chore: tag format_ident with #[clippy::format_args]

[nyurik]

Since 1.85, Clippy supports #[clippy::format_args] attribute - see docs - this way all users of this macro will get the benefits of clippy format arg validations.

The only possible problem is that format_ident!("Foo{bar}") might not be edition2021+ compliant (i.e. may treat that string as verbatim). I suspect it might be ok to modify the macro implementation to always treat a single param macro as a format string because anything with curlies in it is not a valid identifier.

Fixes #290
Related #206

[dtolnay]

Could you share an example of a diagnostic output that this PR would cause clippy to emit?

[nyurik]

From https://github.com/rust-lang/rust-clippy/blob/master/CHANGELOG.md#rust-185 (implemented in rust-lang/rust-clippy#9948) - many existing lints, namely format_in_format_args, recursive_format_impl, to_string_in_format_args, uninlined_format_args, and unused_format_specs will treat format_ident! exactly the same as they treat format!. So any diagnostic output these lints show would simply show up for format_ident!-using code.

Note that the above list is not complete - any Clippy lint that internally calls is_format_macro() will do it too.

[dtolnay]

I know about that, and I have used this feature in dtolnay/anyhow#426. But specifically for quote::format_ident I want to see an actual diagnostic emitted by clippy as a result of this.

[dtolnay]

Specifically, in the following part of the expansion, notice how any $arg would always be evaluated outside the context of a format_args!. It is not obvious that any of the lints you have named would apply to the resulting syntax tree, nor that if they apply, they would produce a suggestion that is meaningful and correct.

quote/src/format.rs

Lines 163 to 165 in ad9fb40

	([$span:expr, $($fmt:tt)*] $arg:expr, $($rest:tt)*) => {
	match $crate::__private::IdentFragmentAdapter(&$arg) {
	arg => $crate::format_ident_impl!([$span.or(arg.span()), $($fmt)*, arg] $($rest)*),

[nyurik]

You are right, it seems current Clippy logic either does not see this macro as a valid candidate for evaluation, or some other thing. I could try to look into Clippy logic itself, or what would you suggest?