[00:04] <CIA-32> ubiquity: cjwatson * r5233 trunk/gui/gtk/ (4 files): Fix whitespace on accessibility properties
[00:28] <CIA-32> ubiquity: cjwatson * r5234 trunk/ (3 files in 3 dirs): Make the migration-assistant tree view expand again (LP: #946123).
[09:31] <tokarbol> I tried the ubuntu-installer GTK mini-iso for precise, but this fails to start Xorg fbdev.
[09:31] <tokarbol> Can you give me a hand on this?
[09:38] <tokarbol> fbdev xorg video driver seems to be in place
[09:41] <tokarbol> https://bugs.launchpad.net/ubuntu/+source/debian-installer/+bug/939278
[09:41] <ubot2`> Launchpad bug 939278 in debian-installer "GTK mini-iso can't start X" [Undecided,New]
[09:41] <tokarbol> oh, thank you, ubot2
[09:41] <tokarbol> that's the one.
[09:41] <tokarbol> :p
[09:44] <ogra_> tokarbol, i dont think anyone ever tests that installer ... we either test the pure text version or ubiquity
[09:45] <tokarbol> I thought so, I don't mind that much
[09:45] <tokarbol> that's why I wanted to investigate it further as I guess you may not have time to spend on this
[09:46] <tokarbol> but I don't know where/how to look for it
[09:46] <tokarbol> it's either Xorg driver problem (crash) or the scripts fail to display anything and the X shuts down
[09:47] <tokarbol> then it's restarted by the /lib... something script
[09:47] <ogra_> well, first of all you should check if the driver is there at all
[09:47] <tokarbol> it is
[09:47] <ogra_> (is there a udeb installed carrying it)
[09:48] <tokarbol> fbdev driver is in place, and that one should suffice
[09:48] <ogra_> all dependencies too ?
[09:48] <ogra_> do you have an xorg log by chance ?
[09:49] <tokarbol> give me a second, I can get you that
[09:49] <tokarbol> pastebin or the bug report?
[09:49] <ogra_> i think the latter (mind you, i'm busy with other stuff, but such info should indeed be on the bug)
[09:53] <tokarbol> ok, it's there
[09:54]  * cjwatson has a look
[10:01] <tokarbol> I think this is either some crash that's not much verbose
[10:02] <tokarbol> or this is a log from a correctly-started-and-stopped X
[10:02] <cjwatson> it's certainly not the latter
[10:02] <tokarbol> but it doesn't say why the X client running stopeed...
[10:02] <tokarbol> no?
[10:02] <cjwatson> EINVAL from FBIOBLANK suggests that it's being run on an incorrectly configured framebuffer
[10:03] <cjwatson> I suppose that isn't necessarily the problem even though it's the last error message in the log
[10:04] <cjwatson> mm, looking at the source, that error doesn't seem to be fatal
[10:05] <tokarbol> I tested this on vbox and an intel gpu
[10:06] <tokarbol> the framebuffer seems to be initialized correctly by the kernel
[10:07] <cjwatson> the server does start up correctly
[10:07] <cjwatson> so yeah, you're right, sorry, it is a log from a correctly-started-and-stopped X
[10:07] <tokarbol> ok - so where should I look for installer error messages?
[10:09] <tokarbol> when I scroll-lock the first console it says
[10:09] <tokarbol> Fatal... server is already active for display 0
[10:10] <tokarbol> ...ddxSigGiveUp: Closing log
[10:10] <tokarbol> Gtk-WARNING **: cannot open display: :0
[10:10] <tokarbol> is it possible that it runs on :0 but cannot access it?
[10:11] <cjwatson> starting the client up by hand seems to work
[10:11] <cjwatson> installer errors should be in /var/log/syslog
[10:12] <tokarbol> which client did you start?
[10:12] <cjwatson> uh, it involves a zillion environment variables :)
[10:12] <cjwatson> DEBIAN_FRONTEND=gtk LANG=C.UTF-8 DISPLAY=:0 MENU=/usr/bin/main-menu /lib/debian-installer/menu
[10:12] <tokarbol> not that bad
[10:14] <soren> Am I supposed to do something special for netcfg to pick up the BOOTIF parameter from the kernel command line when I'm PXE booting? I see it in /proc/cmdline, but I'm still prompted for which interface to use.
[10:14] <soren> This is on precise, btw.
