[Bug libstdc++/102199] New: is_default_constructible incorrect for an inner type with NSDMI

["eyalroz1 at gmx dot com" <gcc-bugzilla@gcc.gnu.org>]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=102199

            Bug ID: 102199
           Summary: is_default_constructible incorrect for an inner type
                    with NSDMI
           Product: gcc
           Version: 12.0
            Status: UNCONFIRMED
          Severity: normal
          Priority: P3
         Component: libstdc++
          Assignee: unassigned at gcc dot gnu.org
          Reporter: eyalroz1 at gmx dot com
  Target Milestone: ---

Stackoverflow discussion: https://stackoverflow.com/q/69050558/1593077
Related LLVM bug: https://bugs.llvm.org/show_bug.cgi?id=38374
GodBolt: https://godbolt.org/z/snPf7Ks4W

Consider the following program:

#include <type_traits>

struct outer {
    struct inner {
        // inner() { }
        unsigned int x = 0;
    };
//    static_assert(std::is_default_constructible<inner>::value,
//          "not default ctorable - inside");
};

static_assert(std::is_default_constructible<outer::inner>::value,
    "not default ctorable - outside");

It compiles. But if we uncomment the first static_assert - it evaluates to
false. Mind you: Not because struct inner is incomplete; it is simply deemed to
not be default-constructible. But - it _is_ default constructible. And if we
add a method to struct outer which default-constructs an inner, it will work.

Also note that if we uncomment the explicit default ctor the definition of
struct inner, both asserts pass.


clang++ seems to exhibit this too (also with -stdlib=libc++). I'm not sure
whether this is an actual bug in the library, or whether the standard mandates
this in some freakish way, but - it's just wrong.