ScalaCheck 1.13.0

[adelbertc]

Proposed fix for #810

This currently compiles, but does not run! When I try running it I get:

[error] Could not run test cats.tests.RegressionTests: java.lang.IncompatibleClassChangeError: Implementing class
[info] DeferredAbortedSuite:
[info] Exception encountered when attempting to run a suite with class name: org.scalatest.DeferredAbortedSuite *** ABORTED *** (0 milliseconds)
[info]   java.lang.IncompatibleClassChangeError: Implementing class

many many times. I thought it might be Discipline so I cloned Discipline, bumped ScalaCheck to 1.13.0, did a publishLocal, and bumped the Cats Discipline version. Still same error. Leaving here for comments/if anyone else has suggestions/wants to give it a shot.

[ceedubs]

I think you will also need a version of scalatest built against scalacheck 1.13

[ceedubs]

See also: scalatest/scalatest#837

[adelbertc]

So steps needed for this to move forward:

• ScalaTest release w/ ScalaCheck 1.13.0
• Discipline with new ScalaTest and new ScalaCheck
• Algebra with new ScalaTest and ScalaCheck

[ceedubs]

Update: @adelbertc has this working with locally published versions of Discipline and Algebra. It even seems to have caught a law-breaking instance! We should be able to move forward with this once there's a stable scalatest release.

Or perhaps we should update to the SNAP (stable snapshot) now and just make sure we wait until there's a stable version before the next cats release? That's actually sounding better to me considering builds are almost all failing.

Edit: I wasn’t thinking about the fact that that needs to happen in both algebra and discipline too. That probably complicates matters a bit.

[ceedubs]

To my understanding, typelevel/scalacheck#218 is a deal-breaker for upgrading to scalacheck 1.13.

Here is my understanding of the situation:

The Cogen support in scalacheck 1.13 is forced to call Option.get because of the fact that scalacheck generators return Option[A] instead of A. This can result in None.get exceptions when using Cogen along with generators that use suchThat/filter. This becomes quite problematic, because many common generators (such as choose, frequency and oneOf) in scalacheck use suchThat.

My understanding is based off of things that @xuwei-k and @non have said. It might not be entirely accurate.

[non]

@ceedubs that's correct although it's a bit more complicated. Even if you remove almost all use of suchThat / filter you still get the failure. I think it has to do with the fact that shrinking can produce invalid Gen instances which will always fail. I'm hoping to get some time this week/weekend to test a proposed fix.

[rossabaker]

Specs2 users: a downgrade to specs2-3.7 gets past the linker errors. That is the latest version based on scalacheck-1.12.x.

[ceedubs]

@adelbertc @non should this be closed in favor of #1345 ?

[non]

Yes I think so.