From: David Abrahams (david.abrahams_at_[hidden])
Date: 2002-07-13 06:36:29
From: "John Maddock" <john_maddock_at_[hidden]>
> > I'd like to suggest two things:
> > 1. A boost config macro we can use to detect this condition
> I'm not against that, but I worry that implementing the 64-bit integer
> specializations is necessarily compiler/standard lib specific.
Of course it is. I don't understand what you're concerned about though.
> option I guess would be to put this code within the standard lib config,
> however that would mean that <boost/config.hpp> would more or less always
> pull in <limits> which may not be such a good thing.
Not neccessarily. It depends if you're willing to forward-declare
> > 2. the implementation details used to implement std::numeric_limits<>
> > boost/detail/limits.hpp should stop using names reserved to the C++
> > implementation and instead be moved to namespace boost::detail in a
> > separate file. Then boost/limits.hpp could generate the missing
> > for the appropriate compilers.
> Yes, we shouldn't really put anything in std::
I have no problem with correcting vendor deficiencies by supplying missing
std:: components, but the implementation details really don't need to live
Boost list run by bdawes at acm.org, gregod at cs.rpi.edu, cpdaniel at pacbell.net, john at johnmaddock.co.uk