DeMorgan for `all`/`any` in attrs

[llogiq]

We should lint attributes of the style #[cfg(all(not(_), not(_), ..))] and #[cfg(any(not(_), not(_), ..))] and suggest #[cfg(not(any(_, _, ..))] and #[cfg(not(all(_, _, ..))], respectively. Bonus points for further simplification.

[JayKickliter]

DeMorgan?

EDIT: I think I remember it from digital circuits.

[llogiq]

Yes. and(not(x), not(y)) == not(or(x, y)) and vice versa. Just replace and with all and or with any.

[clarfonthey]

Don't we have the opposite lint for boolean expressions? I.e. !(x == y && z > w) becomes x != y || z <= w. I think these two go well together.

[goth-turtle]

@rustbot claim

[xFrednet]

Hey @goth-turtle, just a warning. Clippy has some problems dealing with cfg attributes, as they are not preserved as part of the AST/HIR. This behavior is actually defined in the rust reference (IDK why though). I'm pretty sure that there is currently no good way to implement this lint. At some point, we have to find a solution to deal with cfg attributes, as they also mess with some other lints. If you find a solution, feel free to create a PR, but otherwise it might be better to search for a different issue 😅

[goth-turtle]

Ah, thanks for letting me know!