[10:15] <tokarbol> sorry, my proxy connection is just useless
[10:16] <tokarbol> unfortunately, the command gave me an error as well
[10:16] <tokarbol> (tried from second console
[10:16] <tokarbol> )
[10:16] <cjwatson> soren: have a look in syslog, it's kind of picky about the format
[10:17] <cjwatson> tokarbol: I'm trying to build a test image for myself so that I can get a more accurate idea of what's going on
[10:18] <tokarbol> I tried compiling this from the source debian-installer package as well, same results
[10:18] <cjwatson> sure, that wouldn't make the slightest difference
[10:18] <tokarbol> so either something is missing from udeb pkgs
[10:19] <tokarbol> that shouldn't make, but things tend to happen during a build in a dev env
[10:19] <tokarbol> (not this time though)
[10:19] <soren> cjwatson: Will do that, thanks.
[10:23] <soren> cjwatson: Ok, looking at the syslog now... What am I looking for?
[10:24] <cjwatson> does it say "Could not find invalid BOOTIF= entry in /proc/cmdline"?
[10:24] <soren> cjwatson: Should I enable more debugging somehow, perhaps? It doesn't even say BOOTIF anywhere at all.
[10:24] <cjwatson> can I just see the syslog?
[10:24] <soren> cjwatson: Nope. "grep BOOTIF /var/log/syslog" comes up empty.
[10:24] <cjwatson> also can I see your /proc/cmdline?
[10:25] <soren> cjwatson: Let me see if I can extract them somehow.
[10:26] <soren> cjwatson: Before I do this.. Is there more logging I can enable so we have as much to work with as possible?
[10:26] <cjwatson> no
[10:27] <cjwatson> well, in general with installer debugging it's a good idea to add the boot parameter DEBCONF_DEBUG=developer, but I doubt it'll help here
[10:27] <soren> Alright.
[10:29]  * cjwatson throws strace into this GTK image in order to try to get some actual *information*
[10:34] <soren> cjwatson: http://paste.ubuntu.com/871261/ and http://paste.ubuntu.com/871262/
[10:35] <soren> cjwatson: Oh, and the preseed file:
[10:36] <soren> cjwatson: http://paste.ubuntu.com/871263/
[10:38] <cjwatson> Well, that thinks it parsed BOOTIF successfully
[10:38] <cjwatson> Mar  6 10:28:03 netcfg[2449]: INFO: Found interface eth0 with link-layer address 00:25:b5:0b:0c:be
[10:39] <soren> cjwatson: Ok..
[10:41] <cjwatson> oh, wait, try netcfg/choose_interface=auto
[10:41] <cjwatson> that'll make it actually bypass that question based on BOOTIF
[10:42] <cjwatson> by default it only changes the default answer but still asks the question
[10:42] <soren> Ah, yeah, just spotted that, too.
[10:42] <soren> Didn't make it that far down netcfg_get_interface before. :)
[10:43] <soren> I kept assuming it didn't find the right interface of whatever.
[10:46]  * soren never really understood the proportional relationship between the price of a server and the time it takes to boot
[10:49] <soren> cjwatson: Awesome. Seems to have worked like a charm.
[10:50] <soren> cjwatson: Thanks!
[10:58] <cjwatson> tokarbol: hm, I don't think it's even starting the menu
[10:59] <tokarbol> ok...
[11:00] <tokarbol> I manually started Xorg from the second console and I got the menu
[11:00] <soren> cjwatson: Oh, one more thing... With that preseed ( http://paste.ubuntu.com/871263/ ) wouldn't I be getting a statically configured eth0?
[11:01] <cjwatson> how do you mean
[11:01] <cjwatson> ?
[11:02] <tokarbol> who?
[11:02] <soren> /etc/network/interfaces in the installed system is set to configure eth0 using dhcp.
[11:02] <soren> tokarbol: Oh, excellent question :)
[11:02] <cjwatson> I meant soren
[11:02] <soren> Oh, good.
[11:02] <soren> Then I stand by my answer.
[11:03] <cjwatson> soren: mm, possibly, there isn't much provision for making that dynamic save for using either network-manager or biosdevname
[11:03] <cjwatson> you can try biosdevname=1 if it supports your hardware
[11:03] <soren> Sorry, "making that dynamic"?
[11:03] <soren> Let me start over :)
[11:03] <soren> So, BOOTIF chooses the interface.
[11:04] <soren> I want to statically configure that interface.
[11:04] <soren> I thought that preseed file would do that.
[11:05] <soren> ...but for some reason it's choosing dhcp.
[11:05] <soren> ..and I can't work out why.
