Currently, if dump-stats returns an error, the fuzzer (and probably the planner too) don't return a request to the corpus builder. This means that the builder gets stuck in an infinite loop of waiting for the nonexistent requests to turn up.
Ideally, some sort of request should come through, but it should carry some sort of error.
Currently, if
dump-statsreturns an error, the fuzzer (and probably the planner too) don't return a request to the corpus builder. This means that the builder gets stuck in an infinite loop of waiting for the nonexistent requests to turn up.Ideally, some sort of request should come through, but it should carry some sort of error.