Re: Small change to expand_expr CONSTRUCTOR handling

[Graham Stott <graham dot stott at btinternet dot com>]

Hi Richard,

Richard Kenner wrote:
> This was motivated by a large Ada code.  Tested on i686-pc-linux-gnu.
>
> Mon May 19 16:03:53 2003  Richard Kenner  <kenner@vlsi1.ultra.nyu.edu>
>
> 	* expr.c (expand_expr, case CONSTRUCTOR): Put into memory if
> 	constant and EXPAND_CONST_ADDRESS, not just EXPAND_INITIALIZER.
>
> *** gcc/expr.c.may16	Fri May 16 15:50:35 2003
> --- gcc/expr.c	Fri May 16 15:51:45 2003
> *************** expand_expr (exp, target, tmode, modifie
> *** 6713,6717 ****
>   			     TYPE_ALIGN (type)))
>   			&& ! mostly_zeros_p (exp))))
> ! 	       || (modifier == EXPAND_INITIALIZER && TREE_CONSTANT (exp)))
>   	{
>   	  rtx constructor = output_constant_def (exp, 1);
> --- 6713,6719 ----
>   			     TYPE_ALIGN (type)))
>   			&& ! mostly_zeros_p (exp))))
> ! 	       || ((modifier == EXPAND_INITIALIZER
> ! 		    || modifier == EXPAND_CONST_ADDRESS)
> ! 		   && TREE_CONSTANT (exp)))
>   	{
>   	  rtx constructor = output_constant_def (exp, 1);


It looks like you checked in more than just this patch, because
revision 1.540 of expr.c also contains the following changes which
are preventing bootstraps because CONTAINS_PLACEHOLDER_P is undefined.

1.540        (kenner   22-May-03): 	      if (CONTAINS_PLACEHOLDER_P (offset))
1.540        (kenner   22-May-03): 	  else if (CONTAINS_PLACEHOLDER_P (this_offset))
1.540        (kenner   22-May-03): 	  if (CONTAINS_PLACEHOLDER_P (index))
1.540        (kenner   22-May-03): 	  if (CONTAINS_PLACEHOLDER_P (unit_size))

Cheers
Graham