/srv/irclogs.ubuntu.com/2012/04/18/#ubuntu-installer.txt

infinitymfisch: Should be fixed in 2.0.1302:01
mfischinfinity: thanks04:27
cjwatsoninfinity: hm, yeah.  now, if only I could easily test a fix ...09:55
infinitycjwatson: If I had remembered this cycle to fix the &^%@!$ ISOs to boot on my powerstation, I could test.09:55
cjwatsonI wonder why Debian removed kernel-img.conf handling in l-b09:56
infinitycjwatson: I guess it's not too late to do that.09:56
infinitycjwatson: But we switched from mkisofs to something else, right?  So porting Debian's command-line becomes non-trivial.09:56
cjwatsonxorriso's only on x8609:57
cjwatsonwe still use genisoimage on powerpc09:57
infinitycjwatson: Oh, kay.  So just yanking Debian's working command-line args for PPC would work.09:57
cjwatsonshould do09:58
infinitycjwatson: That's my only data point.  "Debian boots, Ubuntu doesn't".09:58
infinityWildly unhelpful.09:58
FourDollarscjwatson: Is there any place to invoke clean_method() in partman-auto/lib/recipes.sh to execute `mv $id/method $id/method-old`?09:58
infinityLike, Ubuntu's so "broken" to this firmware that it doesn't even acknowlege there's a bootable CD in the tray. :P09:58
FourDollarscjwatson: I have verified https://bugs.launchpad.net/ubuntu/+source/partman-efi/+bug/972122 with Ubuntu precise daily and it works.09:58
ubot2Launchpad bug 972122 in partman-efi "Ubiquity pops out a warning message from partman-partitioning on UEFI BIOS when there is a EFI system partition existed." [Medium,Fix released]09:58
FourDollarscjwatson: Somehow it doesn't work on my own customized Ubuntu version.09:59
FourDollarscjwatson: I saw there is a method-old and it has 'efi' inside.09:59
CIA-32debian-installer: cjwatson * r1678 ubuntu/ (5 files in 2 dirs): Update help text translations from Launchpad.10:00
cjwatsonFourDollars: let's not start from a proposed solution10:01
cjwatsonFourDollars: try comparing the output of 'archdetect' in the daily vs. your customised version10:01
cjwatson(I have a private bet on the outcome; we'll see how it compares with reality)10:02
infinitycjwatson: I might poke you later about PPC image generation.  I need sleep.10:02
FourDollarscjwatson: The output of archdetect is "amd64/efi'.10:02
cjwatsonFourDollars: in which image?10:02
FourDollarscjwatson: Both.10:02
* cjwatson loses his bet10:02
cjwatsonFourDollars: tell me more about "it doesn't work"10:03
cjwatsonFourDollars: and please show me any preseeding you're using10:03
FourDollarscjwatson: I saw there is a method-old in EFI system partition folder.10:03
cjwatsonforget about that for now, please10:04
FourDollarscjwatson: It should be a 'method' file normally.10:04
cjwatsonforget about that for now, please10:04
cjwatsonyou're starting from midway down the debugging trail, which is not helpful to me10:04
cjwatsonit's a useful data point, yes, but I need more10:04
FourDollarscjwatson: OK.10:04
cjwatsonAre you doing any preseeding?  Please show me.10:05
FourDollarscjwatson: https://pastebin.canonical.com/64456/10:07
cjwatsonYour preseeding is wrong for EFI; it needs to opt to reuse the existing EFI System Partition.10:08
cjwatson100 150 200 free10:08
cjwatson        $iflabel{ gpt }10:08
cjwatson        $reusemethod{ }10:08
cjwatson        method{ efi }10:08
cjwatson        format{ } .10:08
cjwatsonadd that to partman-auto/expert_recipe, preferably before the entries for the other two partitions (and with appropriate \ etc.)10:08
FourDollarscjwatson: And I use https://pastebin.canonical.com/64457/ to overwrite some settings. (Sorry for the late.)10:09
cjwatsonHmm10:10
cjwatsonwell, that ought to work then, unless there's some weird interaction between $reusemethod and biggest_free, or something10:11
FourDollarscjwatson: https://pastebin.canonical.com/64458/ is /var/log/partman10:12
FourDollarscjwatson: https://pastebin.canonical.com/64459/ is /var/log/syslog10:13
