Re: [PATCH] Fix condition in ira-color.c and lra-spills.c (PR c/61271)

[Vladimir Makarov <vmakarov@redhat.com>]

On 08/21/2014 02:01 PM, Marek Polacek wrote:
> Both ira-color.c and lra-spills.c contain this code, where
> -Wlogical-not-parentheses would warn:
> !FRAME_GROWS_DOWNWARD == STACK_GROWS_DOWNWARD
> I think the condition is semantically right, so I just tweaked it
> into a more common way with !=, so we don't warn on this anymore.
>
> Bootstrapped/regtested on x86_64-linux, ok for trunk?
>
> 2014-08-21  Marek Polacek  <polacek@redhat.com>
>
> 	PR c/61271
> 	* ira-color.c (coalesced_pseudo_reg_slot_compare): Fix condition.
> 	* lra-spills.c (pseudo_reg_slot_compare): Fix condition.
>
> diff --git gcc/ira-color.c gcc/ira-color.c
> index 36c3c87..14958ac 100644
> --- gcc/ira-color.c
> +++ gcc/ira-color.c
> @@ -3850,7 +3850,7 @@ coalesced_pseudo_reg_slot_compare (const void *v1p, const void *v2p)
>    slot_num2 = -ALLOCNO_HARD_REGNO (a2);
>    if ((diff = slot_num1 - slot_num2) != 0)
>      return (frame_pointer_needed
> -	    || !FRAME_GROWS_DOWNWARD == STACK_GROWS_DOWNWARD ? diff : -diff);
> +	    || (FRAME_GROWS_DOWNWARD != STACK_GROWS_DOWNWARD) ? diff : -diff);
>    total_size1 = MAX (PSEUDO_REGNO_BYTES (regno1),
>  		     regno_max_ref_width[regno1]);
>    total_size2 = MAX (PSEUDO_REGNO_BYTES (regno2),
> diff --git gcc/lra-spills.c gcc/lra-spills.c
> index 50f63fc..0d14685 100644
> --- gcc/lra-spills.c
> +++ gcc/lra-spills.c
> @@ -237,7 +237,7 @@ pseudo_reg_slot_compare (const void *v1p, const void *v2p)
>    slot_num2 = pseudo_slots[regno2].slot_num;
>    if ((diff = slot_num1 - slot_num2) != 0)
>      return (frame_pointer_needed
> -	    || !FRAME_GROWS_DOWNWARD == STACK_GROWS_DOWNWARD ? diff : -diff);
> +	    || (FRAME_GROWS_DOWNWARD != STACK_GROWS_DOWNWARD) ? diff : -diff);
>    total_size1 = GET_MODE_SIZE (lra_reg_info[regno1].biggest_mode);
>    total_size2 = GET_MODE_SIZE (lra_reg_info[regno2].biggest_mode);
>    if ((diff = total_size2 - total_size1) != 0)
>
>
Sorry, Marek.  I guess it is wrong.  STACK_GROWS_DOWNWARD is only 0 or 1
in these files (it is achieved by redefinition of STACK_GROWS_DOWNWARD
in the files).  FAME_GROWS_DOWNWARD can be 0 or anything non-zero (even
non-constant) as tm.texi contains

@defmac FRAME_GROWS_DOWNWARD
Define this macro to nonzero value if the addresses of local variable slots
are at negative offsets from the frame pointer.
@end defmac

It works now as all machine description files use 1 as non-zero value.

So the documentation should be changed but it is not wise, error prune,
and right now the macro value can be non-constant during compilation
time theoretically.

So you need another way to transform the expressions to get rid of the
warning.

As I remember the code was actually taken from the old RA and not simple
as it looks.