fix: Hover shows de-aliased type members

[pityka]

Previously metals showed generic type members for arguments of anonymous functions.

This commit recursively de-aliases type references in HoverProvider.

Resolves #6230 for scala 2.13 and scala 3.3.3

[kasiaMarek]

It seems a few tests have failed. @pityka, don't be bothered by macOS-latest jdk-17 unit and Scala CLI integration, tests that failed there are flaky and have nothing to do with your changes.

[rochala]

Hey, firstly, thanks for the contribution you've made.

Your solution is almost correct, but it caused a regression.
I think that the reason is in a recursive dealiasing. We should be more careful of what we're trying to dealias.
https://github.com/scalameta/metals/pull/6251/files#diff-5e56146f7a5bddf85757c9a6408980a990a426b172b4a43f14e49be6d5962571R251-R260
I left you a comment which should help you with finishing the PR.

If you want you can also take the opportunity to rename the method to deepDealias as metalsDealias is super bad name :D

[rochala]

It seems like it's a regression.

[rochala]

We have a method called metalsDealias (it should be named deepDealias) which does similar thing to yours.
The fix should be done directly in that method.
This is a bit more controlled of what we're doing.

I recommend implementing this like that:

  extension (tpe: Type)
    def deepDealias(using Context): Type =
      tpe.dealias match
        case app @ AppliedType(tycon, params) =>
          // we dealias applied type params by hand, because `dealias` doesn't do it
          AppliedType(tycon, params.map(_.metalsDealias))
        case aliasingBounds: AliasingBounds => 
          aliasingBounds.derivedAlias(aliasingBounds.alias.dealias)
        case TypeBounds(lo, hi) => 
          TypeBounds(lo.dealias, hi.dealias)
        case RefinedType(parent, name, refinedInfo) => 
          RefinedType(parent.dealias, name, refinedInfo.deepDealias)
        case dealised => dealised

[pityka]

Hm.. I am getting a strange error in one of the java import tests. I haven't seen this before. Is this something known? This is 3.3.3. Thanks!

exception caught when loading class BoundMethodHandle$Specializer$Factory: java.lang.AssertionError: assertion failed: failure to resolve inner class:
externalName = java.lang.invoke.ClassSpecializer$Factory,
outerName = java.lang.invoke.ClassSpecializer,
innerName = Factory
owner.fullName = java.lang.invoke.ClassSpecializer
while parsing /modules/java.base/java/lang/invoke/BoundMethodHandle$Specializer$Factory.class
Mar 27, 2024 11:13:19 PM scala.meta.internal.pc.CompilerAccess handleError
SEVERE: assertion failed: failure to resolve inner class:
externalName = java.lang.invoke.ClassSpecializer$Factory,
outerName = java.lang.invoke.ClassSpecializer,
innerName = Factory
owner.fullName = java.lang.invoke.ClassSpecializer
while parsing /modules/java.base/java/lang/invoke/BoundMethodHandle$Specializer$Factory.class
java.lang.AssertionError: assertion failed: failure to resolve inner class:
externalName = java.lang.invoke.ClassSpecializer$Factory,
outerName = java.lang.invoke.ClassSpecializer,
innerName = Factory
owner.fullName = java.lang.invoke.ClassSpecializer
while parsing /modules/java.base/java/lang/invoke/BoundMethodHandle$Specializer$Factory.class
at scala.runtime.Scala3RunTime$.assertFailed(Scala3RunTime.scala:8)
at dotty.tools.dotc.core.classfile.ClassfileParser$innerClasses$.classSymbol(ClassfileParser.scala:1124)

[rochala]

Are you getting this on this branch ? Maybe try cleaning the project sbt clean and try again ?

[pityka]

Thank you for fixing the issue in an other PR! Closing this.