2.13 community build regressions

[SethTisue]

since recent Scala version bump (one of these),

• akka has a consistent test failure in ByteStringSpec.scala, see https://scala-ci.typesafe.com/job/scala-2.13.x-integrate-community-build/2577/artifact/logs/akka-build.log
• slick consistently fails to compile, infinite loop during compilation https://scala-ci.typesafe.com/job/scala-2.13.x-integrate-community-build/2578/artifact/logs/slick-build.log

both akka and slick block downstream projects, so fixes may uncover additional failures

[lrytz]

The akka failure is caused by scala/scala#8311, the addition of knownSize to BufferedIterator allows calling size without consuming the iterator.

In 2.13.0:

scala> val it = Vector(1).iterator.buffered; it.size; it.toList
it: scala.collection.BufferedIterator[Int] = <iterator>
res0: List[Int] = List()

In current 2.13.x

scala> val it = Vector(1).iterator.buffered; it.size; it.toList
it: scala.collection.BufferedIterator[Int] = <iterator>
res0: List[Int] = List(1)

@szeiger I guess it's OK for size not to consume the iterator? cc @Jasper-M.

cc @patriknw, the test in question is

• https://github.com/akka/akka/blob/0783855b2c85de1ab7b189785a4f6e640f56da53/akka-actor-tests/src/test/scala/akka/util/ByteStringSpec.scala#L952
• https://github.com/akka/akka/blob/0783855b2c85de1ab7b189785a4f6e640f56da53/akka-actor-tests/src/test/scala/akka/util/ByteStringSpec.scala#L166

[lrytz]

The slick failure is caused by scala/scala#8272

[patriknw]

Thanks for letting us know. Shall we remove or change the test?
cc @raboof

[lrytz]

This would be enough to fix it: lrytz/akka@8275b38

[patriknw]

thanks, I can bring that back to akka/akka in your name

[lrytz]

I'll be famous!

[Jasper-M]

@szeiger I guess it's OK for size not to consume the iterator? cc @Jasper-M.

The same is already true for other iterators which implement an efficient knownSize. On 2.13.0:

scala> val it = Vector(1).iterator; it.size; it.toList
it: Iterator[Int] = <iterator>
res0: List[Int] = List(1)

Anyway, I think the rule is: don't use the iterator after calling size. Because there's no guarantee that it is reusable. But there's also no guarantee that it isn't. Related issue: scala/bug#11224

[lrytz]

https://scala-ci.typesafe.com/job/scala-2.13.x-integrate-community-build/2581/ is running with scala/scala#8401 merged, so slick should pass.

[SethTisue]

green again