Re: [boost] move semantics

From: Howard Hinnant (hinnant_at_[hidden])
Date: 2002-03-13 17:43:20

On Wednesday, March 13, 2002, at 05:26  PM, brangdon_at_[hidden] 
wrote:

> That approach can also be used for relocate, if we have a no-throw 
> default
> constructor:
>
>       Base::Base(relocate from x)
>       {
>           x.notify_destruction();
>           listners_ = relocate from x.listners_;
>           notify_construction();
>       }

You're dancing on the edge of that mine field with 
x.notify_destruction().  Innocent looking maintenance could make that 
call operate on an x that's been partially destructed.

> Earlier (March 9th) you wrote:
>
>    relocate construct is not implementable for those classes that
>    have a base class or member classes that do not support move
>    construct.
>
> I think this is over-stated, that's all. Relocate construct can be
> implemented in terms of relocation of bases or members. It is not
> necessary to have move() to implement relocate().

Agreed.  With the lame duck concept, relocate construct is more 
approachable.

-Howard