Nathan Hammond

> General comment, and not a blocker on this, but using AbsoluteSystemPath is causing us to lose a lot of the type safety we might otherwise get. Calls to things...

Checkpoint, the only open items here: - [Is there a better option for than `reflect` for existence checks?](https://github.com/vercel/turborepo/pull/1991#discussion_r983065811) - [`SafeMkdirAll`'s perf optimization.](https://github.com/vercel/turborepo/pull/1991#discussion_r983077074)

I just noticed that the imported build is redistributing the macOS SDK, we should _probably_ not be doing that.

Can you update docker? When I run this locally I error out of `docker` immediately. This isn't likely a `turbo`-related issue. I'm closing it for now, but if you can...

@yovanoc Looking now.

@tknickman I'm tagging you in on this one; can reproduce, but seems quite strange to me.

There are actually two layers of cache in play here. The ESLint cache is not invalidating depending upon the values inside of `turbo.json`.

After taking a look, https://github.com/eslint/eslint/blob/8cc0bbe440dc5e6af6ef02f00d0514a40ca07c24/lib/cli-engine/lint-result-cache.js#L96-L114 is how ESLint invalidates its cache. If any plugins require input _beyond_ this (which ours does) the ESLint cache will always be stale after first...

What works: - change `turbo.json`, save the target file, run lint. What doesn't work: - change `turbo.json`, run lint.

If we were going to make this work "right" we would need to open an RFC for ESLint As a straw proposal this could be enabling any plugin to define...