Re: RFC: PR 19650

[Dale Johannesen <dalej at apple dot com>]

On Jan 28, 2005, at 12:05 PM, Roger Sayle wrote:

> I think the best route out of this hole is to pass an additional
> "tree arg_type" parameter to the fold_binary_op_with_conditional_arg
> function.  At both call sites in fold, this is simply the value
> "TREE_TYPE (TREE_OPERAND (t, 0))", i.e. the unstripped type of the
> first operand.

Like so, I think you mean (not exactly what you said):

@@ -6466,7 +6474,8 @@ fold (tree expr)
       if (TREE_CODE (arg0) == COND_EXPR || COMPARISON_CLASS_P (arg0))
        {
          tem = fold_binary_op_with_conditional_arg (code, type, arg0, 
arg1,
-                                                    
/*cond_first_p=*/1);
+                                                    /*cond_first_p=*/1,
+                                                    TREE_TYPE 
(TREE_OPERAND (t, 0)));
          if (tem != NULL_TREE)
            return tem;
        }
@@ -6474,7 +6483,8 @@ fold (tree expr)
       if (TREE_CODE (arg1) == COND_EXPR || COMPARISON_CLASS_P (arg1))
        {
          tem = fold_binary_op_with_conditional_arg (code, type, arg1, 
arg0,
-                                                    
/*cond_first_p=*/0);
+                                                    /*cond_first_p=*/0,
+                                                    TREE_TYPE 
(TREE_OPERAND (t, 1)));
          if (tem != NULL_TREE)
            return tem;
        }

> Then inside the function itself we cast "arg",
> "true_value" and "false_value" to this "arg_type", before using them
> as operands to build a tree with code CODE and result type TYPE.
>
>
> + arg = fold_convert (arg_type, arg);
>   if (lhs == 0)
> +   {
> +     true_value = fold_convert (arg_type, true_value);
>       lhs = fold (cond_first_p ? build2 (code, type, true_value, arg)
>                                : build2 (code, type, arg, true_value));
>     }
>   if (rhs == 0)
> +   {
> +     false_value = fold_convert (arg_type, false_value);
>       rhs = fold (cond_first_p ? build2 (code, type, false_value, arg)
>                                : build2 (code, type, arg, 
> false_value));
> +   }

That seems to solve the bootstrap problem and the original failure.  
I'll
run it through more testing.  Thanks.