Dmitry Vyukov

We need to figure it out yet. There should be some way to represent something like a C union. Most likely it should be based on interface{}, but then the...

> So the idea is to simply have something like []MethodAndArgsUnion as input yes > mutation with feedback to get valid sequences? yes also _potentially_ hints/constraints as field tags >...

> Is there a chance that fuzz.F can provide an (additional?) GetRandomData()-based interface like the one I sketched in golang/go#19109 (comment), on top of which a constrained generator library can...

> The disconnect we're seeing here is people thinking of "testing an API called by other parts of my app" vs "testing an untrusted input mechanism". (Modern) fuzzing has appeared...

@thepudds did you implement this in fzgo? I have some memories that somebody implemented this somewhere already...

Yes, the CI failure looks unrelated. There seems to be some change in darwin setup on travis.

> Yes, the CI failure looks unrelated. There seems to be some change in darwin setup on travis. This should be fixed now: https://github.com/dvyukov/go-fuzz/commit/be3528f3a81351d8a438aed216130e1e7da39f7c

Is it possible that the test program has some background goroutines?

Full panic message should contain all goroutines.

Then I think the problem is in `go-fuzz-dep/sonar.go`: ``` go // Sonar is called by instrumentation code to notify go-fuzz about comparisons. // Low 8 bits of id are flags,...