Oli Scherer

> the main problem being that we'd never clean up when a toolchain gets uninstalled. Could we pick a miri folder inside the toolchain? Won't work for distro based miri,...

> Or maybe ship the miri sysroot as part of the miri component? That's not possible as we'd have to ship a miri sysroot per target

> The main problem will be making sure it doesn't try to parse the output as JSON. UI test differentiate between stderr and stdout. stdout is dumped directly while stderr...

Ah yes, I completely forgot about this, thanks! So maybe the first step is to move priroda over here and update it and then keep it updated by building it...

Could we add a way for priroda to skip the current instruction instead of stepping on it? Then we could just interpret until the breakpoint instruction is hit (which errors),...

I think the least intrusive way would be adding an `Option` that priroda can fill in with a handler

Yes, that would be great. We can leave all code in this project, and just split it into crates.

> Checking for lifetimes that are not used again at all sounds like another lint, but I may misunderstand this. I mean to not emit the lint unless the lifetime...

> I think it would be more productive if you came to Zulip t-compiler with a precise problem description and we could then brainstorm a nice solution from scratch, considering...

Just to be sure: are you planning to do all the work that needs to be done for whatever solution we end up accepting?