Re: Side effects between test cases

Hi Tobi,


thanks for the description of your setup.


Regarding your alternative b) you can use ant and JUnit task [1] with 
fork=On and forkMode=perTest and threads=1 which will trade 
"correctness" for total execution time.
If you write a test class for each time-sensitive test each of them will 
be executed in a seperate JVM and cannot interfere with others regarding 
timeout or memory. Note that this setup will increase total execution 
time because a JVM is started for each test.


Stefan


[1] http://ant.apache.org/manual/Tasks/junit.html


Am 28.07.2014 um 09:09 schrieb Tobias Werth [email protected] [junit]:
>
> Hi Stefan,
>
> *Stefan Penndorf [email protected] [junit] <[email protected]> 
> [14/07/25]:
> > I think it's really hard to address those things. I agree with Esko 
> that those
> > concerns are rare in "every-day" projects.
> > [...]
> > Maybe you could explain further why timing is an issue for you. And 
> maybe you
> > can also describe the code and project that causes the issue.
>
> Yes, both of you are probably right and I overestimated the generality
> of my problem.
>
> What we are doing is a bit different: we evaluate student exercises.
> E.g. students have to implement an interface, and we test their code
> with JUnit tests for grading purposes. We also restrict timings since
> the students may code infinite loops or just choose the wrong
> algorithmic complexities.
>
> What might happen currently is: a student might fail on more cases as he
> should fail on. However, we have probably to work around that and
> a) either use that deprecated Thread.stop() method (since we don't test
> multithreaded code)
> b) start each test case in a separate run
>
> Cheers,
> Tobi
>
>