On Thu, 2018-04-19 at 16:15 +0200, Bjoern Voigt wrote:
I wonder, why OpenQA does not detect such inconsistencies (different i586 and x86_64 versions of kernel-default; kernel-default has different versions compared with kernel-devel, kernel-source and others) and why such inconsistent snapshots are automatically released to public.
OpenQA only reports issues for test where somebody wrote code for. Snapshots are automatically released because no test caught an issue. As as long as you don't have kernel-default-devel installed, there is not even something happening on workstations - and zypper dup just works; just that i586 machines got a newer kernel, x86_64 did not. But that it inself does not cause a problem. Even the KMPs in the repo were all 'correcly' build for kernel-default 4.16.1 Having a test that tries to install kernel-default-devel might be a start, but there will be other corner cases hidden. In best case, though, we finally find out why there are so many OBS wokers stalling while building. THIS is the problem that needs to be solved, then such issues jyst magically disappear anyway. Cheers Dominique