Re: [boost] [core] Breaking change to boost::ref in 1.56

Subject: Re: [boost] [core] Breaking change to boost::ref in 1.56
From: Andrey Semashev (andrey.semashev_at_[hidden])
Date: 2014-07-13 06:16:26

On Sunday 13 July 2014 12:48:30 Peter Dimov wrote:
> Andrey Semashev wrote:
> > Perhaps Eric could provide more context on why Boost.Proto needs recursive
> > reference wrappers? Frankly, the requirement looks strange to me.
> 
> Generic code that does
> 
>     boost::reference_wrapper<T> r = boost::ref( t );
> 
> is broken when t is a reference_wrapper.

Ok, but I was also interested in why this is needed by Boost.Proto. I.e. 
whether a reference to a reference to a reference... is the intended behavior 
of the library and whether it's actually desired to collapse the recursion.

I understand that the library doesn't do that currently and we might need to 
revert the change to boost::ref temporarily to fix it for 1.56, but the long 
term plan might be to actually fix Boost.Proto. We might declare the old (non-
collapsing) behavior deprecated for 1.56 in the release notes and provide some 
wrap_reference<T> trait to make the above code work:

  typename boost::wrap_reference<T>::type r = boost::ref(t);

The trait would do a non-collapsing wrapping for 1.56 and collapse references 
for future releases.

> > I'm not strongly opposed to reverting, but the change looks justified
> > enough to me. Allowing dangling reference wrappers is surely not the
> > correct behavior.
> 
> Dangling reference wrappers are prevented by a separate and independent
> change
> 
> https://github.com/boostorg/core/commit/45f7564db29a3bafa5dfd8c41396843493d1
> 378a
> 
> that is (I think) not affected by the removal of the collapsing overloads.

Ok, good. I take it that the motivating example:

  boost::ref(boost::ref(t));

should not compile then?