Regression: Ambiguous implicits should error

Compiler version

• 3.0.0-RC1
• 3.0.1-RC1-bin-20210412-69108bf-NIGHTLY

Minimized code

class Singleton[A]

object Singleton {

  implicit val Unit: Singleton[Unit] =
    new Singleton

  sealed trait Not[A]
  implicit def noSingletonFor[A]: Not[A] = null
  implicit def singletonFor1[A: Singleton]: Not[A] = null
  implicit def singletonFor2[A: Singleton]: Not[A] = null

  implicitly[Singleton    [Unit]] // ok
  implicitly[Singleton.Not[Unit]] // error expected
  implicitly[Singleton.Not[Int ]] // ok
}

Output

It all compiles.

Expectation

The same error as Scala 2:

x.scala:14: error: ambiguous implicit values:
 both method noSingletonFor in object Singleton of type [A]Singleton.Not[A]
 and method singletonFor1 in object Singleton of type [A](implicit evidence$1: Singleton[A]): Singleton.Not[A]
 match expected type Singleton.Not[Unit]
  implicitly[Singleton.Not[Unit]] // error expected
            ^
1 error

[smarter]

You can still cross-compile by defining a NotGiven for Scala 2: https://github.com/typelevel/fs2/blob/main/core/shared/src/main/scala-2/fs2/compat/NotGiven.scala https://github.com/typelevel/fs2/blob/main/core/shared/src/main/scala-3/fs2/compat/NotGiven.scala (something that could be added to https://github.com/scala/scala-collection-compat perhaps?)

[joroKr21]

You can try adding a DummyImplicit to settle the score