Reconsider not writing to known addresses if a pointer contains an unknown element
Opened this issue · 1 comments
michael-schwarz commented
We currently do completely ignore the possibility that any of the known pointers are written as soon as we have ?
in the points-to set. This seems like it is philosophically the wrong thing to do to me.
In my opinion, we should at least potentially update the targets which we have identified.
Lines 1730 to 1748 in a84e07d
sim642 commented
I suspect the SetDomain.Unsupported
code is some legacy thing that never even happens now. AddressSet.top
isn't a lifted Top
but a normal (foldable-over) set with UnknownPtr
in it. And update_one
there just ignores that one when doing Addr.to_mval
.