[Date Prev][Date Next]   [Thread Prev][Thread Next]   [Thread Index] [Date Index] [Author Index]

Re: The future of "rawhide" (was [Fwd: Re: "What is the Fedora Project?"])



On Fri, 2009-10-23 at 20:56 +0000, Colin Walters wrote:
> On Thu, Oct 22, 2009 at 4:50 PM, Jesse Keating <jkeating redhat com> wrote:
> >
> > Rawhide as we know it, /pub/fedora/linux/releases/development/ will
> > remain "rawhide".  We may even change the path to say rawhide, just to
> > catch things up and well I like keeping mirrors on their toes.  Rawhide
> > will be a repository of developmental and experimental packages.  Things
> > being worked on for the future.  It will /not/ be an installable tree,
> > rather it will just be a repository of packages, to be added on to an
> > already stable "base", eg you'd install F12, and enable rawhide to test
> > rawhide.  This will significantly lower the complaints that "rawhide
> > isn't installable".
> 
> So as I understand it there are a number of reasons why rawhide might
> not be installable, but broadly they fall into two major categories:
> 
> * Anaconda
> * Critpath packages
>   - Dependency/rebuild issues
>   - Bugs in %posts (like the user/group one we ran into with dbus)
>   - Core bugs (graphics drivers)
> 
> It seems like we're basically just skipping Anaconda, since you won't
> be able to yum if there are depsolving issues (ok, modulo
> --skip-broken), and for the latter two you don't end up with a
> "working" system.
> 
> Let me do a counter-proposal:
> 
> We simply do not let showstopper regressions in the critpath stay in
> rawhide.  If something in critpath has a showstopper, it halts all
> further commits to the entire critpath until it's resolved (either
> fixed, or reverted).  The definition of "showstopper" might be "AutoQA
> fails".  And since AutoQA will have been doing some basic smoketesting
> of the installer, we have to be producing installer images as a side
> effect.
> 

I... don't see how this helps, other than piss off the rest of the
crit-path maintainers while one thing is broken.

AutoQA will be running at some point, and it can be doing the
qa /before/ things get tagged for rawhide, so if you break deps with
your build, it doesn't get in, unless you force it and then you face the
wrath of releng/qa.  To catch core bugs, we'll need a bit more advanced
autoqa, doing more than just repo level testing but doing actual package
testing.  That will grow over time and again can be done pre-tag.

Your counter proposal also does nothing to help the dual or sometimes
triple role we try to put on "rawhide" the path.

-- 
Jesse Keating
Fedora -- FreedomĀ² is a feature!
identi.ca: http://identi.ca/jkeating

Attachment: signature.asc
Description: This is a digitally signed message part


[Date Prev][Date Next]   [Thread Prev][Thread Next]   [Thread Index] [Date Index] [Author Index]