Nikita Kniazev

I was planning to add more tests to have better coverage, but now I do not know when I will be able to continue the PR. As you may know,...

It could be done in X3 with semantic actions + context variables: ```c++ #include #include #include int main() { namespace x3 = boost::spirit::x3; auto s = "5: 1 2 3...

Semantic actions on a rule definition inhibit automatic attribute propagation. Define your rule placeholder as `x3::rule const refid` to force attribute propagation.

Oh god, there is inconsistency even in Qi itself. ```cpp #include #include #include #include #include template void test(F f) { std::vector v; char s[] = "1:a\n2:b\n3:c\n---\n5:e\n---\n---\n7:h\n---"; f(std::begin(s), std::end(s), v); for...

This is completely mind-blowing: ```cpp #include #include #include #include template void test(F f) { std::vector v; char s[] = "1:a\n2:b\n3:c\n---\n5:e\n---\n---\n7:h\n---"; f(std::begin(s), std::end(s), v); for (auto [x, y] : v) std::cout...

https://svn.boost.org/trac10/ticket/6444 looks like the same problem.

I am reopening it and retargeting to Qi

The regression in MSVC was confirmed and fixed. It is still unclear when it was introduced and which release will have the fix. I am leaving the issue opened to...

Repeated types not being unified is much older thing #610, so I doubt it is the cause. Reversed order was reported in #721 though matching the order does not help...

If the goal is to close the trac tickets, why not just solve the problem by fixing the bugs? Moreover, some of the problems reported and still opened on trac...