[
https://issues.apache.org/jira/browse/CASSANDRA-10517?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14965342#comment-14965342
]
Joel Knighton edited comment on CASSANDRA-10517 at 10/20/15 4:29 PM:
---------------------------------------------------------------------
In summary, this works as designed.
We set a timeout in the ant junit test runner that works on the level of the
forked JVM. A JVM is forked for every test class, and the timeout is global
rather than per-method. Upon timeout, the forked JVM errors out and no more
tests from this class are run.
In order to run all the testcases in a test class, we would need to abandon the
timeouts in the forked JVM and switch to junit annotations.
There's a development philosophy debate attached to this. On the one hand,
running all the tests in every circumstance makes it easier to identify the
timed-out test cases more than one at a time. On the other hand, a failure is
a failure, and a red light will show up for the build in the case of a timed
out.
For now, I think we leave this be and use it as additional impetus to fix our
test problem. In the event that we find timed out tests to be one of the last
remaining issues because of unpredictable runtime behavior in CassCI, we should
consider moving to junit annotations.
EDIT: I'm going to leave this open a bit longer, since I expect there will be
good discussion on this topic, and I'd rather have it centralized.
was (Author: jkni):
In summary, this works as designed.
We set a timeout in the ant junit test runner that works on the level of the
forked JVM. A JVM is forked for every test class, and the timeout is global
rather than per-method. Upon timeout, the forked JVM errors out and no more
tests from this class are run.
In order to run all the testcases in a test class, we would need to abandon the
timeouts in the forked JVM and switch to junit annotations.
There's a development philosophy debate attached to this. On the one hand,
running all the tests in every circumstance makes it easier to identify the
timed-out test cases more than one at a time. On the other hand, a failure is
a failure, and a red light will show up for the build in the case of a timed
out.
For now, I think we leave this be and use it as additional impetus to fix our
test problem. In the event that we find timed out tests to be one of the last
remaining issues because of unpredictable runtime behavior in CassCI, we should
consider moving to junit annotations.
> Make sure all unit tests run on CassCI
> --------------------------------------
>
> Key: CASSANDRA-10517
> URL: https://issues.apache.org/jira/browse/CASSANDRA-10517
> Project: Cassandra
> Issue Type: Sub-task
> Reporter: Jim Witschey
> Assignee: Joel Knighton
> Labels: triage
> Fix For: 3.0.0
>
>
> It seems that some Windows unit tests aren't run sometimes on CassCI, and
> there's no error reporting for this. For instance, this test was introduced
> around the time build #38 would have happened, but has only run in builds
> #50-3 and #64:
> http://cassci.datastax.com/view/cassandra-3.0/job/cassandra-3.0_utest_win32/lastCompletedBuild/testReport/org.apache.cassandra.cql3/ViewTest/testPrimaryKeyIsNotNull/history/
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
