OK, thanks.
On Mon, Nov 11, 2019 at 3:14 PM Jakub Jelinek <ja...@redhat.com> wrote:
> Hi!
>
> The finish_function change to goto cleanup; on DECL_DELETED_FN added
> in the spaceship commit broke the following testcase.
> The problem is that during start_preparsed_function push_nested_class
> pushes a scope, but as ctype is kept NULL when goto cleanup; crosses
> the setting of ctype to something else, pop_nested_class isn't called
> anymore and callers get upset they are in current_binding_level
> of sk_class instead of what they expected.
>
> Fixed thusly, bootstrapped/regtested on powerpc64le-linux, ok for trunk?
>
> 2019-11-11 Jakub Jelinek <ja...@redhat.com>
>
> PR c++/92447
> * decl.c (finish_function): Move ctype initialization before
> DECL_DELETED_FN handling.
>
> * g++.dg/cpp0x/pr92447.C: New test.
>
> --- gcc/cp/decl.c.jj 2019-11-07 09:50:51.000000000 +0100
> +++ gcc/cp/decl.c 2019-11-11 15:29:17.610112820 +0100
> @@ -16803,6 +16803,10 @@ finish_function (bool inline_p)
> }
> }
>
> + /* Remember that we were in class scope. */
> + if (current_class_name)
> + ctype = current_class_type;
> +
> if (DECL_DELETED_FN (fndecl))
> {
> DECL_INITIAL (fndecl) = error_mark_node;
> @@ -16861,10 +16865,6 @@ finish_function (bool inline_p)
> current_function_return_value = NULL_TREE;
> }
>
> - /* Remember that we were in class scope. */
> - if (current_class_name)
> - ctype = current_class_type;
> -
> /* Must mark the RESULT_DECL as being in this function. */
> DECL_CONTEXT (DECL_RESULT (fndecl)) = fndecl;
>
> --- gcc/testsuite/g++.dg/cpp0x/pr92447.C.jj 2019-11-11
> 15:31:28.208138229 +0100
> +++ gcc/testsuite/g++.dg/cpp0x/pr92447.C 2019-11-11
> 15:30:29.314028624 +0100
> @@ -0,0 +1,14 @@
> +// PR c++/92447
> +// { dg-do compile { target c++11 } }
> +
> +template <typename T>
> +void
> +foo ()
> +{
> + struct S { S &operator=(S &&x) = default; const T s{}; };
> +}
> +
> +void bar ()
> +{
> + foo<int>();
> +}
>
> Jakub
>
>
