Regression Analysis - 5.8 WYSIWYG editor settings broke following refactoring (Fixed in 5.9)
What happened During the 5.8 cycle, a number of preferences screens were refactored to work with the standard settings approach which is a very lightweight form with pretty much all elements being driven from settings metadata. One of these settings that were changed was the WYSIWYG editor setting. There was a change in the way that the options for the select list was generated. This meant that the key was changed from the machine name of the option to the value. This had consequences in regards to javascript that runs and having the system detecting what WYSIWYG to load.
How & When was it addressed This was identified and logged in the Lab as an issue which was shown up in Joomla most prominently. It was addressed within two weeks of first being reported, which given the holiday season is probably ok. It was addressed by modifying the settings' usage of option group pseduoconstant to allow for an optional param keyColumn in the same way the DAOs do which then allowed for the switching back to being keyed based on the machine name and not the value.
How did the regression fit with our processes? The PR had been reviewed but there was a number of settings that had been migrated within the cycle and potentially not every single one was checked thoroughly.
Recommendations going forwards In this case a more thorougher code review may have picked up the differences, however given the number of settings changes and the likelihood the reviewer was mostly looking for front end changes rather than doing a full r-run it is likely this was just missed.