Re: Solaris Forte 6u.1 port: limits_test.cpp

From: Gennadiy E. Rozental (rogeeff_at_[hidden])
Date: 2001-09-07 14:23:23

I thought that it is just some problem with SunPro template 
instantiation. This test has a lot of different checks, so it could 
be worth while to know how numerics_limit implemented nor how 
templates are.

Gennadiy.

--- In boost_at_y..., Jens Maurer <Jens.Maurer_at_g...> wrote:
> "Gennadiy E. Rozental" wrote:
> > cvs diff limits_test.cpp
> > Index: limits_test.cpp
> > 
===================================================================
> > RCS file: /cvsroot/boost/boost/libs/config/limits_test.cpp,v
> > retrieving revision 1.7
> > diff -r1.7 limits_test.cpp
> > 69,71c69,71
> > <   const T infinity = lim::infinity();
> > <   const T qnan = lim::quiet_NaN();
> > <   const T snan = lim::signaling_NaN();
> > ---
> > >   const T infinity = (T)lim::infinity();
> > >   const T qnan = (T)lim::quiet_NaN();
> > >   const T snan = (T)lim::signaling_NaN();
> 
> Rejected.  limits_test.cpp only exists to test how usable
> the provided std::numeric_limits<> is.  If, like the Sun
> compiler, it produces an internal compiler error, that's
> a valid test result, to show off to the world.
> 
> And it means that std::numeric_limits<> is probably not
> useful on that compiler.
> 
> Jens Maurer