https://gcc.gnu.org/bugzilla/show_bug.cgi?id=103714

            Bug ID: 103714
           Summary: [11/12 Regression] name lookup in requires-clause
                    finds wrong thing
           Product: gcc
           Version: 11.2.1
            Status: UNCONFIRMED
          Keywords: accepts-invalid
          Severity: normal
          Priority: P3
         Component: c++
          Assignee: unassigned at gcc dot gnu.org
          Reporter: redi at gcc dot gnu.org
                CC: ppalka at gcc dot gnu.org
  Target Milestone: ---

This invalid code started to be accepted at r11-8144


template <int I>
struct derp
{
        constexpr inline static int i = I;

    template <int J>
    void f(derp<J> d) requires(d.i < i)
    {
        __builtin_printf("derp %zu < %zu\n", d.i, i);
        derp<I> d2;
        f(d2);
    }
};

int main()
{
    derp<9> d9;
    derp<8> d8;

    d9.f(d8);
}


If the requires clause is changed to (d.derp<J>::i < i) then it is correctly
rejected, but for some reason (d.i < i) is satisfied.

Reply via email to