I have newtypes around int, as well as classes that contains those newtype.

class IndexA {
    int i;
};
class A {
    IndexA index;
};

// this mimic exactly the hierarchy of `A`
struct IndexB {
    int i;
};
struct B {
    IndexB index;
};

Given that A and B share the exact same binary layout, does the following code contains undefined behavior or is it safe?

std::vector<A> vfoo {...};
std::vector<B> vbar {
    std::move(
        reinterpret_cast<std::vector<B>&>(vfoo)
    )
};

I would have liked to be able to do a zero copy move transformation from std::vector<A> to std::vector<B>. Is it possible?

In my code:

• A and B are really just a wrapper around int, so they have exactly the same binary representation.
• The struct A only contains a mix of IndexA, std::optional<IndexA>, std::variant<IndexA, other types like IndexA>, or std::vector<IndexA or std::optional<IndexA>, …>. I think that I can assume that they have the same representation.
• B is a strict copy of A, where every instances of IndexA are replaced by IndexB. The fields order is kept.

Source: Windows Questions C++