On vendredi 1 février 2019 10:32:19 CET Ben Cooksley wrote: > On Fri, Feb 1, 2019 at 9:15 PM David Faure <fa...@kde.org> wrote: > > I'll try to get around to debugging this, but I'm surprised by one thing: > > doesn't the same issue happen in KIO's own tests? It also has several > > tests > > that make use of KIO slaves directly, skipping KLauncher. > > Indeed, I found it surprising as well. > > Only thing I can think of is that the tests in KIO tell the slaves to > exit before it finishes itself and exits - while the tests in > plasma-integration don't do this.
There's no API in KIO for telling slaves to exit; they are supposed to just do that by themselves, after being idle for a while. > > Ironically, the reason we skip klauncher is exactly to avoid this problem > > in the first place (kdeinit and klauncher would keep running after the > > test ended). > > The CI tooling should eliminate the need to do this for the most part > as long as you don't try to launch them yourself, as it will launch > kdeinit as part of the environment setup process :) > https://invent.kde.org/sysadmin/ci-tooling/blob/master/helpers/run-tests.py# Well, let's try without KDE_FORK_SLAVES then? I just pushed that in https://commits.kde.org/plasma-integration/a94bbdaadff282a26dece12172234b8bec1a29c9 also merged to master. Can you re-enable CI and let me know if it works better - or worse? ;-) -- David Faure, fa...@kde.org, http://www.davidfaure.fr Working on KDE Frameworks 5