[11:06] <cjwatson> oh sorry
[11:07] <cjwatson> the preseed file isn't read until after network configuration
[11:07] <soren> Oh, of course.
[11:07] <cjwatson> anything you need to apply before the preseed file is read needs to go on the kernel command line
[11:07] <soren> *headdesk*
[11:08] <soren> I forget the protocol for loading kernels.. What's the limit on kernel command line length?  Will all that stuff fit?
[11:08] <cjwatson> I think you can drop some of it, one moment
[11:09] <tokarbol> weird, when I run Xorg from the 2nd console, the installer shows the Gtk language selection menu, but my keyboard&mouse are unresponsive, can't even change vt-s
[11:09] <cjwatson> soren: I think you can drop netcfg/dhcp_failed and netcfg/dhcp_options
[11:10] <cjwatson> the rest should fit I think
[11:10] <cjwatson> tokarbol: strace indicates that the shell that's waiting for the X server to finish starting up isn't responding to the SIGUSR1 it gets sent
[11:11] <cjwatson> well, unless strace is lying to me :)
[11:12] <StevenK> strace lying is an interesting concept.
[11:13] <cjwatson> mind you strace is known to affect signal delivery in some cases
[11:14] <cjwatson> so maybe that's bogus
[11:22] <cjwatson> I wonder if this is kickseed's fault, after a fashion
[11:24] <cjwatson> yes!
[11:24] <tokarbol> what?!
[11:24] <cjwatson> rootskel-gtk recently started passing -terminate to the X server, for other reasons
[11:24] <cjwatson> this means that it exits after the first client exits
[11:25] <cjwatson> but kickseed is a client that runs before the main menu does ...
[11:25] <tokarbol> oh
[11:26] <cjwatson> hmm, tricky
[11:26] <tokarbol> ok, I'd never find that
[11:28] <tokarbol> actually, I still can't find it - where is it started from?
[11:30] <cjwatson> /lib/debian-installer.d/S65kickseed
[11:34] <tokarbol> I was looking at the file before... it starts ..../initrd-kickseed which doesn't seem to invoke Xorg, but...
[11:34] <cjwatson> it does via debconf
[11:34] <tokarbol> I'm not that good in this, it seems :)
[11:34] <tokarbol> some db_ command runs Xorg?!?
[11:35] <cjwatson> in order to run debconf you need a debconf frontend; the GTK image is (necessarily) configured for the debconf frontend to be an X client
[11:35] <soren> cjwatson: It seems the kernel command line these days can be up to 2k.
[11:35] <cjwatson> nice
[11:35] <soren> cjwatson: Just FYI.
[11:35] <tokarbol> so a chicken-and-egg problem
[11:36] <cjwatson> nope
[11:37] <cjwatson> X has already been started, it's not chicken-and-egg, but a proper fix would involve moving kickseed to be started from within /lib/debian-installer/menu
[11:37] <cjwatson> however that will take a while so I'm just going to disable kickseed when using the GTK frontend, for now
[11:37] <cjwatson> and leave a TODO comment
[11:40] <CIA-32> kickseed: cjwatson * r286 ubuntu/ (debian-installer.d/S65kickseed debian/changelog): merge from Debian 0.57
[11:40] <tokarbol> it'll do for me, I'm using regular preseed anyway.
[11:41] <CIA-32> kickseed: cjwatson * r287 ubuntu/debian/changelog: releasing version 0.57ubuntu1
[11:41] <cjwatson> tokarbol: ok, will be fixed the next time debian-installer is rebuilt (which happens frequently), thanks
[11:41] <tokarbol> Thank YOU!
[11:43] <cjwatson> that was tricky after only 1.5 cups of coffee :-P
[14:26] <cyphermox> stgraber: back at the office looking at the /e/n/i file that d-i creates; it indeed looks like ifblacklist_migrate.sh just isn't intelligent enough to deal with wifi
[14:29] <stgraber> cyphermox: did you figure which part is failing?
[14:29] <stgraber> quickly looking through the script it's not obvious what part wouldn't match wireless
[14:29] <cyphermox> cjwatson: fyi, when installing with alternate, d-i appears to create /etc/network/interfaces and if you configure wifi; that configuration is left on the installed system and causes NetworkManager to ignore the interface. I'll file the bug now, and I'm looking into fixing ifblacklist_migrate.sh (which is shipped by n-m) to take that into account
[14:29] <cyphermox> stgraber: yeah, found it
[14:29] <cjwatson> OK, thanks
[14:29] <cyphermox> the script only expects no extra lines under iface <iface> inet dhcp, really.
