What happened to std::atomic<X>::value_type?

Question

According to this reference manual

For every std::atomic<X> (whether or not specialized), std::atomic<X>::value_type is X.

But if I try using such type I get a compilation error.

I've tried it with g++ 8.2.1:

$ g++ -std=c++11 test.cc 
test.cc: In function ‘int main()’:
test.cc:6:23: error: ‘value_type’ is not a member of ‘std::atomic<int>’
     std::atomic<int>::value_type x = 0;

And with clang 6.0.1

$ clang -std=c++11 test.cc 
test.cc:6:23: error: no type named 'value_type' in 'std::atomic<int>'
    std::atomic<int>::value_type x = 0;
~~~~~~~~~~~~~~~~~~^

The afore mentioned reference manual specifies also says that...

specification was substantially rewritten to resolve numerous issues in particular, member typedefs value_type and difference_type are added

The P0558R1 specification however seems not to forbid the existence of value_type.

Any idea what's going on?

Edit

A colleague of mine made me realize that P0558R1 is just a proposal.

Answer 1

You are explicitly using C++11. If we look at page 1119 of the last draft of the C++11 standard, there is no mention of value_type for std::atomic:

template <class T> struct atomic {
    bool is_lock_free() const volatile;
    bool is_lock_free() const;
    void store(T, memory_order = memory_order_seq_cst) volatile;
    void store(T, memory_order = memory_order_seq_cst);
    T load(memory_order = memory_order_seq_cst) const volatile;
    T load(memory_order = memory_order_seq_cst) const;
    operator T() const volatile;
    operator T() const;
    T exchange(T, memory_order = memory_order_seq_cst) volatile;
    T exchange(T, memory_order = memory_order_seq_cst);
    bool compare_exchange_weak(T&, T, memory_order, memory_order) volatile;
    bool compare_exchange_weak(T&, T, memory_order, memory_order);
    bool compare_exchange_strong(T&, T, memory_order, memory_order) volatile;
    bool compare_exchange_strong(T&, T, memory_order, memory_order);
    bool compare_exchange_weak(T&, T, memory_order = memory_order_seq_cst) volatile;
    bool compare_exchange_weak(T&, T, memory_order = memory_order_seq_cst);
    bool compare_exchange_strong(T&, T, memory_order = memory_order_seq_cst) volatile;
    bool compare_exchange_strong(T&, T, memory_order = memory_order_seq_cst);

    atomic() = default;
    constexpr atomic(T);
    atomic(const atomic&) = delete;
    atomic& operator=(const atomic&) = delete;
    atomic& operator=(const atomic&) volatile = delete;
    T operator=(T) volatile;
    T operator=(T);
};

It is similarly absent in the C++14 draft.

cppreference just fails to mention "since C++17" for value_type.

Edit: It has been pointed out that the addition of value_type was in the form of a defect report and should be applied retroactively to implementations of C++11. As such, cppreference is not actually wrong, the DR just has not been implemented in the given compiler versions.

Answer 2

Given the nature of P0558R1, I expected it to be eventually implemented retroactively in previous standard modes as a de facto defect report and documented it as such. The paper performs major surgeries on the non-member function templates that depend on the presence of these typedefs. Similar surgeries have generally been applied retroactively by implementers. As a data point, the only major implementation known to have fully implemented P0558R1 (MSVC) does so regardless of language version.

cppreference's target is a hypothetical complete and correct implementation of each C++ standard plus all defect reports and clarifications applicable to that standard.