I agree completely. I think as a group we need to be on board for always having green tests; it has been shown time and again to be one of the best ways of achieving stable, reliable software.
I've tried a few times to get tests running locally for the bigger projects like plasma-workspace, plasma-desktop, and kwin but they seem to have serious issues; if I remember correctly failing even though they pass on CI, as well as getting hung and not progressing. Not a locale issue, since my system is en_us. If anyone has insight into how we can fix that and make them easier and more reliable to run locally it would help a lot. As far as the CI goes, I think a good solution is making test passing mandatory for merging. I started our migration to this already: https://discuss.kde.org/t/test-failures-fatal-by-default-in-kde-ci/31670 My work on that has fallen a bit by the wayside since I was hired as Plasma Software Engineer and I have been putting most of my time into KISS, but I will try to squeeze in the time to work at it.. On Thu, Jul 24, 2025 at 11:40 AM Nicolas Fella <nicolas.fe...@gmx.de> wrote: > Hi, > > for the last few weeks the CI pipelines in plasma-workspace and > plasma-desktop have been broken, and we have become accustomed to ignoring > them. Needless to say that's bad, and it needs to change. > > In plasma-workspace for most of the time it was only clipboardtest that's > failing, but since recently it looks like more of them are failing. > The following tests FAILED: > 33 - servicerunnertest (Failed) > 54 - clipboardtest (Failed) > 55 - devicenotifiertest (Failed) > 68 - kcm_colors_test (Failed) > > 69 - kcm_cursortheme_test (Failed) > > Why clipboardtest started failing in the first place isn't clear to me. > Some of the other tests could be explained by recent CI changes. > > In plasma-desktop kcm_kded_test and kcm_keys_test. Both are failing > because of newly introduced warnings from Kirigami code, and despite the > problem being openly visible and me pointing out the problem in chat no fix > was made. > > Furthermore, when I try to investigate failures locally I find myself > unable to even run then relevant appium tests locally ( > https://invent.kde.org/sdk/selenium-webdriver-at-spi/-/issues/16), and > for a long time I remember having trouble with tests being broken when > running locally due to e.g. being sensitive to non-english locale. > > Note that I don't want to blame any particular person(s) here. This > appears to be partly a social problem in that we collectively got so used > to flaky tests that we started ignoring broken pipelines. That's not a > healthy situation to be in though. > > Cheers > > Nico > > -- Cheers, Kristen
