Warn on useless ?Sized anti-bound

[rileyshahar]

What it does

Consider the following (admittedly silly) code:

fn id<T: Copy + ?Sized>(t: &T) -> &T {
    t
}

The author clearly meant for this method to work on unsized T, but the Copy bound implicitly implies Clone, which implies Sized, so the compiler will deduce that T is Sized. Right now, it compiles and passes clippy silently. Instead, clippy should warn that the ?Sized anti-bound is useless, and suggest removing one of the two bounds.

Here's a silly-but-slightly-less-so example:

unsafe fn deref<T: Copy + ?Sized>(ptr: *const T) -> T {
    *ptr
}

Here, someone might have this function without the ?Sized anti-bound (Copy is required to move out of the pointer), realize they want to call it with an unsized type, and modify it like so. (In my case, I was trying to construct an example of ?Sized and pointer dereferencing, and ended up surprised for a few minutes that this compiled.)

A similar case is #1463, i.e. bounds where one implies the other, for example T: Copy + Clone. That seems potentially less bad, because there's no issue with the code's intent; it's just an unnecessarily complex bound.

Lint Name

useless_anti_bound

Category

suspicious

Advantage

• Clearer intent
• Removes useless code

Drawbacks

None I can think of.

Example

fn id<T: Copy + ?Sized>(t: &T) -> &T {
    t
}

Could be written as:

fn id<T: Copy>(t: &T) -> &T {
    t
}