On 8/5/21 12:52 AM, John Maddock via Boost wrote:
On 05/08/2021 01:53, Andrey Semashev via Boost wrote:
On 8/5/21 2:14 AM, Gavin Lambert via Boost wrote:
On 5/08/2021 7:03 am, Andrey Semashev wrote:
I might add that including <type_traits> just to test got a feature macro is not good.
No need for that - we already include <version> (which is lightweight) and all that's needed in this case. Both <version> and is_constant_evaluated are C++20, so you're very unlikely to have the latter without the former.
OK - it's all crystal clear now ... NOT. I don't see a consensus here. I'm liking the idea of implementing boost::is_constant_evaluated() back to C++11 which a default value of false if there is not compiler support. Along with a boost config style feature macro so we'd have a unified model. Personally, I've never found compile times to be a big problem. I'm not saying that there is no compile time cost - just that it doesn't seem so much that it's been worth worrying about. Robert Ramey