For minutes of previous meetings, please see FoundationsTeam.

Present

• RobbieWilliamson - chair

• ColinWatson

• EvanDandrea

• JamesWestby

• LarsWirzenius

• MichaelVogt

• MikeTerry

• MuharemHrnjadovic

• ScottJamesRemnant

• SteveLangasek

Apologies

• MatthiasKlose - OEM sprint

Agenda

• Karmic
  • i486/i586 [Keybuk]
  • Status Updates on Alpha 3 Deliverables

    • https://wiki.ubuntu.com/FoundationsTeam/ReleaseStatus/Karmic

• Sponsorship queue
• Objectives
• Good news

  • Hardy and Jaunty are LSB 4.0 compliant

• AOB

Outstanding actions from last meeting

Actions from this meeting

Activity reports

Colin Watson

• Removed nasty kernel version hardcoding from armel CD build scripts(the replacement code is nearly as nasty, but at least shouldn't require frequent changes).
• Ran into udevd crash (#396957) and spent some time helping to track it down.
• Reviewed/approved specs:
  • foundations-karmic-multiarch-support
• Sponsorship:
  • libnice #392225.
  • python-pylibacl #391574.
  • scim-bridge #305452.
  • gparted #398099.
  • xscreensaver #398007.
• Merged: dpkg, grub-installer, partman-auto
• Archive admin half-day: cleared syncs (plus some minor tool fixes), did a bunch of not-built-from-source removals, adjusted some kernel component overrides, cleared priority-mismatches, cleared binary NEW
• Feature work:
  • foundations-karmic-server-installer-improvements:
    • Finished and committed LVM configuration rearrangement.
    • Merged mdcfg into partman-md, partly because we've needed to do
      • that for ages anyway and partly because it'll make the RAID configuration rearrangement much easier. Patch sent upstream for review.
    • Started work on RAID configuration rearrangement.
• Adjusted finish-install for upstart 0.6.
• Built i486/i586 live CDs out of live filesystems delivered by Adam.
• Updated my parted swap naming patch; accepted upstream.
• Reviewed relevant mail thread and added Muharem to the archive administration team. Filed RT ticket for shell access.
• Caught up on technical-board mail actions.
• Fixed hdparm #396837 (crash when HDIO_DRIVE_CMD(identify) fails).
• Fixed partman-auto #287571 (create_primary_partitions sometimes attempts to create last partition larger than free space).
• TB meeting.

Evan Dandrea

• Investigated using wubi's tools/pywine to unit test the windows frontend and backend for usb-creator.
• Reviewed the oem-config and ubiquity merge a number of times,sharing input and patches with mterry.
• Dug at 397366, but have not found a suitable solution yet.
• Merged UNR changes to Wubi for NCommander.
• Undid the merging of bin/usb-creator-{gtk,kde}, cleaning up the code a bit along the way.
• Released a new Wubi and umenu for 8.04.3.
• Sorted out some final travel details for Debcamp.
• Debugged a bug in partman that turned out to be the fault of my desktop running 9.04 rather than karmic.
• Worked on the slideshow specification. Mostly finished the ubiquity code. Had a discussion with mpt about some aspects of the

design. Tried to start a conversation with Gerry, but he's on holiday until later this month.

• Followed up with a community member (on ubuntu-installer@) who wanted an option to not wipe a partition on install in ubiquity.

James Westby

Distributed Development

• Importers down for 4 days due to LP API breaking. Back up now.
• Cleaned up and improved the documentation somewhat.
• Investigated some of the failures.

Daily Builds

• No progress this week.

Kerneloops

• Worked on the kerneloops changes.
• Hit a snag that apport may not submit from the same machine that the problem happened on, so need to be a bit more intrusive in to kerneloops than first thought.

Distro Tasks

• Reviewed some packages on REVU.
• "Half-day" archive admin duties on Monday.

Other

• Chasing up confirmation for DebConf/sprint flights (again).

Lars Wirzenius

Michael Vogt

AppCenter

• Code cleanup in gnome-app-install (AppCenter spec)

• experiment with a apt-xapian-index plugin for the app-install-data-ubuntu data, not happy yet
• Performance analysis in g-a-i startup

software-properties

• Bug triage/fixing in software-properties
• Ensure daily apt cron job is executable when software-properties changes it (#390319)
• Port gnome-control-center to the new ubuntu-system-service
• Software-properties: ensure only valid lines can be added with add-apt-repository

synaptic

• Integrate apt-xapian-index --update into synaptic
• Debug problem with a-x-i and synaptic (constantly rebuilding), send patch to enrico
• Merge synaptic from debian, upload new version to ubuntu

unattended-upgrades

• Add pm-utils hibernate hook to ensure that unattended-upgrades is finished before going into final sleep
• Merge unattended upgrades from the mvo branch, upload to karmic

update-manager/update-notifier

• Show number of available updates in update-manager
• Work on update-notifier to ensure that logrotated logs are considered for the auto-launch code
• Upgrade testing
• Test update-manager landscape upgrade support and set spec to beta-available
• Fix main-all auto-upgrade tester profile
• Port update-notifier to use GAppInfo instead of g_spawn (#46258)
• Fix crash in update-manager refresh_updates_count()
• upload new version

kernel-crashdump

• update wiki because crash behaviour on sysreq-c changed in 2.6.31
• work on the apport-retracer
• Push apport changes for the kernel crashdump stuff to lp:~mvo/apport/kernel-crashdump
• add crash_signature support plus test
• Discuss apport behaviour for local retracing (#392074)

apt

• Work on more dynamic apt cdrom method(lp:~mvo/apt/libudev-dlopen)
• Debug failure in main-all auto-upgrade test, turns out that for really huge install lists there is a (hard to trigger) bug in libapt
• Add check to apt.cron.daily to ensure that timestamps in the future get killed

