forward declared in 'interval.hpp'; but besides that, the code that
uses it in 'io.hpp':

line 58:  interval<T, Policies> r_wide = widen(r, eps / 2.0);

cannot bound 'eps / 2.0' which is of type double with 'const T&', so a
specialization is not found.
The fix requires a simple cast:

  interval<T, Policies> r_wide = widen(r, static_cast<T>(eps / 2.0));

With the above changes, ALL 11 tests compiled and run successfully both in
debug and release modes with bcc5.5.1

Regards,

Fernando Cacciola