[00:37] pkgsel: cjwatson * r112 ubuntu/debian/ (63 files in 2 dirs): [00:37] pkgsel: Ask whether the user wants to install updates automatically, using [00:37] pkgsel: either unattended-upgrades or landscape-client. This question is only [00:37] pkgsel: asked at a high enough priority to be displayed on server or netboot [00:37] pkgsel: installations, but may be preseeded on other installations. [00:40] cjwatson: Thanks for the ubuntustudio tasksel changes. I was going to ask you about that but Cory beat me to it. I'll test them after the next daily run to be sure things work how we would like them to. [00:40] pkgsel: cjwatson * r113 ubuntu/debian/changelog: releasing version 0.20ubuntu4 [00:40] ok, sounds good [00:40] TheMuso: how goes dmraid? anything I can help with? I've not been keeping up, but you seem to have been making plenty of progress ... [00:42] cjwatson: Practically done. Just got to fix up grub-installer to properly take os-prober's output of OSs on dmraid arrays, but thats more a bug than a feature, so just got to finish packaging and upload dmraid event code, and upload dmraid changes with some udev additions. [00:42] Then, it is just a matter of waiting for dmraid/partman-dmraid to be moved to main. [00:43] One other thing I need to do is sort out partman-auto and recepies. Haven't yet got to that, so if anything, I may actually need help with that. [00:43] In terms of getting it done sooner. [00:43] is there a bug for dmraid promotion? [00:43] As now 2.6.27 is in/going into the archive, I now have to get alsa 1.0.17 in before FF. [00:43] cjwatson: Yes there is. There has been since hardy, and it is set back to new ready for examination. [00:44] let me dig it up. [00:44] bug 191557? [00:44] and 191564 === superm1 is now known as superm1|away [00:45] are the specific issues that were raised in 191557 fixed now? (notification on defective component, and rebuilding a degraded array) [00:47] Re 191557, no work has appeared from upstream concerning rebuilding degraded arrays, but as for detecting an array that has a problem, the extra package bits I'm working on for this can do that. I have also put a script into dmraid to sanity check an array, so that it is not brought up in a degraded state unless the user asks. [00:48] From what I can see, there is very little to no movement with dmraid upstream, unless there is a vcs repo that is being updtaed that I don't know about, but a red hat employee is the upstrea maintainer. [00:49] TheMuso: I don't think 2.6.27 is settled yet [00:49] cjwatson: oh ok. [00:51] then I'll hold off on my uploads then... [00:54] hmm, though it does look like linux is at 2.6.27 now [00:54] but the discussion with mdz on ubuntu-devel didn't seem to reach a clear conclusion [00:54] I'd say put the alsa (and pulseaudio?) packages in a PPA for now so that we can try them out with 2.6.27 [00:54] Yeah I gathered. I'll wait. [00:55] Yep I'll do that once my dmraid work is done. [00:55] drop me a /msg or e-mail once you're comfortable with dmraid and friends moving to main, and I'll sort that out [00:55] Ok. === superm1|away is now known as superm1 [02:35] cjwatson: Actually, the changes to partman-auto are much much simpler than I thought. [05:26] partman-basicfilesystems: TheMuso * r569 ubuntu/ (debian/changelog init.d/autouse_swap): * init.d/autouse_swap: We no longer need to skip sataraid partitions. [05:27] partman-basicfilesystems: TheMuso * r570 ubuntu/debian/changelog: releasing version 60ubuntu2 === superm1 is now known as superm1|away [06:50] console-setup: cjwatson * r75 ubuntu/ (50 files in 2 dirs): Fix property counts in Goha and legacy BDF fonts. [07:16] Doing netinstall of hardy, can I use an NFS mount for the pool? [07:16] I'm thinking I probably need to anna-install the nfs client first (from an http mirror). [07:16] I'm afraid there's no nfs-retriever yet [07:17] been meaning to write one for years [07:17] cjwatson: oh, I have to do the full d-i before I can nfs install? [07:17] so no, you can't [07:17] cjwatson: intersting; rh's kickstart seems to heavily favour NFS as the pool method. [07:17] I realise that, but I think in modern environments HTTP is not too much of an imposition? [07:18] cjwatson: I just have to work out how to drive apache [07:18] The box that's mirroring is in the dmz, so I have to be super careful [07:18] Since I can't just firewall 80 like I can NFS [07:18] I probably