Trouble with Closure

[Alex-128]

Hi, me again! :)

I've been able to use BA-DUA for data-flow coverage SBFL on the standard Defects4J projects (Chart, Lang, Math, Time), but with Closure it becomes so slow that it's unusable. I see that in the articles on Jaguar, Closure wasn't included as a benchmark project. Did the same sort of problem occur? Is there a way to fix it? I also saw that you used a subset of faulty versions from your benchmarks, specifically the versions with complete data-flow spectra. Do you have the specification of complete faulty versions available somewhere? Thank you for all the help you've already given me!

Kind regards
Alexandra van der Spuy

[andrioli]

Hi Alexandra,

Digging on my emails I didn't find a reason why Closure was not include as a benchmark project. I'll ask the other group members to check if anyone recall why and if someone have the specification of complete faulty versions.

But also on my emails I found a discussion about Closure running slow and unstable. The discussion conclusion was high memory consumption and high GC overhead. In the same discussion we suggested to increase the maximum Java heap size (e.g., -Xmx1536m) and looks that this fixed the problem. Can you test it ?