Re: [trunk][patch] fix a bug in fold_call_expr and one in record_temporary_equivalences_from_stmts_at_dest

[Diego Novillo <dnovillo at google dot com>]

On 4/30/08 9:54 AM, Rafael Espindola wrote:

> --- a/gcc/builtins.c
> +++ b/gcc/builtins.c
> @@ -10553,6 +10553,7 @@ fold_call_expr (tree exp, bool ignore)
>  		  if (CAN_HAVE_LOCATION_P (realret)
>  		      && !EXPR_HAS_LOCATION (realret))
>  		    SET_EXPR_LOCATION (realret, EXPR_LOCATION (exp));
> +		  return realret;
>  		}
>  	      return ret;
>  	    }

OK.

> diff --git a/gcc/tree-ssa-threadedge.c b/gcc/tree-ssa-threadedge.c
> index d0d8fb2..71aa91b 100644
> --- a/gcc/tree-ssa-threadedge.c
> +++ b/gcc/tree-ssa-threadedge.c
> @@ -226,6 +226,7 @@ record_temporary_equivalences_from_stmts_at_dest (edge e,
>    for (bsi = bsi_start (e->dest); ! bsi_end_p (bsi); bsi_next (&bsi))
>      {
>        tree cached_lhs = NULL;
> +      tree rhs;
>  
>        stmt = bsi_stmt (bsi);
>  
> @@ -252,6 +253,18 @@ record_temporary_equivalences_from_stmts_at_dest (edge e,
>  	  || TREE_CODE (GIMPLE_STMT_OPERAND (stmt, 0)) != SSA_NAME)
>  	continue;
>  
> +      rhs = GIMPLE_STMT_OPERAND (stmt, 1);
> +
> +      /* The result of __builtin_object_size depends on all the arguments
> +	 of a phi node. Temporally using only one edge produces invalid
> +	 results. */

s/Temporally/temporarily/ ?

OK, but this needs a code snippet example.  It's not immediately obvious 
what the problem is here.


> +      if (TREE_CODE (rhs) == CALL_EXPR)
> +	{
> +	  tree fndecl = get_callee_fndecl (rhs);
> +	  if (DECL_FUNCTION_CODE (fndecl) == BUILT_IN_OBJECT_SIZE)
> +	    continue;

fndecl can be NULL here.


Diego.