need one of those not-really-XML gibberish config entries in apache.conf [07:19] cjwatson: mainly I wanted to use NFS because I expected it to be faster than http [07:24] really? HTTP is a very efficient protocol [07:25] I didn't actually benchmark [07:26] Seems to me though that nfs will be faster and have less overhead for the server than a (heavyweight) httpd [07:26] For something like thttpd, shrug. [07:26] I've implemented HTTP; I'd be surprised if you could get significantly better throughput using NFS [07:27] in any case given that d-i doesn't support it HTTP will be an order of magnitude less trouble [07:27] cjwatson: does d-i support rsync? [07:27] console-setup: cjwatson * r76 ubuntu/debian/ (changelog control): Switch from console-tools-udeb to kbd-udeb, in line with Debian. [07:27] twb: no [07:28] it supports FTP, but that typically has worse file transfer performance than HTTP [07:28] FTP has a huge startup overhead compared to http [07:29] yes, it's a crap file transfer protocol :) [07:29] Not to mention that it can't handle NAT [07:29] honestly, just use HTTP. It's not that hard [07:29] cjwatson: right now I'm using http with my ISP's server, while I wait for someone to sort out the local mirror's http export [07:32] HTTP's main performance weakness is the fairly large headers; however, since d-i only fetches somewhere on the order of 1500 files and their mean size is relatively large compared to the header, I don't think that will make a huge difference [07:32] Hmm, it didn't ask me if I wanted a separate /home partition [07:32] that is correct [07:32] use expert mode if you want that [07:32] Was that removed? OK. [07:32] or manual partitioning [07:33] it was never there [07:33] Debian's has/had it [07:33] sure, but it was never in Ubuntu [07:33] Fair enough. [07:34] Debian is blithely ignoring the fact that a user has no reasonable basis on which to make the decision of how much disk to allocate to each, there's no particularly reasonable default, and if they get it wrong at installation time (when they can be expected to have least knowledge) it's very hard to change later [07:34] Debian is perhaps justified in doing so because their users are on average more knowledgeable [07:34] but I've never felt that it makes sense for Ubuntu to present such a question [07:35] Plus we kind of remove the need for separate /home partitions. [07:35] Apologies for inserting myself into the conversation. [07:37] When you click on LVM LV entries in the partman, it's a noop. It might be useful if a popup said "don't do that, choose `configure lvm' instead" [07:37] which entries exactly? (what's the text?) [07:38] cjwatson: after you create an LV like "root", it's the entry about the one for the filesystem. [07:39] Sorry, I've already clicked past it. [07:39] did you create the LV using "Configure the Logical Volume Manager"? [07:39] cjwatson: yes. [07:40] it shouldn't be a no-op, then, if I'm understanding you correctly (without exact text it's hard to be sure); selecting the LV should let you e.g. create a filesystem on it [07:40] For real disks, clicking on it says "write a new partition table for this?" [07:41] cjwatson: nono, creating a filesystem and mountpoint is on the next line [07:41] oh, right [07:41] (To clarify: this is netinst d-i on 8.04, without gtk) [07:41] Er, pxe boot, not a "netinst" type CD [07:42] yes, I agree. please file a bug on partman-base [07:43] the real problem is http://bugs.debian.org/cgi-bin/bugreport.cgi?bug=247342, though [07:43] namely, that you get two lines rather than one for an LV in the first place [07:44] * twb does "bts subscribe" [07:44] Sorry, I'm too lazy to get a new/find out my launchpad password [07:45] evand, cjwatson, with the d-i changes I've made for dmraid, and with such changes making partman-dmraid just about redundant, I don't see why we couldn't get ubiquity installing to dmraid arrays, unless parts of ubiquity itself need to be worked on to allow this. [07:46] Stupid question: if "md" is "multiple disks", what does "dm" stand for? [07:46] device mapper [07:46] Ah. [08:15] ubiquity: evand * r2781 ubiquity/ (4 files in 4 dirs): [08:15] ubiquity: Add before and after graphical partition bars to the autopartitioning [08:15] ubiquity: page. [11:25] cjwatson: I'm wondering what your thoughts are about getting rid of the dmraid=true kernel command-line, as well as the sataraid warning