问题
Working in scala 2.11.12, JDK 1.8.0_131, I have been able to replicate a thread safety bug observed in Apache Spark with the following code, in which I repeatedly check with multiple threads whether Option[Int]
can be matched via <:<
to Option[_]
:
package stuff
import java.util.concurrent.{Executors, Future}
import scala.collection.mutable.ListBuffer
object Main {
val universe: scala.reflect.runtime.universe.type = scala.reflect.runtime.universe
import universe._
def mirror: universe.Mirror = {
universe.runtimeMirror(Thread.currentThread().getContextClassLoader)
}
def localTypeOf[T: TypeTag]: `Type` = {
val tag = implicitly[TypeTag[T]]
tag.in(mirror).tpe.dealias
}
def matcher[T: TypeTag]: Boolean = {
val typ = localTypeOf[T]
typ.dealias match {
case t if t <:< localTypeOf[Option[_]] =>
true
case _ =>
false
}
}
def main(args: Array[String]): Unit = {
val executor = Executors.newFixedThreadPool(5)
try {
val futures = new ListBuffer[Future[_]]()
for (i <- 1 to 10) {
futures += executor.submit(new Runnable {
override def run(): Unit = {
if (Main.matcher[Option[Int]]) {
println("ALL OK")
} else {
throw new Exception("THIS SHOULD BE IMPOSSIBLE!!!!!!")
}
}
})
}
futures.foreach(_.get())
} finally {
executor.shutdown()
}
}
}
This code should always print "ALL OK", but sometimes (~5% chance) it actually throws the "THIS SHOULD BE IMPOSSIBLE" error, with the following stacktrace:
Exception in thread "main" java.util.concurrent.ExecutionException: java.lang.Exception: THIS SHOULD BE IMPOSSIBLE!!!!!!
at java.util.concurrent.FutureTask.report(FutureTask.java:122)
at java.util.concurrent.FutureTask.get(FutureTask.java:192)
at stuff.Main$$anonfun$main$2.apply(Main.scala:81)
at stuff.Main$$anonfun$main$2.apply(Main.scala:81)
at scala.collection.immutable.List.foreach(List.scala:392)
at scala.collection.generic.TraversableForwarder$class.foreach(TraversableForwarder.scala:35)
at scala.collection.mutable.ListBuffer.foreach(ListBuffer.scala:45)
at stuff.Main$.main(Main.scala:81)
at stuff.Main.main(Main.scala)
Caused by: java.lang.Exception: THIS SHOULD BE IMPOSSIBLE!!!!!!
at stuff.Main$$anonfun$main$1$$anon$1.run(Main.scala:75)
at java.util.concurrent.Executors$RunnableAdapter.call(Executors.java:511)
at java.util.concurrent.FutureTask.run(FutureTask.java:266)
at java.util.concurrent.ThreadPoolExecutor.runWorker(ThreadPoolExecutor.java:1142)ALL OK
at java.util.concurrent.ThreadPoolExecutor$Worker.run(ThreadPoolExecutor.java:617)
at java.lang.Thread.run(Thread.java:748)
- Why?
- Is there something I can change in my program to fix this behavior?
- If this is an issue with scala, is it fixed in any future releases?
回答1:
- Why? Turns out to be a known thread-safety bug in scala reflection itself https://github.com/scala/bug/issues/10766
- Is there something I can change in my program to fix this behavior? After trying out a couple things, it works to wrap each
<:<
call in a synchronized block. Still interested in knowing if anyone has a more elegant way to do this. - If this is an issue with scala, is it fixed in any future releases? No, it is present in all current versions of scala (through
2.13.0-M5
) (see above scala issue).
来源:https://stackoverflow.com/questions/55150590/thread-safety-in-scala-reflection-with-type-matching