TreeMap.transformTree operation not being applied after inlining

[hughsimpson]

Compiler version

3.0.2

Minimized code

File 1:

package bug

import scala.quoted.*

object Macros { inline def mk[T]: T = ${ MacroImpl.mk[T] } }

object MacroImpl { def mk[T](using Quotes, Type[T]): Expr[T] = new Maker[T].make }

class Maker[T](using Quotes, Type[T]) {
  import quotes.reflect.*

  def make: Expr[T] = new TreeMap {
    override def transformTerm(tree: Term)(owner: Symbol): Term = {
      tree match {
        case Block(stats, Typed(expr, tpt)) =>
          Block.copy(tree)(transformStats(stats)(owner), Typed(expr, TypeTree.of[T]))
        case other => super.transformTerm(tree)(owner)
      }
    }
    override def transformStatement(tree: Statement)(owner: Symbol): Statement = {
      tree match {
        case c @ ClassDef(name, cstr, lsel, self, body) =>
          ClassDef.copy(c)(name, cstr, lsel :+ TypeTree.of[T], self, body)
        case other => super.transformStatement(tree)(owner)
      }
    }
  }.transformTree('{ new {} }.asTerm)(Symbol.spliceOwner).asExprOf[T]
}

File 2:

package bug

import Macros._

trait Foo {}

object M extends App {
  val x = mk[Foo]
}

Output

[info] running bug.M
[error] (run-main-0) java.lang.ExceptionInInitializerError
[error] java.lang.ExceptionInInitializerError
[error] 	at java.base/java.lang.J9VMInternals.ensureError(J9VMInternals.java:184)
[error] 	at java.base/java.lang.J9VMInternals.recordInitializationFailure(J9VMInternals.java:173)
[error] 	at bug.M.main(Main.scala)
[error] 	at java.base/jdk.internal.reflect.NativeMethodAccessorImpl.invoke0(Native Method)
[error] 	at java.base/jdk.internal.reflect.NativeMethodAccessorImpl.invoke(NativeMethodAccessorImpl.java:62)
[error] 	at java.base/jdk.internal.reflect.DelegatingMethodAccessorImpl.invoke(DelegatingMethodAccessorImpl.java:43)
[error] 	at java.base/java.lang.reflect.Method.invoke(Method.java:566)
[error] Caused by: java.lang.ClassCastException: bug.M$$anon$1 incompatible with bug.Foo
[error] 	at bug.M$.<clinit>(Main.scala:10)
[error] 	at bug.M.main(Main.scala)
[error] 	at java.base/jdk.internal.reflect.NativeMethodAccessorImpl.invoke0(Native Method)
[error] 	at java.base/jdk.internal.reflect.NativeMethodAccessorImpl.invoke(NativeMethodAccessorImpl.java:62)
[error] 	at java.base/jdk.internal.reflect.DelegatingMethodAccessorImpl.invoke(DelegatingMethodAccessorImpl.java:43)
[error] 	at java.base/java.lang.reflect.Method.invoke(Method.java:566)

Expectation

My expectation was that I would be able to use this approach to build some stub objects for various traits -- the expr tree certainly looks like what I was hoping for -

{
  final class $anon() extends bug.Foo

  (new $anon(): bug.Foo)
}

• however it didn't. This is quite possibly not a bug and simply a case of me abusing the API to try to do what was never intended, but it had the feel of one. Attempting to add new statements to the body in a ClassDef.copy call also caused some additional failures, but I've excluded that from this example to keep things more minimal, and in case this isn't a bug but a misunderstanding of the feature.

[hughsimpson]

Did anyone ever figure out whether this was actually a bug, or if I was just doing it wrong? Would be lovely to know how to implement my intended macro

[nicolasstucki]

This looks wrong. The code is creating a tree and symbol representing that anonymous class. But then the tree map changes the tree but keeps the original symbol. There is a mismatch between the two definition that the compiler won't be able to reconcile.

The macro should use Symbol.newClass with the correct signature and then construct the tree from it with ClassDef.apply.

[hughsimpson]

Seems like it was a skill issue on my part then, I'll close this