Avoid using unexpected cfgs in user code

[Urgau]

This PR adds:

• the ability to declare a feature without cfg checking
• and mark every feature with missing corresponding target feature cfgs as such

This is a prerequisites to rust-lang/rust#132577 which wants to enable reporting of the unexpected_cfgs lint in external macros and is unfortunately triggering on the macros call generated by stdarch, like is_x86_feature_detected!, ...

I have computed the list of missing target feature cfgs by comparing:

• rg "[ ]+@feature: .*: \"(.*)\";" -r '$1' --no-filename crates/std_detect/src/detect/ | sort | uniq
• with tests/ui/check-cfg/well-known-values.stderr¹ (from the main Rust repo)

(actual.txt, referenced.txt, diff-actual-referenced.txt)

Footnotes

1. https://github.com/rust-lang/rust/blob/e8c698bb3bdc121ac7f65919bd16d22f6567a3f1/tests/ui/check-cfg/well-known-values.stderr#L177 ↩

[rustbot]

r? @Amanieu

rustbot has assigned @Amanieu.
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

[Amanieu]

Could we have some way of checking, in stdarch CI, that when rustc adds support for a feature we don't forget to re-enable the flag here?

[Urgau]

Could we have some way of checking, in stdarch CI, that when rustc adds support for a feature we don't forget to re-enable the flag here?

Sure, I just pushed (with the small fix) a commit that add a test function as part of the feature! macro. That automatically generated test function denies both the unexpected_cfgs, and unfulfilled_lint_expectations lint such as the for the known target feature cfgs it will trigger the unexpected_cfgs lint if it isn't marked with without cfg check: true and if it's an unexpected cfg it will trigger the lint which is expected with an #[expect] attribute.

I have successfully tested it locally with the modification from upstream.