hengchu / fuzzi-model Goto Github PK

currently, for some reason, rnm may report different path constraint lengths depending on which path we took. this is very problematic...

let's implement a pretty-printer of the embedded code and diagnose the issue by hand

otherwise a constraint may get dropped but then reappear again because the intermediate set of constraints already didn't contain it

using the parameters generated from this snippet of code:

would this be a problem in practice?

In the python impl, all symbolic results that match a given concrete results are grouped together. This may not be sound, because two symbolic results may come from totally different distributions, and the merged set of constraints may not make any sense.

symbolic grouping should be done based on alpha equivalence of the provenance. this ensures the output comes from the same underlying distribution.

instead of adding all of the privtree stuff, just add a constructor for unreachable code, and throw away bad branches in symbolic execution with SymExecError

The guard condition in symbolic unions explode. Need to investigate why