[PKG-Openstack-devel] Bug#835177: Bug#835177: aodh: FTBFS with eatmydata (build hangs)

[Thomas Goirand]

On 10/11/2016 05:41 PM, Santiago Vila wrote:
> On Tue, 11 Oct 2016, Thomas Goirand wrote:
> 
>> The problem was python-fixtures, which is maintained within the PKG
>> OpenStack, but for some reason, the Maintainer: field still has Robert
>> Collins, which is why it didn't appear in the PKG OpenStack QA page, and
>> I missed the upload to unstable.
> 
> That was really only half of the problem.
> 
> The other half was not building the package in a clean sid chroot.
> Hence my suggestion to always upload in source-only form.
> 
> Thanks.

Santiago,

Before throwing stone at me, you should consider the amount of package
we maintain within the OpenStack PKG group:
https://qa.debian.org/developer.php?login=openstack-devel@lists.alioth.debian.org

It is impossible maintain 400+ interacting packages the way you would
with your single pet package.

For the every day work, I do build all of the OpenStack packages into a
clean Jessie chroot within the OpenStack CI (which produces a
jessie-backport repository) using Gerrit. All of that is automated.
Within that, I know that, within the subset of OpenStack packages,
(build-)dependencies are correct (there's no way to cheat this system).

I would very much like to build the same way for Sid, using sbuild also.
Unfortunately, given the fact that I need to manually trigger the
builds, and given the number of packages, it is not feasible within a
reasonable amount of time. And I really wanted the migration from
Experimental (containing OpenStack Newton) to Sid (which had OpenStack
Mitaka) to take the least amount of time, to avoid brokenness which such
move would inevitably do. The shortcut I took was the only way to move
all packages from Experimental to Unstable. I just wish we had
automation within Debian to do it, unfortunately, this isn't the case.

If considering all of the packages as a single set, everything is fine
though, and all (build-)dependencies are satisfied. Python-fixtures
really is an exception here, that I wouldn't have missed if it appeared
in the group's QA page.

The only way I see to fix this, is deploying the same kind of CI/CD we
have designed in the OpenStack infrastructure, but using Sid, and
allowing this to upload to the archive directly. It is one of the
projects I would like to achieve within Debian (and the DSA machines).
Hopefully, this will happen one day. I'm not sure if I'll succeed
convincing the FTP masters that it is safe to have a CI to upload
packages on the behalf of maintainers, but that's another story.

Hoping you enjoyed reading the insights,
Cheers,

Thomas Goirand (zigo)