Pushing from updates-testing to updates

Hans de Goede j.w.r.degoede at hhs.nl
Fri Jun 1 17:26:11 UTC 2007


Will Woods wrote:
> On Fri, 2007-06-01 at 10:56 -0400, Jeff Sheltren wrote:
>> Is there a policy (or guideline) in place for pushing packages from  
>> updates-testing to the updates repo?
> 
> We talked about this a bit in yesterday's QA meeting. Here's a first
> draft of our proposal.
> 
> Pushing from updates-testing to updates currently requires intervention
> from a rel-eng person (i.e. someone who has the official signing key).
> But how do they know when a package is ready to go? 
> 
> Each update in Bodhi's Testing Updates list (i.e. the stuff in
> updates-testing) should have a "QA Verified" flag. If the flag is set,
> the package should be considered ready to go live.
> 
> How does the flag get set? One of two ways:
> 
> 1) The QA team (i.e. pretty much anyone who can use bodhi and knows how
> to do basic package sanity checking) will go through that list and
> verify the packages in the updates. 
> 
> 2) After a week in updates-testing, the flag is automatically set - this
> matches the traditional updates-testing behavior for Core.
> 
> "Verifying" a package means doing basic sanity testing - install the
> packages, make sure there's no broken dependencies, make sure binaries
> run without segfaulting, etc. We'll post package verification guidelines
> on the wiki for discussion and refinement.
> 
> Does that sound reasonable?
> 

It does sound reasonable, however I don't see any QA done flag in Bodhi atm, is 
this on the todo list?

Also I would like to see a way to shortcircuit this, for for example trivial 
fixes fixing serious bugs. Rationale:
-Since Fedora X is a stable release, the build environment for packages should
  not change significantly, and thus a mere rebuild against the normal stable
  environment should have a very small cance of introducing regressions
-If the patch is trivially correct and thus can be verified at the source
  level, the patch should have a very small chance of introducing regressions
  too
So if an update fixes a serious bug with a trivial patch, then the importance 
of getting the bugfix to users outways the very small chance for regressions. 
(much like security fixes)

Notice that this was already discussed and agreed upon some weeks ago (I can 
dig up the thread if want me too), now all we need is a mechanism to make this 
happen.

Regards,

Hans




More information about the Fedora-maintainers mailing list