Some tests seems to be unpredictable (maybe depeds on arch)
While adding fixes from !10 (closed) and #110 (closed), I still cannot build folks for Fedora rawhide, due to failed 'test' run at the end of the package build. I made two builds, which are available here [1][2].
One eds test fails on timeout, one bluez test on SIGTRAP, but only on some architectures, not on all, which makes me wonder whether the tests are really deterministic, or, better said, whether there could be done any tweaks to make them fail in a less ratio. There's probably some timeout argument which can be used somewhere to increase it from 30 seconds to 45 or 60. The SIGTRAP once on s390x might be harder, depending why it failed (but not in both builds, which is confusing).
I also tried an 0.12.1 build with unpatched version [3] (for Fedora 30, thus old evolution-data-server) and it fails as well, thus I'd say it's not an evolution-data-server issue (but I've no proof of it, other than trying the same folks release with different evolution-data-server).
[1] https://koji.fedoraproject.org/koji/taskinfo?taskID=34977648
[2] https://koji.fedoraproject.org/koji/taskinfo?taskID=34977880
[3] https://koji.fedoraproject.org/koji/taskinfo?taskID=34978884