Dependent-Test-Aware Regression Testing Techniques

This repository contains tools for dependent-test-aware regression testing techniques.

More details about the algorithms that are supported can be found in its paper and website.

Usage

The dependent-test-aware regression testing techniques are meant to be used on two versions of a subject. We will refer to the older version as the firstVers and the later version as the subseqVers.

The algorithms of the techniques require three main steps:

1. Setting up the metadata needed for the regression testing algorithms on the firstVers
2. (Optional) Computing the test dependencies for the regression testing algorithms on the firstVers
3. Running the regression testing algorithms on the subseqVers

Setup

The contents of this repository is tested on a Ubuntu machine with JDK 8. To install JDK 8, you may try the following.

sudo apt-get install openjdk-8-jdk
export JAVA_HOME=/usr/lib/jvm/java-1.8.0-openjdk-amd64/

You can see whether the setup was successful or not by seeing the text files under logs/.

Relevant contents of this directory

• setup.sh main script to setup the metadata needed for the regression testing algorithms (Step 1)
• compute-deps.sh main script to compute dependencies for the regression testing algorithms (Step 2)
• run.sh main script to run the regression testing algorithms (Step 3)
• example.sh example script for how to run the three main scripts
• dt-impact-tracer source code of the enhanced and unenhanced regression testing algorithms

Prerequisites

• The firstVers and subseqVers must be installed and its dependencies are copied into the target/dependency directory
• All tests in the original test order must pass. Step (1) will generate an original order to run and any test that doesn't pass in the original order is then skipped in Steps (2) and (3)
• It is possible that rerunning the setup script will result in tests having different coverage (e.g., one test can cover different paths in different test runs) and consequently result in different regression testing orders. The coverage of tests that achieved the results in our paper is available here
• The tools are intended to work with tests that are compatiable with JUnit version 4.12

General use case

To use the dependent-test-aware regression testing algorithms on any Maven-based, Java project one would need to do the following.

1. Compile and save the dependencies of the firstVers and subseqVers of any project (e.g., running mvn install dependency:copy-dependencies)
2. Setup the metadata by running bash setup.sh <path_to_firstVers_module> <algorithm_label> <path_to_subseqVers_module>
3. (Optional) Compute dependencies by running bash compute-deps.sh <path_to_firstVers_module> <algorithm_label> <path_to_subseqVers_module>
4. Running the regression testing algorithm by running bash run.sh <path_to_firstVers_module> <algorithm_label> <path_to_subseqVers_module>

The options for algorithm_label can be found in Tables 1-3 of our paper.

Example use case

The example.sh in the repository runs the enhanced algorithms on kevinsawicki/http-request (M9). E.g., running bash example.sh t2 runs (test prioritization, statement, relative) on M9 and it generally takes about 33 minutes to run. The script will then generate the following in the current directory:

• firstVers directory containing the firstVers of M9 (d0ba95c)
• secondVers directory containing the subseqVers of M9 (ef89ec6)
• logs directory containing the logs for running the various steps including building the two different versions of M9
• lib-results directory containing the results of the three steps. Specifically, the directory contains
  • PRIORITIZATION-ORIG-LIB-STATEMENT-RELATIVE-CONTAINS_DT-GIVEN_TD-true.txt contains the results of each test, the order the tests ran, and number of dependent tests observed in the enhanced T2 algorithm
  • prio-DT_LIST-lib-statement-relative.txt contains the computed dependencies used to enhance the T2 algorithm
  • PRIORITIZATION-ORIG-LIB-STATEMENT-RELATIVE-CONTAINS_DT-OMITTED_TD-true.txt contains the results of each test, the order the tests ran, and number of dependent tests observed in the unenhanced T2 algorithm
  • lib-orig-order contains the original order used. When running this order, all tests are observed to have passed
  • lib-ignore-order contains the tests that failed in the original order and are ignored for the algorithms
  • lib-orig-time.txt contains the time each test took to run in the original order
  • sootTestOutput-orig contains the coverage of each test
  • PRIORITIZATION-ORIG-LIB-STATEMENT-RELATIVE-FIXED_DT-OMITTED_TD-true.txt contains debugging information from computing dependencies

Note that the provided compute-deps.sh script can compute dependencies for test selection algorithms using the orders of test prioritization and parallelization as described in the our paper, but they do not merge all of the dependencies from all test prioritization and parallelization algorithms as we did in the paper. Instead, S1 and S4 relies on P1 with 16 machines to compute dependencies, while S2, S3, S5, and S6 relies on T1, T2, T3, and T4 (respectively) to compute dependencies. This change is to reduce the amount of time test selection algorithms may take to compute dependencies on new projects. To merge dependencies as we did in the paper, one can simply run compute-deps.sh on all prioritization and parallelization algorithms, and then merge their output (e.g., prio-DT_LIST-lib-statement-absolute.txt from T1 and prio-DT_LIST-lib-statement-relative.txt from T2) into one file.

Cite

If you use any of this work, please cite our corresponding ISSTA paper:

@inproceedings{LamETAL2020ISSTA,
    author      = "Wing Lam and August Shi and Reed Oei and Sai Zhang and Michael D. Ernst and Tao Xie",
    title       = "Dependent-Test-Aware Regression Testing Techniques",
    booktitle   = "ISSTA 2020, Proceedings of the 2020 International Symposium on Software Testing and Analysis",
    month       = "July",
    year 	= "2020",
    address 	= "Virtual Event",
    pages       = "298--311"
}