FourDollarsThat is my question in the beginning.10:14
FourDollarsIs there any place to invoke clean_method() in partman-auto/lib/recipes.sh to execute `mv $id/method $id/method-old`?10:14
cjwatsonSTOP10:14
cjwatsonI want to debug this myself10:14
cjwatsonplease don't keep taking me down this10:14
FourDollarsOK. Sorry.10:14
cjwatsoncan you insert 'set -x' at the top of /lib/partman/lib/recipes.sh before starting the installer, and then get me the resulting /var/log/syslog, please?10:14
FourDollarscjwatson: Maybe I can use 'break=init' kernel parameter to modify /root//lib/partman/lib/recipes.sh .10:16
cjwatsonJust run it interactively.10:16
cjwatsonmethod-old represents what was autodetected before autopartitioning.  Your results tell us that a partition was autodetected as efi, but no amount of fiddling with clean_method will tell us anything about why the autopartitioning recipe isn't reusing that partition.10:16
cjwatsonSo that's a distraction.10:16
cjwatsonclean_method is called unconditionally at the top of /bin/autopartition.  This is correct.10:17
cjwatsonIt's also not the problem.10:17
jibeloem mode is brkoen today: bug 98473610:17
ubot2Launchpad bug 984736 in ubiquity "oem-config not installed after initial installation in OEM mode" [Undecided,New] https://launchpad.net/bugs/98473610:17
jibel*broken10:17
cjwatsonjibel: Do you know when it last worked?10:18
jibelcjwatson, yesterday10:18
cjwatsonIt doesn't even seem to have tried to install oem-config-gtk, although I see the log for apt-install running.10:18
cjwatsonNo indication of why.10:19
jibelwant a run in debug mode ?10:19
cjwatsonWouldn't hurt10:19
jibelack10:19
cjwatsonNot sure whether it will tell me anything10:19
FourDollarscjwatson: The output of `sh -x /lib/partman/lib/recipes.sh` is '+ unnamed=0'10:21
cjwatsonThat's no use at all.10:25
cjwatsonYou didn't do what I asked10:25
cjwatsonI don't want the output of running that with 'sh -x'.  That file is sourced by lots of bits of the installer.  You can't run it outside the installer.10:25
cjwatsonWhen I said "Just run it interactively", I meant boot the CD interactively and run the installer by typing 'ubiquity'.10:25
FourDollarsSorry.10:27
jibeldebug log attached.10:37
cjwatsonjibel: could I have the syslog from the debug run too?10:40
jibelcjwatson, from alternate i386 http://paste.ubuntu.com/935260/ all these kde packages looks weird10:40
jibelsure, I'll attach it10:40
cjwatsonNot sure that signifies much.10:40
cjwatsonThough I'll see if I can track down why.10:41
cjwatsonIt shouldn't have stopped it from even trying.10:42
jibelsyslog attached10:44
stgrabercjwatson: I'm happy to have a look at that one if you're busy with something else10:45
cjwatsonWhich one?10:45
stgrabercjwatson: oem10:45
stgrabercjwatson: bug 98473610:46
ubot2Launchpad bug 984736 in ubiquity "oem-config not installed after initial installation in OEM mode" [Undecided,New] https://launchpad.net/bugs/98473610:46
cjwatsonIf you could, yes please10:47
stgrabercjwatson: hmm, actually I think I figured it out already, it's bad timing between livefs and cd build10:48
stgrabercjwatson: /pool contains oem-config 2.10.13 but the livefs contains ubiquity 2.10.1210:49
cjwatsonThat would indeed make sense10:49
stgrabercjwatson: can you maybe start a rebuild to confirm?10:49
stgraberI guess we want to test the new ubiquity anyway, so a rebuild would give us that10:49
cjwatsonsure, running10:49
cjwatsonoem-config-gtk pulling in a pile of KDE bits is odd; 'apt-get install oem-config-gtk' in a minimal chroot reproduces it, but 'apt-get install oem-config-gtk ubiquity-frontend-gtk' doesn't10:50
cjwatsonI blame apt, but maybe there's some way to work around it10:50
cjwatsonHmm.  The right workaround is probably to restructure the packaging, which I'm not doing now.  Maybe tweak the apt-get command we're using, then.10:51
