`if_assert`

What it does

It should deny or warn when a predicate is checked by both an assert (or any similar built-in macro) and any conditional control-flow construct, in any order

Related: #8567

Advantage

• It removes a redundant check that's certainly a bug
• Less confusion for readers
• No change in code-gen, as the optimizer could catch the dupe and eliminate it

Drawbacks

• False negatives because of custom macros
• Cannot be auto-fixed, because intent isn't clear (wrong?)

Example

fn f<T: std::fmt::Debug + PartialEq>(a: T, b: T) {
    if a == b {
        assert_eq!(a, b);
        // do something
    }
}

fn g<T: std::fmt::Debug + PartialEq>(a: T, b: T) {
    assert_eq!(a, b);
    if a == b {
        // do something
    }
}

Could be written as:

fn f<T: std::fmt::Debug + PartialEq>(a: T, b: T) {
    if a == b {
        // do something
    }
}

fn g<T: std::fmt::Debug + PartialEq>(a: T, b: T) {
    assert_eq!(a, b);
    // do something
}

Start with related issue #8567 and the two examples in this issue. Define the lint behavior for both assertion-before-conditional and conditional-before-assertion cases, then verify that it reports redundant predicate checks while accounting for the stated false negatives around custom macros.