function that takes only literal integers

Question

I\'m looking for a way to simulate certain overloaded GCC built-ins in C++. The built-ins are similar to these:

__builtin_foo(char *a, signed int b);
__buil         


        

Answer 1

Here is a portable C++11 solution so that your function (macro actually, sorry) accepts only integer literals and triggers a compile-time error instead:

constexpr int operator "" _literal(unsigned long long i)
{
    return i;
}

#define m_builtin_foo(integer) builtin_foo(integer ## _literal)

User-defined literals accept only literals (hence their name). Therefore, if you make your macro paste a user-defined literal to what is passed to it, it should only accept the literals accepted by the corresponding user-defined literal.

However, that's rather ugly and I guess that could be error-prone. Not sure how though.

---

@MichaelAnderson pointed in the comments that m_builtin_foo(i+1) would still work. That's right. @Angew suggested wrapping the return type of _literal in a wrapper incompatible with integer arithmetic and adding explicit conversions. He was also right, here is the more complete solution:

struct wrapper
{
    constexpr wrapper(int n):
        value{n}
    {}

    explicit constexpr operator int() const
    {
        return value;
    }

    int value;
};

constexpr wrapper operator "" _literal(unsigned long long i)
{
    return { int(i) };
}

#define m_builtin_foo(integer) builtin_foo(int(integer ## _literal))