[00:01] <jenders> hi folks
[00:01] <jenders> in the preseeding line, "d-i mirror/suite string ____"
[00:01] <jenders> is it possible to track a specific point release instead of a codename
[00:02] <jenders> it seems as if we're getting version drift in our deployments and I'm trying to track it down
[00:02] <jenders> I'm wondering if perhaps after debootstrap an update is performed
[00:02] <jenders> and since we're tracking more or less 'stable' we're getting packages that are not in the snapshot of the latest LTS
[00:02] <jenders> is this a valid theory?
[00:03] <jenders> I should note, we also have, 'd-i pkgsel/update-policy select none
[00:03] <jenders> so that we can manually vett security updates
[00:04] <jenders> I also notice this option: d-i pkgsel/upgrade select none
[00:05] <jenders> guidance much appreciated
[00:15] <cjwatson> jenders: I'm afraid the archive infrastructure doesn't make it possible to pin to a point release at the moment
[00:16] <jenders> oh no
[00:16] <cjwatson> the only way to do that is to take a mirror snapshot at the appropriate time
[00:16] <jenders> is this because of old-release.archive.... etc
[00:16] <cjwatson> no
[00:16] <cjwatson> it's because Launchpad doesn't have any database model of which packages went into the point release
[00:16] <jenders> ah
[00:16] <jenders> well crap
[00:16] <cjwatson> https://bugs.launchpad.net/launchpad/+bug/701595
[00:16] <jenders> heh
[00:16] <ubot2`> Launchpad bug 701595 in launchpad "representation of Ubuntu point releases" [Low,Triaged]
[00:17] <cjwatson> I think those people with burning needs for this sometimes work around the problem by using the released DVD image as a quasi-mirror
[00:17] <cjwatson> it's not a great answer but it can help
[00:17] <jenders> could this be faked by pointing preseeding at a full mirror
[00:17] <jenders> oh i see
[00:49] <CIA-70> ubiquity: cjwatson * r5195 trunk/ (3 files in 3 dirs):
[00:49] <CIA-70> ubiquity: Restore old fallback code in case /cdrom/casper/filesystem.size doesn't
[00:49] <CIA-70> ubiquity: exist (LP: #557388).
[01:19] <jenders> cjwatson: does the following mechanism for gating updates sound about right:
[01:19] <jenders> You need to create a full mirror of the repository at a given point in time. Aim /etc/apt/sources.list for all hosts at this mirror and then sudo apt-get update && sudo apt-get upgrade on each host. To perform updates you would need to update the full mirror while keeping a backup of previous versions in the mirror, probably with rsync without --delete. The repo metadata would probably also need to be updated to include both new and older ver
[01:19] <jenders> and then pinning could be done either at the host level or by removing a .deb from the repo?
[01:20] <cjwatson> rsync without --delete - not enough if you want clients to be able to stick at older point releases, you'd need to create new dists/ trees
[01:21] <cjwatson> ah, maybe that's part of what got cut off by the IRC line length limit
[01:21] <jenders> http://pastebin.com/5Tq6L7Ns
[01:22] <jenders> without creating new dists, I could pin at the host level, yea?
[01:22] <jenders> (or remove the package from the repo and regen the metadata)
[01:22] <cjwatson> I guess ... it's going to be a chunk of work, consider *why* you need to pin at point releases
[01:22] <jenders> because my coworkers don't trust security updates not to break existing infrastructure
[01:23] <cjwatson> then they shouldn't trust point releases either
[01:23] <jenders> and want to vett each and every package -_-
[01:23] <cjwatson> the only significance of the point releases is really that we spin new images
[01:23] <cjwatson> there isn't significant extra QA for them over and above what's already done for any stable update
[01:23] <jenders> I guess point release is misnomer, it's really just trusting the upstream repo at any given time
[01:24] <cjwatson> I can't really advise you on pinning, it's not one of my strong points
[01:24] <jenders> and being able to gain control over updates
[01:24] <cjwatson> for some reason I've never been able to fit apt pinning into my head
[01:25] <jenders> you and me both, i use Priority: 1001 to pin packages
[01:25] <cjwatson> sounds like you sort of want to build something a bit like snapshot.debian.org where you get a snapshot dist for every mirror pulse
[01:25] <cjwatson> good luck ;-)
[01:25] <jenders> ha
[01:26] <jenders> my coworkers are pushing for centos, although I'm unsure if it offers this feature set
[01:26] <jenders> I'd rather ust stick with Ubuntu or Debian proper
[01:27] <jenders> but then we still have to wrangle partman-auto/expert_recipe's ;)
[01:29] <cjwatson> it's clearly (to me) possible with the Debian-format archive structure, but requires a reasonably smart mirror
[01:29] <cjwatson> there is something called "debmarshal" that purports to offer this
[01:29] <cjwatson> http://code.google.com/p/debmarshal/wiki/DebmarshalRepositoryMirrorSetup
[01:29] <cjwatson> sounds pretty close to what you're asking for
[01:30] <jenders> interesting
[01:30] <jenders> I'll investigate
[01:30] <cjwatson> I've not used it, that's just from a bit of searching
[01:54] <CIA-70> ubiquity: cjwatson * r5196 trunk/gui/gtk/stepKeyboardConf.ui: Sync up deduce_layout button description in glade file with debconf template.
