Re: [boost] [move] problem with msvc and BOOST_MOVABLE_BUT_NOT_COPYABLE

Subject: Re: [boost] [move] problem with msvc and BOOST_MOVABLE_BUT_NOT_COPYABLE
From: Thomas Klimpel (Thomas.Klimpel_at_[hidden])
Date: 2010-05-28 12:29:13

Ilya Sokolov wrote:
> Thomas Klimpel wrote:
> > It seems my answer to Oliver wasn't clear enough. This is not a
> > problem of Boost.Move,
> 
> Yes, it is (arguably). The problem is that error is reported too late.

>From my point of view, there is not much difference between link-time and compile-time. An error at run-time would be too late.
 
> > but a simple programmer error in "X X::create()" that would have a
> > significant performance penalty even if the compiler/linker
> > would accept the code.
> 
> There is no implementation of copy ctor in the original example.
> How compiler/linker could accept the code? What "performance penalty"
> you are talking about?

I'm talking about the BOOST_COPYABLE_AND_MOVABLE case. Most compilers won't do RVO for code written like "X X::create()".

> Probably, but that is not important for me.
>  I like the following benefits:
> - one macro less
> - uniformity between boost::movable_only and boost::noncopyable

You're probably right that a macro "BOOST_MOVABLE" just containing the conversion operators would be enough for the non-optimized mode. The BOOST_MOVABLE_BUT_NOT_COPYABLE case would derive privately from boost::movable_only and implement move constructor and move assignment operator, the BOOST_COPYABLE_AND_MOVABLE would just implement the correct move constructor and assignment operator.

Regards,
Thomas