that gets displayed. If it was decided to remove it, partman-dmraid could be retired, and the remaining partman-dmraid bits could be merged into partman-base. THoughts? [11:29] I'm all for removing it if it works well without, although I'm less sure about merging in partman-dmraid [11:29] what are the contents of that udeb now? [11:33] All partman-dmraid has is the warning, and setting the sataraid flag for dmraid devices. I figured that when partman-base creates /var/lib/partman/devices and its subdirs for devices, that if it detects a dmraid device, it sets the sataraid flag there. [11:33] partman-dmraid also added the commit option to the top of the manual partitioning window, which is now redundant due to my libparted changes. [11:35] normally I'd say that it's better to have an extra module than to make partman-base even larger and more confusing, but if it's literally just setting a single flag, then I guess that isn't worth a separate module [11:35] what consumes the sataraid flag? [11:37] Prior to some recent changes in other parts of partman, it was used to disable features. After all of this is done, the only thing that will use the sataraid flag is partman-auto, to prevent skipping over dmraid mapper devices. [11:38] sounds sensible to merge it all into partman-base, then [11:39] ok great. Will go ahead and tie up the loose ends. [12:40] cjwatson: Yay Oem worked :) well on Ubuntu I'll try on Kubuntu now :) [12:40] oh good [12:43] partman-base: TheMuso * r103 ubuntu/ (debian/changelog init.d/parted): [12:43] partman-base: * init.d/parted: Set the sataraid flag for dmraid arrays. This code was [12:43] partman-base: originally in partman-dmraid, which is being retired. [12:45] partman-base: TheMuso * r104 ubuntu/debian/changelog: releasing version 121ubuntu5 [12:52] partman-auto: TheMuso * r269 ubuntu/ (debian/changelog lib/auto-shared.sh): [12:52] partman-auto: * lib/auto-shared.sh: Consider device-mapper devices/dmraid arrays with [12:52] partman-auto: the sataraid flag set as usable for automatic partitioning. [12:53] partman-auto: TheMuso * r270 ubuntu/debian/changelog: releasing version 78ubuntu2 [13:19] cjwatson: Kubuntu Alt 64bit just failed due to the adept breakage but it isn't showing up on http://cdimages.ubuntu.com/kubuntu/daily/current/report.html :( [13:20] cjwatson: http://paste.ubuntu.com/40893/ [13:21] sometimes the report doesn't manage to figure out installability just exactly the same way as apt [13:21] anyway, please talk to the Kubuntu people [13:22] and see http://people.ubuntu.com/~ubuntu-archive/testing/intrepid_probs.html which does report this [13:22] I believe it's being worked on [13:23] cjwatson: Hrm looks like we won't have dmraid event monitoring for intrepid... I am trying to build part of the code here, and it is somewhat broken, as well as not being able to find changelogs for any of the bits needed anywhere on the net or in tarballs... [13:48] hey - does anyone know why d-i insists on creating an extended partition when i'm specifying three partitions as primary in an expert_recipe? [13:48] this is the recipe: 512 512 512 ext3 $primary{ } $bootable{ } method{ format } format{ } use_filesystem{ } filesystem{ ext3 } mountpoint{ /boot } . 512 2048 4096 linux-swap $primary{ } method{ swap } format{ } . 2048 10000 1000000000 xfs $primary{ } method{ format } format{ } use_filesystem{ } filesystem{ xfs } mountpoint{ / } . [13:49] /boot and swap end up as primary, but / ends up as extended... any ideas? [13:49] cjwatson: thanks again I've informed riddell the adept issue is minor and should be a doddle to fix and the mysql one is being worked on :) [13:49] jamesb`: do you mean logical rather than extended? [13:49] yeah, sorry [13:50] jamesb`: I think I'd have to see syslog and partman logs [13:50] jamesb`: are you telling it to use the whole disk, or just some existing free space? [13:50] cjwatson: the whole disk [13:51] are the logs saved anywhere after install? [13:51] the logs should let me trace through what's goin on [13:51] going [13:51] yeah, /var/log/installer/ [13:52] though I'm curious as to why it's important that / be primary [13:52] usually I've found it best to use logical partitions wherever possible [13:52] as it weakens the constraints that partitioners may find themselves having to solve later [13:53] (I understand that your BIOS probably wants at least one partition to be primary and