RECOVSLICING

Here is a case where data dependencies were recovered using RecovSlicing.

Recovering data dependencies with RecovSlicing ultimately enabled the successful identification of the root cause. 

The bug originates from the Defects4j dataset, specifically from Cli 11.

Defects4j citation: https://github.com/rjust/defects4j

RecovSlicing UI

Components of the UI

In the RecovSlicing Perspective in Eclipse, there are 5 components acting as the visual aid of RecovSlicing, namely:

1. Buggy Trace View: Execution Trace (Buggy Version)

2. Correct Trace View: Execution Trace (Working Version)

3. Compare View: Source Code of the Program

4. Step Properties View: Variables in the Focal Step

5. Console: Logging of Debugging Information, LLM Prompts and Responses

You can use this 'Recover All Data Dependency' button to generate trace:

Or you can use this 'Recov Slicing Evaluation' button to batch run dataset:

How to use RecovSlicing

Before using the 'Recover All Data Dependency' button, you need to set configuration in preferences. The specific operation is as follows: you can click Eclipse-> preferences-> Microbat Debugging, set your Java (1.8) home path; step limit (the maximum trace length you want to generate, in the experiment we use 100,000); variable layer (the number of variable expansion layers, here we set it to 1)

Next, you need to set configurations in RecovSlicing Setting. In Experiment Settings, check Enable Tracerecov and Enable Tracerecov in Auto Root Cause Localization. In LLM Model Setting, select your LLM Model and set the API Key. 

Finally, set the path and other information of the project you need to use in Tregression (Defects4j).

If you want use the 'Recov Slicing Evaluation' button, you need prepare your dataset, and then open eclipse, click File -> Open File in eclipse, select running_configs/run_configs.txt in the repo, and click Open. Before running you need select the configuration file to be executed, and then click the Execute button in the toolbar.