stgraberyeah, I noticed that a few time and I think that was also the source of the problem ogra_, infinity had with the ac100 image pulling kde10:52
stgraberI usually test ubiquity locally by installing and running it in a container, having to install with "ubiquity ubiquity-frontend-gtk" to avoid pulling half of kde with it10:53
* ogra_ doesnt actually know what the source of that prob was ... its one of these things i discovered and before i could look infinity said "or that was my fault, fixed already!"10:54
ogra_(which it indeed was)10:54
CIA-32ubiquity: cjwatson * r5383 trunk/ (debian/changelog finish-install.d/01oem-config-udeb):10:56
CIA-32ubiquity: Install ubiquity-frontend-$frontend along with oem-config-$frontend to10:56
CIA-32ubiquity: improve dependency resolution; with the packaging the way it is at the10:56
CIA-32ubiquity: moment, apt may sometimes select ubiquity-frontend-kde to resolve10:56
CIA-32ubiquity: ubiquity's dependency on a frontend before realising that oem-config-gtk10:56
CIA-32ubiquity: already depends on ubiquity-frontend-gtk.10:56
cjwatsonstgraber: well, for ubiquity, you can just install ubiquity-frontend-gtk directly10:56
cjwatsonnow, why on earth didn't my panel workaround wowrk10:57
cjwatson*work10:57
cjwatsonoh, because ubiquity 2.10.13 isn't actually on the images yet11:03
cjwatsonheh, yeah, jibel noticed the same I see11:04
cjwatsonFourDollars: any luck with that?11:05
FourDollarscjwatson: No. :(11:05
cjwatsonFourDollars: setup problems?11:07
FourDollarscjwatson: I am still thinking of how to insert `set -x` to collect the information for you.11:08
cjwatsonFourDollars: Boot the CD into a live session.  Edit the file with a normal editor, as root.  Save and exit the editor.  Run 'ubiquity'.11:13
FourDollarscjwatson: With my custmized CD, there is no live session mode. :(11:14
cjwatsonOh11:14
cjwatsonWell, then, break=bottom, then:11:14
cjwatsonsed -i '1s/^/set -x; /' /root/lib/partman/lib/base.sh11:15
FourDollarsmaybe executing `sed -i '2 a\set -x\n' /root/lib/partman/lib/recipes.sh` works.11:15
FourDollarscjwatson: haha11:16
cjwatsonIt'll work on the first line, since that file has no #! line11:16
FourDollarsCool~ :D11:16
cjwatsonwell, thinking about it, you probably want it after . /usr/share/debconf/confmodule11:16
cjwatsonso  sed -i '3s/^$/set -x/' /root/lib/partman/lib/base.sh11:17
cjwatsonor your version with the right line breaks inserted, but a and i do tend to be a bit of a pain to get right11:17
jibelthe slideshow has been removed from end user config in oem mode or is it a bug ?11:19
cjwatsonCould be the same problem as earlier; try with today's refreshed images once they finish building11:19
cjwatsonIt's not been intentionally removed as far as I know11:20
FourDollarscjwatson: Here it is. https://pastebin.canonical.com/64462/ :)11:22
FourDollarscjwatson: /var/log/syslog11:22
FourDollarsrsyslogd drops messages. :(11:25
cjwatsongrr11:27
cjwatsonin that case:11:27
cjwatsonsed -i '3s,^$,set -x; exec 2>>/tmp/partman.shx,' /root/lib/partman/lib/base.sh11:28
cjwatsonwhich should send it all to /tmp/partman.shx11:28
FourDollarsThx11:29
cjwatsonsigh, livefs build failure due to langpacks11:29
* cjwatson spots an entirely unrelated display bug in FourDollars' logs11:31
cjwatsonApr 18 09:33:02 ubuntu ubiquity: /lib/partman/update.d/58default_visuals: 33: printf: Illegal option -æ11:31
FourDollarsWhich log?11:32
* FourDollars seems to see this message before.11:33
cjwatsonhttps://pastebin.canonical.com/64459/11:33
FourDollarscjwatson: ubiquity keeps busy for a long long time.11:33
FourDollars:(11:33
CIA-32partman-base: cjwatson * r1364 ubuntu/ (debian/changelog update.d/default_visuals visual.d/type):11:34
CIA-32partman-base: Be more careful about printing translated visuals, in case a translation11:34
CIA-32partman-base: starts with "-" (which the zh_TW translation of "FREE SPACE" does).11:34
cjwatsonFourDollars: Doing anything at all?11:34
