Recently, I've seen several various panics in chalk (the most recent one here). Should I just try to manually minimize them to a rust-analyzer test-case?
Or should we invest into some kind of tracing/bug-reproducing architecture for Chalk? I imagine if I could just dump everything chalks looks at into JSON, the minization would be faster then from-source minimization?
I think both sound useful
But I do think that investment would pay off big time
e.g. that infinite loop bug was fairly easy to fix once we narrowed it down
but getting it to reproduce was a lot of the time
I did start building something like that a while ago... the code has probably rotted by now though, and I didn't go very far anyway