Hi folks, I’m trying to reactivate the ConcurrentLaws tests in incubator but I got no luck so far, i...

# arrow-contributors

[aballano]

Hi folks, I’m trying to reactivate the ConcurrentLaws tests in incubator but I got no luck so far, it seems there’s a flaky scenario specifically within the cancellableReceivesCancelSignal test, any help here would be greatly appreciated.

Also, I noticed that incubator provides its own instances of genK and eqK for IO -> https://github.com/arrow-kt/arrow-incubator/blob/master/arrow-mtl-data/src/test/kotlin/arrow/mtl/test/generators/GenK.kt#L20 so I’d also like to ask if there’s a specific reason not to use the provided in fx-test // cc @Jannis

[simon.vergauwen]

Needs an update PR after the IO `Gen`/`Eq` were merged to arrow-fx-test

[simon.vergauwen]

That’s just outdated code from the arrow-test re-org

[aballano]

I just used the fx-test instances and we’ve got more laws failing 😕

[simon.vergauwen]

Whut!?

[aballano]

arrow.mtl.test.OptionTTest.Applicative Laws: apEval consistent with ap

[aballano]

Property failed for
Arg 0: OptionT(value=<http://arrow.fx.IO|arrow.fx.IO>$Pure@7042e567)
Arg 1: OptionT(value=<http://arrow.fx.IO|arrow.fx.IO>$RaiseError@6fd57ab6)
after 31 attempts
Caused by: expected: true but was: false

[simon.vergauwen]

Wow weird. That should not fail

[simon.vergauwen]

Why is it failing? Hard to read with IO$Pure

[aballano]

it seems the gen in mtl was only giving pure instances of IO, not errors

[aballano]

so the tests are failing due to that not taken into account it seems

[simon.vergauwen]

Shouldn’t impact the apEval test

[Jannis]

No it may, I think that I implemented apEval in the wrong way

[Jannis]

Will check and make a pr if so

[Jannis]

Well yes it is wrong, will make a pr in the next hour or so. The annoying part with this is stacksafety because we can't thread eval through an arbitrary F so this will fail for stack unsafe F. That's usually not a problem, but still annoying

[simon.vergauwen]

Wasn’t that the whole point of going through Eval?

[Jannis]

It is for all applicatives that are not transformers

[Jannis]

Only the monad transformer types and effect types have this problem

[simon.vergauwen]

Right, okay so you’re saying that apEval cannot be stack-safe for any F for a transformers instance?

[simon.vergauwen]

Right

[Jannis]

Yeah because you usually need to flatMap through F first

[simon.vergauwen]

Yep

[Jannis]

Oh right: I already fixed both of the issues here: https://github.com/arrow-kt/arrow-incubator/pull/76/ I removed the io genK parts and fixed the optiont issue there^^

[simon.vergauwen]

I can review that later today

[Jannis]

Thanks 👍

[aballano]

aaaaaand the PR is green!

https://github.com/arrow-kt/arrow-incubator/pull/85