Thanks for looking into this! I think we can put in the fix and remove one
of the tests, yes.
@Robert What do you think? I think you initially added this test a loooong
while back.
On Mon, 9 Jan 2017 at 20:11 Alexey Demin <diomi...@gmail.com> wrote:
> Hi,
>
> I am trying make small review for slow test and I found small issue:
>
> NonReusingReOpenableHashTableITCase
>
> testSpillingHashJoinWithMassiveCollisions
> testSpillingHashJoinWithTwoRecursions
>
>
> for testSpillingHashJoinWithTwoRecursions exist description
>
> /*
> * This test is basically identical to the
> "testSpillingHashJoinWithMassiveCollisions" test, only that the number
> * of repeated values (causing bucket collisions) are large enough to make
> sure that their target partition no longer
> * fits into memory by itself and needs to be repartitioned in the
> recursion again.
> */
>
> but he incorrect, because code of both test fully equal,
> one difference line very similar on bug after refactoring with inserting
> recordReuse
>
> testSpillingHashJoinWithMassiveCollisions
> 353 while ((record = buildSide.next(record)) != null) {
>
> (f51f1b4 19.03.14, 1:17 Aljoscha Krettek* Change MutableObjectIterator to
> allow immutable objects)
>
>
> Aljoscha, can we remove one test and fix buildSide.next(record) to
> buildSide.next(recordReuse) ?
>
> P.S. I started review because we have a lot of failing test due to cpu
> time limit
>
> Thanks,
> Alexey
>
