I'm afraid there's not much we can do in Pax Exam about incompatible
changes in Surefire...
But there _is_ some homework to do regarding the test container
lifecycle and container strategies of Pax Exam.
I've observed some sporadic failures both in the master and
exam3-milestones branches which need further investigation.
Some of these effects have been fixed in PAXEXAM-256 and PAXEXAM-324.
Some other issues seem to occur with Equinox only, which uses some
additional cache directories not influenced by setting the official
org.osgi.framework.storage property.
I don't really want to resort to framework specific properties in Native
Container, but there might be no way around that. Then again, I may be
wrong, I just don't know enough about Equinox internals.
Anyway, the scenario with multiple containers started up front should no
longer be an issue with the temp directory fixes in 2.4.0 - each
container will get its storage directory in that case.
Best regards,
Harald
Am 13.01.2012 17:38, schrieb Sanjeeb Sahoo:
In some earlier version of surefire plugin (I have noticed it for
version 2.4.3), surefire used to instantiate all the runners upfront in
order to create a test suite and then invoke them serially. With this
kind of behavior, we will run into issues when
EagerSingleStagesReactorFactory strategy is used as this will mean
multiple containers started simultaneously up front. In case of
NativeTestContainer, that could lead to use of the same OSGi storage
area! There can be other side effects as well. Fortunately, this
behavior of surefire has changed. But, my question really is what's the
guarantee that it will stay this way?
Thanks,
Sahoo
_______________________________________________
general mailing list
[email protected]
http://lists.ops4j.org/mailman/listinfo/general
