[00:22] <udit99> an anyone tell me how to generate a preseed file from an existing ubuntu installation?...like just get a sample answers file from whatever option I chose at install time
[00:35] <cjwatson> udit99: https://help.ubuntu.com/8.04/installation-guide/i386/preseed-creating.html
[00:44] <CIA-1> partman-base: cjwatson * r99 ubuntu/ (debian/changelog lib/base.sh):
[00:44] <CIA-1> partman-base: Add support for explicit 'b' or 'B' suffixes on partition sizes to force
[00:44] <CIA-1> partman-base: interpretation as bytes.
[00:46] <CIA-1> partman-base: cjwatson * r100 ubuntu/ (test/conversions debian/changelog debian/rules lib/base.sh): Add a small test suite for longint2human and human2longint.
[01:23] <giosue_c> could someone tell me what the differences between $dist $dist-proposed $dist-security $dist-updates
[01:24] <giosue_c> I was noticing that the xubuntu hardy seeds require stuff out of both hardy and hardy-updates
[01:26] <TheMuso> giosue_c: $dist-proposed is used for testing packages for stable release updates. $dist-updates is where all tested packages are placed when a stable release update is made available. $dist-security is for security updates.
[01:26] <giosue_c> so are the seeds updated as things are promoted to dist-updates?
[01:28] <giosue_c> for example if a new kernel comes out... it will be in $dist-security or $dist-updates and will have a higher version number.  The seeds then have to reference that higher number.
[01:45] <TheMuso> giosue_c: The seeds do not reference versions of packages. The seeds only reference package names, and the infrastructure is told whether to fetch packages from -updates or -security as appropriate.
[01:51] <giosue_c> but the kernel module seeds have their versions in the package names.
[01:59] <TheMuso> giosue_c: Yes but they have metapackages to refer to them, linux-generic, linux-restricted-modules-generic, etc.
[02:04] <giosue_c> TheMuso: Ok.  I'm trying to get everything to make sense to me.  I'm looking at the installer seed in platform.hardy...  I see a line like this:
[02:04] <giosue_c>  * Kernel-Version: 2.6.24-19-generic
[02:04] <giosue_c> Then i noticed that all the modules for that kernel are in hardy-updates
[02:05] <giosue_c> sooo..  I guess the seeds are updated any time the kernel version is promoted?
[02:24] <TheMuso> giosue_c: Yes for the installer they have to be updated.
[10:43] <soren> cjwatson: So, you had some other options for implementing the server seed thing?
[10:58] <S[h]O[r]T> I'm trying to understand how I would go about loadgin a "preseed/run" script as described here https://help.ubuntu.com/7.04/installation-guide/powerpc/preseed-contents.html#preseed-network. So that I can re-load the network configuration with a static IP as it talks about
[10:59] <S[h]O[r]T> But i'm not sure what I need to exactly place into my preseed to load the script with the arguments it mentions
[11:02] <cjwatson> S[h]O[r]T: does https://help.ubuntu.com/7.04/installation-guide/powerpc/preseed-advanced.html help? you put the script in the same directory as the preseed file, and just give its file name
[11:05] <S[h]O[r]T> why yes it does :D gonna give it a go right now
[11:12]  * S[h]O[r]T thinks it worked
[11:12] <S[h]O[r]T> have to wait till the install is over to find out
[12:33] <CIA-1> casper: cjwatson * r518 casper/ (debian/changelog scripts/casper): Fix mount argument ordering for klibc mount.
[12:34] <CIA-1> casper: cjwatson * r519 casper/debian/changelog: releasing version 1.135
[12:38] <cjwatson> soren: Right. The problem here is that there's no way to preseed tasksel into selecting a given task but still displaying the task question, because it uses the seen flag a bit oddly.
[12:38] <cjwatson> soren: I think it would be a good thing to fix that anyway ...
[12:38] <cjwatson> (ubuntustudio has wanted it in the past)
[12:39] <cjwatson> the alternative I had in mind does bad things to netboot installations so on reflection probably isn't a great idea
[12:39] <cjwatson> soren: I don't think there should be something that installs ubuntu-minimal without ubuntu-standard, personally.
[12:43] <soren> cjwatson: Well, nor do I.
[12:43] <soren> cjwatson: However...
[12:43] <soren> if you read the recent thread on ubuntu-server about this stuff, there seems to be two major camps:
[12:44] <soren> The ones who want a *really* lean system and the ones who would want the server seed as well.
[12:44] <cjwatson> people who want ubuntu-minimal on its own can use debootstrap
[12:45] <soren> Also, having it only install minimal, would remove the need for a JeOS ISO altogether.
[12:45] <cjwatson> do the ones who want a really lean system really want a system without cron?
[12:45] <soren> Yes.
[12:45] <cjwatson> or iptables?
[12:45] <soren> Yes.
[12:45] <cjwatson> or ssh?
[12:45] <soren> I'm not kidding.
[12:45] <soren> Oh, yes.
[12:45] <cjwatson> then I have to ask "why?"
[12:45] <soren> ssh in particular seemed to annoy them.
[12:45] <cjwatson> we can't work on "seemed to annoy"
[12:45] <cjwatson> we need coherent, justifiable reasons
[12:46] <soren> Something along the lines of not wanting any client tools at all on their servers.
[12:46] <cjwatson> every option we add here adds overhead, translation requirements, confusion
[12:46] <cjwatson> I think it should be the server team's job to weed out the "requirements" that just don't make sense
[12:46] <soren> Yes. That's why I would like to just have one option that installs the server seed, and one option that install minimal, and nothing in between.
