run new Vector through 2.13 community build

[SethTisue]

testing scala/scala#8534

cc @lrytz @szeiger

[SethTisue]

https://scala-ci.typesafe.com/job/scala-2.13.x-integrate-community-build/3220/

SUCCEEDED: 133 is promising

specs2

https://scala-ci.typesafe.com/job/scala-2.13.x-integrate-community-build/3220/artifact/logs/specs2-build.log

[specs2] [error] x for a List
[specs2] [error]  'Vector(1: java.lang.Integer, 2: java.lang.String): scala.collection.immutable.Vector1' != 'Vector(1: java.lang.Integer, 2: java.lang.String): scala.collection.immutable.Vector' - g2.e4 (NotNullStringsSpec.scala:51)
[specs2] [error] Actual:   ...ector[1]
[specs2] [error] Expected: ...ector[]

just an overly sensitive test, but then the failure took out a wide swath of the build

scala-parallel-collections

[scala-parallel-collections] failing seed for ParVector[Int].slices must be equal is wfklajcDD9byN_2eXRPLPJtVKmxtCzIwIudHEl0yyAC=
[scala-parallel-collections] failing seed for Parallel collections.ParVector[Int].slices must be equal is wfklajcDD9byN_2eXRPLPJtVKmxtCzIwIudHEl0yyAC=
[scala-parallel-collections] [info] ! Parallel collections.ParVector[Int].slices must be equal: Exception raised on property evaluation

@szeiger perhaps scala-parallel-collections need to be modified as a consequence of your changes?

scala-collection-laws

https://scala-ci.typesafe.com/job/scala-2.13.x-integrate-community-build/3220/artifact/logs/scala-collection-laws-build.log

multiple failures

offhand, it looks like most of the failures might be caused by scala/scala#8554 rather than by the Vector changes? not sure, I leave it to @szeiger (and perhaps @Ichoran) to dig into it

[SethTisue]

made a replacement-vector branch in this repo with that specs2 test excluded

new run queued: https://scala-ci.typesafe.com/job/scala-2.13.x-integrate-community-build/3222/

[Ichoran]

@SethTisue - No, the error log shows an match exception in VectorSliceBuilder. I'm pretty sure it's Vector.

That there are so many errors is because I assumed that Vector was a safe battle-tested collection that I could use as ground truth for everything else.

[SethTisue]

3222 now has just the two failures (scala-parallel-collections, scala-collection-laws)

[szeiger]

The line number is a bit suspicious:

[scala-collection-laws] , scala.MatchError: 1 (of class java.lang.Integer)
[scala-collection-laws] scala.collection.immutable.VectorSliceBuilder.result(Vector.scala:1151)

The failing match must be the one in line 1152.

[szeiger]

Modifying VectorTest.testSlice to use sizes 0 to 100 yields:

[error] Test scala.collection.immutable.VectorTest.testSlice failed: java.lang.AssertionError: expected:<Range 36 until 66> but was:<Vector(36, 37, 38, 39, 40, 41, 42, 43, 44, 45, 46, 47, 48, 49, 50, 51, 52, 53, 54, 55, 56, 57, 58, 59, 60, 61, 62, 63, 32, 33)>, took 0.465 sec
[error]     at scala.collection.immutable.VectorTest.$anonfun$testSlice$3(VectorTest.scala:345)
[error]     at scala.collection.immutable.Range.foreach$mVc$sp(Range.scala:190)
[error]     at scala.collection.immutable.VectorTest.$anonfun$testSlice$2(VectorTest.scala:341)
[error]     at scala.collection.immutable.Range.foreach$mVc$sp(Range.scala:190)
[error]     at scala.collection.immutable.VectorTest.$anonfun$testSlice$1(VectorTest.scala:340)
[error]     at scala.collection.immutable.Range.foreach$mVc$sp(Range.scala:190)
[error]     at scala.collection.immutable.VectorTest.testSlice(VectorTest.scala:336)

It's not an exception but probably the same bug. The exception is caused by a VectorSliceBuilder result having a maxDimension of 1 despite a size > 32. The bug is probably in rebalancing or earlier in keeping track of size of dimension during building.

[szeiger]

I updated the PR with a fix for a slicing bug (or two, depending on how you want to count them). Based on the nature of these bugs I'm not sure how they could lead to the MatchError. Can we do a new run of the special community build branch?

I used tested every possible slice of every Vector up to a length of 1000, including verifying the structure and comparing the results. I didn't enable it in the test suite by default because it takes several minutes to run.

If the MatchError persists I'll have to try to reproduce and simplify it in a different way. Slicing from a valid Vector should be absolutely solid now. My best guess at the moment is that an invalid Vector was produced by a bad slice and a consecutive slice ran into the other error mode.

[SethTisue]

Can we do a new run of the special community build branch?

I'll do that tonight or tomorrow morning, but it takes a long time. You'd get a first round of feedback much faster by running scala-collections-laws yourself locally, outside of the community build.

[SethTisue]

queued: https://scala-ci.typesafe.com/job/scala-2.13.x-integrate-community-build/3234/

[SethTisue]

3234 fixed the failures. so likely we're good, but some other repos had intermittent failures, re-run queued to be sure: https://scala-ci.typesafe.com/view/scala-2.13.x/job/scala-2.13.x-integrate-community-build/3235/

[szeiger]

https://scala-ci.typesafe.com/view/scala-2.13.x/job/scala-2.13.x-integrate-community-build/3239/ succeeded