[02:47] <CIA-70> ubiquity: cjwatson * r5197 trunk/ (7 files in 4 dirs): Fix some more uses of deprecated python-apt APIs.
[03:05] <CIA-70> ubiquity: cjwatson * r5198 trunk/ (bin/ubiquity-dm debian/changelog): ubiquity-dm: Try openbox after openbox-lubuntu (LP: #888107).
[03:37] <CIA-70> ubiquity: cjwatson * r5199 trunk/ (77 files in 3 dirs): Thomson SA is now called Technicolor SA (LP: #856992).
[09:18] <damo22> hi im trying to create another livecd with sshd enabled by default, im having trouble using UCK to acheive this, i get error messages like "(initramfs) unable to find a medium containing a live filesystem"
[11:22] <CIA-70> ubiquity: cjwatson * r5200 trunk/ (3 files in 3 dirs): Limit maximum length of username editing widgets to 32 (LP: #831319).
[11:27] <CIA-70> ubiquity: cjwatson * r5201 trunk/ubiquity/tz.py: whitespace
[12:22] <CIA-70> ubiquity: cjwatson * r5202 trunk/ (debian/changelog ubiquity/plugins/ubi-timezone.py):
[12:22] <CIA-70> ubiquity: UTF-8-encode the syslog message when a geoname lookup fails, since it
[12:22] <CIA-70> ubiquity: might include non-ASCII characters entered by the user (LP: #928891).
[14:06] <CIA-70> ubiquity: cjwatson * r5203 trunk/ (debian/changelog debian/rules tests/run):
[14:06] <CIA-70> ubiquity: Run test suite under xvfb-run by default when running it from the
[14:06] <CIA-70> ubiquity: command line. 'tests/run --no-xvfb' inhibits this behaviour for cases
[14:06] <CIA-70> ubiquity: where it's useful to see the UIs constructed by the test suite.
[14:25] <CIA-70> ubiquity: cjwatson * r5204 trunk/ (4 files in 4 dirs):
[14:25] <CIA-70> ubiquity: Stop manually inserting newlines in hostname and username errors. GTK+
[14:25] <CIA-70> ubiquity: 3 seems to do a reasonable job of wrapping these by itself now.
[14:25] <cjwatson> ev: ^- joy
[14:25] <cjwatson> Well, ish.  There's still a vertical alignment problem, but I think that's separate ...
[14:26] <cjwatson> oh, hmm, you were specifically trying to avoid growing and shrinking ... /me looks into this some more
[14:26] <cjwatson> need to sort out translating those strings and this is a prereq, though.
[14:28] <cjwatson> maybe a computed-once slightly larger allocation for {host,user}name_error would do it.
[14:28] <ev> yeah, making that page not jump all over the place as soon as things start appearing and disappearing has proven to be a bit tricky
[14:31] <cjwatson> there's some bug in the layout of username_error that doesn't apply to hostname_error - it's getting a massive vertical allocation as soon as it appears
[14:31] <cjwatson> but the fact that it's different for those two widgets suggests it's a bug in our layout
[14:38] <cjwatson> maybe it just needs a vertical box around it and expand="False"
[14:38] <cjwatson> wonder if I can get the test suite to expose this somehow
[16:04] <CIA-70> ubiquity: cjwatson * r5205 trunk/ubiquity/plugins/ubi-usersetup.py: Adjust how ubiquity.misc functions are looked up to make mocking easier.
[16:33] <cjwatson> ev: well, I have a failing test now at least, and FWIW reverting to the manually inserted newlines doesn't make a lot of difference - so I'm going to leave that patch in place
[16:34] <ev> hm, actually
[16:34] <ev> just a thought, but maybe it's worth us using a GtkFixed here
[16:34] <cjwatson> possibly, though I think we could do the same by setting allocations manually?
[16:34] <cjwatson> but yeah, it's a backstop I guess
[16:34] <ev> true
[16:36] <CIA-70> ubiquity: cjwatson * r5206 trunk/ (tests/test_gtkui.py ubiquity/frontend/gtk_ui.py): Move addition of network_change watch to customize_installer, so that there's one fewer thing to mock out in tests.
[16:37] <cjwatson> my current test is http://paste.ubuntu.com/844599/ (with the actual assertions deliberately removed for inspection purposes but they should be fairly obvious)
[23:54] <CIA-70> ubiquity: superm1 * r5207 ubiquity/ (bin/ubiquity-dm debian/changelog):
[23:54] <CIA-70> ubiquity: Have two separate failsafe attempts for 'fbdev' and 'vesa'. It's
[23:54] <CIA-70> ubiquity: possible that /dev/fb0 will exist but be backed by a VGA framebuffer
[23:54] <CIA-70> ubiquity: causing X to exit non-zero.