How can you reduce risk when implementing agile development in your SDLC?
In the simplest terms, regression testing is a form of software testing that confirms or denies the functionality of software components after changes to a system. A defect that emerges after a change to the software is known as a regression bug. In the event that a change is made to a software component and a regression bug is discovered and fixed, more regression testing is required to confirm that resolving this issue does not in turn cause other issues.
Unfortunately, it is all too common for regression testing to focus solely on the success rate of tests, whether automated or manual, without fully evaluating if this indicates the software is functioning as expected. In some instances, making changes to fix bugs can uncover functional problems that were not addressed by the previous test suite, making exploratory testing that much more beneficial. Furthermore, when developers make too many changes to the code base in too short a timespan, this often leads to instability of the software, the introduction of numerous bugs, and much higher fault rates.
It should, therefore, come as no surprise that weaving an exploratory testing phase into the regression testing process is a simple, yet powerful, step that can greatly improve the quality of the software. In many cases, a human will catch an unknown issue, even where a test case indicates no such issue exists.
For example, imagine a web project with an assortment of standard registration fields that users can fill out. At some point during regression testing, a bug is discovered that was preventing the nickname field from being displayed. Even after that bug is fixed in the code base and nickname is now visible, it's often the case that other elements of the application are broken after this fix, since many previous developer decisions could've been made in the time since the nickname field first vanished.
Even more likely, but just as problematic, is the possibility that test cases for the previously-hidden nickname field simply don't exist. Since the field was all but lost up until this recent discovery and bug fix, it now requires someone to closely examine all related test cases and perhaps create new ones or alter existing cases to also account for this "new" field.
Better known as a masked defect, even fairly simple examples of these types of issues can be problematic for the application development process. In cases where defects alter the behavior to such a degree that it is effectively hidden from the developers, it's common for test cases to fail to notice these issues.
For this reason, it can be beneficial to perform a crowdtesting pass before a major release. More often than not, testers may reveal masked defects, or even point to improvements, that would dramatically alter the end-result of the system and how it functions, saving a great deal of unnecessary or even wasted coding time during the subsequent development.
Regression testing is ideally performed after any changes are made to the code base. Regression tests are also typically executed anytime a previously discovered issue has been fixed.
The frequency of regression testing will vary from project to project, but most projects are served well by performing regression testing on a schedule: after every change, at the end of every day, weekly, bi-weekly, etc. Generally speaking, the more often regression testing can occur, the more issues can be discovered and resolved, and the more stable and production-ready the application becomes.
On the other hand, it's important to note the drawbacks of regression testing as it relates to testing frequency. In particular, when regression testing implementations focus primarily on automated functional tests, there can be a tendency for tests of this nature to be rather volatile and unstable, often requiring a good deal of personal intervention on the part of developers or QA members to babysit the test suite and confirm the performance and the results.
For this reason, a final suite of regression tests should also include a phase of exploratory testing before the system goes into production. Through crowdtesting or other methods, exploratory testing during regression testing empowers individual testers with the responsibility and freedom to examine, tinker, and unveil issues or avenues of inquiry that were not considered previously.
There are typically three different methods for approaching regression testing on a project:
Regardless of the method of implementation outlined above that you choose -- or if you opt for a hybridization of methods -- there are a handful of general best practices to following when implementing regression testing: