I think it's butter to return a Result for invalid add

[quanweiZhou]

https://github.com/BurntSushi/chan/blob/master/src/wait_group.rs#L57

    pub fn add(&self, delta: i32) {
        let mut count = self.0.count.lock().unwrap();
        *count += delta;
        assert!(*count >= 0);
        self.0.cond.notify_all();
    }

[BurntSushi]

I disagree. But in any case, it doesn't matter. Did you miss the first line of the README? https://github.com/BurntSushi/chan#this-crate-has-reached-its-end-of-life-and-is-now-deprecated

[quanweiZhou]

@BurntSushi I think it can constrain the user's usage, but as a lib there should not trigger the panic, the Result can be returned and the user decides by himself.

[BurntSushi]

but as a lib there should not trigger the panic

This is wrong. It is perfectly idiomatic to set a precondition for library APIs. See RefCell::borrow or str::split_at.

See also: https://blog.burntsushi.net/unwrap/

[BurntSushi]

And this isn't just Rust. Take a look at popular C libraries like PCRE2. There are innumerable ways to provoke UB by giving PCRE2 API functions invalid inputs. Those are preconditions and they are strictly worse than what Rust tends to: panic.

[quanweiZhou]

Thanks for your reply.