Boost logo

Boost :

From: Jesse Jones (jejones_at_[hidden])
Date: 2000-10-19 20:24:37


>Jesse Jones wrote:
>
> > ... The standard says:
> >
> >1) There are four signed integral types: signed char, short int, int, and
> >long int. [3.9.1]
> >
> >2) Each cv-unqualified type has three cv-qualified "versions" of its
>type:
> >a const version, a volatile version, and a const volatile version.
>However
> >these are distinct types. [3.9.3]
> >
> >Miller seems to think this can be read as allowing const int to be a
>signed
> >integral type, but I still don't buy it.
>
>He wasn't just reading those sentences, but also looking at the other
>material about value representation.

If you approach the text without preconceptions I think you have to say
that const int's are not signed integral types. Miller knows what he wants
the text to mean and he seems to be using the value representation
discussion to bolster this viewpoint, but I don't think this is justified.

>Anyhow, I'll press on with a defect report.
>
>But the result will almost certainly be that const int is a signed integral
>type, and so forth.

Good deal. Does this mean that boost:: is_standard_signed_integral should
be renamed is_signed_integral? :-)

  -- Jesse


Boost list run by bdawes at acm.org, gregod at cs.rpi.edu, cpdaniel at pacbell.net, john at johnmaddock.co.uk