Debian Installer team monthly meeting minutes (20060128 meeting)
The ninth Debian Installer team meeting was held from 17:00UTC to 18:09UTC
on Saturday January 28th 2006.
There were about 73 people connected to the channel during the meeting
and 13 of them spoke during the meeting at least once.
The full log of the meeting is available at
Most of the meeting discussions were targeted at the D-I etch beta2
Please refer to
http://lists.debian.org/debian-boot/2006/01/msg00559.html for the
The transition of 2.6.15 kernel packages to testing is what determines
the whole schedule, so the discussion was initially focussed on this.
An upload of 2.6.15-4 packages to unstable is planned around this week-end.
(2/2 note: did not happen yet)
A dependency loop exists with udev but is not release critical, from
the RM point of view, so this should not prevent 2.6.15-4 to enter
testing. Frans confirmed later with Steve Langasek this is a RC issue but
couldn't find the bug anymore.
Maximilian Attems mention a leak with md (drivers?) which supposedly
has been fixed in lkml. Whether this is RC or not needs to be
In short, the transition of 2.6.15-4 has to be fasttracked and we have
to remind all our fellow developers that this is a blocker for D-I
No blocker issue remains. There is work on klibc and stuff that needs choices and decisions (see the meeting log for the details) but this is anyway
not a blocker for us as there are working options for all arches with
some recent work on base-installer.
Status of udebs transition
Almost all general AND arch-specific udebs have been rebuilt and uploaded.
Some work remains for m68k and hppa. Frans will post a reminder to the
porters. hppa needs to switch from 2.6.14 to 2.6.15 and upload its
arch-specific udebs (post-meeting update: Dann Frazier did the
uploads, except kernel that will be handled by Bdale Garbee). m68k
needs kernel udeb updates and its arch-specific udebs as well.
base-installer will be rebuilt and uploaded to fix the last discovered
untranslatable strings and add the needed translations.
An iso-codes upload could trigger some interest in rebuilding
localechooser and then add some more translations for country
names. choose-mirror could benefit from this as well. Christian will
first check whether he can do the NMU for iso-codes (usually well
accepted by its maintainer).
A request to hint locales (and therefore) should be sent as its
2.3.5-10 version is needed by localechooser. There is no blocker
except the presence of a udeb requiring a request from
us. (post-meeting update: this has been done by Frans)
The main blocker is obviously the kernel transition. See above.
Other issues exist:
Some work should still be done on localization-config to integrate the
stage1. Konstantinos Margaritis will be unavailable for months, thanks
to the Greek Army. He's currently seeking for someone to act as l-c
co-maintainer during that time. Anyway, even without l-c, the
localized installs work pretty well as X now successfully detects the
needed keymap. So, this issue is not a blocker.
A few UTF-8 issues currently exist. The weirdest one seems to be
corrupted display (Mojibake) with languages *not* using UTF-8 (such as
Korean) when debconf screens are shown during pkgsel runs. The
workaround is to switch all languages to UTF-8 (which is pretty much
the case now).
These UTF-8 issues have to be coordinated. Miroslav Kure volunteered
to follow all such issues and summarize them on the wiki, for instance
There seems to be some brokeness with aptitude on hppa. The issue has
to be made clearer right now. Julien Louis will do an install report
of the issue he encountered.
Both hppa and m68k are currently uninstallable because they are being
ignored for testing migrations and bugs are preventing some packages in
base from being updated which causes debootstrap to fail...
All these are actually not really blockers except the issue of the
netinst image (and, of course, the kernel transition).
Finalize the schedule
Nothing really worth mentioning. The blocker is the kernel
transition. Frans schedule is still pertinent.
Heavy testing can already begin, by using the daily builds (the
dailies point to sid_d-i, which is currently appropriate).
Official testing and filling the release-checklist file in
installer/doc/devel will begin when all udebs have been migrated to
testing and the daily links have been switched for etch_d-i.
This ends discussions directly related to beta2.
Most issues are summarized on http://wiki.debian.org/DebianInstallerCrypto.
Max Vozeler explains that the main blocker to build and upload a udeb
are the missing required udebs for gnupg and busybox.
The request for a gnupg udeb hasn't got any echo from the gnupg
maintainer up to now. The patch seems pretty much not invasive, though.
The busybox changes didn't get the agreement of busybox maintainer.
In short, partman-crypto will obviously not make it for beta2. It
would be nice to make it a target for the next release and push the
blockers a little.
Summary of Extremadura meeting
The meeting has been very productive and the D-I thanks again the
Junta de Extremadura and its representatives who did setup great
facilities for the meeting to be as successful as possible.
Davide Viti will soon post a full report of the meeting.
The VT switch bug has been fixed (it was one of the main blockers) and
most changes needed in upstream GTKFB were done by Mike Hemmel.
The G-I build system now builds a businesscard ISO image in addtion to
the netboot one.
So, now, updates of the libraries can start and Frans will begin to
file bugs to the relevant packages. libd-i and zlib seem to be the
The needed changes to dpkg-dev have been made and thus the udeb
dependency problems should soon go away.
We forgot to fix the next meeting date.
It will be on a Wednesday near the end of February, preferrably after
beta2 release so that it can be the next release cycle launch.
Wednesday February 22th 21:00UTC could be a good target.