>LGTM. > >I wonder if in general we could support producing an error when ADDR(foo) >is used and we still don't know the address of foo. > >Cheers, >Rafael I am also thinking about this right now :) I think we need something like this to find cases like PR32031. George.