Updated to Cats Effect 3.7, Scala Native 0.5, and dotty-cps-async 1.0…

[djspiewak]

Also changed the name to cats-effect-direct. When we release I'll update the repo and the readme.

[djspiewak]

Okay this is super fun. I'm observing the following two issues. First, on Java 8 (and only Java 8), the Scala Native ++2.13 test seems to produce the following:

  | => cat StackTrace_PrintStackTrace
  at stackOverflowHandler
  at _sigtramp
  at _SM46cats.effect.cpsinternal.AsyncAwaitStateMachineD6tryGetL17scala.util.EitherL16java.lang.ObjectEO
  at _SM48cats.effect.AsyncAwaitSuite$stateMachine$async$1D5applyL17scala.util.EitheruEO
  at _SM48cats.effect.AsyncAwaitSuite$stateMachine$async$1D5applyL16java.lang.ObjectL16java.lang.ObjectEO
  at _SM46cats.effect.cpsinternal.AsyncAwaitStateMachineD21$anonfun$onComplete$8L16java.lang.ObjectL16java.lang.ObjectuEPT46cats.effect.cpsinternal.AsyncAwaitStateMachine
  at _SM56cats.effect.cpsinternal.AsyncAwaitStateMachine$$Lambda$8D5applyL16java.lang.ObjectEO
  at _SM19cats.effect.IOFiberD7runLoopL14cats.effect.IOiiuEPT19cats.effect.IOFiber
  at _SM19cats.effect.IOFiberD24asyncContinueSuccessfulRuEPT19cats.effect.IOFiber
  at _SM19cats.effect.IOFiberD3runuEO
  at _SM31cats.effect.unsafe.WorkerThreadD3runuEO
  at _SM39scala.scalanative.runtime.NativeThread$D16threadEntryPointL38scala.scalanative.runtime.NativeThreaduEPT39scala.scalanative.runtime.NativeThread$
  at _SM39scala.scalanative.runtime.NativeThread$D24$anonfun$threadRoutine$1L28scala.scalanative.unsafe.PtrL28scala.scalanative.unsafe.PtrEPT39scala.scalanative.runtime.NativeThread$
  at _SM49scala.scalanative.runtime.NativeThread$$$Lambda$1G17$extern$forwarder
  at ProxyThreadStartRoutine
  at _pthread_start

I haven't been able to reproduce this on more recent versions of the JVM, so that's a fun one.

The second issue I'm hitting (also on 2.13) seems to be not JVM specific (thankfully…) and it results in the following test error:

==> X cats.effect.AsyncAwaitSuite.async[OptionT[IO, *]] - work on successes 0.210s scala.MatchError: Ptr@70001135d8 (of class scala.scalanative.unsafe.Ptr)

This appears to be nondeterministic.

It's worth noting that we don't have any SN specific stuff here, so something very odd is going on.

[durban]

Something is very wrong here: Trying this branch locally (2.13, x86_64 linux, JVM 17) I get a segfault in cats.effect.AsyncAwaitSuite. If I comment out there every test except the first one (async[IO] - work on success), then it fails with a scala.MatchError: class scala.util.Right (of class java.lang.Class), i.e., it gets essentially a classOf[Right[_]] (that is, a java.lang.Class instance, and not a scala.util.Right instance). (It's this match: https://github.com/djspiewak/cats-effect-cps/blob/c79c36fa75d53218a2880be7d16761698fc12fb9/core/shared/src/main/scala-2/cats/effect/cpsinternal/AsyncAwaitStateMachine.scala#L89.) The static type of that argument is Either[F[AnyRef], (F[Unit], AnyRef)], so this should be impossible.

[djspiewak]

I think it's fair to say this is surfacing a bug in Scala Native.