On Sat, 7 Mar 2020 at 11:00, Andrew Haley <aph@xxxxxxxxxx> wrote: > > On 3/6/20 5:45 PM, Jonathan Wakely wrote: > > On Fri, 6 Mar 2020 at 16:17, Chris Hall <gcc@xxxxxxx> wrote: > > > >> FWIW: clang gets this right, and where the Standard says a parameter > >> must be an _Atomic(foo_t)* [for a standard atomic_xxx()], clang rejects > >> foo_t* arguments. > > > > It's not clear to me that C actually requires it to be rejected, or if > > it's just undefined (in which case GCC's decision to accept it and do > > the obvious thing is OK). > > Except in pedantic mode. If the standard doesn't actually forbid it (which I don't think it does) then it's up to the implementation if and when it diagnoses it. Doing so for -pedantic might be reasonable though. > I remember there was a move to actually permit > this in C++: > > http://www.open-std.org/jtc1/sc22/wg21/docs/papers/2014/n4013.html > > This seems obviously right to me... Yeah. The feature is in C++20, via the std::atomic_ref<T> template. You wrap an ordinary non-atomic variable in an atomic_ref and can perform atomic operations on it. For this to work, the implementation has to be able to accept non-atomic variables in its __atomic_xxx functions (though not necessarily in the public atomic_xxx ones).
