Stable update procedures

Luke Macken lmacken at
Sat Jun 6 12:08:39 PDT 2009

On Thu, Jan 29, 2009 at 11:50:08PM +0000, James Westby wrote:
> What are the rules for what is allowed?

I think Rahul did a good job of explaining some of the details of our
updates policy, so I'll briefly explain a little bit about how our
updates management tool works, since I wrote it.

> What is the process for doing it?

After packages are built in our buildsystem Koji[0], the developer can
then push them into Bodhi[1], our updates management system.
Here the maintainer is encouraged to reference bugs that this update
fixes, as well as a description of the update, along with some other

> What QA is done on the updates?

We encourage package maintainers to push their updates into our
'updates-testing' repository first, but there is nothing stopping them
from pushing updates straight to the stable repository.

Bodhi employs the concept of 'karma' for each update.  This allows for
testers to easily provide feedback by +1/-1ing updates.  When submitting
an update, the maintainer can choose to allow 'karma automatism' and can
set high/low karma thresholds for their update.  This means that
maintainers can say "Automatically push this update to the stable
repository after it reaches a karma of 5", or "Unpush this update if it
reaches a karma of -3".  We do also allow for anonymous feedback,
however it does not effect the karma of the update.

I ran some statistics[2] a few months ago on how many updates go to
testing, and for how long.

We currently do not have any automated QA procedures for our updates,
but there is a solution in the works[3].

> How are regressions caught?

As I mentioned above, we croud-source a lot of our QA, which has been
working out pretty well so far.

> How are regressions handled?

Regressions in updates-testing builds are easily mitigated by just
unpushing the update.  Regressions to stable updates will require
another update to fix it, since we do not unpush stable updates.


[0]: |
[1]: |

More information about the Distributions mailing list