On 13 Jun 2015 at 11:27, Edward Diener wrote:
For each feature test, we would need both (e.g.) __cpp_constexpr and __cpp_constexpr_and_implemented_correctly. The chance of just (e.g.) __cpp_constexpr being reliable enough is low, IMO, particularly for non-trivial feature use (e.g. Boost).
What you are saying is that we can use SD-6 but that we shouldn't trust that whoever implements it is doing it correctly.
Both GCC and clang have been lenient in letting you use some minor C++ 14 constexpr when in C++ 11 constexpr mode (albeit usually with a warning). VS2015 will not be lenient. You get guaranteed as much constexpr as is minimally required to implement C++ 11, anything past that is gravy. I've tried feeding test code into the unreleased full constexpr web MSVC compiler and so far for me at least it's working. I would like to assume someone in Microsoft has tried flipping BOOST_CONSTEXPR on in Boost with their latest release, but maybe STL can tell us more on that. One ballsy move might be for Boost to turn on BOOST_CONSTEXPR unilaterally for VS2015, and let the bug reports work their magic :) Niall -- ned Productions Limited Consulting http://www.nedproductions.biz/ http://ie.linkedin.com/in/nialldouglas/