15:24:17biketurns out we do still allow implicit downcasts from T_O to Symbol_O, and it's done in a few places in cando. so that's probably-unnecessary runtime overhead. i'll see if i can fix that
15:27:22bikerecently the _Type field in FFNonbond_O was changed from a Symbol_sp to T_sp. But there are still a couple places that implicitly assume it's a symbol, like FFAngleDB_O::findTerm
18:57:32bikeso, i have been doing all this cleanup on the basis that if i spend another week staring at mmtk crashes without a break i'll lose it, but while doing it i just found somewhere we're apparently tagging a cons as a general. paranoia checks are good every once in a while, i guess.