Aiding Software Developers to Maintain Developer Tests

Abstract

Abstract-Unit and integration tests can be invaluable during software maintenance as they help to understand pieces of code, they help with quality assurance and they build up confidence amongst developers. Unfortunately then, previous research has shown that unit tests do not always co-evolve nicely with the production code, thus leaving the software vulnerable. This paper presents TestNForce, a tool that helps developers to identify the unit tests that need to be altered and executed after a code change, thereby reducing the effort needed to keep the unit tests in sync with the changes to the production code. In order to evaluate TestNForce, we perform a user study that evaluates the adequacy, usefulness and completeness of TestNForce .

When, how, and why developers (do not) test in their IDEs

“…The weak correlations between test churn and test executions (RQ1.3), and production churn and test executions (RQ1.4) suggest an explanation: developers simply do not assert for every change that their tests still run, because "this change cannot possibly break the tests." Even when the modifications to production or test code get larger, developers do not necessarily execute tests in the IDE more often [43]. These observations could stem from a development culture that embraces build failures and sees them as part of the normal development life-cycle, especially when the changes are not yet integrated into the main development line.…”

Studying Fine-Grained Co-evolution Patterns of Production and Test Code

“…We prefer a dynamic solution over a static analysis approach because it is more precise as pointed out by Van Rompaey and Demeyer [11]. The key idea behind our approach is to run each test case separately, thereby identifying all the entities from the production code addressed by the test, similarly to the approach used [12]. To retrieve the covered entities (e.g., Java classes) we process test coverage information gathered with Cobertura 1 .…”

“…In response to observations on the lack of co-evolution, Hurdugaci and Zaidman [12] and Soetens et al [18] proposed ways to stimulate developers to co-evolve their production and test code by offering specialized tool support.…”

Measuring Test Case Similarity to Support Test Suite Understanding

“…This system is used as an example in the book "Growing Object-Oriented Software, Guided by Tests" by Freeman et al [6] to describe TDD techniques. The software and the related tests are explained in detail in this book and are publicly available 9 . The system comprises approximately 2,000 lines of code.…”