FourDollarscjwatson: yup11:35
FourDollarsIt seems too many log.11:35
FourDollars/tmp/partman.shx grows to 40M .11:35
CIA-32partman-base: cjwatson * r1365 ubuntu/debian/changelog: releasing version 153ubuntu411:35
cjwatsonGreat, lots of data :)11:35
FourDollars45 M11:35
cjwatsonfeel free to bzip2 it or whatever11:35
FourDollars47M11:36
* FourDollars is wondering when it will stop11:36
FourDollars56M...11:37
cjwatsonIf it runs out of memory we can try again with something more limited, but to start with I'd like to see everything if possible11:37
FourDollarsFortunately, it has 8GB memory in this machine. :)11:39
FourDollarsErr... it seems to take a long long time to complete or fail.11:42
FourDollarsI will put /tmp/partman.shx in some where about 14 hours later11:43
* FourDollars needs to leave for his dinner.11:43
cjwatsonthanks11:44
infinitycjwatson: I've never understood why link_in_boot isn't the default for all arches.  For the ones where it doesn't matter, it's tidy, and for the ones where /boot is special, well, it's special.12:19
cjwatsonI remember thinking that in like breezy and intending to change it at some point.12:19
cjwatsonBut I only ever seem to think about it just before a release.12:19
infinityHeh.12:20
gemacjwatson: a user added logs to bug 934614, the entent of the problem seems to be bigger than we initially thought12:27
ubot2Launchpad bug 934614 in grub2 "reinstall of precise breaks grub with invalid arch independent ELF magic" [High,Incomplete] https://launchpad.net/bugs/93461412:27
* cjwatson drowns :-/12:27
gemacjwatson: there are some people in PS that have also hit the prboelm12:27
gemabut they are struggling to get their systems booting12:28
gemaMr.H was kind enough to attach results for the boot-info-script12:28
cjwatsonwell, really?  that log shows all kinds of boot loaders installed.12:28
gemaso it doesn't help?12:29
cjwatsonThat log seems to support my previous analysis, actually.12:30
gemaok12:30
cjwatsonyou can see "Grub2 (v1.99) is installed in the MBR of /dev/sda" at the top - that's a BIOS version, but then there's also /efi/boot/bootx64.efi on /dev/sdb112:32
cjwatson(well, no, the latter is the USB stick, sorry)12:33
cjwatsonActually it's not even clear that Mr. H's logs have anything to do with this bug.  I've followed up.12:34
cjwatsonI dunno.  Looks entirely different to me really.12:37
jibelto verify bug 745960, I do a default alternate installation with lvm encrypted on a Mac and check the system reboots, is it all or anything else to check ?13:01
ubot2Launchpad bug 745960 in grub2 "Cannot boot GRUB on Intel Mac after installing to LVM" [High,Incomplete] https://launchpad.net/bugs/74596013:01
jibelreboots meaning: ubuntu boots and any existing version of osx boots too13:02
cjwatsonjibel: sounds reasonable13:11
cjwatsonusing the amd64+mac image13:12
jibelindeed13:12
pmatulisdoes anyone have experience installing PPA-based packages with pxe/preseed?13:29
brendandpmatulis, yes13:45
pmatulisbrendand: feel like sharing?13:48
pmatulis:)13:48
brendandpmatulis, what have you tried?13:51
brendandpmatulis, you know about late_command?13:52
pmatulisbrendand: yes, i thought there may be a better way than late_command.  i heard that apt-setup might work.  i havne't started on it yet13:53
stgrabercjwatson: shouldn't finish-install.d/01oem-config-udeb install oem-config-slideshow-ubuntu? or am I missing something?13:54
stgrabercjwatson: bug 98488313:54
ubot2Launchpad bug 984883 in ubiquity "Ubuntu Precise alternate OEM: No slideshow during end user setup" [Undecided,New] https://launchpad.net/bugs/98488313:54
stgraber(rdepends for oem-config-slideshow-ubuntu is empty and I can't figure out what would install it if not finish-install.d/01oem-config-udeb on alternate)13:55
stgraberwith ubiquity ubi-usersetup does the apt-install call for frontend-gtk + slideshow13:55
