Re: concerns about the state of buildds for jessie
On 2015-02-12 22:18, Philipp Kern wrote:
> On 2015-02-12 13:02, Holger Levsen wrote:
>>> Possible avenues include updating the forks and working on making the
>>> forks
>>> no longer necessary.
>> are these forks maintained in VCSs?
>
> http://anonscm.debian.org/cgit/buildd-tools/sbuild.git/log/?h=buildd-0.64
>
> Historically forks have been needed because fixes in stable are hard. If
> stuff breaks in testing or unstable you usually need to fix it quicker
> than with a point release. A point could be made that the changes should
> be pushed to stable instead.
>
> As far as I know there's also still no builddadm-maintainable puppet
> tree. (Partly my fault I acknowledge, because I hoped to be able to do
> rabbitmq, but failed working against a black box I don't understand.) If
> we could ship the relevant helper scripts through Puppet (and unify
> configuration) we could also make most of the fork moot and just
> cherry-pick new versions from testing.
>
> Kind regards
> Philipp Kern
>
>
For reference, Ansgar (CC'ed) have started to merge the branches (see [1]).
I believe his short-term plan was to merge the general stuff into
master, have that uploaded to unstable and then do a buildd specific
version with the remaining Debian-specific parts.
In the long term, we should definitely aim for having exactly one
implementation of sbuild. We might need to optimise some of the
processes for this to work. I believe DSA are usually happy with
pulling from either proposed-updates and backports, so we should be able
to find a suitable solution for getting timely fixes available on the
Debian machines.
Thanks,
~Niels
[1] https://anonscm.debian.org/cgit/users/ansgar/sbuild.git/log/
Reply to: