On Thu, Jul 25, 2019 at 06:30:10PM +0200, Paolo Bonzini wrote: > On 25/07/19 18:20, Sean Christopherson wrote: > > On Thu, Jul 25, 2019 at 06:10:37PM +0200, Paolo Bonzini wrote: > >> On 25/07/19 18:09, Sean Christopherson wrote: > >>>> This investigation confirms it is a new test code failure on stable-rc 5.2.3 > >>> No, it only confirms that kvm-unit-tests/master fails on 5.2.*. To confirm > >>> a new failure in 5.2.3 you would need to show a test that passes on 5.2.2 > >>> and fails on 5.2.3. > >> > >> I think he meant "a failure in new test code". :) > > > > Ah, that does appear to be the case. So just to be clear, we're good, right? > > Yes. I'm happy to gather ideas on how to avoid this (i.e. 1) if a > submodule would be useful; 2) where to stick it). As a starting point, what about adding "stable" branches for each kernel release to kvm-unit-tests, e.g. linux-5.2.y? I assume we'd need something similar for the submodules anyways.