that might as well be /boot) [13:53] no particular reason, mostly just to keep things the same as our existing machines [13:54] just waiting for the install to finish and i'll grab the logs [14:52] cjwatson: i've got the partman and syslog available now if you want to check them out [14:53] you can grab them from http://dneg.com/~tech/installer/ [15:12] hw-detect: TheMuso * r84 ubuntu/ (debian/changelog debian/disk-detect.templates disk-detect.sh): [15:12] hw-detect: * disk-detect.sh: Do not check the kernel command line for any option [15:12] hw-detect: to enable dmraid support. If functional dmraid arrays are found, they [15:12] hw-detect: will be activated. [15:12] hw-detect: * debian/disk-detect.templates: Remove disk-detect/dmraid/enable template. [15:14] hw-detect: TheMuso * r85 ubuntu/debian/changelog: releasing version 1.63ubuntu2 [15:16] cjwatson: When you get a chance, feel free to look over dmraid for promotion to main. If there are any issues, ping me and I'll attend to them first thing in the morning/during the meeting. As I said, no dmraid event monitoring due to broken code which needs addressing by upstream. [15:16] Now I'm off to bed. [15:26] jamesb`: blink, /dev/vda? [15:29] cjwatson: its a kvm vm - i've tested the same preseed on real hardware and the behaviour is the same though [15:31] pkgsel: cjwatson * r114 ubuntu/debian/ (changelog postinst): Check /cdrom/.disk/info before unmounting /cdrom. [15:32] pkgsel: cjwatson * r115 ubuntu/debian/changelog: releasing version 0.20ubuntu5 [15:46] jamesb`: hmm, that's odd, it seems to think there are more partitions to be created after that last one. Could I see your whole preseed file, please? (feel free to strip passwords from it obviously) [15:51] sure - http://dneg.com/~tech/ubuntu.cfg [16:01] cjwatson, so I've been told I should help with the design of Landscape setup during server installation [16:01] cjwatson, but that's all I've been told. Has anything been implemented yet? If not, is it intended for Intrepid? [16:08] mpt: OK, so I *very* recently added the core stuff to pkgsel: that's a question asked just before doing bulk package installation that asks you how you want to manage updates [16:08] mpt: the broad design of that question is as specified by Mark [16:09] __Choices: No automatic updates, Install security updates automatically, Manage system with Landscape [16:09] _Description: How do you want to manage upgrades on this system? [16:09] Applying updates on a frequent basis is an important part of keeping your [16:09] system secure. [16:10] . [16:10] By default, updates need to be applied manually using package management [16:10] tools. Alternatively, you can choose to have this system automatically [16:10] download and install security updates, or you can choose to manage this [16:10] system over the web as part of a group of systems using Canonical's [16:10] Landscape service. [16:10] we haven't put the glue together for the last bit yet, but I believe the idea is that if you select "Landscape" here then the landscape-client package will ask for your Landscape ID [16:11] assuming that you have one [16:11] http://bazaar.launchpad.net/~ubuntu-core-dev/pkgsel/ubuntu/revision/112 if you want the whole thing [16:11] right, I gather you'd be able to leave it blank [16:11] thanks [16:12] I don't know a whole lot about how Landscape works - mdz just gave me the design in person [16:12] Well, I was wondering if anyone would choose "Manage system with Landscape" because they want to start using Landscape, and therefore don'g have an ID yet. [16:12] although the wording is my fault and now's an excellent time to improve it [16:12] I don't know how you go about getting an ID. Presumably you have to sign up for support [16:13] (or a trial) === superm1|away is now known as superm1 [16:15] this was a "just one more thing I can do" late last night kind of deal, so I haven't actually tried it yet [16:15] ok [16:16] cjwatson, will it be present in the latest daily server ISO? [16:16] part of the "late last night" bit was that it was *cough* a bit busted [16:16] I fixed the bustedness I saw today, but that won't be in an ISO yet [16:17] ok [16:17] I had to mess around a bit to get that question to appear only in netboot or server installations [16:21] jamesb`: (I'll get back to you, BTW, still trying to figure out what's going on) [16:21] dendrobates: so, as I said in e-mail, we