Replace-if-let-with-match assist generates non-exhaustive match

[fmease]

Given the following code:

fn main() {
    if let Some('!') = todo!() {} else {}
}

The assist replace_if_let_with_match currently generates the following match expression:

fn main() {
    match todo!() {
        Some('!') => {}
        None => {}
    }
}

… which is non-exhaustive leading rustc to emit the error E0004 since it misses the case Some(_). I expected the assist to generate something akin to this:

fn main() {
    match todo!() {
        Some('!') => {}
        _ => {}
    }
}

… since I think the assist is meant to map if-lets to semantically equivalent matches. If that's not a requirement, feel free to close this issue.

---

rust-analyzer version: ed4b312 2021-10-11 stable

[flodiebold]

Duplicate of #8690, can you try the current nightly?

[Veykril]

#10491 only fixed the specific issue there actually, so this one should still be a problem on current nightly.

[fmease]

The same thing happens on nightly.

rust-analyzer version: 87d5ef8 2021-10-19 nightly

What's more, the first example in #8690 (the one involving slice patterns, not the Result<Option<_>, _> one) still generates a non-exhaustive match meaning #10491 only partially addressed the aforementioned issue.

[k-nasa]

I tried cover this case by test, but it passed. k-nasa#1
Next, I tried this case on my editor, generates non-exhaustive match.

The phenomenon confuse me. 🤔

[profetia]

I just did some investigations. It seems when run this case by test, there is something wrong with types in make_else_arm so it just fell back to wildcard directly. While I tried it on IDE, the types problem did not occur and due to some bugs, maybe uncovered corner cases, in #10491, non-exhausitive match happened.

[profetia]

I will try if I can fix this. It seems to be easy.

@rustbot claim

[lnicola]

Fixed in #18797