Google published the results of a research of the time spent by engineers on updating open source software packages used in the company’s internal products. The data comes from a survey of 346 Google employees involved in package updates. Questions related to how much time an engineer spends preparing an updated version of a package and whether the associated labor costs can be predicted using metrics for assessing the complexity of maintaining source code.
In most cases, preparing an update took up to 4 hours, but in some situations the process dragged on for days or months. The main indicators of complexity that influence labor costs during maintenance are the presence of separately supported patches that are not transferred to upstream, the number of dependencies of the package and the size of the developer community in the upstream project. In 49% of cases, delays in preparing an update are associated with the need to adapt their patches to changes in new releases, in 37% of cases – with the appearance of changes in the upstream project that break compatibility, in 31% – with changes in Google code that break compatibility, and in 29% – with the overall complexity of the package.
