Niklas Mischkulnig

What that linked section is trying to say: `parcel serve` and `parcel watch` by default don't use any of your explicit `targets` but uses a different one with `last 1...

In Node (which introduced the `exports` field that Next.js/Webpack/Turbopack are imitating), this is actually the intended behavior. It always uses the first entry in the array. The reason it currently...

(Responding to https://github.com/vercel/next.js/issues/68805#issuecomment-2289039278) Due to various implementation detail intricacies in pnpm, that does indeed work because that external package is available even though it wasn't actually declared anywhere as a...

> How does your implementation of Webpack currently accomplish this? This all works in Webpack correctly. Webpack simply doesn't have these error messages, the output is/would be the same. >...

> So in addition to using NPM, your repro actually [adds the packages to Webpack externals](https://github.com/mischnic/turbopack-externals-repro/blob/main/next.config.js?rgh-link-date=2024-08-17T00%3A18%3A45Z#L9), which we are in control of in userland and I would be happy to...

Do you know/remember why you are externalizing `drizzle-kit`, `pino`, `pino-pretty` and `graphql`?

Great, I'll take a look! > SQLite is a brand new database adapter that we support and it looks like it's currently incompatible with Turbopack You mentioned this in the...

> OK - we have a repo using Postgres set up that demonstrates the issue. > [jmikrut/nextjs-68805-pg](https://github.com/jmikrut/nextjs-68805-pg) The latest Turbopack in canary 120 compiles that just fine when not making...

> Oh nice, so pg no longer causes issues? I'm not sure if I hit all code paths at runtime, but apparently it works. > What about `mongoose` and `mongodb`...

Please add a `turbopack-tests` unit test for that case of the pattern ending up with a blackslash