[dataflow][PATCH] Fix a fall-out from latest merge with mainline (take 2)

[Kenneth Zadeck]

Seongbae Park wrote:
> This time it's a proper fix. Bootstrapped and regtested on x86-64.
> A quick comparison against latest mainline shows no regression
> but I haven't completely set up this new test environment yet
> so I might have extra regressions in my mainline test result.
>
> Also, given that this is just a merge, I'm not sure the following
> changelog
> is accurate - should I just say it's a merge with the mainline ?
>
> 2007-03-20  Seongbae Park <seongbae.park@gmail.com>
>
>        * combine.c (setup_incoming_promotions): Use the argument FIRST
>        instead of recomputing it.
> ------------------------------------------------------------------------
>
> Index: combine.c
> ===================================================================
> --- combine.c	(revision 123093)
> +++ combine.c	(working copy)
> @@ -1307,14 +1307,11 @@ init_reg_last (void)
>  static void
>  setup_incoming_promotions (rtx first)
>  {
> -  rtx first;
>    tree arg;
>  
>    if (!targetm.calls.promote_function_args (TREE_TYPE (cfun->decl)))
>      return;
>  
> -  first = get_insns ();
> -
>    for (arg = DECL_ARGUMENTS (current_function_decl); arg;
>         arg = TREE_CHAIN (arg))
>      {
>   
ok for commit
kenny