Pattern matching regression warning in inline quotes with sealed class

[soronpo]

Compiler version

v3.1.0 snapshot of latest master

Minimized code

Macro.scala

import scala.quoted.*

sealed class Foo()
inline def hh(): Unit = ${ interpMacro() }

private def interpMacro()(using Quotes): Expr[Unit] =
  import quotes.reflect.*
  '{
    val res: Either[String, (Foo, Foo)] =
      Right((new Foo, new Foo))
    val (a, b) = res.toOption.get
  }

Main.scala

@main def main: Unit = hh()

Output

[warn]  1 |@main def main: Unit = hh()
[warn]    |                       ^^^^
[warn]    |                       match may not be exhaustive.
[warn]    |
[warn]    |                       It would fail on pattern case: (_, _)

Expectation

No warning.

[dwijnand]

How latest? Does it include 094a3c5, because we added a @unchecked there.

[soronpo]

I tried a git pull now and I'm on the updated master

[nicolasstucki]

The match for val (a, b) = ... seems to lose its @unchecked annotation when inlining.

We have

private[<empty>] def interpMacro()(using x$1: quoted.Quotes): 
      quoted.Expr[Unit]
     = 
      {
        import x$1.reflect.*
        '{
          {
            val res: Either[String, Tuple2[Foo, Foo]] = 
              Right.apply[Nothing, (Foo, Foo)](
                Tuple2.apply[Foo, Foo](new Foo(), new Foo())
              )
            val $1$: (Foo, Foo) = 
              res.toOption.get:(Foo, Foo) @unchecked match 
                {
                  case Tuple2.unapply[Foo, Foo](a @ _, b @ _) => 
                    Tuple2.apply[Foo, Foo](a, b)
                }
            val a: Foo = $1$._1
            val b: Foo = $1$._2
            ()
          }
        }.apply(x$1)
      }

but we get

@main() def main: Unit = 
      {
        val res: Either[String, Tuple2[Foo, Foo]] = 
          Right.apply[Nothing, (Foo, Foo)](
            Tuple2.apply[Foo, Foo](new Foo(), new Foo())
          )
        val $1$: (Foo, Foo) = 
          (res.toOption.get:(Foo, Foo) match 
            {
              case Tuple2.unapply[Foo, Foo](a @ _, b @ _) => 
                Tuple2.apply[Foo, Foo](a, b)
            }
          )
        val a: Foo = $1$._1
        val b: Foo = $1$._2
        ()
      }:Unit

[nicolasstucki]

The regression happen on aeb3db1. It is strange because that commit only touched the Unapply definition but the missing @unchecked is in a Typed in the scrutinee.

Will revert this commit and look into the actual issue later.