ACS-Math-25: it is not a test-suite adequate repair

ASSERT-KTH / drr

Tool & data on the correctness of Defects4 patches generated by program repair tools http://arxiv.org/pdf/1909.13694

http://arxiv.org/pdf/1909.13694

Creative Commons Attribution Share Alike 4.0 International

10 stars 6 forks source link

Closed monperrus closed 6 years ago

monperrus commented 6 years ago

ACS-Math-25: it is not a test-suite adequate repair, only one failing test is repaired, one remains (checkMissingFastMathClasses):

does Defects4J info gives the same list of originally failing tests?
is it because of the usage of Java8 (We may ony use Java7 for the whole experiment)?

tdurieux commented 6 years ago

defects4j is not yet compatible with java8... The branch that adds java8 supports: https://github.com/rjust/defects4j/tree/java-8-support

monperrus commented 6 years ago

I propose to strictly to Java7 because I guess that most of the experiments were made with Java7.

tdurieux commented 6 years ago

It would surprise me a LOT if this patch is invalid since it is almost the same as the human one: http://program-repair.org/defects4j-dissection/#!/bug/Math/25

and personally I don't see how the generated patch can be wrong(if the human patch is correct)

monperrus commented 6 years ago

The problem was indeed with the JDK version