[01:37] <CIA-2> os-prober: TheMuso * r223 ubuntu/debian/changelog: releasing version 1.28ubuntu1
[03:06] <KarlStrings> Hi.  I have had some serious issues with setting up preseed and the PXE installer.  I had a lot of issues with the documentation...  I would like to fix it.
[13:24] <CIA-2> debian-installer: cjwatson * r989 ubuntu/ (build/config/armel.cfg debian/changelog):
[13:24] <CIA-2> debian-installer: Move armel to 2.6.27-9, since it's currently being built as a mainline
[13:24] <CIA-2> debian-installer: architecture.
[17:02] <superm1> evand, i'm assuming you are doing an ubiquity upload prior to alpha 1, correct?
[17:04] <evand> superm1: ja
[17:04] <superm1> evand, okay just making sure :)
[17:11] <CIA-2> oem-config: cjwatson * r564 oem-config/ (d-i/manifest debian/changelog debian/control):
[17:11] <CIA-2> oem-config: Automatic update of included source packages: console-setup 1.28ubuntu1,
[17:11] <CIA-2> oem-config: localechooser 2.08ubuntu1, tzsetup 1:0.24.
[17:12] <CIA-2> oem-config: cjwatson * r565 oem-config/debian/changelog: releasing version 1.54.0
[17:15] <evand> 1.54.0?  Should I make the next bump 1.54.1, or was that a typo?
[17:15] <evand> cjwatson: ^
[17:15] <cjwatson> deliberate, see changelog, don't need to bump configure.ac every time any more
[17:16] <cjwatson> we'll have the 1.54.* series cover jaunty, I think
[17:16] <evand> ah, awesome
[17:16] <cjwatson> I'd have used a more sensible number but didn't feel 2.0 was appropriate and so was a bit stuck
[17:16] <cjwatson> I did the same configure.ac thing for ubiquity, where the numbering is more sensible
[17:18] <evand> I've been trying to do the same for usb-creator, but the intersection of cdbs and Python's distutils seems to be filled with broken glass and piles of salt.
[17:18] <evand> I do have it at least autogenerating off of debian/changelog for --version though.
[17:19] <cjwatson> distutils has never really done it for me
[17:19] <cjwatson> I at least understand the autotools reasonably well, although I appreciate many people find them hairy
[17:19] <evand> ah, indeed
[17:20] <cjwatson> at some point I must make ubiquity's build system a bit more consistent
[17:20] <cjwatson> the half-autotools half-dh_install thing is not really all that great
[17:23] <beowabbit> I need to specify UIDs for a bunch of accounts created during install (ntp, snmp, openldap, etc.) so that file ownership and backup are portable across machines.
[17:25] <beowabbit> So basically, I want to provide a relatively fully populated passwd file starting early in the install.  (This is an unattended install with a preseed, started with pxelinux.)
[17:25] <cjwatson> beowabbit: I have to go out right now, but if you stick around for a bit I can help you out when I get back if nobody beats me to it
[17:25] <beowabbit> Is there a good way to do that?
[17:26] <beowabbit> cjwatson: Awesome, thanks.  I'll hang out all day.
[17:27] <CIA-2> oem-config: evand * r566 oem-config/ (3 files in 3 dirs):
[17:27] <CIA-2> oem-config: Use an iconview as the language selection widget. Only show scrollbars
[17:27] <CIA-2> oem-config: for the language selection widget if the window is going to be larger
[17:27] <CIA-2> oem-config: than the desktop.
[17:28] <cjwatson> woo, thanks
[17:29] <beowabbit> (I tried creating my own custom base-passwd package and putting it in a custom repository, available to d-i, but it seemed to be ignored.)
[17:30] <evand> I'm glad to have finally fixed that
[20:25] <cjwatson> beowabbit: I think the easiest way (not trivial, but not too hard either) would be to set preseed/early_command to a script that writes out a script in /usr/lib/post-base-installer.d/ and makes it executable; that inner script would do 'chroot /target adduser ...'
[20:28] <cjwatson> beowabbit: debootstrap is the thing that installs the base system (including base-passwd), and it only knows how to look at a single repository, which is why your custom package was ignored. That said you could install it from such a post-base-installer script if you wanted to do it that way; however base-passwd is still an anomalous package that prompts on a tty for confirmation of added users, so that probably won't ...
[20:28] <cjwatson> ... work as well for you as you might like
[21:13] <CIA-2> oem-config: cjwatson * r567 oem-config/debian/changelog: note that Evan's iconview change is for the GTK frontend
[21:25] <beowabbit> cjwatson: Thanks!  That's exactly what I was looking for.
[21:27] <beowabbit> cjwatson: I didn't know about the existence of /usr/lib/post-base-installer.d .
[21:27] <beowabbit> cjwatson: Do you know if "in-target" works yet at that point?
[21:33] <cjwatson> yes
[21:33] <cjwatson> (it does)
[21:33] <beowabbit> cjwatson: I guess I could have the early_command grab the SSH host keys and drop a script into /usr/lib/post-base-installer.d to copy them into /target too, right?  (Just making sure I understand when in the installation all of these things get run: early_command before any changes are made, and post-base-installer.d scripts after drives have been partitioned and a very skeletal OS is installed.)
[21:34] <cjwatson> early_command runs just after "retrieving installer components"
[21:34] <cjwatson> post-base-installer.d: correct
[21:34] <beowabbit> Great, thanks!
[21:34] <cjwatson> yes, you could also have early_command wget the script from somewhere if that's more convenient than writing it all directly into the early_command
[21:34] <cjwatson> and wget whatever pieces it needs
[21:35] <beowabbit> Yeah, that's what I'd do.  (That's what I do in late_command.)
[21:55] <TheMuso> cjwatson: Which disks will get the GTK d-i?
[22:00] <cjwatson> dunno yet
[22:00] <cjwatson> probably all with d-i to start with and we'll see; it's optional anyway
[22:01] <cjwatson> it'll probably end up depending on space
[22:03] <TheMuso> cjwatson: Right, I know we may consider it for studio, since users have complained about no GUI installer.
[22:04] <TheMuso> And space is not an issue.
[22:04] <cjwatson> it's not going to be like ubiquity, of course
[22:04] <TheMuso> Oh I understand that.
[22:04] <cjwatson> it's a generated UI depending on what the debconf question structure looks like, not a fully designed UI
[22:04] <TheMuso> Yep I know.
[22:04] <cjwatson> ok, cool
[22:04] <TheMuso> I've also thought that depending on what alternate disks it can be added to, that longer term, it may be a better bet for an accessible installer.
[22:05] <TheMuso> But that adds a lot of extra baggage.
[23:04] <beowabbit> cjwatson: Thanks again!
[23:04]  * beowabbit takes off