T *p = ::operator new(sizeof(T));
And then
new (p) T;
I wonder how the syntax is working as the placement new is defined as:
void* operator new(std::size_t, void*)
The second argument void* seems point to the constructor to the object, so how cpp connect the new (p) T to the void*？
In gcc libsupc++ the implement is simply:
// Default placement versions of operator new.
_GLIBCXX_NODISCARD inline void* operator new(std::size_t, void* __p) _GLIBCXX_USE_NOEXCEPT
{ return __p; }
_GLIBCXX_NODISCARD inline void* operator new[](std::size_t, void* __p) _GLIBCXX_USE_NOEXCEPT
{ return __p; }
So is it compiler responsible for translate the syntax new(*T) T to the calling of a placement operator new ?
>Solution :
Is it compiler’s responsibility to translate the syntax new(*T) T to
the calling of a placement operator new?
Yes, exactly.
Placement new can not be implemented as a library function, it needs special handling from the compiler anyway. The actual implementation that you saw in the libsupc++ source code can be misleading here, as most of the crucial properties of a placement new call are not reflected by that implementation.