Week in review

[Jesse Keating]

- Had releng meeting, talked about summit results of rawhide
improvements and pungi changes.  Also discussed using PHX machines for
composes, spins for F9, trac

- Chatted with Mike McGrath about releng compose needs in PHX and tried
to come up with a strategy for providing us equipment but not having it
sit idle.  I opened a ticket and we will experiment with a few different
options.

- Talked with anaconda folks about some things leaving anaconda, like
pkgorder and splittree.py.  These have found their way into pungi and
have been tested OK.  Will look at rewriting some of those if necessary

- Started looking more into the M part of MRG for use in Fedora
infrastructure.  Pluggable architecting with keeping tools from having
too much domain knowledge and instead relying upon a communications bus
with subscribers and topics seems like a very powerful way to do about
building up features within our infrastructure.  I've got some materials
that I'll be going through over the next few weeks and possibly setting
up a test instance to play with a bit.

- Started work on getting pungi to gather debuginfo rpms.  Since no
communication can happen to the buildsystem itself this proves to be a
bit... interesting.

- Investigated why no images showed up in rawhide Tuesday, realized it
was because anaconda guys did a build without splittree/pkgorder and I
hadn't done a pungi build yet.  Fixed.

- Did more work on fedora-release to be useful for secondary arches.
This time the approach was as spot suggested, drop the key files with
arch names, refer to them in repo config files with $basearch.  A test
package was created and handed over to Dennis Gilmore who gave it thumbs
up.  Will build rawhide/updates for 9 today.

- Investigated a potential yum bug brought up by jpackage/jboss folks.
Was able to confirm it was a bug and get a test patch from the yum
developers.  Test patch worked, will build an updated F9 package to
resolve the issue.

- Read through John Poelstra's rawhide report from FUDCon and made some
minor edits, as well as filed some pungi tickets as a result of the
conversations.

- Spent a lot of time debugging more multilib issues with arch vs noarch
packages, this time in the 'pattern' path of yum install.  After
stepping through a lot of code I realized where the problem was and
called up James Antill to discuss potential fixes.  Tested various
patches from James to resolve the issue.

- Helped David Walluck through some mock configuration issues for
building jboss / jpackage contents.  Issues surrounded how to include
only specific packages from a given repo and how to add more to the
default buildroot.  This led to a discussion with yum/pykickstart folks
about giving yum install the ability to understand syntax like "install
@group packagea packageb @group2".  This would make it easier to
customize mock chroots and do other types of package work from the
command line.  Tested a patch James Antill created.

- Fixed up an update that was being hidden by an older n-v-r of the same
package.  Not sure how bodhi allowed them both to be shipped at the same
time.

- Discussed comps validation with Will Woods and Nicolas Mailhot; tools
to use, strategies on when to use them, etc...

- Investigated why there were no rawhide images on Wed, discovered that
when anaconda folks removed the anaconda-runtime package they forgot to
fix the repoquery and yumdownloader commands to not look for it.
Committed a fix to anaconda.

- Prepared F-9 testing update for fedora-release, with the ia64 key in
it and things setup to work for secondary arches.

- Played with the xmlrpc interface to testopia a bunch, working out ways
to autofile test runs each night for rawhide.  Ran into a couple
problems with the current script and made some changes.  Will discuss
with Bill Peck and David Malcolm long term changes both in our scripting
and in upstream API

- Ordered a couple OpenPGP smart cards to play with for the Fedora
signing server.

- Spent a good chunk of an afternoon trying to get xenner to work, only
to completely fail at it.  Resorted to full virt with kvm instead.