now have a question that will prompt for the type of automatic updates you want, of which one choice is Landscape [16:22] cjwatson: thanks, i appreciate it. let me know if there's any more info you need or if there's anywhere to start digging to investigate myself [16:22] dendrobates: we need to get that choice bound to a sensible action; currently it's bound to "install landscape-client", which is apparently not quite right [16:23] dendrobates: so we need to arrange for landscape-client to ask for the ID if and only if the user chose to manage their system using Landscape, right? [16:23] cjwatson: could it be to install landscape client at a certain debconf priority/ [16:23] there is no such concept in the installer [16:23] cjwatson: ahh, right. [16:23] sorry :) [16:23] cjwatson: so that is why I planned on doing it a different way. [16:24] however, we could easily arrange to set some dummy template in landscape-client that you can use to determine the priority at which you ask the question [16:24] that would be my usual suggestion [16:24] db_get landscape-client/ask-for-id; if [ "$RET" = true ]; then priority=high; else priority=medium; fi; db_input "$priority" landscape-client/id [16:24] something like that [16:24] cjwatson: if you are selecting landscape to update, it can be assumed that you have an account, no? [16:25] 16:12 Well, I was wondering if anyone would choose "Manage system with Landscape" because they want to start using Landscape, and therefore don'g have an ID yet. [16:25] I think we need to assume that some people will choose it out of curiosity [16:26] cjwatson: so db_get landscape-client/ask-for-id in debconf sounds like the best bet. [16:26] dendrobates: can I see the current packaging? [16:27] cjwatson: chinstrap:~rclark/uploads/ [16:28] _Description: Account Name: [16:28] Short lowercase identifier of the Landscape account this computer [16:28] will be assigned. [16:28] is that expected to be an identifier you already know? [16:28] or is it analogous to a hostname? [16:30] dendrobates: please change debian/landscape-client.postinst to put '. /usr/share/debconf/confmodule' at the very top of the file, right under 'set -e'; non-intuitively, sourcing that file has the effect of re-execing the whole script from the start, so you usually want to source it right up top [16:30] (you could do the $1 = configure test before it to save resources in corner cases, I suppose) [16:31] cjwatson: I think account name is already known, but I will check with the landscape guys. [16:31] also, please remove db_stop; it doesn't do what you want, and can do bad things in the installer context [16:31] cjwatson: will do. [16:33] postrm: ! deluser --version >/dev/null 2>&1 || exec deluser landscape [16:33] lose the "exec" - you have stuff after it which won't get run if you exec deluser [16:34] dendrobates: anyway, give me a template name you want to use, and add it to debian/templates with Type: boolean and an untranslated Description (i.e. Description: not _Description:) [16:35] oh, and also Default: false [16:35] I'll arrange to set it to true [16:35] cjwatson: ok, give me a few minutes. [16:39] cjwatson, so the prompt that asks for your Landscape ID could direct you to http://landscape.canonical.com/ if you don't have one already [16:40] other than that, I'm not sure there's anything for me to do here [16:42] dendrobates: ^- that's one for your side of this I think [16:45] jamesb`: OK, I'm a bit mystified here. Could you run some debugging for me? Comment out the debconf/priority line in your preseed file so that we get an opportunity for some interaction; add DEBCONF_DEBUG=developer to the kernel parameters; then, when it asks you for a hostname, press alt-f2, start a shell, 'nano /bin/perform_recipe', and put 'set -x' on the second line [16:45] jamesb`: then continue, and I'll need those logs again [16:45] cjwatson: okay, i'll do that and let you know when the logs etc are ready [16:46] that should give me a shell trace of where it's applying the partitioning recipe, and I ought to be able to see whatever's broken from that [16:46] thanks [16:48] no problem [16:49] cjwatson: landscape-client/register_system [16:56] cjwatson: do you want this in the postinst? db_get landscape-client/ask-for-id; if [ "$RET" = true ]; then priority=high; else priority=medium; fi; db_input "$priority" landscape-client/id [16:57] well, that was