Re: [linuxtools-dev] Fwd: [Hudson] Hudson build is still unstable: linux

[Bernd Hufmann <bhufmann@xxxxxxxxx>]

Hi Roland

us too, we try to have the tests independent from each other. But
sometimes if they work in a given order then we don't notice the
dependencies.

/Bernd

On Thu, Mar 14, 2013 at 1:45 PM, Roland Grunberg <rgrunber@xxxxxxxxxx> wrote:
>> In noticed that some JUnit tests of our build have been failing since
>> yesterday. I focused on the LTTng/TMF/CTF test plug-ins. I noticed
>> that the order of test methods executed within a JUnit test file
>> changed for some reason. And it looks like that some test are
>> dependent on other tests to be executed before. Then I was wondering
>> why did the order change. I noticed that JUnit version 4.11 is used
>> instead of 4.10 and in that release the way changed how the order of
>> execution is determined (see [1]). There is a fix using an annotation
>> provided by JUnit Test 4.11, however this would require every
>> designer
>> has to upgrade to 4.11. Since Eclipse Juno comes with 4.10 this is
>> not
>> really an option right now. So in the end we need to make all test
>> cases independent from each other.
>>
>> I'm not sure if the test failures for GCov are caused by this but it
>> is worth considering it.
>>
>> [1]
>> http://randomallsorts.blogspot.ca/2012/12/junit-411-whats-new-test-execution-order.html
>
> Thanks for looking into this! I was starting to suspect JUnit was doing
> something weird with the test orderings.
>
> Generally, I try to avoid having tests that must execute in some defined
> order (I've seen these kinds of issues in the past), but I believe for
> GCov, the tests only depend on some code from another test that can
> easily be pushed into the setup phase.
>
> Thanks Again,
> --
> Roland Grunberg
> _______________________________________________
> linuxtools-dev mailing list
> linuxtools-dev@xxxxxxxxxxx
> https://dev.eclipse.org/mailman/listinfo/linuxtools-dev