https://kotlinlang.org logo
Channels
Powered by
#arrow-contributors
  • a

    AdrianRaFo

    01/17/2019, 5:29 PM
  • s

    simon.vergauwen

    01/18/2019, 12:55 PM
    I cannot reproduce the failing test. Ran for 
    Iteration: 9737. result = [6, 3, 2, 4, 1, 5] && order.toList() = [1, 2, 3, 4, 5, 6]
  • s

    simon.vergauwen

    01/18/2019, 12:57 PM
    Seems like a race condition to me where 
    Thread.sleep(240)
    finishes before 
    Thread.sleep(200)
    .
  • s

    simon.vergauwen

    01/18/2019, 12:58 PM
    Any other ideas? I’ve seen this issue come up couple of times before.
  • s

    simon.vergauwen

    01/18/2019, 12:58 PM
    A semantic blocking 
    IO.sleep
    wouldn’t help here, would it?
  • s

    simon.vergauwen

    01/18/2019, 12:59 PM
  • j

    Jannis

    01/18/2019, 1:11 PM
    I have had a similar thing a while ago when I had a IO test fail randomly. I think this may be down to running the tests in parallel (I think they do because of https://github.com/arrow-kt/arrow/blob/master/build.gradle#L167). If this test runs at the same time with another test that is using some sleep as well then this would make sense. Stopped there because I did not have time to further confirm or deny that 😕
  • s

    simon.vergauwen

    01/18/2019, 1:13 PM
    So replacing 
    Thread.sleep
    with 
    IO.sleep
    might solve this issue since we don’t block the Thread anymore, right?
  • j

    Jannis

    01/18/2019, 1:19 PM
    Likely (if that was the issue). I think it's worth a try, but it's a bit hard to prove things work even if it passes. Can't see any other reason for those tests failing tho
  • j

    Jannis

    01/18/2019, 1:20 PM
    I mean as long as there are no long running and/or thread blocking tests anywhere this should be fine
  • p

    pakoito

    01/18/2019, 1:26 PM
    we stopped doing parallel tests to avoid these deadlocks a while ago
  • p

    pakoito

    01/18/2019, 1:26 PM
    they may have popped back up again after the migration
  • p

    pakoito

    01/18/2019, 1:26 PM
    using default dispatchers that rely on number of cores could land you on a dev server that had 1 core and then no parallelism for you
  • p

    pakoito

    01/18/2019, 1:27 PM
    which caused these timing issues
  • j

    Jannis

    01/18/2019, 1:32 PM
    ah ok, then is that line from the build gradle not in effect or something? o.O also on a machine with 1 core can the parallel tests even succeed properly?
  • p

    pakoito

    01/18/2019, 1:33 PM
    it should because we moved all code to not use CPU-count-bound schedulers
  • p

    pakoito

    01/18/2019, 1:33 PM
    may have skipped some tho
  • a

    AdrianRaFo

    01/18/2019, 4:22 PM
    Hi, Could I get another 👀 here https://github.com/arrow-kt/arrow/pull/1255? Thanks
  • s

    simon.vergauwen

    01/22/2019, 11:28 AM
  • s

    simon.vergauwen

    01/22/2019, 2:40 PM
    @AdrianRaFo is this what you fixed in #1255?
  • s

    simon.vergauwen

    01/22/2019, 2:40 PM
    Copy code
    arrow.effects.IOTest > parallel execution makes all IOs start at the same time FAILED
    java.lang.AssertionError: expected: [1, 2, 3, 4, 5, 6] but was: [1, 2, 3, 4, 6, 5]
        at io.kotlintest.DslKt.equalsError(dsl.kt:149)
        at io.kotlintest.DslKt.shouldBe(dsl.kt:46)
        at arrow.effects.IOTest$23.invoke(IOTest.kt:254)
        at arrow.effects.IOTest$23.invoke(IOTest.kt:27)
arrow.effects.IOTest > parallel execution preserves order for synchronous IOs FAILED
    java.lang.AssertionError: expected: [1, 2, 3, 4, 5, 6] but was: [1, 2, 3, 4, 6, 5]
        at io.kotlintest.DslKt.equalsError(dsl.kt:149)
        at io.kotlintest.DslKt.shouldBe(dsl.kt:46)
        at arrow.effects.IOTest$24.invoke(IOTest.kt:280)
        at arrow.effects.IOTest$24.invoke(IOTest.kt:27)
  • a

    AdrianRaFo

    01/22/2019, 3:19 PM
    Yes that's what I tried to fix but it's random and persistent
  • r

    raulraja

    01/23/2019, 6:29 PM
  • a

    AdrianRaFo

    01/23/2019, 6:45 PM
    I guess this https://github.com/arrow-kt/arrow/pull/1255 should be first to be sure all laws and test are passing
  • s

    simon.vergauwen

    01/23/2019, 6:45 PM
    Ye it should
  • r

    raulraja

    01/23/2019, 10:29 PM
    @AdrianRaFo is that ready to merge?
  • a

    AdrianRaFo

    01/23/2019, 10:39 PM
    I think so just needs two approvals
  • r

    raulraja

    01/23/2019, 10:40 PM
    travis is failing
  • a

    AdrianRaFo

    01/23/2019, 11:35 PM
    Yes, I found the problem I'm fixing it
  • a

    AdrianRaFo

    01/24/2019, 1:16 PM
    It's green now
    👍 1
1...707172...115Latest