Re: Implement C++20 constexpr <algorithm>, , and <array>

["Ed Smith-Rowland via libstdc++" <libstdc++ at gcc dot gnu dot org>]

On 3/18/19 6:18 PM, Jonathan Wakely wrote:
> On 17/03/19 22:54 -0400, Ed Smith-Rowland via libstdc++ wrote:
> > Greetings,
> >
> > This patch implements C++20 p0202 - Add Constexpr Modifiers to 
> > Functions in <algorithm> and <utility> Headers
> >
> > and C++20 p1023 - constexpr comparison operators for std::array.
> >
> >
> > The patch is large because of the testsuite additions. Basically, the 
> > algorithms and the array comparison operators are all marked 
> > constexpr for C++20.  This has been built and tested on x86_64-linux.
>
> As discussed on IRC< the tests need to be run with -std=gnu++2a to
> ensure everything we test (which is not exhaustive, but is the best we
> can do) still works in C++2a mode.
>
> The indentation went bad here:
>
>   template<typename _FIter1, typename _FIter2, typename _BinaryPredicate>
> -    _FIter1
> +    _GLIBCXX20_CONSTEXPR
> +   _FIter1
>     find_first_of(_FIter1, _FIter1, _FIter2, _FIter2, _BinaryPredicate);
>
>
> I would have to look through old emails but I think there's a reason
> the function objects like _Iter_comp_to_val take their arguments by
> non-const reference.
The _Iter_comp_to_val functions need non-const reference o'loads to 
accommodate mutable functors.  These things now have both const and 
nonconst o'loads.
> I'm very surprised that none of the algos that dispatch to
> __builtin_memove need changes, because those optimizations won't work
> in constant expressions. I would expect to have to use
> std::is_constant_evaluated to disable the optimizations when used in
> constant expressions

I'm also actually surprised that __builtin_memove works also.  In other 
strange news, I had to write a std::is_constant_evaluated branch around 
__builtin_memcmp even though *that* *is* supposed to work for constexpr.

Ed