Alexander added a comment - 19/Mar/12 8:34 AM It's very similar to https://issues.scala-lang.org/browse/SI-5579

Paul Phillips added a comment - 20/Mar/12 2:40 AM With the intermediate assignment, it seems to infer "Success[_ >: Some [Int] with None.type <: Option [Int] ]" as the result of the match. Then it decides this is incompatible with the typevar. Without the intermediate assignment, it is able to connect the dots between ?A and the match. I guess my question is, why doesn't it simplify Success[_ >: Some [Int] with None.type <: Option [Int] ] to Success[Option [Int] ]. Of course if you do that by hand with type annotations, everything works.

Paolo G. Giarrusso added a comment - 21/Mar/12 1:04 AM - edited In my experience using wildcards a lot more than appropriate, the error message seems normal. It just says that you cannot pass a BoundedWildcardType as a type parameter, as it often happens (in my experience). Reading the source, this does not make sense since recursion should end here: https://github.com/scala/scala/blob/d8ba5d091e5641553b438ef9930a6023a2709dcd/src/compiler/scala/reflect/internal/Types.scala#L5347 However, it needs to pass a couple of layers before. For the error to happen, in fact, it is required a double mismatch: Success vs Result and type variable vs bounded wildcard. Either making Success covariant or exchanging Success with Result either at the Apply.apply signature or at its caller hides the bug. <console>:13: error: no type parameters for method apply: (f: Int => Result[A])Apply[A] in object Apply exist so that it can be applied to arguments (Int => Success[_ >: Some[Int] with None.type <: Option[Int]]) --- because --- argument expression's type is not compatible with formal parameter type; found : Int => Success[_ >: Some[Int] with None.type <: Option[Int]] required: Int => Result[?A] def goo : Apply[Option[Int]] = Apply { i =>

Jason Zaugg added a comment - 10/Feb/13 10:43 AM - edited Here's another look at this: % ~/code/scala2 cat sandbox/t5585.scala class Result[+A]   case class Success[A](x: A) extends Result[A]   class Apply[A]   object Apply { def apply[A](f: Int => Result[A]): Apply[A] = new Apply[A] }   object TestUnit { def goo : Apply[Option[Int]] = Apply.apply { i => val p = (0: Int) match { case 1 => Success(Some(1)) case _ => Success(None) } p }   // // OK //     def f[A](r: Result[A]): Apply[A] = ??? f { (0: Int) match { case 1 => Success(Some(1)) case _ => Success(None) } }: Apply[Option[Int]]   f { val p = (0: Int) match { case 1 => Success(Some(1)) case _ => Success(None) } p }: Apply[Option[Int]]     case class H[-A, +B](b: B) def h[A](r: H[Int, Result[A]]): Apply[A] = ??? h { val p = (0: Int) match { case 1 => Success(Some(1)) case _ => Success(None) } H(p) }: Apply[Option[Int]]     // // KO //   def g[A](r: Function1[Int, Result[A]]): Apply[A] = ??? g { val p = (0: Int) match { case 1 => Success(Some(1)) case _ => Success(None) } (i: Int) => p }: Apply[Option[Int]]   } % qbin/scalac sandbox/t5585.scala   % git diff diff --git a/src/compiler/scala/tools/nsc/typechecker/Typers.scala b/src/compiler/scala/tools/nsc/typechecker/Typers.scala index dc5491a..00af937 100644 --- a/src/compiler/scala/tools/nsc/typechecker/Typers.scala +++ b/src/compiler/scala/tools/nsc/typechecker/Typers.scala @@ -2837,7 +2837,7 @@ trait Typers extends Modes with Adaptations with Tags { // } val formals = vparamSyms map (_.tpe) val body1 = typed(fun.body, respt) - val restpe = packedType(body1, fun.symbol).deconst.resultType + val restpe = body1.tpe.deconst val funtpe = typeRef(clazz.tpe.prefix, clazz, formals :+ restpe) // body = checkNoEscaping.locals(context.scope, restpe, body) treeCopy.Function(fun, vparams, body1).setType(funtpe) An understanding of `packedType` seems important here. https://github.com/scala/scala/commit/225fac5

Paolo G. Giarrusso added a comment - 12/Jun/13 12:31 AM > In my experience using wildcards a lot more than appropriate, the error message seems normal. It just says that you cannot pass a BoundedWildcardType as a type parameter, as it often happens (in my experience). I should clarify: with "normal" I meant "a bug I see all the time", not "the right thing to do". At least, that's what I should have meant.