Dmitry Vyukov

> Thing is ... I don't know how to execute regression testing with go-fuzz :P Such mode does not exist in go-fuzz. Should be possible to add. But there must...

> So what I want is to expose an endpoint that returns the stats of a particular fuzz instance as json (or any other structure format for what I care)....

> Where does that come from ? From `-v=1` flag. But we can change the other output line as well.

Hi @Komosa, Thanks for looking into this. Go-fuzz already has a scoring system and chooses inputs according to score (see updateScores function in hub.go). So I think we just need...

@Komosa This is hard questions. The only more-or-less objective way to tune such things is to measure impact on real fuzzing targets (how does coverage progress? at what coverage level...

Particular machine/env does not matter as long as you compare before/after the change in the same conditions. > Of course I plan to automate such tests. It would be great...

Yes, corpus is number of different, interesting inputs. But each new input is exponentially harder to obtain, so linear grow is exponentially better. I think, that cover field is a...

What would be correct links?

No, because you did not accept my pull request :) https://github.com/akaros/go-akaros/pull/1 But seriously, unfortunately that won't help as we need Go 1.9+.

From @brho: ``` Short version: we're still sorting out the best, full-proof way to handle userspace passing bad pointers to the kernel for delicate operations. ... this will happen for...