From: Stefan F. <ste...@we...> - 2012-02-09 13:15:14
|
> > I would advocate against building UI units (as regression tests) right > now because > - units would have to be adjusted too frequently (UIs are currently > altered far more frequently than the game engine) > - the number of regressions has been very low > - additional technology is required to do UI units right > - UI units wouldn't have helped with javac/jvm issues as those > reported in this thread > Frederick: one could argue about the ratios between regressions and frequency of changes in the UI and you are right that the recent bug would not have been catched, however that was not the reason for me to bring up the topic: The GUI is not yet covered by any test, so it has marginal gain is largest here to add tests there. And for each release I do some manual testing (e.g. if Rails still starts and allows to finish an auction and load a game). This is the only manual task in the release process and I would be glad to automate that as well. Have you any experience in the choice of a Java/Swing ui framework? I have only limited experience with Mercury testing (from my professional past), but there should be much more focused/integrated open source tools. Maybe Brett has some recommendations too? What I have seen recently was Cucumber's approach, but most likely that is an overkill, but I am curious about it, especially as it seems to be (in all respects) language-agnostic. Stefan |