28 May
2010
28 May
'10
9:51 p.m.
On 28/05/2010 23:30, Ilya Sokolov wrote:
Ion Gaztañaga wrote: What do you think?
It's an option, I just was trying to say that we get a similar guarantee that a simple rvalue reference based implementation could offer. If we should add more guarantees because the compiler diagnostics are not good enough, I'm open to suggestions. Adding a base class is not something I like because in some C++03 (Msvc I think, empty base classes are not optimized with multiple inheritance), although maybe the benefits outweigh the problems. Thanks for your suggestions, Ion