Don't lint let_unit_value when needed for type inferenece

[Jarcho]

fixes: #1502

Pinging @dtolnay. I think this is enough to fix the issue. Do you have a good list crates to test this on?

changelog: Don't lint let_unit_value when needed for type inference

[rust-highfive]

r? @llogiq

(rust-highfive has picked a reviewer for you, use r? to override)

[llogiq]

I don't know which crates suffered this problem, but code-wise this looks good to me.

[bors]

☔ The latest upstream changes (presumably #8602) made this pull request unmergeable. Please resolve the merge conflicts.

[llogiq]

Sorry for letting this wait for so long, but I lack insight into actual cases in the wild, so I'm not sure if they are all caught

However, as even if this misses some cases it is still a good improvement, r=me with a rebase.

[Jarcho]

This will now handle block expression and branches when checking if type inference is needed.

The only thing I can think of that would also be a false positive would be a something like:

trait Foo {
    type T;
    fn get_t(&self) -> Self::T;
    fn use_t(&self, _: T);
}
struct Bar;
impl Foo for Bar {
    type T = ();
    ...
}

fn f(bar: Bar) {
    let x = bar.get_t();
    bar.use_t(x);
}

But I'm pretty sure that already doesn't lint.

[bors]

☔ The latest upstream changes (presumably #8660) made this pull request unmergeable. Please resolve the merge conflicts.

[llogiq]

Ok, I might be offline for a while, but r+ when rebased. Thanks and again sorry for the wait.

@bors delegate+

[bors]

✌️ @Jarcho can now approve this pull request

[Jarcho]

@bors r+

[bors]

📌 Commit 70f7c62 has been approved by Jarcho

[bors]

⌛ Testing commit 70f7c62 with merge 0bc93b6...

[bors]

☀️ Test successful - checks-action_dev_test, checks-action_remark_test, checks-action_test
Approved by: Jarcho
Pushing 0bc93b6 to master...