Rewrite InProcessExecutors

[tokatoka]

• We should use builder pattern instead of current constructors that has 90% same code
• Redesign crash, timeout hooks.

[addisoncrump]

Can we change this to "rewrite executors"? There is too much technical debt here.

[riesentoaster]

While we're at it, I'm not sure why run_target gets a reference to the Fuzzer. Last I checked, this was only used in some InProcessExecutors to retrieve observers, and I think you already have/can get a reference to those by directly passing the observers to the executor during creation. This is just a cyclical dependency, which is never great. Stuff like passing self to a function on another struct raises some alarm bells for me:

https://github.com/AFLplusplus/LibAFL/blob/c5b7c7c23514c79c5b5df50d6c577b2d23fec430/libafl/src/fuzzer/mod.rs#L936

[tokatoka]

While we're at it, I'm not sure why run_target gets a reference to the Fuzzer

this is for crash handler. When you crash in a inproc handler, you have to continue execution from a signal handler. This signal handler has to know the pointer to 4 things, the executor, the fuzzer, the state, and the event manager.

We write the pointer to these object before executions, so we need the references.

Last I checked, this was only used in some InProcessExecutors to retrieve observers, and I think you already have/can get a reference to those by directly passing the observers to the executor during creation.

In the crash handler, we need to get fuzzer.objectives() so that's why we need fuzzer.

[riesentoaster]

In the crash handler, we need to get fuzzer.objectives() so that's why we need fuzzer.

The only thing ever needed from the fuzzer are the objectives, right? In that case: Why not just pass the OT instead of the entire fuzzer?

[riesentoaster]

So I've tried it, and at least the libafl crate builds: #2892

[riesentoaster]

Alright, it works. Ready to merge from my side.

This doesn't mean this issue is to be closed, but at least the executors are a bit more disentangled from fuzzer.