Missing "unreachable pattern" warning on char ranges

[Nadrieril]

I tried this code (playground):

match 'a' {
    '\u{0}'..='\u{D7FF}' | '\u{E000}'..='\u{10_FFFF}' => {},
    '\u{D7FF}'..='\u{E000}' => {}, // should be unreachable
}

The first pattern by itself is exhaustive, because it covers all the valid chars. The second pattern looks like it includes more chars than the first, but it doesn't, because those extra chars are all invalid. So the second pattern should be marked unreachable.

This fails on all rust versions on the playground, and I think it's been that way since we implemented exhaustive_integer_patterns. I discovered it while working on the exhaustiveness algorithm.

@rustbot modify labels: +A-exhaustiveness-checking

[leonardo-m]

Are such intervals ever going to change in the UNICODE world?

[Nadrieril]

Hm, good question, I have no idea. But they are already hardcoded in the compiler since the following is exhaustive:

match 'a' {
    '\u{0}'..='\u{D7FF}' | '\u{E000}'..='\u{10_FFFF}' => {}
}

So this issue is more about making the two consistent. If unicode changes it would change the above too.

[Dylan-DPC]

Current error:

warning: multiple patterns overlap on their endpoints
 --> src/main.rs:4:5
  |
3 |     '\u{0}'..='\u{D7FF}' | '\u{E000}'..='\u{10_FFFF}' => {},
  |     --------------------   -------------------------- this range overlaps on `'\u{e000}'`...
  |     |
  |     this range overlaps on `'\u{d7ff}'`...
4 |     '\u{D7FF}'..='\u{E000}' => {}, // should be unreachable
  |     ^^^^^^^^^^^^^^^^^^^^^^^ ... with this range
  |
  = note: you likely meant to write mutually exclusive ranges
  = note: `#[warn(overlapping_range_endpoints)]` on by default

[Dylan-DPC]

Closing this as the warning message now indicates that the 2nd arm is unreachable.

[Nadrieril]

That's an unrelated lint, it does not indicate that the 2nd arm is unreachable (only my comment in the code does). This still shows the issue. (thx so much for triaging tho!)