https://gcc.gnu.org/bugzilla/show_bug.cgi?id=118939
--- Comment #16 from Richard Earnshaw <rearnsha at gcc dot gnu.org> ---
(In reply to Eric Botcazou from comment #13)
> Possible kludge to work around the questionable mechanism:
>
> diff --git a/gcc/config/arm/arm.cc b/gcc/config/arm/arm.cc
> index 59b41e3d046..2d51874a05b 100644
> --- a/gcc/config/arm/arm.cc
> +++ b/gcc/config/arm/arm.cc
> @@ -30887,7 +30887,17 @@ arm_frame_pointer_required (void)
> && arm_except_unwind_info (&global_options) == UI_TARGET
> && cfun->can_throw_non_call_exceptions)
> {
> - HOST_WIDE_INT size = get_frame_size ();
> + HOST_WIDE_INT size;
> +
> + /* ??? We need to return the same answer during IRA and LRA,
> + the problem being that get_frame_size does not do so, so
> + use the offsets computed by arm_compute_frame_layout to
> + get back to the early IRA value. */
> + if (lra_in_progress)
> + size = cfun->machine->stack_offsets.locals_base
> + - cfun->machine->stack_offsets.soft_frame;
> + else
> + size = ROUND_UP_WORD (get_frame_size ());
>
> /* That's irrelevant if there is no stack adjustment. */
> if (size <= 0)
That can't be right. If the frame size has changed, then the frame needs to be
laid out again and any earlier layout assumptions revisited.
