RFC: Mass rebuild of Fedora Extras before FC5 and how to handle orphaned packges for FC5

[Josh Boyer]

On Sun, 2006-01-15 at 15:27 -0500, Jeff Spaleta wrote:
> On 1/15/06, Josh Boyer <jwboyer at jdub.homelinux.org> wrote:
> > 1) Create a FE5 blocker bug.
> > 2) Open a rebuild bug for every package in devel and add it to #1.
> > 3) Maintainers rebuild their packages, fixing issues as they encounter
> > them.
> > 4) Close the bugs as they are completed.
> 
> And what happens if maintainers fail to kick off rebuilds?  Or there

We'd be able to tell because their bugs would still be open.  That would
allow others to help out in those situations.

> some sort of cascade such that an underlying dependancy package needs
> to be rebuilt at the same time as another package but package A and
> package B are maintained by different people?

Bugzilla has fields to allow this.  Package maintainers should know what
their packages depend on.  Mark package B's bug as being blocked by A's.

> 
> I've been in conversations with at least 2 people in #fedora now about
> weird oddness associated with extras-development rebuild attempts
> under mock where a chain of packages needed to be rebuilt together or
> else the rebuilt results failed.

I find that bizarre for cases where one is truly just rebuilding.  Have
bugs been filed?  Do we know why this oddness occurs?

> 
> I really think an effort needs to be made to do a mass-rebuild and let
> notify maintainers based on the failures in that coordinated mass
> rebuild.  The mass-rebuild tree doesn't necessarily need to be the
> normal public tree. But I think a mass rebuild should be attempted and
> the failures cataloged to see exactly how bad the gcc change has been
> for Extras.

To my knowledge, the buildsys doesn't have such a "non-public" tree.  I
know a scratch repo was suggested, but I don't think it has come to
fruition as of yet.

> How long would it really take Core development to get its packageset
> rebuilt if each Core maintainer was individually responsible for
> rebuilds after the gcc changes?  Instead of having a coordinated
> rebuild like Jesse pushed through the meatgrinder?

That's all well and good, but who are you going to get to pioneer this
effort?  Not to take away from Jesse's efforts, but that was his job.
There aren't many of us that get paid to work on Extras in that
manner :).

Seriously, if we can get a few people to help drive and work through a
mass rebuild then I'm all for it.  I'll even help.  But I still think it
should be tracked in bugzilla whether or not a single person is doing it
or each maintainer.  That way, if changes _are_ needed and a maintainer
is MIA for some reason, we at least have some history of what happened
and where the problems were.

josh