[
https://issues.apache.org/jira/browse/CASSANDRA-16852?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17398363#comment-17398363
]
Andres de la Peña commented on CASSANDRA-16852:
-----------------------------------------------
I see that since CASSANDRA-16557 {{build-test}} is not a dependency of {{ant
jar}} in neither 3.0 or 3.11, while it's still a dependency on 4.0 and trunk,
where the multiplexer works smoothly.
The multiplexer relies on the CircleCI {{build}} job building the tests with a
call to {{ant clean jar}}, so the runners can skip the {{build-test}} target. I
haven't checked yet why {{build-test}} was removed from the {{jar}}
dependencies, but my initial guess is that we can either restore it or run
{{ant build-test}} in the CircleCI workflow. The later can be done either in
the {{build}} job or in the {{run_repeated_utest}} job.
> testsome fails in the CircleCI multiplexer because the class JStackJUnitTask
> can't be found
> -------------------------------------------------------------------------------------------
>
> Key: CASSANDRA-16852
> URL: https://issues.apache.org/jira/browse/CASSANDRA-16852
> Project: Cassandra
> Issue Type: Bug
> Components: CI
> Reporter: Ekaterina Dimitrova
> Assignee: Andres de la Peña
> Priority: Normal
> Fix For: 3.11.x
>
>
> As [~adelapena] explained on
> [CASSANDRA-16827|https://issues.apache.org/jira/browse/CASSANDRA-16827?focusedCommentId=17398060&page=com.atlassian.jira.plugin.system.issuetabpanels%3Acomment-tabpanel#comment-17398060],
> the problem might be in the flag _-Dno-build-test=true_ that the multiplexer
> uses to not run _build-test_ in every iteration, relying on that this has
> already been done by the build task. This is a performance optimisation that
> can be especially effective in fast tests.
> It would be ideal to have the initial multiplexer behavior restored, but we
> can always run _ant build-test_ at the beginning of every multiplexer runner
> while still not running it on every iteration. This seems enough to keep the
> multiplexer working but it requires some extra builds and therefore higher
> costs.
> The goal of this ticket is to find the root of this failure and fix the
> multiplexer either by restoring its initial flow or adding _ant build-test_
> at the beginning of every multiplexer runner, if there is no more effective
> solution.
--
This message was sent by Atlassian Jira
(v8.3.4#803005)
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
