I think removing the INIT_FINI_PREP assignment in the aix5 arm of the 
case statement is better, but it will have the same effect.  Sorry I 
didn't catch that when making the suggestion.