Previously it didn't matter whether we looked through a TARGET_EXPR in
constexpr evaluation, but now that we have constexpr destructors it does.
On IRC I mentioned the idea of clearing TARGET_EXPR_CLEANUP in
digest_nsdmi_init, but since this initialization is expressed by an
INIT_EXPR, it's better to handle all INIT_EXPR, not just those for a member
initializer.
Tested x86_64-pc-linux-gnu, applying to trunk.
* constexpr.c (cxx_eval_store_expression): Look through TARGET_EXPR
when not preevaluating.
---
gcc/cp/constexpr.c | 6 ++++++
gcc/testsuite/g++.dg/cpp2a/constexpr-new10.C | 19 +++++++++++++++++++
2 files changed, 25 insertions(+)
create mode 100644 gcc/testsuite/g++.dg/cpp2a/constexpr-new10.C
diff --git a/gcc/cp/constexpr.c b/gcc/cp/constexpr.c
index 806d3ab2cff..5fe6d0277b6 100644
--- a/gcc/cp/constexpr.c
+++ b/gcc/cp/constexpr.c
@@ -4577,6 +4577,12 @@ cxx_eval_store_expression (const constexpr_ctx *ctx,
tree t,
}
new_ctx.ctor = *valp;
new_ctx.object = target;
+ /* Avoid temporary materialization when initializing from a TARGET_EXPR.
+ We don't need to mess with AGGR_EXPR_SLOT/VEC_INIT_EXPR_SLOT because
+ expansion of those trees uses ctx instead. */
+ if (TREE_CODE (init) == TARGET_EXPR)
+ if (tree tinit = TARGET_EXPR_INITIAL (init))
+ init = tinit;
init = cxx_eval_constant_expression (&new_ctx, init, false,
non_constant_p, overflow_p);
if (ctors->is_empty())
diff --git a/gcc/testsuite/g++.dg/cpp2a/constexpr-new10.C
b/gcc/testsuite/g++.dg/cpp2a/constexpr-new10.C
new file mode 100644
index 00000000000..500a3240c8f
--- /dev/null
+++ b/gcc/testsuite/g++.dg/cpp2a/constexpr-new10.C
@@ -0,0 +1,19 @@
+// PR c++/91369
+// { dg-do compile { target c++2a } }
+
+struct S {
+ constexpr S (int* i) : s{i} {}
+ constexpr ~S () { delete s; }
+ int *s;
+};
+
+struct T { S t = { new int }; };
+
+constexpr auto
+foo ()
+{
+ T b;
+ return true;
+}
+
+static_assert (foo ());
--
2.18.1
