On Mon, Sep 21, 2020 at 9:55 AM Miro Hrončok
<mhroncok(a)redhat.com> wrote:
>
> On 21. 09. 20 15:40, Mark Wielaard wrote:
>> Hi,
>>
>> I couldn't build elfutils because of an annobin bug that showed up on
>> ppc64le. Nick was nice enough to fix it and push a new annobin version:
>>
https://bodhi.fedoraproject.org/updates/FEDORA-2020-c78fce452e
>> So I could build elfutils again:
>>
https://bodhi.fedoraproject.org/updates/FEDORA-2020-06dafb46c1
>> But now I am getting notices about the ELN build failures:
>>
https://koji.fedoraproject.org/koji/buildinfo?buildID=1613859
>> Which is clearly because it doesn't have the new annobin package.
> In this case, the build fails, which is the better outcome.
>
> However recently, I've rebuild redhat-rpm-config with this change:
>
>
https://src.fedoraproject.org/rpms/redhat-rpm-config/c/0d621460
>
> And later I've built python3.9 with the new macro to actually deliver the
change.
>
> In order to make it work, I've run:
>
> $ koji wait-repo f34-build --build=redhat-rpm-config-172-1.fc34
> $ koji wait-repo eln-build --build=redhat-rpm-config-172-1.eln103
>
> before building python3.9 in rawhide.
>
> But I wondered: In case I would have not run the ELN waitrepo, would ELN
> possibly rebuild python3.9 before the new redhat-rpm-config was available in the
> ELN buildroot?
Yes, this race is possible if the time it takes ELN to complete the
rebuild of the first package is longer than the time it takes the
second package to build and then also trigger the ELN build. It's
unlikely to occur often, but it is a possibility. I'm not sure how to
avoid it since without a side-tag we can't know which builds might be
dependent on one another.
For those grouped builds that use a side-tag, we're going to be
monitoring those and rebuilding them in the same order as they are
submitted. We can also add a waitrepo to our behavior to ensure that
we don't start newer builds before the prior ones succeed.
Thanks for bringing this up!
I think I have another build ordering issue. Recently libevent had a
soname bump and deps were built in a side tag. However, in ELN
openmpi-4.0.5-2.eln103 (which was the release bump for the libevent
rebuild) was built before libevent-2.1.12-2.eln103 was. Now I'm getting
spammed every couple of hours with netcdf and vtk ELN build failures due
to broken deps on libevent (and others).
Are we now stuck with bumping the release for openmpi again for the ELN
build?
Orion
--
Orion Poplawski
Manager of NWRA Technical Systems 720-772-5637
NWRA, Boulder/CoRA Office FAX: 303-415-9702
3380 Mitchell Lane orion(a)nwra.com
Boulder, CO 80301