[C++ PATCH] Fix cxx_eval_store_expression (PR c++/89336)

[Jakub Jelinek]

On Mon, Feb 18, 2019 at 03:01:14PM -1000, Jason Merrill wrote:
> But it's not clear to me that the standard actually allows this.  I don't
> think changing the active member of a union in the mem-initializer for
> another member is reasonable.

There is in [expr.const]/2:

an lvalue-to-rvalue conversion (7.1) that is applied to a glvalue that refers to a non-active member of a
union or a subobject thereof;

an assignment expression (8.18) or invocation of an assignment operator (15.8) that would change the
active member of a union;

in C++17 it seems, so maybe my union testcases are accepts-invalid.
This has been introduced in P0137R1 and removed again in P1330R0.  Does that
mean e.g. following is valid in C++14, invalid in C++17 and valid again in
C++20?  Or has one of the above papers changed retroactively previous
standards?

// PR c++/89336
// { dg-do compile { target c++14 } }

constexpr int
foo ()
{
  union U { int a; long b; };
  union V { union U u; short v; };
  V w {};
  w.u.a = w.v = w.u.b = 5L;
  return w.u.a;
}

static_assert (foo () == 5, "");

constexpr int
bar ()
{
  union U { int a[5]; long b; };
  union V { union U u; short v; };
  V w {};
  w.v = 5;
  w.u.a[3] = w.u.a[1] = w.v;
  return w.u.a[1] + w.u.a[3];
}

static_assert (bar () == 10, "");

struct Z { int x, y; };

constexpr Z
baz ()
{
  union W { Z a; long long w; };
  W w {};
  w.a = { 5, 0 };
  w.a = { (int) (w.w = 17LL + w.a.x), 2 };
  return w.a;
}

static_assert (baz ().x == 22, "");
static_assert (baz ().y == 2, "");


	Jakub