testsuite allocators patch

[Jonathan Wakely]

On 23/07/14 22:33 +0200, François Dumont wrote:
>    I have a small question regarding some code next to the one I am 
>modifying in this patch. I can see lines like:
>
>      propagating_allocator() noexcept = default;
>
>    When using a default implementation shouldn't we let the compiler 
>decide if it should be noexcept or not depending on the member fields 
>or base class default constructors ?

Stating it explicitly means you get an error if the default
implementation is not noexcept. That can be useful, to ensure you
don't silently start getting a throwing constructor by mistake because
of a change to a base class.

I'm not sure if I added the noexcept above, but if I did that might
have been what I was intending it to do. I don't remember.

I'll review the rest of the patch ASAP. Did you test it with no other
changes in your tree, and run the entire testsuite?