> Regression tests opr_sanity and sanity_check are now failing.

Um, Bruce, I've said several times that I didn't think Perchine's large
object changes should be applied until someone had actually reviewed
them.