undefined behavior from reinterpret_cast

[changkhothuychung]

https://github.com/bemanproject/inplace_vector/blob/main/include/beman/inplace_vector/inplace_vector.hpp#L364-L367
Suppose T is a user-defined type, wouldn't this reinterpret_cast result in an undefined behavior? A solution would be to use std::launder.

template <class __T, size_t __N> struct __aligned_storage2 {
  alignas(__T) byte __d[sizeof(__T) * __N];
  constexpr __T *__data(size_t __i) noexcept {
    __IV_EXPECT(__i < __N);
    return reinterpret_cast<__T *>(__d) + __i;
  }
  constexpr const __T *__data(size_t __i) const noexcept {
    __IV_EXPECT(__i < __N);
    return reinterpret_cast<const __T *>(__d) + __i;
  }
};

[wusatosi]

Sorry I am not familiar with std::launder, Can you elaborate on why this would result in undefined behavior?

[jbab]

There is no undefined behaviour here, https://en.cppreference.com/w/cpp/language/reinterpret_cast#Explanation, paragraph 5.
Basically we put Ts in and get Ts out. Also, the proper alignment is ensured.

Of course, it takes more than the quoted code to prove the point. The implementation of inplace_vector has to make sure that we really only put Ts in. This is easy to review, though.