`extract_int` overflow difference between Qi and X3

[Kojoley]

#include <boost/spirit/home/x3/support/numeric_utils/extract_int.hpp>
#include <boost/spirit/home/qi/numeric/numeric_utils.hpp>
#include <iostream>

int main()
{
    namespace x3 = boost::spirit::x3;
    namespace qi = boost::spirit::qi;

    char const* s = "70000", * it = s, * const end = s + std::strlen(s);
    short x;
    std::cout << std::boolalpha;
    std::cout << "qi:\n";
    x = -1; it = s; std::cout << qi::extract_int<short, 10, 1, -1>::call(s, end, x) << ' ' << x << '\n';
    x = -1; it = s; std::cout << qi::extract_int<short, 10, 1, 5>::call(s, end, x) << ' ' << x << '\n';
    std::cout << "x3:\n";
    x = -1; it = s; std::cout << x3::extract_int<short, 10, 1, -1>::call(s, end, x) << ' ' << x << '\n';
    x = -1; it = s; std::cout << x3::extract_int<short, 10, 1, 5>::call(s, end, x) << ' ' << x << '\n';
}

Output:

qi:
false 7000
false 7000
x3:
false -1
true 7000

https://wandbox.org/permlink/TnmsMtZcTMvFAfgr

Is this intended? From my point of view all the parsers must fail and even output variable be untouched (false -1 on all cases).

[djowel]

I think it's documented ( somewhere :-) ) that the attribute value is undefined on a failed parse, for the sake of efficiency. OTOH, if it won't affect efficiency, then the qi behavior is preferred.

[Kojoley]

I think it's documented ( somewhere :-) ) that the attribute value is undefined on a failed parse, for the sake of efficiency.

That's understandable and I fully agree.

But my question was why X3 succeeds parsing with MaxDigits=5? Is this intended or a regression?

[djowel]

Oh duh. Looks like a regression. Not sure yet.