Closed flyingmutant closed 2 years ago
After some early experiments, the integration is relatively straightforward and "works", but is practically useless without ability to customize the mutator -- rapid
itself is much better at getting the good coverage. Looks like without ability to teach the fuzzer how to do some kind of structure-aware mutations, there is no point in the integration.
Maybe we can use black magic to get hold of the raw coverage data, and play the fuzzing game ourselves?
Closing this after merging an initial implementation -- let's track improvements as a separate issues.
https://go.dev/blog/fuzz-beta