SI-9080 Replace Cloneable/Serializable traits with type aliases

[soc]

No description provided.

[soc]

Looks like we need to bootstrap the compiler between changing Definitions and removing the files...

[soc]

Last discussion: https://groups.google.com/d/topic/scala-internals/3pIe7Uxqwyo/discussion

[adriaanm]

@soc, we like this, but RC1 is very close. Can you propose a plan to get this green by the end of next week without releasing another STARR?

[soc]

I don't see one. Definitions assumes real, existing Scala classes. This has to be fixed before the Scala classes can be removed in favor of aliases. Another thing that I would like to see is dropping the package when printing the type names.

[retronym]

I don't see one. Definitions assumes real, existing Scala classes

@soc What steps could I take to reproduce the problem?

[soc]

@retronym Replace the classes with aliases in the same step, without building a new STARR. Compiler will crash.

[retronym]

I see, the STARR compiler fails with:

[debug]     /Users/jason/code/scala/build/quick/classes/library
scala.reflect.internal.MissingRequirementError: class scala.Serializable in compiler mirror not found.
    at scala.reflect.internal.MissingRequirementError$.signal(MissingRequirementError.scala:17)
    at scala.reflect.internal.MissingRequirementError$.notFound(MissingRequirementError.scala:18)
    at scala.reflect.internal.Mirrors$RootsBase.$anonfun$getModuleOrClass$4(Mirrors.scala:54)
    at scala.reflect.internal.Mirrors$RootsBase.getModuleOrClass(Mirrors.scala:54)
    at scala.reflect.internal.Mirrors$RootsBase.getModuleOrClass(Mirrors.scala:66)
    at scala.reflect.internal.Mirrors$RootsBase.getClassByName(Mirrors.scala:102)
    at scala.reflect.internal.Mirrors$RootsBase.getRequiredClass(Mirrors.scala:105)
    at scala.reflect.internal.Mirrors$RootsBase.requiredClass(Mirrors.scala:108)
    at scala.reflect.internal.Definitions$DefinitionsClass.SerializableClass$lzycompute(Definitions.scala:371)
    at scala.reflect.internal.Definitions$DefinitionsClass.SerializableClass(Definitions.scala:371)

[soc]

@retronym This one should compile, but as mentioned, I think the "package" in scala.package.Foo is really ugly, and we should not print it in general (not only in the case of Clonable/Serializable).
Didn't have the time to find the right place to eliminate it, though.

[smarter]

You probably need to get rid of scala.Serializable in https://github.com/scala/scala/blob/2.12.x/src/library/scala/runtime/LambdaDeserializer.scala too

[soc]

I fixed the issue in scalap/caseClass, although I don't even want to know why it's scala.Serializable once and java.io.Serializable a few lines down:

case class CaseClass[A <: scala.Seq[scala.Int]](i: A, s: scala.Predef.String) extends scala.AnyRef with scala.Product with scala.Serializable {
...
object CaseClass extends scala.AnyRef with java.io.Serializable {
...

[soc]

@retronym What do you think of the need for SerialVersionUIDs for the serialization tests?

I added a few of them to make things compile, but I think either all or none of the collection classes should get the annotations.
I think we don't guarantee serialization compatibility between major versions, but in this case I don't think that any breakage is possible (which would be in favor of adding annotations, and not changing the test values in the tests ...)

Opinions?

Edit: Forgot the test in question, it is https://github.com/scala/scala/blob/2.12.x/test/files/run/t8549.scala#L117
Many of the collection tests changed, and it shows very nicely that we don't have much test coverage for all those specialized 0/1/2/3/4-item collections.

[soc]

@smarter Thanks! I simplified things, could you have a look whether I should go further and get rid of this completely? As I understand, there now no benefit anymore of doing it, as the only marker we add is already added automatically.

[smarter]

@retronym can probably answer this better than me

[soc]

Can we disable scaladoc generation for this one?
My nerves won't survive adding another work-around just for one commit.

[error] /home/jenkins/workspace/scala-2.12.x-validate-test/src/library/scala/Cloneable.scala:14: Cloneable is already defined as type Cloneable in package object scala
[error] trait Cloneable extends java.lang.Cloneable
[error]       ^
[error] /home/jenkins/workspace/scala-2.12.x-validate-test/src/library/scala/Serializable.scala:14: Serializable is already defined as type Serializable in package object scala
[error] trait Serializable extends Any with java.io.Serializable
[error]       ^
[info] No documentation generated with unsuccessful compiler run
[error] two errors found

[adriaanm]

Let's continue work on this, but I think we should do this under a bigger umbrella of library cleanups in 2.13. 2.12.0-RC1 is too close.

[soc]

Well, it's just waiting for a merge. After that, we need a new bootstrap, and then the final change, removing the traits, can be merged, fixing the scaladoc error.

[soc]

Any news on this?

[adriaanm]

No, this will have to go into 2.13. I don't want the RC cycle to slip further.