Fix prioritization of givens over implicits (#21226)

- Fix a typo in `Applications#compare#isAsGood#isGiven` which always 
used `alt1`, to determine if the alternatives passed to `isAsGoodValueType`
 were givens.

- Update `isAsGoodValueType` to not prefer givens over extensions, by
negating the `isGiven` parameter, letting extensions and givens now have
the same priority level as far as that rule is concerned.

- Modify `given`/`implicit` definitions from the `Namer` and the
`PPrint` community-project to resolve ambiguity errors introduced by the changes.

Author: EugeneFlesselle

community-build/community-projects/PPrint

@@ -1830,10 +1830,8 @@ trait Applications extends Compatibility {
             isAsGood(alt1, tp1.instantiate(tparams.map(_.typeRef)), alt2, tp2)
           }
         case _ => // (3)
-          def isGiven(alt: TermRef) =
-            alt1.symbol.is(Given) && alt.symbol != defn.NotGivenClass
           def compareValues(tp1: Type, tp2: Type)(using Context) =
-            isAsGoodValueType(tp1, tp2, isGiven(alt1), isGiven(alt2))
+            isAsGoodValueType(tp1, tp2, alt1.symbol.is(Implicit), alt2.symbol.is(Implicit))
           tp2 match
             case tp2: MethodType => true // (3a)
             case tp2: PolyType if tp2.resultType.isInstanceOf[MethodType] => true // (3a)
@@ -1851,7 +1849,7 @@ trait Applications extends Compatibility {
      *     available in 3.0-migration if mode `Mode.OldImplicitResolution` is turned on as well.
      *     It is used to highlight differences between Scala 2 and 3 behavior.
      *
-     *   - In Scala 3.0-3.5, the behavior is as follows: `T <:p U` iff there is an impliit conversion
+     *   - In Scala 3.0-3.5, the behavior is as follows: `T <:p U` iff there is an implicit conversion
      *     from `T` to `U`, or
      *
      *        flip(T) <: flip(U)
@@ -1870,15 +1868,15 @@ trait Applications extends Compatibility {
      *     for overloading resolution (when `preferGeneral is false), and the opposite relation
      *     `U <: T` or `U convertible to `T` for implicit disambiguation between givens
      *     (when `preferGeneral` is true). For old-style implicit values, the 3.4 behavior is kept.
-     *     If one of the alternatives is a given and the other is an implicit, the given wins.
+     *     If one of the alternatives is an implicit and the other is a given (or an extension), the implicit loses.
      *
      *   - In Scala 3.5 and Scala 3.6-migration, we issue a warning if the result under
      *     Scala 3.6 differ wrt to the old behavior up to 3.5.
      *
      *  Also and only for given resolution: If a compared type refers to a given or its module class, use
      *  the intersection of its parent classes instead.
      */
-    def isAsGoodValueType(tp1: Type, tp2: Type, alt1isGiven: Boolean, alt2isGiven: Boolean)(using Context): Boolean =
+    def isAsGoodValueType(tp1: Type, tp2: Type, alt1IsImplicit: Boolean, alt2IsImplicit: Boolean)(using Context): Boolean =
       val oldResolution = ctx.mode.is(Mode.OldImplicitResolution)
       if !preferGeneral || Feature.migrateTo3 && oldResolution then
         // Normal specificity test for overloading resolution (where `preferGeneral` is false)
@@ -1896,7 +1894,7 @@ trait Applications extends Compatibility {
 
         if Feature.sourceVersion.isAtMost(SourceVersion.`3.4`)
             || oldResolution
-            || !alt1isGiven && !alt2isGiven
+            || alt1IsImplicit && alt2IsImplicit
         then
           // Intermediate rules: better means specialize, but map all type arguments downwards
           // These are enabled for 3.0-3.5, and for all comparisons between old-style implicits,
@@ -1911,8 +1909,8 @@ trait Applications extends Compatibility {
               case _ => mapOver(t)
           (flip(tp1p) relaxed_<:< flip(tp2p)) || viewExists(tp1, tp2)
         else
-          // New rules: better means generalize, givens always beat implicits
-          if alt1isGiven != alt2isGiven then alt1isGiven
+          // New rules: better means generalize, givens (and extensions) always beat implicits
+          if alt1IsImplicit != alt2IsImplicit then alt2IsImplicit
           else (tp2p relaxed_<:< tp1p) || viewExists(tp2, tp1)
     end isAsGoodValueType
 

compiler/src/dotty/tools/dotc/typer/Applications.scala

@@ -1106,7 +1106,7 @@ class Namer { typer: Typer =>
   class ClassCompleter(cls: ClassSymbol, original: TypeDef)(ictx: Context) extends Completer(original)(ictx) {
     withDecls(newScope(using ictx))
 
-    protected implicit val completerCtx: Context = localContext(cls)
+    protected given completerCtx: Context = localContext(cls)
 
     private var localCtx: Context = uninitialized
 

compiler/src/dotty/tools/dotc/typer/Namer.scala

@@ -1,4 +1,4 @@
-//> using options -Xmax-inlines:33
+//> using options -Xmax-inlines:35
 
 import scala.deriving.Mirror
 import scala.compiletime._

tests/pos/i13044.scala

@@ -6,7 +6,8 @@ class NT(t: Tup):
 object NT:
   extension (x: NT)
     def app(n: Int): Boolean = true
-  given Conversion[NT, Tup] = _.toTup
+  given c1: Conversion[NT, Tup] = _.toTup
+  implicit def c2(t: NT): Tup = c1(t)
 
 def test =
   val nt = new NT(Tup())

tests/pos/i19715.scala

@@ -0,0 +1,33 @@
+
+trait Functor[F[_]]:
+  def map[A, B](fa: F[A])(f: A => B): F[B] = ???
+trait Monad[F[_]] extends Functor[F]
+trait MonadError[F[_], E] extends Monad[F]:
+  def raiseError[A](e: E): F[A]
+trait Temporal[F[_]] extends MonadError[F, Throwable]
+
+trait FunctorOps[F[_], A]:
+  def map[B](f: A => B): F[B] = ???
+implicit def toFunctorOps[F[_], A](target: F[A])(implicit tc: Functor[F]): FunctorOps[F, A] = ???
+
+class ContextBounds[F[_]: Temporal](using err: MonadError[F, Throwable]):
+  def useCase = err.raiseError(new RuntimeException())
+  val bool: F[Boolean] = ???
+  def fails = toFunctorOps(bool).map(_ => ())  // warns under -source:3.5, // error under -source:3.6
+
+class UsingArguments[F[_]](using Temporal[F])(using err: MonadError[F, Throwable]):
+  def useCase = err.raiseError(new RuntimeException())
+  val bool: F[Boolean] = ???
+  def works = toFunctorOps(bool).map(_ => ()) // warns under -source:3.5
+
+
+object Minimization:
+
+  trait A
+  trait B extends A
+
+  def test1(using a1: A)(using    b1: B) = summon[A] // picks (most general) a1
+  def test2(using a2: A)(implicit b2: B) = summon[A] // picks (most general) a2, was ambiguous
+  def test3(implicit       a3: A, b3: B) = summon[A] // picks (most specific) b3
+
+end Minimization