Hi all, a bit of a meta-question.
If I have a big mess of heavily generic code for which rust-analyzer isn't correctly inferring types, it is acceptable to create a GitHub issue whose repro instructions are basically just "go download this big mess of code from here" ?
I've done this a few times already, but I'm not sure if it's annoying to the team to get things that aren't minimal repros.
well, it's going to make it more likely that your bug will get looked at if you have a small repro, but any actual repro is already helpful ;)
The biggest risk is that someone from the r-a team spends time figuring out what is going wrong, and then determines it is a dupe of a known issue
it can be really helpful to try to minimize the repro though, you can often remove large swathes of dependencies and code. keep in mind that the repro doesn't need to _work_. you can remove lots of code and replace most function bodies by
That is a good point.
Commenting out deps in Cargo.toml and module declarations is also a good start
but you could also post your big repro first, maybe I or someone else will immediately recognize it as a known problem. If it's involving lots of traits etc. that's unlikely though
I wonder if we could create an LSP record/replay feature
We can already log the LSP messages and we could presumably replay them to a new server instance.
This question became irrelevant in this case, since I was indeed able to come up with a simple repro: https://github.com/rust-analyzer/rust-analyzer/issues/7864
But it is good to know the expectation for the future :)