[PATCH]: Fix typos in long double builtins in convert_to_real()

[Kaveh R. Ghazi]

As noted in question #3 here:
http://gcc.gnu.org/ml/gcc/2004-03/msg00621.html

there are a few typos in convert_to_real which I've fixed with the
patch below.  (We're supposed to be looking for the `long double'
versions of these builtins and some of them weren't.)

Bootstrapped on sparc-sun-solaris2.7, no regressions.

(This was noted by inspection, however a specific testcase is
forthcoming after I polish it up a bit.)

Patch below ok for mainline?

		Thanks,
		--Kaveh


2004-03-16  Kaveh R. Ghazi  <ghazi@caip.rutgers.edu>

	* convert.c (convert_to_real): Fix typos in `long double'
	builtins.
	
diff -rup orig/egcc-CVS20040315/gcc/convert.c egcc-CVS20040315/gcc/convert.c
--- orig/egcc-CVS20040315/gcc/convert.c	Sun Mar 14 20:01:14 2004
+++ egcc-CVS20040315/gcc/convert.c	Tue Mar 16 19:29:30 2004
@@ -174,9 +174,9 @@ convert_to_real (tree type, tree expr)
   if (optimize
       && (((fcode == BUILT_IN_FLOORL
 	   || fcode == BUILT_IN_CEILL
-	   || fcode == BUILT_IN_ROUND
-	   || fcode == BUILT_IN_TRUNC
-	   || fcode == BUILT_IN_NEARBYINT)
+	   || fcode == BUILT_IN_ROUNDL
+	   || fcode == BUILT_IN_TRUNCL
+	   || fcode == BUILT_IN_NEARBYINTL)
 	  && (TYPE_MODE (type) == TYPE_MODE (double_type_node)
 	      || TYPE_MODE (type) == TYPE_MODE (float_type_node)))
 	  || ((fcode == BUILT_IN_FLOOR