Hi,

On Mon, Aug 17 2026, Josef Melcr wrote:
> Hi,
> internal functions and their calls are not represented in the call
> graph, so the if statement removed by this patch will never go through.
> While I am making an effort to introduce some internal calls into the call
> graph, it is for optimization purposes, so the check would be in the way
> even in that case.
>
> Bootstrapped and regtested on x86_64-linux by replacing the return with
> __builtin_abort without any failures.
>
> I did encounter one segfault when compiling gfortran.dg/pr46844.f,
> however it was in the frontend and it was happening even with a clean
> tree, so it was likely some misconfiguration issue.  I did more tests
> on different machines and they were all okay, just mentioning this for
> the sake of transparency.
>
> OK for master?
>
> Best regards,
> Josef
>
> gcc/ChangeLog:
>
>       * ipa-prop.cc (ipa_compute_jump_functions_for_edge): Remove
>       dead if statement checking for internal calls.

OK,

thanks.

Martin


>
> Signed-off-by: Josef Melcr <[email protected]>
> ---
>  gcc/ipa-prop.cc | 2 --
>  1 file changed, 2 deletions(-)
>
> diff --git a/gcc/ipa-prop.cc b/gcc/ipa-prop.cc
> index 668932ec329..3f9f45b02c8 100644
> --- a/gcc/ipa-prop.cc
> +++ b/gcc/ipa-prop.cc
> @@ -2501,8 +2501,6 @@ ipa_compute_jump_functions_for_edge (struct 
> ipa_func_body_info *fbi,
>    if (flag_devirtualize)
>      vec_safe_grow_cleared (args->polymorphic_call_contexts, arg_num, true);
>  
> -  if (gimple_call_internal_p (call))
> -    return;
>    if (ipa_func_spec_opts_forbid_analysis_p (cs->caller))
>      return;
>  
> -- 
> 2.55.0

Reply via email to