[mejrs]

It's an error to have #[inline]#[rustc_force_inline]fn blah(){}, but they both lower to the same InlineAttr, so this would error and then have an attribute list of multiple inline attrs, which breaks the invariant that a list of attributes has no duplicate variants

[rustbot]

Some changes occurred in compiler/rustc_attr_parsing

cc @jdonszelmann, @JonathanBrouwer

[rustbot]

r? @jackh726

rustbot has assigned @jackh726.
They will have a look at your PR within the next two weeks and either review your PR or reassign to another reviewer.

Use r? to explicitly pick a reviewer

Why was this reviewer chosen?

The reviewer was selected based on:

• Owners of files modified in this PR: compiler
• compiler expanded to 77 candidates
• Random selection from 19 candidates

[JonathanBrouwer]

r? me

[rust-bors]

☔ The latest upstream changes (presumably #164894) made this pull request unmergeable. Please resolve the merge conflicts by rebasing.

[JonathanBrouwer]

which breaks the invariant that a list of attributes has no duplicate variants

I'm leaning towards that that's fine, since one of the attributes is rustc internal and the broken invariant is not really observable. Imo this change makes the code less readable. What is the concrete advantage of this change?

Another approach would be replacing rustc_inline_force with #[inline(rustc_force)] or similar, then we can have only one attribute.

View changes since this review

[mejrs]

I wanted to add a (debug) assertion to check for no duplicates (except doc comments) in #164260, but hit it for a rustc_force_inline test instead.

Another approach would be replacing rustc_inline_force with #[inline(rustc_force)] or similar, then we can have only one attribute.

The downside of having #[inline(rustc_force)] (or #[inline(force)]) is that this sort of looks like it's on a path towards stabilization, which IIRC it is very much not.

[JonathanBrouwer]

I wanted to add a (debug) assertion to check for no duplicates (except doc comments) in #164260, but hit it for a rustc_force_inline test instead.

Another way would be to also add a similar exception for inline, or to make it a delayed bug such that it is allowed when there are errors.

The downside of having #[inline(rustc_force)] (or #[inline(force)]) is that this sort of looks like it's on a path towards stabilization, which IIRC it is very much not.

Yeah I proposed rustc_force specifically so it would be clear that it's internal. Feel free to suggest another name that makes this even clearer.