In the context of GUI testing, identifying robust locators (i.e., attributes to unambiguously identify on-screen widgets to be used in test sequences) is still considered an unsolved challenge by the researchers’ community. The frequent variation of attributes between different releases of the System Under Test (SUT) leads in fact to testing fragility, i.e., test case failing because of invalidated locators. Recent studies have highlighted the benefits of adopting multi-locator approach, i.e., the combination of multiple locators to enhance the robustness of widget identification.

The objective of this work is to provide insights into the composition of Android applications, assessing the characteristics of different layout-based properties and their suitability to be used as locators for widgets in the context of GUI-based testing.

We investigated the state of the practice by analysing the distribution of widget values within 30 real apps selected from the Google Play Store. For those apps, we selected two different versions to examine how they evolved over time from both visual and structural perspectives.

The results of our analysis showed that providing robust GUI testing over multiple releases of mobile SUTs is difficult, as identifying a single attribute or technique (either coordinate-, property-, or visual-based) capable of locating visual elements is often not sufficient due to missing values, variability, and instability of attributes.