failing build tests

Martin Gräßlin mgraesslin at
Mon Apr 24 18:53:01 UTC 2017

Am 2017-04-24 18:26, schrieb Jonathan Riddell:

> In master kscreenlocker and plasma-integration are also broken while
> kwin is pleasingly green

In case of kwin I must say that Plasma 5.9 had never been green, but the 
same state on master when it branched was green. To me it looks like a 
false positive and I decided to not spend time on it when I first saw 
it. Unfortunately we do not even see what's wrong as the all the 
relevant information is not shown. "The rest of the test output was 
removed since it exceeds the threshold of 307200 bytes." before there is 
any relevant information.

In case of plasma-integration the situation is similar. There is hardly 
any difference to master and it's failing nevertheless.

It might be a good idea to say the projects have to be green, but the 
test stability of our CI system is just not good enough for it (yet). We 
still have a bunch of tests failing if the system is under high load. 
E.g. if scripty did a run it is more likely that the tests in KWin fail. 
Also if krita is being built at the same time as kwin, the tests in kwin 
will fail. And if one pushes to Plasma 5.9 and then merge to master one 
can be quite certain that the tests will fail randomly.

If you want to have them green, just make sure no other build is running 
and trigger recompile till it's green.


More information about the Plasma-devel mailing list