==== misc =====

• app-install-data-ubuntu update (review new desktop files etc)

• Add new script "add-apt-repository" to python-software-properties(ApturlPolicy spec)

• Debug/workaround problem in gksu with the latest libgtk (with seb)
• Make "include-links" the default in python-central and push to lp:~mvo/+junk/python-central
• improve spelling suggestions in command-not-found
• Port ubuntu-system-serivce to polkit-1

sponsoring

• computer-janitor, gnome-app-install
• calctool (new ver), brasero (new ver), update-manager (bugfix)

Mike Terry

Ubiquity

• Finally pushed the oem-config merge to trunk! Thanks to Colin and Evan for spotting and fixing some final issues.
• Created a branch for translated timezones and filed a merge request. It's very fancy, but there needs to be a bit of discussion about how to sort the unicode lists.
• Started looking at writing some unit tests

Rsyslog

• MIR for rsyslog was approved
• MIR for librelp was not, as there was a possible out-of-bounds error

in librelp that needs to be fixed. I emailed the librelp folks about it.

Muharem Hrnjadovic

My Launchpad past caught up with me :P in the last week I was mostly sorting out package set and archive permission issues that would stand in the way of the archive reorganisation:

• done: Fix hard-coded archive references in package set based archive permissions (https://bugs.edge.launchpad.net/bugs/397490)

• in progress: Expose package set related archive permissions on the Launchpad API; the 3 branches that implement this are currently being reviewed. (https://bugs.edge.launchpad.net/soyuz/+bug/397494)

• in progress: Package sets should be associated with a distro series; this has been analysed for impact and a database patch was prepared but it's highly unlikely these changes are going to land this week. (https://bugs.edge.launchpad.net/bugs/399186)

Scott James Remnant

• i486/i586 performance comparisons; the result was that the big performance gain we saw comes from gcc 4.4. There is little changed between -march=i486 and -march=i586 when we *also* force -mtune=generic as we do (generic is in fact tuning for P4-era processors). In fact, the only thing enabled by -march=i586 -mtune=generic over -march=i486 -mtune=generic is the cmpxchg8b instruction, which libc would use if it wasn't explicitly #ifdef'd out. It may be prudent to do the switch anyway in case there are other optimisations later down the line.

Upstart

• Upstart 0.6.0 was finally released! This is a new stable version based off the D-Busified 0.5 branch I'd been working on for the past couple of years in between management duties.
• After the release, was still having issues with the "expect fork" and "expect daemon" code [which still uses ptrace()], debugged and applied a patch that seems to fix it well enough until we move to the netlink code.
• Created a "release" bzr branch of Upstart; this contains common history with both upstart trunk and libnih trunk, merged together in the correct layout - and then has the tarball unpacked over the top.

• Created an Ubuntu packaging branch from that; I can "bzr pull" to update to the latest release, and I can "bzr merge -c" to cherry-pick from trunk!

• Some issues with building Upstart on ARM, the test suite triggers a compiler bug and either hangs, runs out of memory or hits an Internal compiler error. I love my test suite.

udev

• Updated to 145
• Did some gardening of the bugs list, since I hadn't been focussing on it for a while

util-linux

• Got the branches all back in sync between Lamont and I
• Added a commit for Colin's recent upload
• Updated the Debian and Ubuntu trees to 2.15.1, uploaded to Ubuntu
• Ported our hwclock udev rules to Debian and made the start rule of the init script conditional on udev not running (agreed during Debian Boot sprint).

Kernel

• The Upstart test suite strikes again! This time it identified a regression in inotify since it was entirely rewritten upstream to be based on fsnotify
• Discussed the bug with upstream maintainer, it's being caused by merging duplicate events for a directory based on only the mask - and not checking the filename (ie. multiple CREATE, CHANGE or DELETE events for different files in the same directory are merged into one for the first file)
• Patches supplied to the kernel team since they just missed rc3

Misc

• A fair amount of discussion over the gdm vs. getty bug, where both end up sharing the same vt. This is fundamentally caused by getty VTs being hardcoded while X is free to chose the first it thinks is free. I'm still in favour of moving getty to be not hardcoded, so our configuration would be "start N gettys" and they'd get the lowest N numbered VTs after X

Steve Langasek

Release management

• Ubuntu 8.04.3
  • finalizations of several pending SRUs
  • push fix for open-iscsi bug #236640
  • candidate images available with final kernel
  • image rerolls for wubi, umenu fixes
  • discussions with desktop team regarding bug #207072
  • identify and implement fix for bug #207072
• phone call to discuss the escalation process for Support SRU bugs

Packages

• clean up acpi-support dependencies

Specs

• continued work with Debian on multiarch spec

Misc

• Monday archive duties

• Preparing DebConf talks

• work with Dustin on pam_motd