poll the fuzzer for possible tweakable parameters (such as fuzzer weights, but also maybe flags);
randomly set some of them to false;
feed that information into the fuzzer.
This would be a very rudimentary first approximation of swarm testing.
This sort of thing could be done by the fuzzer itself, but I'm putting it in here for a few reasons:
the tester now actively feeds the fuzzer configuration, and so using that feed to pass it swarmed parameters sounds like it could actually be quite an interesting use of that property;
eventually, one could envision the tester using analysis over previous runs to shape its configuration of the fuzzer (not sure if I'll get this done any time soon, but);
whatever we do for the fuzzer would then generalise to other fuzzers in the future, so long as they expose a similar interface;
idk I prefer having as much infrastructure in Go as possible :D
Effectively, I want the fuzz runner to:
This would be a very rudimentary first approximation of swarm testing.
This sort of thing could be done by the fuzzer itself, but I'm putting it in here for a few reasons: