Guillaume Martres

> It turns out that Dimi's attempted hylolib port had lots of problems with ambiguities, and they all went away with https://github.com/lampepfl/dotty/pull/19395, which includes this PR as the only change...

> I know there are repl / scala / scalac sbt tasks, but I highly prefer working in zsh versus sbt shell For what it's worth, if you have sbt...

I think it'd be useful to get this in 3.5.0-RC1 for testing purposes, we'd then revert it before the release if the sip committee votes against it.

I'd like to formally appeal that decision. As can be seen in https://github.com/scala/improvement-proposals/pull/84, there is support among the SIP committee members for not unnecessarily delaying this. Delaying makes sense if...

I've now moved the new logic under an experimental language feature so that it can be merged without SIP approval.

> You're not supposed to need a new in Scala 3, whether you're in annotations or not. So IMO this is a real regression. This was actually an intended change,...

There's a few things going on here: - When we read a Java annotation from a classfile, we pretend it has a constructor whose arguments are all the methods defined...

Another (unlikely) way for the existing code to do the wrong thing would be the case of a java annotation where the author decided to swap the order of the...

I believe the forSome here comes from the result type inferred by Scala 2 for the classesOf method and should disappear if an explicit result type of `Seq[Class[_]]` is used....

> It could be made more helpful by including the offending line in project or dependency. Agreed, but here again this might be tricky to reconstruct from the scala 2...