Subject: Re: [SG10] First conceivably-complete draft recommendation
From: John Spicer (jhs_at_[hidden])
Date: 2013-06-05 13:30:37
Regarding the conditionally-supported constructs, for 7.5p2 the document says "Probably no single macro would make sense for this.". I agree with this, but I wonder if the same reasoning applies to linkage specifications in 14p4? In both cases, it would only let you know that some unspecified linkage other than C and C++.
We could probably just not provide a macro, but if we wanted to give guidance perhaps something like __cond_has_XYZ_linkage (where XYZ would be kind of linkage supported (or the closest version usable in a macro name)).
A similar issue exists for 7.6.1p3 for attribute namespaces. Knowing that the implementation supports some unspecified set of attribute namespaces does not seem very useful.
On Jun 4, 2013, at 7:38 PM, "Nelson, Clark" <clark.nelson_at_[hidden]> wrote:
> I have updated the draft based on feedback from (and following) the teleconference. (At this point, I have gotten author feedback for only about 1/3 of the C++14 papers.)
> The introductory text now covers all of the important points I have considered and we have discussed (as far as I remember). Please review it carefully; if it says anything it shouldn't, or doesn't say anything it should, please let me know. In particular, I have tried to avoid my own personal bias in talking about WG21's pace and practices, but I'd appreciate opinions about where I may have failed. I've also explicitly mentioned the more vexing problems we're still facing. There should probably be more examples; help is solicited in that department.
> The "Recommendations" section is intended to be able to function as a separate document if necessary.
> In the table for C++14 features, the main changes are that I have included alternate name recommendations I have received from authors (and others); we still need to pick the ones we think best. It was also pointed out that we should probably pick singular or plural forms, and use them consistently. (Regulars in Core know that use of plural instead of singular forms is a point about which I'm generally picky. This time, I have no recourse but to plead distraction.)
> I turned Tom's list of C++11 features into a table, and added initial proposals for macro names -- in some cases including alternatives. (I'll add the links to the papers later.) In the long run, we probably need to consider Boost's complete list of configuration flags; if so, I would definitely need help with that.
> I also surveyed the working paper for conditionally-supported behavior, and added a table with my findings. Obviously, this is a topic that's going to need further discussion.
> Clark Nelson Vice chair, PL22.16 (ANSI C++ standard committee)
> Intel Corporation Chair, SG10 (WG21 study group for C++ feature-testing)
> clark.nelson_at_[hidden] Chair, CPLEX (WG14 study group for C parallel language extensions)
> Features mailing list
SG10 list run by firstname.lastname@example.org