c++: constant direct-initialization [PR108243]

When testing the proposed patch for PR71093 I noticed that it changed the
diagnostic for consteval-prop6.C.  I then noticed that the diagnostic wasn't
very helpful either way; it was complaining about modification of the 'x'
variable, but it's not a problem to initialize a local variable with a
consteval constructor as long as the value is actually constant, we want to
know why the value isn't constant.  And then it turned out that this also
fixed a missed-optimization bug in the testsuite.

	PR c++/108243

gcc/cp/ChangeLog:

	* constexpr.cc (cxx_eval_outermost_constant_expr): Turn
	a constructor CALL_EXPR into a TARGET_EXPR.

gcc/testsuite/ChangeLog:

	* g++.dg/cpp2a/consteval-prop6.C: Adjust diagnostic.
	* g++.dg/opt/is_constant_evaluated3.C: Remove xfails.

1 files changed, 15 insertions, 1 deletions

diff --git a/gcc/cp/constexpr.cc b/gcc/cp/constexpr.cc
index 58187a4..4cf9dd7 100644
--- a/gcc/cp/constexpr.cc
+++ b/gcc/cp/constexpr.cc
@@ -8651,7 +8651,21 @@ cxx_eval_outermost_constant_expr (tree t, bool allow_non_constant,
 	}
       if (!object)
 	{
-	  if (TREE_CODE (t) == TARGET_EXPR)
+	  if (TREE_CODE (t) == CALL_EXPR)
+	    {
+	      /* If T is calling a constructor to initialize an object, reframe
+		 it as an AGGR_INIT_EXPR to avoid trying to modify an object
+		 from outside the constant evaluation, which will fail even if
+		 the value is actually constant (is_constant_evaluated3.C).  */
+	      tree fn = cp_get_callee_fndecl_nofold (t);
+	      if (fn && DECL_CONSTRUCTOR_P (fn))
+		{
+		  object = CALL_EXPR_ARG (t, 0);
+		  object = build_fold_indirect_ref (object);
+		  r = build_aggr_init_expr (type, r);
+		}
+	    }
+	  else if (TREE_CODE (t) == TARGET_EXPR)
 	    object = TARGET_EXPR_SLOT (t);
 	  else if (TREE_CODE (t) == AGGR_INIT_EXPR)
 	    object = AGGR_INIT_EXPR_SLOT (t);