[14:30] <cyphermox> in the case of wireless you'll have at the very least wpa-ssid; then perhaps wpa-psk
[14:30] <cyphermox> (or some wep values too... perhaps I'll need to look at netcfg to figure out exactly which ones)
[14:30] <stgraber> do we have cases where we have "dhcp" and something useful added by netcfg we actually want to kepp?
[14:30] <stgraber> *keep
[14:31] <cyphermox> I don't know
[14:31] <cyphermox> I'd be tempted to say that on a "desktop" install, you'd almost never want to keep the config
[14:31] <stgraber> my guess is that we only really care about static and so just checking for dhcp should be enough to know we should drop the interface from /e/n/i
[14:31] <cyphermox> but of course I have a vested interest in NetworkManager ;)
[14:31] <cyphermox> possibly
[14:32] <stgraber> ideally we'd have a interfaces-to-nm script that'd take care of generating a profile but that can be 12.10 work
[14:32] <cyphermox> stgraber: however in that case we still need to fix the grep/sed magic to remove all the lines under that iface line; up until the next iface
[14:32] <stgraber> for 12.04 I think it's safe to say that any automatically configured interface (so in our case, .*dhcp) gets dropped
[14:32] <cyphermox> guess so
[14:33] <cyphermox> I'm fully behind that, any reason why that might not be a good idea?
[14:33] <stgraber> from what I know of netcfg I think it's safe
[14:33] <cyphermox> I mean... other cases are already taken care of by the preseed yes
[14:34] <stgraber> right and we have the netcfg/network-manager preseed variable to turn off the script anyway
[14:34] <cyphermox> that's what I meant :)
[17:33] <jibel> bug 947738 seems new and reported twice today
[17:33] <ubot2`> Launchpad bug 947738 in ubiquity "precise failed to install: unable to initialize policy plugin" [High,Confirmed] https://launchpad.net/bugs/947738
[17:56] <stgraber> unable to open /etc/sudoers??? that seems weird
[17:57] <stgraber> could be that sudo isn't setuid for some reason
[17:57] <cjwatson> sudo hasn't changed recently ...
[17:58] <stgraber> yeah but neither did that part of ubiquity...
[17:58]  * stgraber grabs today's daily
[17:59] <cjwatson> doing a test run now
[17:59] <cjwatson> overlayfs bug maybe?  but just guessing
[17:59] <cjwatson> actually overlayfs shouldn't be involved, should it
[18:00] <cjwatson> looks setuid at least after the initial copy
[18:01]  * cjwatson leaves all this running and goes for dinner
[18:35] <cyphermox> cjwatson: (I'm leaving for lunch now) when you have time, I'd like your opinion on https://bugs.launchpad.net/ubuntu/+source/network-manager/+bug/948217; I'm trying to fix these reported cases of NetworkManager not managing connections after an alternate install; seems like ifblacklist_migrate.sh was missing something
[18:35] <ubot2`> Launchpad bug 948217 in network-manager "wireless devices are not managed when installing from alternate" [High,In progress]
[19:07] <CIA-32> ubiquity: stgraber * r5235 ubiquity/ (debian/changelog ubiquity/frontend/gtk_ui.py): Fix bug 830946 by setting the Vte font to Ubuntu Mono 8
[19:07] <ubot2`> Launchpad bug 830946 in ubiquity "Nothing displayed on embedded terminal." [Medium,Triaged] https://launchpad.net/bugs/830946
[19:41] <CIA-32> ubiquity: stgraber * r5236 ubiquity/debian/ (76 files in 2 dirs): Change ubiquity/text/webcam_heading based on suggestion from bug 946123. Run debconf-updatepo to refresh all the .po files.
[19:41] <ubot2`> Launchpad bug 946123 in ubiquity "Text Box space inadequate in 'Import Accounts' page." [Medium,Fix committed] https://launchpad.net/bugs/946123
[19:42] <stgraber> oops, wrong bug number ... should have been bug 926493
[19:42] <ubot2`> Launchpad bug 926493 in ubiquity "[UIFe] The installer still says your picture will be used on the login screen" [Low,Fix committed] https://launchpad.net/bugs/926493
[19:43] <stgraber> at least debian/changelog is right (commit message was caused by copy/paste from the wrong paragraph of the changelog)
[22:22] <stgraber> ouch, it took 15min for dmraid to scan all my SAN volumes before starting partman ;)
[22:24] <stgraber> at least fiber-channel and multipath seem to still work in d-i