On Wed, Apr 12, 2017 at 1:44 PM, Barak Korren <bkorren(a)redhat.com> wrote:
On 12 April 2017 at 13:33, Piotr Kliczewski
<pkliczew(a)redhat.com> wrote:
>
> I need to merge them and release before I can make the engine build to
> succeed and if
> the are api changes or dependency changes like in this thread I do not
want
> my changes
> to be used without the engine update.
>
> Sorry but we can't workaround it without making sure that we are using
> correct versions.
So this is a case where change A in project X needs change B in project
Y. We will eventually allow some syntax of specifying something like
this (probably in the commit message like OpenStack) to ensure they are
merged and tested together. In the meantime the closest approximation we
have is merging such changes as simultaneously as possible.
It is not possible due to maven central taking time to propagate the
artifact.
Consider what it means when an API-breaking change is merged into vdsm
master for example without the required engine-side change. It means
that from that point on, all further vdsm changes will need to wait
until the engine change is merged.
>> If we limit the CI system to only use tagged stuff it becomes useless.
>
> I makes sense for OST.
As I've commented before, you need to be clear what you mean when you
say 'OST'. If you clone the OST repo and run it, or use the manual job
you end up using the releases repos or (for master) the 'tested' repos.
Which seems not to be the case for the last failure.
The point of the 'experimental' flow is to tell you if the sum total of
oVirt's 'master' branches (or more general, branches by version) is stable
or not. Not running it on the latest patches would not make sense.
--
Barak Korren
bkorren(a)redhat.com
RHCE, RHCi, RHV-DevOps Team
https://ifireball.wordpress.com/