Planning the Fedora 11 Mass Rebuild

[Jesse Keating]

We have a perfect storm brewing for Fedora 11.  Multiple features
require rebuilds of some/all of our packages.

https://fedoraproject.org/wiki/Features/ArchitectureSupport
  All i386 packages becoming i586 packages.

https://fedoraproject.org/wiki/Features/gcc4.4
  All gcc compiled packages rebuilt.

https://fedoraproject.org/wiki/Features/StrongerHashes
  All rpms rebuilt for sha256 rpm hashes

Add them up and we have to rebuild everything.  Unfortunately each of
these have their own set of prerequisites before we can start building
them.  I'm estimating 5 days worth of koji time (background builds) to
get through everything.  Combine that with Feature freeze coming on
March 3rd, Beta freeze on March 10th, we need to start these builds
pretty darn quickly.

On Monday, this will be the hot topic of the Fedora Release Engineering
Meeting.  Releng has taken the task of coordinating the various above
features to the point where we can build once for all of them.  We will
be asking feature owners or responsible parties, as well as maintainers
of dependent software/systems to attend the meeting and help coordinate.

We'll also be discussing how to enact the mass rebuild.  In the past,
we've given maintainers a window to build their own packages before the
script started, as well as given maintainers a way to opt-out of the
script in order to do the build themselves.  This time, given the time
crunch and multiple features being juggled, I'm requesting that we start
the script as soon as the buildsystem/buildroots are ready, without
giving maintainers a chance to do it on their own.  I am still
interested in giving maintainers the ability to opt-out of the script,
and will be looking for suggestions on how to enable this.

As for the script itself, I plan to use rpmdev-bumpspec rpm the
rpmdevtools package to bump the spec, and then issue a background build
in koji.  The master script will do each step (cvs checkout, spec bump,
check in, build) per package before moving on to the next package.  This
should minimize the chances of cvs collisions with maintainers not aware
that their package is being built.  However due to the nature of
background koji builds, the scripted build may wind up superseding a
"newer" build a maintainer has done.  I'll be running some scripts to
detect this occurrence and react accordingly.

Before we start the script I'm going to ask Matt Domsch to kick off
another Fails to Build From Source run to see what state our packages
are in.  It is vitally important that any of these failures be fixed
ASAP.  Proven Packagers, its your time to shine (:  I'll also be running
the rpmdev-bumpspec across packages to view the output and identify
problem cases.  (help with this would be appreciated).

So again, if you wish to play a part in this task, or otherwise have
opinions on how it should be done, please reply here and/or show up at
the releng meeting Monday in #fedora-meeting at 1800 UTC.

-- 
Jesse Keating
Fedora -- Freedom² is a feature!
identi.ca: http://identi.ca/jkeating
-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: application/pgp-signature
Size: 197 bytes
Desc: This is a digitally signed message part
Url : http://lists.fedoraproject.org/pipermail/devel/attachments/20090213/27d2bd06/attachment.bin