A feedback-based debugger for interactively recommending suspicious step in buggy program execution.

Microbat is a feedback-based debugger which aims to locate bugs by interactively recommending suspicious program steps with developers' feedback. Given a buggy program, Microbat records its execution trace and allow developers to make light-weight feedback on trace steps, such as correct-step, wrong-variable-value, wrong-path, and unclear. Microbat reasons and analyzes the feedback along with program information to recommend a suspicious step for further inspection and feedback. Such a debugging process continues until the bug is found. A short demonstration of Microbat is available in https://linyun.info/microbat/index.html.

We support four types of feedback, i.e., correct-step, wrong-variable-value, wrong-path, and unclear. Based on these types of feedback, we iteratively and interactively recommend suspicious steps on recoreded trace.

Our evaluation on Microbat shows promising results. The detailed evaluation results can be checked at https://linyun.info/microbat/index.html

If you need to reference our technique, please use the following citations:

• Yun Lin, Jun Sun, Yinxing Xue, Yang Liu, and Jinsong Dong. Feedback-based Debugging. The 39th ACM SIGSOFT International Conference on Software Engineering (ICSE 2017), pp. 393-403.
• Yun Lin, Jun Sun, Lyly Tran, Guangdong Bai, Haijun Wang, and Jinsong Dong. Break the Dead End of Dynamic Slicing: Localizing Data and Control Omission Bug. The 33rd IEEE/ACM International Conference on Automated Software Engineering (ASE 2018), pp. 509-519.

Our debugger can be divided into two parts: trace collector and the bug inference engine. Two parts are presented in terms of an Eclipse plugin.

1. After you clone the git repository, you need to load the microbat repository into eclipse (as the tool is manifested by Eclipse plugin). We recommend that the user should import the project through "Git perspective".

2. Given our trace collector is implemented through Java instrumentation technique, the user need to run microbat.tools.JarPackageTool. Please modify the DEPLOY_DIR by your $eclipse_root_folder\dropins\junit_lib. After running the code, you will generate an instrumentator.jar under the DEPLOY_DIR folder. More details can be refer to https://github.com/llmhyy/microbat/wiki/Compile-Runtime-Agent.

3. Remember to replace the instrumentaor.jar in the lib folder under microbat project.

4. You may run the code as an Eclipse Application then.

• In the running Eclipse Application, specify the configuration in Perspective>>Microbat Debugging. Here, you need to specify (1) which eclipse project you are going to debug; (2) where is the JDK library for running your Java program; (3) Step Limit (e.g., 10000); (4) Variable Layer (e.g., 2); and (5) Lanuch Class. Lanuch class is supposed to be the class containing main() method. Alternatively, you can also speicify a Junit test case.

Alternatively, we also support users (i.e., programmers) to run Microbat with JUnit4, JUnit5 and TestNG test cases. In this case, we need to package the project of "microbat_test_runner" and export it as a jar file "testrunner.jar". Moreover, we provide five jar files, i.e., junit.jar, org.hamcrest.core.jar, junit-platform-console-standalone-1.0.0.jar, testng-6.0.jar and junit-platform-runner-1.0.0.jar.

Place the testrunner.jar, the five jar files in junit_lib mentioned above into the path eclipse_root_folder\dropins\junit_lib\

We can use the following configuration to have the trace of a test method.