Diagnostics for label traits

[SpecificProtagonist]

Objective

Diagnostics for labels don't suggest how to best implement them.

error[E0277]: the trait bound `Label: ScheduleLabel` is not satisfied
   --> src/main.rs:15:35
    |
15  |     let mut sched = Schedule::new(Label);
    |                     ------------- ^^^^^ the trait `ScheduleLabel` is not implemented for `Label`
    |                     |
    |                     required by a bound introduced by this call
    |
    = help: the trait `ScheduleLabel` is implemented for `Interned<(dyn ScheduleLabel + 'static)>`
note: required by a bound in `bevy_ecs::schedule::Schedule::new`
   --> /home/vj/workspace/rust/bevy/crates/bevy_ecs/src/schedule/schedule.rs:297:28
    |
297 |     pub fn new(label: impl ScheduleLabel) -> Self {
    |                            ^^^^^^^^^^^^^ required by this bound in `Schedule::new`

Solution

diagnostics::on_unimplemented and diagnostics::do_not_recommend

Showcase

New error message:

error[E0277]: the trait bound `Label: ScheduleLabel` is not satisfied
   --> src/main.rs:15:35
    |
15  |     let mut sched = Schedule::new(Label);
    |                     ------------- ^^^^^ the trait `ScheduleLabel` is not implemented for `Label`
    |                     |
    |                     required by a bound introduced by this call
    |
    = note: consider annotating `Label` with `#[derive(ScheduleLabel)]`
note: required by a bound in `bevy_ecs::schedule::Schedule::new`
   --> /home/vj/workspace/rust/bevy/crates/bevy_ecs/src/schedule/schedule.rs:297:28
    |
297 |     pub fn new(label: impl ScheduleLabel) -> Self {
    |                            ^^^^^^^^^^^^^ required by this bound in `Schedule::new`

[alice-i-cecile]

Looks like CI is mad about the attribute. What's different vs the existing usages?

[SpecificProtagonist]

@LikeLakers2 Would you happen to know what the correct formulation for the expect & allow attribute here is?

I though I found a workaround by putting the trait impl into a module inside a const _: () = {…, but that doesn't work for doctests because super accesses the parent module, not parent scope.

[LikeLakers2]

@SpecificProtagonist I'm not actually sure what the solution for this would be.

See, those lint attributes should be applying, but rustc clearly isn't applying them as we expect.

Normally, my next step would be to move the lint attributes up a level. For example, if I have a module, and inside that module is a struct that I'm trying to place those lint attributes on... then I move those lint attributes to apply to the entire module.

Only problem is, there's not exactly an encapsulating item here... So you'd probably have to make an encapsulating item here.

But you seem to have tried that - and I don't know how to fix it otherwise without causing some churn. If you're okay with some churn, here's a couple ideas:

• You could put the #[diagnostic::do_not_recommend] behind a cfg_attr that will only include it on nightly... But then you'd have to come back and change it once it's stable.

• You could also modify the root Cargo.toml to include lints.rust.unknown_or_malformed_diagnostic_attributes = "allow", but I don't know how the maintainers would feel about that.

Otherwise, I might suggest giving up on this until #[diagnostic::do_not_recommend] is stable.

[SpecificProtagonist]

Thanks! I'm going with the last option of removing the do_not_recommend for now and adding it back when stable.