2.10.0-M6: spurious SCOPE_EXIT warnings

[scabug]

this is a less severe form of #6049. #6049 was closed because the back end no longer crashes, but there are still spurious warnings.

so e.g. borrowing Miguel's minimized code from #6049:

Welcome to Scala version 2.10.0-M6 (Java HotSpot(TM) 64-Bit Server VM, Java 1.6.0_33).

scala> :paste
object O {
  while(true) {
    var j = 0
    try j
    catch { case t: Throwable => }
  }
}
// Exiting paste mode, now interpreting.

<console>:11: warning: Visited SCOPE_EXIT before visiting corresponding SCOPE_ENTER. SI-6049
           try j
           ^
defined module O

at #6049 Miguel said this warning could also be a sign of issues with "whether the debugger shows the variable in question"

[scabug]

Imported From: https://issues.scala-lang.org/browse/SI-6191?orig=1
Reporter: @SethTisue
Affected Versions: 2.10.0-M6

[scabug]

@SethTisue said:
warning still present in 2.10.0-M7. the warning references #6049, should probably reference this ticket instead

[scabug]

Jeff Olson (jdolson) said:
FYI, I'm getting this warning with the following code:

import java.util.concurrent.ExecutorService
import java.util.Queue

class Spinner[-T](f: T => Unit, execService: ExecutorService, queue: Queue[T]) {
  @volatile private[this] var runFlag = true

  def start() {
    execService.execute(new Runnable() {
      def run() {
        while (runFlag) {
          val item = queue.poll()
          if (item != null) {
            try {
              f(item)
            }
            catch {
              case t: Throwable => t.printStackTrace()
            }
          }
        }
      }
    })
  }

}

Running scalac [Scala version 2.10.0 (Java HotSpot(TM) 64-Bit Server VM, Java 1.6.0_29)] gives the following:

test.scala:12: warning: Visited SCOPE_EXIT before visiting corresponding SCOPE_ENTER. SI-6049
          if (item != null) {
          ^
one warning found

[scabug]

Lars Hupel (larsrh) said:
This also happens in the scalaz codebase and I'm unsure what to do with it. Is this warning bogus or should I change the code until it disappears?

[scabug]

@nilskp said:
Yep, just hit this one. Not sure if I should care. Is the byte code broken or what's going on?

[scabug]

@magarciaEPFL said:
This amounts to some debugging information missing, about the scope of a method variable. The debugger will show that variable as having a scope either too large or too small, or might not display it at all. The bytecode instructions are fine.

[scabug]

@som-snytt said:
I also just noticed GenASM report this on itself. Just a bit of irony.

...jvm/GenASM.scala:133: warning: Visited SCOPE_EXIT before visiting...

I agree that the warning is a distraction. Shouldn't it sit behind -Xdev?

[scabug]

@retronym said:
scala/scala#2182

[scabug]

@som-snytt said:
OK, then you don't need mine. I see the semantics of devWarning is Xdev || Ydebug.

That might be more useful, on the theory that if a dev sees the SCOPE_EXIT enough to be annoyed, they might be motivated to remedy it.

[scabug]

@JamesIry said:
2.10.2 is about to be cut. Kicking down the road and un-assigning to foster work stealing.

[scabug]

@adriaanm said:
Debug info still needs to be improved, but the warning is nog longer emitted by default.

[scabug]

@SethTisue said:
fix version is...?

[scabug]

@adriaanm said:
2.11.x? This didn't seem urgent?

[scabug]

@SethTisue said:
the ticket is closed as fixed. essentially, I'm asking what branch the fix has already been made on. (I'm not suggesting a 2.10.x backport)

[scabug]

@adriaanm said:
Oh, I was thinking we were fixing it in the future. Looks like an accidental close -- I really need SI-autocorrect.

[SethTisue]

fixed by new back end in 2.12