almann
commented
3 years ago Some interesting references here:
Specifically, fuzzing the creation of Element and fuzzing the round-tripping of those sequences would be a great start. Also, fuzzing the parsing of ion-tests into the former would also be great.
The most insidious bugs we've found are data races, and so running a massive fuzz test in parallel, even though it may not be deterministic for those kinds of bugs would be useful in at least giving us some confidence around thread-safety.
238 really makes me think we should be fuzz testing the library. @rgantt has experience with this and we should port some of the work he has done in this space.