Re: [boost] Re: [review results] hash functions accepted

From: Jonathan Wakely (cow_at_[hidden])
Date: 2005-03-24 05:10:10

On Wed, Mar 23, 2005 at 10:01:21AM -0800, Eric Niebler wrote:

> 
> Jonathan Wakely wrote:
> >On Wed, Mar 23, 2005 at 04:06:09PM +0000, Joaquin M Lopez Munoz wrote:
> >
> >
> >>"I can't think of any way that this extension could
> >>break a conforming program, considering that users
> >>are not permitted to forward-declare standard
> >>library components"
> >>
> >>Is this really so? Why cannot I forward declare a
> >>stdlib component?
> >
> >
> >Opening namespace std for anything except specialisations of standard
> >library templates (with user-defined types as parameters) is illegal.
> >
> 
> 
> OK, but that has never stopped us (Boost) from doing it before. 
> config/suffix.hpp defines std::min and std::max for non-compliant std 
> libs. lambda/detail/operator_return_type_traits.hpp forward-declares 
> std::complex<>. I don't see anything particularly wrong with that, 
> except that perhaps forward-declarations of std types should be moved 
> into the config sub-project to better handle the variety in different 
> std libs.

I was never suggesting Boost shouldn't do it, (I also pointed out that
it's done already for BOOST_NO_STDC_NAMESPACE,) I was just giving a
likely rationale for the comment in the DR, which is what I thought
Joaquin was asking about.  The fact that you can get away with it safely
in practice doesn't mean it's legal.

I think I'll quit while I'm behind :)

jon