Generated `#[deny(safe_packed_borrows)]` is not compatible with `#![forbid(future_incompatible)]`

[yoshuawuyts]

The compiler's built-in #[deny(future_incompatible)] seems to be incompatible with pin-project-internal. It seems #[deny(safe_packed_borrows)] will be disallowed in the future, but pin-project generates code that uses that lint under certain circumstances. Given the Rust 2021 edition may turn all 2018 warnings into errors, this may become a problem.

I'm not exactly sure what triggers this, but adding the lint to the current main of tide surfaces the error. Though I feel I've seen this happen before when working on other crates. I'm not sure what the right outcome here is, but thought it might be important enough to raise. Thanks!

#34 seems related.

Screenshot

[Aaron1011]

It looks like you're actually using pin_project_lite.

This looks like a bug in rustc - it's emitting an error about this #[deny(safe_packed_borrows)] getting 'overruled'. I think that error is only supposed to occur when a #[warn] or #[allow] get overruled.

[Aaron1011]

I think it would be fine for pin-project-lite to use #[forbid], though.

[taiki-e]

I definitely do not recommend using forbid on the lints other than unsafe_code lint. They will constant improvement/change, and code that was not warned at the release time may become unavailable. Especially for lint groups like future_incompatible, it's possible that a new lint will be added.

That said, as @Aaron1011 said, this can be easily fixed by switching #[deny(safe_packed_borrows)] to #[forbid(safe_packed_borrows)], so I'll do that.

---

This looks like a bug in rustc - it's emitting an error about this #[deny(safe_packed_borrows)] getting 'overruled'. I think that error is only supposed to occur when a #[warn] or #[allow] get overruled.

I guess this is intentional behavior, as it should be possible to allow lint in scopes where lint is denied, but it's certainly not intuitive...

[taiki-e]

Fixed in pin-peoject 0.4.24 and pin-project-lite 0.1.8

[yoshuawuyts]

They will constant improvement/change, and code that was not warned at the release time may become unavailable.

I hadn't given this enough thought, and you're absolutely right. I'll have to revise this. Thank you!