Plan for tomorrow's (20070604) Release Engineering meeting

[Josh Boyer]

On Sun, 2007-06-03 at 17:29 +0200, Patrice Dumas wrote:
> On Sun, Jun 03, 2007 at 11:08:23AM -0400, Jesse Keating wrote:
> > /topic RELENG-Meeting - Updates Policy - JesseKeating
> 
> About update policy, first I think that it should be easy to do 
> something like:
> 
> make go-to-release
> 
> which would use the changelog entry (or something similar). If I recall
> well this is already more or less agreed.

Yes.  I think Luke is working on a bodhi command line client (or will
be) that could be wrapped in the makefiles.

> It should also be possible for a maintainer to decide what amount of
> time the package should stay in the testing repo. The default would be
> something reasonable (one week?). The rel-eng team could have the 
> possibility to block the update at any time if they find something is 
> abnormal.
> 
> For example 
> make go-to-release 1
> would make the package stay 1 day in testing repo.

That would be the point that needs the most discussion.

> And there should be automatic checks for broken deps, for duplicated
> provides, for upgrade path.

That is planned.  Just needs coding.  I think everyone agrees that would
be a very good thing to have.

> > /topic RELENG-Meeting - Freeze Policy - JesseKeating
> 
> On that topic, I think that during a freeze there shouldn't be a need to 
> mail to rel-eng to push to the frozen collection. It should be a
> specific target, so for example to push to the frozen collection one
> should do
> 
> make build
> make go-to-freeze
> 
> Then the rel-eng would see that (together with the changelog entry), and 
> block if something seems dubious, or if it is too late in the freeze. It 

Their ability to do that, under those circumstances, might be limited.
Rel-eng doesn't immediately know if it's too invasive, which is why a
request is sent.  Not all changelogs are descriptive.  But perhaps
something better could be done, yes.

Also remember there will be a point in time where things are completely
frozen so that ISOs can be made.  During that (relatively) short period
of time, the push feature would be disabled.

josh