example code, but that sort of idea, yes [16:58] I'm not sure which questions you actually want to ask [16:58] cjwatson: except in the installer, none. [16:58] right now, the db_get calls you have will just retrieve information from the database; you'll need 'db_input "$priority" question/name || true; db_input "$priority" other/question/name || true; db_go' or something like that to actually have them asked [16:59] with the priority conditional above they will be asked in expert mode outside the installer and asked in either normal or expert mode in the installer, which I think is the right model personally [17:01] cjwatson: ok, good. we are trying to avoid every desktop user being asked a question when they upgrade. [17:02] yes [17:02] cjwatson: I have avioded debconf, and now I wish I hadn't, would make this all easier. [17:02] I mean, if you'd rather, you could also do 'if [ "$RET" = true ]; then db_input ... || true; db_input ... || true; ... db_go; fi' [17:02] i.e. only ask the question at all in the installer, regardless of default priority [17:03] it's really not as scary as it looks at this level, but takes a little getting used to I know [17:03] maybe I should give an open week talk on it or something at some point [17:24] pkgsel: cjwatson * r116 ubuntu/debian/ (changelog postinst): [17:24] pkgsel: Preseed landscape-client/register_system to true if Landscape is [17:24] pkgsel: selected. [17:29] cjwatson: partman and syslog are here: http://dneg.com/~tech/log.debug/ - let me know if you need anything else [17:44] jamesb`: ah, I was looking in the intrepid source and it's fixed there; no wonder I couldn't figure it out [17:44] partman-auto 74: [17:44] * Don't create the last partition as a logical partition when all partitions [17:44] are defined as primary and there are sufficient primary partitions [17:44] available. Closes: #413505. [17:44] jamesb`: do you need a fix urgently, or is it sufficient to know that it's fixed in later releases? [17:46] cjwatson: its not crazy urgent - would be nice to have though. i guess i can try patching it in myself if its updated in a newer version [17:48] http://bazaar.launchpad.net/~vcs-imports/partman-auto/trunk/revision?start_revid=1138 [18:00] cheers === persia_ is now known as persia [18:23] cjwatson: I made the landscape debconf changes and uploaded the package to chinstrap. Can you take a quick look before I move on? [18:25] dendrobates: I didn't think we were going to ask landscape-client/register_system, only have it be preseeded by the installer [18:26] dendrobates: if that's the case, (a) it should have Description not _Description (untranslated) (b) the landscape.canonical.com link should be moved somewhere that's actually visible [18:26] cjwatson: oops, you did say that. [18:26] dendrobates: the indentation in landscape-client.postinst is pretty foul, and you *must* have a space after [ or it will break [18:27] dendrobates: you don't appear to actually ask the questions anywhere [18:29] dendrobates: chinstrap:~cjwatson/landscape-client-1.0.17/debian/landscape-client.postinst is more what I had in mind [18:29] ok [18:29] I haven't tested this, but I guess you hadn't either :) [18:30] to test the question-asking, you could do this before installing the package: echo landscape-client landscape-client/register_system boolean true | sudo debconf-set-selections [18:31] cjwatson: not yet, unfortunately I'm trying to do about 6 things at once and doing none of them well. I will fix it all up, in the end though. [18:58] pkgsel: cjwatson * r117 ubuntu/debian/changelog: releasing version 0.20ubuntu6 [23:07] evand, would you mind taking a look at bug 261676 sooner then alpha5? it's what will pull DKMS onto the dvds, so I wanted to make sure that we could include dkms as an item to install in the preseed and that it actually ends up working? [23:08] superm1: will do. This week has been insanely busy, but I should have much more time post FF [23:09] evand, sure, just so long as can try a dvd image prior to alpha5. we had a mad rush closer to the end of last time fighting daily after daily on some things, so sooner is better. [23:15] debian-installer: cjwatson * r954 ubuntu/ (4 files in 3 dirs): Move mainline architectures to 2.6.27-1 kernels. [23:17] noted [23:19] debian-installer: cjwatson * r955 ubuntu/debian/changelog: releasing version 20080522ubuntu11 === superm1 is now known as superm1|away