cjwatsonstgraber: Yes, I think it should.  Compare r451014:02
cjwatsonpmatulis: apt-setup/local0/repository etc. - see the installation guide14:03
cjwatsonpmatulis: https://help.ubuntu.com/11.10/installation-guide/i386/preseed-contents.html#preseed-apt14:03
pmatuliscjwatson: thank you14:04
CIA-32ubiquity: stgraber * r5384 ubiquity/ (debian/changelog finish-install.d/01oem-config-udeb): Also install oem-config-slideshow-ubuntu when doing an OEM install in d-i14:06
stgraberjibel_: can you confirm that ^ isn't new? I see no indication that this is a recent breakage.14:08
cjwatsonbeen a problem since the slideshow was introduced AFAIK14:08
cjwatsoni.e. natty14:09
jibelstgraber, it is new, previously I had bug 96104614:20
ubot2Launchpad bug 961046 in ubiquity "oem-config: Content of the slideshow shifted to the bottom right" [Medium,Fix released] https://launchpad.net/bugs/96104614:20
jibelwell, that was a desktop image, not alternate14:21
cjwatsonjibel: the missing slideshow is alternate-specific14:25
jibeland I missed it since natty :(14:45
jibelnow I'm depressed14:45
stgraberwell, apparently so did our users ;)14:47
CIA-32user-setup: cjwatson * r259 ubuntu/ (debian/changelog user-setup-apply):14:47
CIA-32user-setup: user-setup-apply: After zeroing to-be-encrypted swap devices, attempt to14:47
CIA-32user-setup: format and enable the encrypted device.14:47
CIA-32ubiquity: cjwatson * r5385 trunk/ (debian/changelog tests/test_misc.py ubiquity/misc.py):16:33
CIA-32ubiquity: Don't default to installing GRUB to a removable device with /cdrom16:33
CIA-32ubiquity: mounted from it, even if /cdrom's filesystem isn't iso966016:33
CIA-32ubiquity: (LP: #984989).16:33
CIA-32user-setup: cjwatson * r260 ubuntu/debian/changelog: bug ref16:37
CIA-32user-setup: cjwatson * r261 ubuntu/debian/changelog: releasing version 1.42ubuntu216:38
=== xnox_ is now known as xnox
* cjwatson returns to the encrypted home problem, this time fuelled by champagne22:07
cjwatson'cos that'll help22:07
antarusoh dear22:07
infinityHah.22:26
* cjwatson considers http://paste.ubuntu.com/936126/ plus http://paste.ubuntu.com/936127/22:32
cjwatsonor something like that22:32
stgrabercjwatson: you import but don't use subprocess in the python part, the shell part looks good (didn't spend much more than 30s on it though), assuming /etc/crypttab in the livefs is correct (I'd have thought /target/etc/crypttab would be what we want but I never really looked at ecryptfs in ubiquity)22:38
cjwatsonoh yeah, I can ditch the import.  no, deliberately using /etc/crypttab in the live filesystem here because the target is mid-copy22:40
stgrabernevermind, I looked at the patch out of order ;)22:40
stgraberyeah, /etc/crypttab makes sense now22:40
cjwatsonmight need to make sure /etc/fstab is right when calling ecryptfs-setup-swap though22:41
cjwatsonit'll need to match the target system temporarily22:41
cjwatsonhorrible, but ...22:41
CIA-32apt-setup: cjwatson * r232 ubuntu/ (debian/changelog generators/01setup): If apt-setup/multiarch is empty, remove /etc/dpkg/dpkg.cfg.d/multiarch.22:55
CIA-32apt-setup: cjwatson * r233 ubuntu/debian/changelog: releasing version 1:0.55ubuntu322:56
CIA-32user-setup: cjwatson * r262 ubuntu/ (debian/changelog user-setup-apply):23:22
CIA-32user-setup: * user-setup-apply:23:22
CIA-32user-setup:  - Fix mkswap/swapon device name.23:22
CIA-32user-setup: cjwatson * r263 ubuntu/ (debian/changelog user-setup-apply):23:30
CIA-32user-setup: If OVERRIDE_ALREADY_ENCRYPTED_SWAP is set in the environment, copy23:30
CIA-32user-setup: /etc/crypttab from the installer environment rather than re-creating23:30
CIA-32user-setup: and re-zeroing swap (LP: #979350).23:30
cjwatsonMaybe it's better to just copy /etc/crypttab straight in.  I don't really like copying crypttab and fstab around at wildly different times.23:34

Generated by irclog2html.py 2.7 by Marius Gedminas - find it at mg.pov.lt!