On Tue, 29 Jul 2025, Jakub Jelinek wrote:
> Hi!
>
> In the PR119483 r15-9003 change we've allowed musttail calls to noreturn
> functions, after all the decision not to normally tail call noreturn
> functions is not because it is not possible to tail call those, but because
> it screws up backtraces. As the following testcase shows, we've done that
> only for functions not declared [[noreturn]]/_Noreturn but later on
> discovered through IPA as noreturn. Functions explicitly declared
> [[noreturn]] have (for historical reasons) volatile FUNCTION_TYPE and
> the FUNCTION_DECLs are volatile as well, so in order to support those
> we shouldn't complain on ECF_NORETURN (we've stopped doing so for musttail
> in PR119483) but also shouldn't complain about TYPE_VOLATILE on their
> FUNCTION_TYPE (something that IPA doesn't change, I think it only sets
> TREE_THIS_VOLATILE on the FUNCTION_DECL). volatile on function type
> really means noreturn as well, it has no other meaning.
>
> Fixed thusly, bootstrapped/regtested on x86_64-linux and i686-linux,
> ok for trunk/15.2?
OK.
Thanks,
Richard.
> 2025-07-29 Jakub Jelinek <ja...@redhat.com>
>
> PR middle-end/121159
> * calls.cc (can_implement_as_sibling_call_p): Don't reject declared
> noreturn functions in musttail calls.
>
> * c-c++-common/pr121159.c: New test.
> * gcc.dg/plugin/must-tail-call-2.c (test_5): Don't expect an error.
>
> --- gcc/calls.cc.jj 2025-06-04 17:21:01.000000000 +0200
> +++ gcc/calls.cc 2025-07-28 11:58:35.205769000 +0200
> @@ -2589,7 +2589,8 @@ can_implement_as_sibling_call_p (tree ex
> return false;
> }
>
> - if (TYPE_VOLATILE (TREE_TYPE (TREE_TYPE (addr))))
> + if (TYPE_VOLATILE (TREE_TYPE (TREE_TYPE (addr)))
> + && !CALL_EXPR_MUST_TAIL_CALL (exp))
> {
> maybe_complain_about_tail_call (exp, _("volatile function type"));
> return false;
> --- gcc/testsuite/c-c++-common/pr121159.c.jj 2025-07-28 12:09:37.031226990
> +0200
> +++ gcc/testsuite/c-c++-common/pr121159.c 2025-07-28 12:09:24.069394221
> +0200
> @@ -0,0 +1,17 @@
> +/* PR middle-end/121159 */
> +/* { dg-do compile { target musttail } } */
> +/* { dg-options "-O2 -fdump-tree-optimized" } */
> +/* { dg-final { scan-tree-dump-times "foo \\\(\[^\n\r]*\\\); \\\[tail
> call\\\] \\\[must tail call\\\]" 1 "optimized" } } */
> +
> +[[noreturn, gnu::noipa]] void
> +foo (void)
> +{
> + for (;;)
> + ;
> +}
> +
> +void
> +bar (void)
> +{
> + [[gnu::musttail]] return foo ();
> +}
> --- gcc/testsuite/gcc.dg/plugin/must-tail-call-2.c.jj 2023-03-08
> 21:44:15.986550516 +0100
> +++ gcc/testsuite/gcc.dg/plugin/must-tail-call-2.c 2025-07-29
> 08:48:30.736998832 +0200
> @@ -55,5 +55,5 @@ volatile fn_ptr_t fn_ptr;
> void
> test_5 (void)
> {
> - fn_ptr (); /* { dg-error "cannot tail-call: " } */
> + fn_ptr ();
> }
>
> Jakub
>
>
--
Richard Biener <rguent...@suse.de>
SUSE Software Solutions Germany GmbH,
Frankenstrasse 146, 90461 Nuernberg, Germany;
GF: Ivo Totev, Andrew McDonald, Werner Knoblich; (HRB 36809, AG Nuernberg)
