First project, test never ends

[nilskp]

So, after watching the ScalaDays presentation, I decided to take a brand new project can port it from Gradle to Mill.

However, the test, which runs fine under Gradle, never completes in Mill.

The project (with working Gradle build) is here:
https://github.com/bulletproof-ninja/nostrrent

And the Mill branch is here:
https://github.com/bulletproof-ninja/nostrrent/tree/build/mill

[lefou]

Sounds like #6224, which should be fixed in 1.1.0-RC3. But it looks like you're already on 1.1.0-RC3. Can you show us the output of jstack?

[lihaoyi]

Should be fixed in the main branch by #6348

There were also some other issues with your build file that I have fixed for you below

import mill.scalalib.*
import mill.javalib.*

object `package`
extends SbtModule:
  outer =>

  def scalaVersion = "3.7.4"
  def jvmVersion = "25"

  def mvnDeps =
    mvn"org.scala-lang:scala3-library_3:${scalaVersion()}" ::
    mvn"com.github.plokhotnyuk.jsoniter-scala:jsoniter-scala-core_3:$vJsoniter" ::
    mvn"com.typesafe.scala-logging:scala-logging_3:$vScalaLogging" ::
    mvn"ch.qos.logback:logback-classic:$vLogback" ::
    mvn"org.ngengine:bech32:$vBech32" ::
    mvn"org.eclipse.jetty.ee10:jetty-ee10-servlet:$vJetty" ::
    mvn"org.eclipse.jetty:jetty-server:$vJetty" ::
    mvn"org.eclipse.jetty:jetty-util:$vJetty" ::
    mvn"org.scalatra:scalatra-jakarta_3:$vScalatra" ::
    mvn"com.frostwire:jlibtorrent:$vJLibTorrent" ::
    mvn"com.frostwire:jlibtorrent-linux-x86_64:$vJLibTorrent" ::
    mvn"com.frostwire:jlibtorrent-windows:$vJLibTorrent" ::
    mvn"fr.acinq.secp256k1:secp256k1-kmp:$vSecp256k1" ::
    mvn"fr.acinq.secp256k1:secp256k1-kmp-jvm:$vSecp256k1" ::
    mvn"fr.acinq.secp256k1:secp256k1-kmp-jni-jvm:$vSecp256k1" ::
    Nil

  def compileMvnDeps = super.compileMvnDeps() ++ {
    mvn"com.github.plokhotnyuk.jsoniter-scala:jsoniter-scala-macros_3:$vJsoniter" ::
    Nil
  }

  def javacOptions = super.javacOptions() ++ {
    "-source" :: jvmVersion ::
    "-target" :: jvmVersion ::
    Nil
  }

  def jvmId = s"zulu:$jvmVersion"

  def repositories = super.repositories() ++ {
    "https://dl.frostwire.com/maven" ::
    Nil
  }

  object test
  extends SbtTests
  with TestModule.ScalaTest:

    def mvnDeps = outer.mvnDeps() ++ {
      mvn"org.scalatest:scalatest_3:$vScalaTest" ::
      mvn"com.vladsch.flexmark:flexmark-all:$vFlexmark" ::
      mvn"org.bitcoin-s:bitcoin-s-core_2.13:$vBitcoinS" ::
      Nil
    }

    def testParallelism = false
    def testSandboxWorkingDir = false

  // Dependency versions:
  def vJsoniter = "2.38.4"
  def vScalaLogging = "3.9.4"
  def vLogback = "1.5.20"
  def vBech32 = "1.0.3"
  def vJetty = "12.1.4"
  def vScalatra = "3.1.2"
  def vJLibTorrent = "2.0.12.7"
  def vSecp256k1 = "0.21.0"

  def vScalaTest = "3.2.19"
  def vFlexmark = "0.64.8"
  def vBitcoinS = "1.9.11"

[nilskp]

Thanks. A couple of questions:

1. You removed the Mill version comment. I assume because Mill will auto-update and always use the latest?
2. Do I even need mvn"org.scala-lang:scala3-library_3:${scalaVersion()}", or will that be included automatically anyway?
3. I guess it's a naming choice, but I've never used SBT and I'm not using it now. It kinda bothers me that my Mill build have to be called something with SBT, when it's got nothing to do with SBT. It's just Scala with Maven convention, which SBT happened to copy.

Thanks again!

[lefou]

1. It probably will, but I'd recommend to explicitly state the version you want to use.

2. It will be added by default, no need to explicitly add it, unless you need to do special stuff.

3. The name is a bit unlucky. All SbtModule does is applying the SBT Layout, which is in fact the Maven Layout plus the extra src/main/scala and src/test/scala source dirs. If you don't like it, you can either rename it when importing or define it yourself. You could also adopt the Mill layout, which uses a flat src dir for all source files.

[nilskp]

Ok, the latest version 1.1.0-RC3-30-8575ac does not fix this on my end. I should add that I'm running this under WSL2/Ubuntu, if that makes a difference.
mill.jstack.1.txt
mill.jstack.2.txt

Stuck.Mill.Test.av1.mp4

I've attached the jstack dumps and a video showing the behavior.

[lefou]

I assume:

• The cause of the issue is some non-daemon thread, that isn't properly stopped at the test end
• The fix Add explicit sys.exit to ZincWorkerMain to ensure process terminates #6348 should catch this case, but is not yet in the latest available binary snapshot

Other hints:

• jps is a shorter alternative to get the process ID of a java process

[lihaoyi]

@nilskp 1.1.0-RC3-30-8575ac does not contain the latest commits, will ping you when there's a release that does

[nilskp]

Could you address my questions? Thanks.