Impls and impl items inherit dead_code lint level of the corresponding traits and trait items

[mu001999]

https://github.com/rust-lang/rust/blob/master/compiler/rustc_passes/src/dead.rs#L360-L361 won't insert assoc items into the live set, so that impl items cannot be marked live.

This PR lets impls and impl items can inherit lint levels of the corresponding traits and trait items.

Fixes #144060

r? @petrochenkov

[compiler-errors]

This behavior seems really non-local and pretty broad. #[allow(dead_code)] on a trait definition suppresses all dead code warnings in all impls with this change?

[mu001999]

#[allow(dead_code)] on a trait definition suppresses all dead code warnings in all impls with this change?

@compiler-errors yes, and I think this makes sense. We would have only one trait definition but many impls, letting #[allow(dead_code)] on traits propagate to impls is convenient. Otherwise we need to add #[allow(dead_code)] also on each impls which use unused items.

[petrochenkov]

r? @compiler-errors

[petrochenkov]

Hmm, why did rustbot assign it to me again?
I'll just block this on #144863 now.
@rustbot blocked

[mu001999]

Hmm, why did rustbot assign it to me again?

I tried reassigning, but rustbot didn’t respond after waiting a while, so I deleted the comment. 😂

[bors]

☔ The latest upstream changes (presumably #144863) made this pull request unmergeable. Please resolve the merge conflicts.

[mu001999]

@rustbot ready

[tgross35]

@cjgillot gentle ping for a review here

[wesleywiser]

Nominated the issue for lang team discussion, let's see what they think before merging.

[mu001999]

oh, I missed the non-local in the note from @compiler-errors

and if the trait is non-local, the dead-code lint level won't spread to the local impls. cc @wesleywiser

we only care about local traits, see https://github.com/rust-lang/rust/pull/144113/files#diff-719b4f36bbbf3b3bf48683429408f5d27e5633f70c1144a93554bd536d5bf655R786

[tmandry]

@rfcbot reviewed

[traviscross]

How does what's proposed in this PR compare with @nikomatsakis' model in #144060 (comment)?

[mu001999]

How does what's proposed in this PR compare with @nikomatsakis' model in #144060 (comment)?

For examples given by @nikomatsakis:

pub trait Foo { }
impl Foo for u32 { }

struct PrivateType;
impl Foo for PrivateType { } // <-- warns as dead, even though Foo is public

struct AnotherPrivateType;
impl Foo for AnotherPrivateType; { } // <-- warns as dead, even though Foo is public

works like before.

#[allow(dead_code)]
pub trait Foo { }
impl Foo for u32 { }

struct PrivateType;
impl Foo for PrivateType { } // <-- no warning, trait is "allowed"

struct AnotherPrivateType;
impl Foo for AnotherPrivateType; { } // <-- no warning, trait is "allowed"

This one is related to this PR. And it will work as expected after this PR.

pub trait Foo { }
impl Foo for u32 { }

struct PrivateType;
#[allow(dead_code)]
impl Foo for PrivateType { } // <-- no warning, impl is allowed

struct AnotherPrivateType;
impl Foo for AnotherPrivateType; { } // <-- warns as dead, even though Foo is public

works like before.

[traviscross]

Thanks. OK.

@rfcbot reviewed

[rust-rfcbot]

🔔 This is now entering its final comment period, as per the review above. 🔔

[nikomatsakis]

@rfcbot reviewed

Nice work @mu001999 !

[rust-rfcbot]

The final comment period, with a disposition to merge, as per the review above, is now complete.

As the automated representative of the governance process, I would like to thank the author for their work and everyone else who contributed.

This will be merged soon.

[mu001999]

@wesleywiser I think the lable S-blocked can be removed now, right?

[mu001999]

And friendly ping @cjgillot

[tgross35]

@wesleywiser I think the lable S-blocked can be removed now, right?

And friendly ping @cjgillot

Fyi you can use @rustbot label -S-blocked to adjust labels yourself. @rustbot reroll allows you to ask for a new reviewer given it's been a while (but give Camille another week or two since they self-assigned and I just adjusted the labels now.)

[mu001999]

Fyi you can use @rustbot label -S-blocked to adjust labels yourself.

thanks, glad to know that ;)

[mu001999]

@cjgillot friendly ping

[wesleywiser]

r? compiler

[mati865]

The idea sounds good to me (based on the discussions), but I know nothing of this code.

@rustbot reroll