Does Closure use a different Fault Localization strategy compared with other projects in Defects4J

[Kaka727]

Hi @tdurieux
I modified the source code of Arja according to your pull request and now I can run it.
Thanks!
I find another strange question. I can get the results among five projects except Closure in Defects4J. The log message is :

I am very confused about that. We donot transfer a parameter named gzoltarDataDir. So according to the source code of ARJA, it should call GZoltarFaultLocalizer2 to perform FL. Why in the trace is GZoltarFaultLocalizer and why it fails.

[tdurieux]

We did not use a different Fault Localization strategy. But I spent a bunch of in Nopol to configure gzoltar correctly for Closure. This project requires a really clean classpath to run the test.

[Kaka727]

Does your pull request in arja reveal this configuration change?
If not, how should I make the corresponding operation. I'm very confused.

[tdurieux]

I thought we faced the same issue during our executions but it is not the case, sorry about that.

I dont know what is happening exactly. It is related to the classpath or the java version, ...
It can be because two different versions of JUnit is present in the classpath, or the java release version,...

[Kaka727]

Just now I downloaded the docker image and run the command: docker run -it --rm -v /home/wp/Desktop/results:/results tdurieux/repairthemall Arja -b Defects4J -i Closure_1
The error is the same as I met in my own environment.

I don't know why. It is very strange.