c++: constexpr init of union sub-aggr w/ base [PR105491]

Here ever since r10-7313-gb599bf9d6d1e18, reduced_constant_expression_p
in C++11/14 is rejecting the marked sub-aggregate initializer (of type S)

  W w = {.D.2445={.s={.D.2387={.m=0}, .b=0}}};
                     ^
ultimately because said initializer has CONSTRUCTOR_NO_CLEARING set,
hence the function must verify that all fields of S are initialized.
And before C++17 it doesn't expect to see base class fields (since
next_initializable_field skips over them), so the presence thereof
causes r_c_e_p to return false.

The reason r10-7313-gb599bf9d6d1e18 causes this is because in that
commit we began using CONSTRUCTOR_NO_CLEARING to precisely track whether
we're in middle of activating a union member.  This ends up affecting
clear_no_implicit_zero, which recurses into sub-aggregate initializers
only if the outer initializer has CONSTRUCTOR_NO_CLEARING set.  After
that commit, the outer union initializer above no longer has the flag
set at this point and so clear_no_implicit_zero no longer recurses into
the marked inner initializer.

But arguably r_c_e_p should be able to accept the marked initializer
regardless of whether CONSTRUCTOR_NO_CLEARING is set.  The primary bug
therefore seems to be that r_c_e_p relies on next_initializable_field
which skips over base class fields in C++11/14.  To fix this, this patch
introduces a new helper function next_subobject_field which is like
next_initializable_field except that it never skips base class fields,
and makes r_c_e_p use it.  This patch then renames next_initializable_field
to next_aggregate_field (and makes it skip over vptr fields again).

	PR c++/105491

gcc/cp/ChangeLog:

	* call.cc (field_in_pset): Adjust after next_initializable_field
	renaming.
	(build_aggr_conv): Likewise.
	(convert_like_internal): Likewise.
	(type_has_extended_temps): Likewise.
	* class.cc (default_init_uninitialized_part): Likewise.
	(finish_struct): Likewise.
	* constexpr.cc (cx_check_missing_mem_inits): Likewise.
	(reduced_constant_expression_p): Use next_subobject_field
	instead.
	* cp-gimplify.cc (get_source_location_impl_type): Adjust after
	next_initializable_field renaming.
	(fold_builtin_source_location): Likewise.
	* cp-tree.h (next_initializable_field): Rename to ...
	(next_aggregate_field): ... this.
	(next_subobject_field): Declare.
	* decl.cc (next_aggregate_field): Renamed from ...
	(next_initializable_field): ... this.  Skip over vptr fields
	again.
	(next_subobject_field): Define.
	(reshape_init_class): Adjust after next_initializable_field
	renaming.
	* init.cc (build_value_init_noctor): Likewise.
	(emit_mem_initializers): Likewise.
	* lambda.cc (build_capture_proxy): Likewise.
	* method.cc (build_comparison_op): Likewise.
	* pt.cc (maybe_aggr_guide): Likewise.
	* tree.cc (structural_type_p): Likewise.
	* typeck2.cc (split_nonconstant_init_1): Likewise.
	(digest_init_r): Likewise.

gcc/testsuite/ChangeLog:

	* g++.dg/cpp0x/constexpr-union7.C: New test.
	* g++.dg/cpp0x/constexpr-union7a.C: New test.
	* g++.dg/cpp2a/constinit17.C: New test.

1 files changed, 26 insertions, 10 deletions

diff --git a/gcc/cp/decl.cc b/gcc/cp/decl.cc
index 0fa758f..872b02d 100644
--- a/gcc/cp/decl.cc
+++ b/gcc/cp/decl.cc
@@ -6384,20 +6384,36 @@ static tree reshape_init_r (tree, reshape_iter *, tree, tsubst_flags_t);
 
 /* FIELD is an element of TYPE_FIELDS or NULL.  In the former case, the value
    returned is the next FIELD_DECL (possibly FIELD itself) that can be
-   initialized.  If there are no more such fields, the return value
-   will be NULL.  */
+   initialized as if for an aggregate class.  If there are no more such fields,
+   the return value will be NULL.  */
 
 tree
-next_initializable_field (tree field)
+next_aggregate_field (tree field)
 {
   while (field
 	 && (TREE_CODE (field) != FIELD_DECL
 	     || DECL_UNNAMED_BIT_FIELD (field)
 	     || (DECL_ARTIFICIAL (field)
-		 /* In C++17, don't skip base class fields.  */
-		 && !(cxx_dialect >= cxx17 && DECL_FIELD_IS_BASE (field))
-		 /* Don't skip vptr fields.  We might see them when we're
-		    called from reduced_constant_expression_p.  */
+		 /* In C++17, aggregates can have bases.  */
+		 && !(cxx_dialect >= cxx17 && DECL_FIELD_IS_BASE (field)))))
+    field = DECL_CHAIN (field);
+
+  return field;
+}
+
+/* FIELD is an element of TYPE_FIELDS or NULL.  In the former case, the value
+   returned is the next FIELD_DECL (possibly FIELD itself) that corresponds
+   to a subobject.  If there are no more such fields, the return value will be
+   NULL.  */
+
+tree
+next_subobject_field (tree field)
+{
+  while (field
+	 && (TREE_CODE (field) != FIELD_DECL
+	     || DECL_UNNAMED_BIT_FIELD (field)
+	     || (DECL_ARTIFICIAL (field)
+		 && !DECL_FIELD_IS_BASE (field)
 		 && !DECL_VIRTUAL_P (field))))
     field = DECL_CHAIN (field);
 
@@ -6595,7 +6611,7 @@ reshape_init_class (tree type, reshape_iter *d, bool first_initializer_p,
   if (base_binfo)
     field = base_binfo;
   else
-    field = next_initializable_field (TYPE_FIELDS (type));
+    field = next_aggregate_field (TYPE_FIELDS (type));
 
   if (!field)
     {
@@ -6762,10 +6778,10 @@ reshape_init_class (tree type, reshape_iter *d, bool first_initializer_p,
 	  if (BINFO_BASE_ITERATE (binfo, ++binfo_idx, base_binfo))
 	    field = base_binfo;
 	  else
-	    field = next_initializable_field (TYPE_FIELDS (type));
+	    field = next_aggregate_field (TYPE_FIELDS (type));
 	}
       else
-	field = next_initializable_field (DECL_CHAIN (field));
+	field = next_aggregate_field (DECL_CHAIN (field));
     }
 
   /* A trailing aggregate element that is a pack expansion is assumed to