[12:47] <soren> It was a very tiresome debate on the mailing list. Truly.
[12:47] <cjwatson> this is the trouble with mailing lists ;-)
[12:47] <soren> I ended up concluding to myself that they don't really want a server distro, but just a platform of some sort. That's the only way I can make it make sense in my head.
[12:47] <cjwatson> design decisions should not be taken by democracy
[12:48]  * soren nods frantically
[12:49] <cjwatson> I can see the argument about removing the need for JeOS, though be careful as that means you need to ship the -virtual kernel and stuff
[12:49] <cjwatson> and usplash
[12:49] <cjwatson> well, minimal install: as you wish
[12:50] <soren> Yes, I'll be putting the -virtual kernel on there.
[12:50] <soren> I also plan to extend the CPU detection code in the install a bit.
[12:50] <cjwatson> ok
[12:50] <soren> e.g. to choose the virtual kernel based on CPUID..
[12:50] <cjwatson> as long as you extend the test suite too :)
[12:51] <soren> :)
[12:51] <cjwatson> so, tasksel. are you OK with having a go at that?
[12:51] <soren> Someone suggested adding the -generic kernel to the server ISO, too.
[12:51] <soren> ...since that doesn't require PAE.
[12:51] <soren> So the installer would need to be taught to make decisions based on that as well.
[12:51] <soren> I'm not sure about that, though.
[12:52] <soren> It's another 100MB or so, I imagine.
[12:52] <cjwatson> one of the problems with the -server mailing list seems to be that nobody actually checks facts
[12:52] <soren> Hahah :)
[12:52] <cjwatson> cjwatson@antimony:~$ grep linux-generic cdimage/www/full/ubuntu-server/daily/current/intrepid-server-i386.list
[12:52] <cjwatson> /pool/restricted/l/linux-meta/linux-generic_2.6.26.3.4_i386.deb
[12:52] <cjwatson> it's been there forever
[12:52] <soren> Oh, really? Well, then it's just the installer that needs a bit of love.
[12:52] <cjwatson> now, the default server preseeding rather stomps all over that
[12:52] <soren> Yeah.
[12:53] <cjwatson> and the reason for that is to make netboot sensible
[12:53] <cjwatson> netboot should prefer generic to server, I think (no information about intended profile) while the CD needs to prefer server to generic where possible
[12:53] <soren> Right. It's the "where possible" that needs a bit of readjustment. :)
[12:55] <soren> It /is/ rather counterintuitive that you can install the system just fine (since that uses the generic kernel), but you can't boot it when you're done, if you have a non-pae capable system.
[12:56] <cjwatson> yes, it's a bug
[12:57] <cjwatson> that was a rather cheesy base-installer hack. Fabio made me do it
[12:57] <cjwatson> ;-)
[12:57] <soren> Oh, it's on purpose?
[12:57] <cjwatson> err, in the sense that I knew I was introducing the bug, yes
[12:57] <soren> Ah :)
[12:57] <cjwatson> not in the sense that I think it's a good idea
[12:58] <cjwatson> it's done by simply overriding base-installer's idea of the right kernel to install and saying "always install -server"
[12:58] <soren> Well, if a big Italian dude was standing next to me telling me to do things... y'know..
[12:58] <soren> :)
[12:58] <soren> Right.
[12:59] <soren> Hmm... Why /is/ the generic kernel .deb on the server cd?
[13:00] <soren> Is there any way (without dropping to a shell) to actually install it?
[13:00] <cjwatson> why as in mechanism, or why as in reason?
[13:00] <cjwatson> the mechanism is that it's in the boot seed
[13:00] <cjwatson> reason: surely server people are not typically scared of shells
[13:01] <soren> Heh :)
[13:01] <soren> That's true.
[13:02] <soren> Ok, so to get back to the server seed..
[13:03] <soren> What can we do instead of installing the a new server task? Install an ubuntu-server meta package through pkgsel?
[13:05] <cjwatson> we could fix tasksel
[13:05] <cjwatson> honestly, I think that's the best approach
[13:05] <cjwatson> pkgsel/include is supposed to be reserved for the user and I don't think we should use it in CD preseeding
[13:06] <soren> Ok.
[13:07] <soren> So the bug is... what? If the question hasn't been seen, it clears the list of selected packages, and if it has been seen, the user doesn't get to choose additional tasks?
[13:07] <cjwatson> the bug is that it does not distinguish preseeded-and-marked-seen from preseeded-and-not-marked-seen
[13:08] <cjwatson> if tasksel/first is marked seen, then it should just do what it says without asking the question (it does this now)
[13:08] <cjwatson> if tasksel/first is preseeded but not marked seen, then it should use that as the default but ask the question
[13:08] <soren> Right.
[13:09] <cjwatson> tasksel/tasks is potentially different because that's what's used when you re-run tasksel, and seen flag handling is not really what you want there
[13:09] <cjwatson> but for tasksel/first it should use installer-style seen flag handling
[13:09] <soren> is any of this something you'll be looking at?
[13:10] <cjwatson> not this month ...
[13:11] <soren> Ok.
[13:11] <soren> I'll try to squeeze it in myself, then.
[17:33] <CIA-1> ubiquity: cjwatson * r2694 ubiquity/debian/changelog: line length
[17:37] <CIA-1> ubiquity: cjwatson * r2695 ubiquity/debian/ (changelog ubiquity.install-any): Update file name for apt-setup 1:0.32.
[17:38] <CIA-1> ubiquity: cjwatson * r2696 ubiquity/debian/ (changelog ubiquity.install-any): Add apt-setup local generator to improve preseeding facilities.