On 04/09/2015 08:13 PM, Nelson, Clark wrote:
> I want to publish a revision of SD-6 in the pre-meeting mailing (deadline
> tomorrow). I'm not knocking myself out to achieve any particular state;
> I just want it to be as complete as I can reasonably achieve in the
> available time.
>
> But I have made some changes on which people might like to comment, so I
> have posted a new revision on the wiki:
>
> http://wiki.edg.com/twiki/pub/Wg21lenexa/SG10/sd-6.html
>
> Over the weekend, I'll be traveling to the WG14 meeting, so I want to
> submit this for the mailing in just about 24 hours. So if you want to
> review this before it is published, please don't delay.
>
> The most important change is that I have added examples for some library
> features.
>
> Now, the fact that I have dared even to attempt to write examples for C++
> library features should give anyone who knows me only moderately well some
> idea of how serious I am about considering and recording the rationale for
> all of our recommendations going forward.
>
> A big part of my desire to publish this is to let the proposers of the new
> features review the rationale/example. So I'm not depending on anybody in
> SG10 reviewing them. But if you have some time -- and especially if you're
> up for a laugh -- you might want to look them over. (The examples for N3911
> and N4258 are by Walter and Pablo, not me, so don't look to them for humor
> value.)
>
> The other significant changes include:
>
> Deleting the names of people from whom I am hoping to get a contribution
> of example or rationale. :-)
>
> I think that we really should distinguish cases where we change the
> value of a macro from those where we introduce a new macro. I have
> introduced a convention to do that, which is explained in the document.
> For examples, see N4267, N3928 and N3652. I'm interested in feedback on
> whether what I have done is enough (or too much).
>
> I have decided that if a name is workable in the library, it is
> unambiguous enough for a macro. So for N4279 and N4258, I have added new
> proposals about the spelling of the macro name, to better match the (or
> at least some) library name that it represents. Conversely, for N4169 and
> N3911, I have retracted proposals that I made before, of longer names to
> better disambiguate.
>
> Thanks in advance for all feedback.
>
> --
> Clark Nelson Chair, PL22.16 (ANSI C++ standard committee)
> Intel Corporation Chair, SG10 (C++ SG for feature-testing)
> clark.nelson_at_[hidden] Chair, CPLEX (C SG for parallel language extensions)
> _______________________________________________
> Features mailing list
> Features_at_[hidden]
> http://www.open-std.org/mailman/listinfo/features
>
In addition to the example for enumerator attributes add an example for
namespace attributes:

namespace tr1
#if __cpp_namespace_attributes
[[deprecated("These utilities have been moved to namespace std according
to C++11 and TR29124")]]
#endif
{
...
}

My rationale for two feature tests is that the semantics of the
particular attribute deprecated hasn't changed - it gives the same warning.
The scope of the attribute has widened but this is due to an enhancement
to the C++11 attribute feature itself.
This feature could in principle be used for other attributes.
In the case of g++ it is used that way (although in this case the
attribute if also file scoped) but the principle remains.
At least in my mind. ;-)

Ed