[PING patch]: Preparations for call abi switching via attribute.

[Hans-Peter Nilsson]

> From: Kai Tietz <Kai.Tietz@onevision.com>
> Date: Wed, 7 May 2008 10:50:34 +0200

> 2008-04-29  Kai Tietz  <kai,tietz@onevision.com>

s/,/./?  Or are you trying to avoid being harvested?

>         * config/cris/cris.h: Add comment to #if 0 clause.

Nah, it should be "adjust #if 0:d RETURN_IN_MEMORY code". :)
But, I'm ok with removing the #if 0:d code and changing the
comment to "Let's try returning structures in registers".  Or
rather, remove it all; changing the ABI now would probably be
worse in total.

> +++ gcc/gcc/config/cris/cris.h
> @@ -924,10 +924,11 @@ struct cum_args {int regs;};
>  /* Node: Aggregate Return */
>  
>  #if 0
> +/* FYN, if fixing change it to use TARGET_RETURN_IN_MEMORY instead.  */

Whats a FYN?  (Or rather, don't add this comment.)  (Please
avoid introducing new TLA's.  Google finds a certain Danish
island and "Florida Yards and Neighborhoods Program".)

>  /* FIXME: Let's try this some time, so we return structures in registers.
>     We would cast the result of int_size_in_bytes to unsigned, so we will
>     get a huge number for "structures" of variable size (-1).  */
> -#define RETURN_IN_MEMORY(TYPE) \
> +#define RETURN_IN_MEMORY(TYPE, FNTYPE) \
>   ((unsigned) int_size_in_bytes (TYPE) > CRIS_MAX_ARGS_IN_REGS * UNITS_PER_WORD)
>  #endif

Shouldn't that also be a
s/RETURN_IN_MEMORY/TARGET_RETURN_IN_MEMORY/?  No need to have a
separate way of editing just because it's #if 0:d. :)

Still, better scrap all inside #if 0.

I'd also be fine if you just skip it and leave it to me.

Thanks for tending to us little targets.

brgds, H-P