Re: c++11

[coypu%SDF.ORG@localhost]

On Tue, Nov 08, 2016 at 07:28:42AM +0000, David Holland wrote:
> On Fri, Nov 04, 2016 at 11:32:03PM +0000, Taylor R Campbell wrote:
>  >    On Fri, Nov 04, 2016 at 07:43:43PM +0100, Joerg Sonnenberger wrote:
>  >    > The real problem is that it doesn't solve the problem of mixing runtime
>  >    > libraries. In the GCC of GCC 5.1+, you even end up with broken ABI
>  >    > differences when using -std=c++11 vs not.
>  >    
>  >    Do you have a suggestion how to handle this then?
>  > 
>  > Sounds like the problem is if two libraries use c++11, and one of them
>  > is happy with GCC 4.9 but the other needs GCC 5.1, pkgsrc will build
>  > the first with GCC 4.9 and the second with GCC 5.1 and then they will
>  > fail to work together in the same address space.
>  > 
>  > Is that right?  If so, it sounds like having logic, as we do now, that
>  > works per-package to choose a C++ compiler is no good, and instead we
>  > need to choose a C++ compiler for an entire set of packages at once to
>  > satisfy them all.
> 
> This problem also exists with native vs. pkgsrc versions of X
> libraries.
> 
> That is to say, it would be good to come up with a scheme for dealing
> with this kind of nonlocal reasoning...
> 
> -- 
> David A. Holland
> dholland%netbsd.org@localhost


I think this issue is secondary to haveing a proper framework to say 'this
package requires C++11', and more like an inevitabile issue of allowing
multiple GCC versions from pkgsrc.