> Michael's single_set explanation makes sense, but I have no idea what > the right thing to do about it would be. One possible - but expensive - approach would be to run flow beforehand. Then all the sets of the status register that are not used will get a REG_UNUSED note, and single_set can suceed for the other set.