Re: [boost] [1.53.0 beta] [type_traits] is_POD

Subject: Re: [boost] [1.53.0 beta] [type_traits] is_POD
From: Jeffrey Lee Hellrung, Jr. (jeffrey.hellrung_at_[hidden])
Date: 2013-01-28 03:12:03

On Sun, Jan 27, 2013 at 3:38 PM, Cromwell Enage <sponage_at_[hidden]> wrote:

> John Maddock wrote:
>
> > Cromwell Enage wrote:
> > > I found out that Boost.TypeTraits defines a metafunction called is_POD
> that is essentially a
> > > template alias of is_pod, but it isn't documented. In particular, the
> documentation needs to mention
> > > that is_POD must be overridden in the same cases that is_pod must be
> overridden.
> >
> > It's depricated, it shouldn't be used in new code, and isn't documented
> for a reason.  There may just
> > possibly be some (very) old code still using it, but I'll add a note to
> the header to mention that it's
> > depricated.
>
>
> In <boost/iterator/iterator_facade.hpp>, the
> boost::detail::use_operator_brackets_proxy metafunction is defined in terms
> of boost::is_POD.  I don't know if the concern that the associated comment
> mentions can be addressed by replacing with boost::has_trivial_copy and/or
> boost::has_nothrow_copy.
>

Sounds like something I can take a pass at, both replacing is_POD with
is_pod and determining whether has_trivial_copy is sufficient (I'm guessing
has_nothrow_copy is probably not desirable; without looking at the code, I
seem to remember the point was to try to detect if the value_type was cheap
enough to copy).

- Jeff