[tree-ssa] BIND_EXPRs

[law@redhat.com]

I could have sworn I had regression tested the changes I posted
earlier today.  Alas, clearly they're causing a regression.

Specifically certain hunks of code are unprepared for a 
BIND_EXPR with no apparent use being turned into an empty statement.

This change reverts that optimization.  It fixes the regressions
introduced this morning.

	* c-simplify.c (c_build_bind_expr): Revert change from earlier today.
Index: c-simplify.c
===================================================================
RCS file: /cvs/gcc/gcc/gcc/Attic/c-simplify.c,v
retrieving revision 1.1.4.47
diff -c -3 -p -r1.1.4.47 c-simplify.c
*** c-simplify.c	21 Apr 2003 16:47:29 -0000	1.1.4.47
--- c-simplify.c	21 Apr 2003 22:36:52 -0000
*************** c_build_bind_expr (block, body)
*** 367,377 ****
    c_simplify_stmt (&BIND_EXPR_BODY (bind));
    gimple_pop_bind_expr ();
  
-   /* If the block has an empty body and no decls, then just
-      return an empty statement.  */
-   if (BIND_EXPR_BODY (bind) == empty_stmt_node && decls == NULL_TREE)
-     return empty_stmt_node;
- 
    return bind;
  }
  
--- 367,372 ----