https://gcc.gnu.org/bugzilla/show_bug.cgi?id=118169
Jakub Jelinek <jakub at gcc dot gnu.org> changed:
What |Removed |Added
----------------------------------------------------------------------------
CC| |jason at gcc dot gnu.org
--- Comment #18 from Jakub Jelinek <jakub at gcc dot gnu.org> ---
Maybe just special case VOID_TYPE_P in there?
--- gcc/cp/constexpr.cc.jj 2025-01-02 11:47:10.290500486 +0100
+++ gcc/cp/constexpr.cc 2025-01-03 17:28:23.879757899 +0100
@@ -8870,14 +8870,17 @@ cxx_eval_outermost_constant_expr (tree t
/* Turn off -frounding-math for manifestly constant evaluation. */
warning_sentinel rm (flag_rounding_math,
ctx.manifestly_const_eval == mce_true);
- tree type = (object
- ? cv_unqualified (TREE_TYPE (object))
- : initialized_type (t));
+ tree type = initialized_type (t);
+ if (object && !VOID_TYPE_P (type))
+ type = cv_unqualified (TREE_TYPE (object));
tree r = t;
bool is_consteval = false;
if (VOID_TYPE_P (type))
{
- if (!constexpr_dtor)
+ if (constexpr_dtor)
+ /* Used for destructors of array elements. */
+ type = cv_unqualified (TREE_TYPE (object));
+ else
{
if (cxx_dialect < cxx20)
return t;
is a partial reversion of that patch, which keeps previous behavior for the
VOID_TYPE_P (initialized_type (t)) case (dunno about cv_unqualified vs. nothing
for the dtors case) but still fixes the constexpr-prvalue{2,3}.C ICEs.
