Imported From: https://issues.scala-lang.org/browse/SI-9331?orig=1
Reporter: @retronym
See #8039

@retronym said:
See also: scala/scala#3519, which has since touched the code that introduced the regression.

@retronym said:

scalac-hash 46e8eceecc5b5f6b49075550e54f035126fc256a~1 -Xdev sandbox/test.scala
topic/require-java-8 ~/code/scala scalac-hash 46e8eceecc5b5f6b49075550e54f035126fc256a -Xdev sandbox/test.scala
warning: !!! isSubType2(Proxy[PP], P, _) compares HK type with proper type
warning: !!! isSubType2(Proxy[PP], P, _) compares HK type with proper type
sandbox/test.scala:9: error: kinds of the type arguments (Proxy[PP]) do not conform to the expected kinds of the type parameters (type P) in class Stuff.
Proxy[PP]'s type parameters do not match type P's expected parameters:
trait Proxy has one type parameter, but type P has one
case class Stuff[+P[PP <: Proxiable] <: Proxy[PP]](content: P[Data])
           ^
two warnings found
one error found

@adriaanm said:
It looks to me like it's forming a type it shouldn't form: Proxy[PP] means nothing in the body of canEqual (PP is not in scope)

    <synthetic> def canEqual(x$1: Any): Boolean = x$1.$isInstanceOf[Stuff[Proxy[PP]]]();

@retronym said:
Here's a smaller test case:

import scala.language.higherKinds
trait Proxy[+T]
case class Stuff[+P[PP] <: Proxy[PP]]()

The most self contained workaround seems to be manually implementing canEqual.

import scala.language.higherKinds
trait Proxiable {
  def source: String
}
trait Proxy[+T <: Proxiable] {
  def source: String
}
case class Data(source: String) extends Proxiable
case class Stuff[+P[PP <: Proxiable] <: Proxy[PP]](content: P[Data]) {
  def canEqual(x$1: Any): Boolean = x$1 match { case _: Stuff[_] => true; case _ => false}
}

@retronym said:
See also #8039, which would make it possible to write (or synthesize) isInstanceOf[Stuff[_]] without getting a kind error.

@retronym said:
WIP: scala/scala@2.11.x...retronym:ticket/9331

Fred (fred) said:
I'm the author of the ML report.

I saw this didn't make it in 2.11.7 and was wondering if there is anything I could do to help get it in 2.11.8?

Thank you

@adriaanm said:
It looks like we just need to submit the PR in time for 2.11.8 based on the WIP, and convince ourselves it's safe to do so in a 2.11.x release. I've gone over the patch a couple of times now, and it looks like it should be good to go.

Fred (fred) said:
FYI, I built my own scala distribution based on scala 2.11.7 into which I merged the above mentioned patch. Using it, I could successfully compile everything... and tests pass.

Fred (fred) said:
Could this be considered for merging into scala 2.12?

We would love not to be stuck with scala 2.10 and akka 2.3 until late 2016 (planned release of scala 2.11.9 according to http://scala-lang.org/news/2.11.8/).

Thanks

@adriaanm said:
Fix for 2.11.9: scala/scala#5633