[00:22] cjwatson, welcome back. [00:22] evand, cjwatson We are going to be going with an explicit echo "d-i languagechooser/language-name ..." | debconf-set-selections [00:23] the main thing that I was concerned about (for my information), was why my offline testing didnt work [00:24] when I did the debconf-set-selections and ran "oem-config" in an already-installed OS, it never worked. But when I did the exact same script as part of post-install script it worked [05:40] cjwatson: I've had problems recently with Gutsy, it seems by default /etc/network/interfaces last line, "iface eth0 inet dhcp" is commented out. Any idea what's going on? [05:40] I vaguely suspect NM because it has caused problems for me on Debian. [05:41] (i.e. prejudice) [09:37] twb: it's intentionally commented out if (and only if) network-manager is going to be installed, because otherwise they interfere [10:40] cjwatson re city selector interface in ubiquity, I was looking this video http://www.youtube.com/watch?v=YrFz7IxXppU (1m10s) [10:44] xivulon: you sure that's the right one? that's 3m20s and is about kwin composite effects [10:44] yep go to 1min 10 sec [10:44] it's just the idea of the magnifier, does not need composite to be implemented [10:45] if you have 2 images of different resolution should be easy to put togethter [10:45] ah. somebody with time is welcome to try it :-) [11:02] once I am in a good shape with wubi i'll give it a go, that should be pygtk correct? [11:15] yes [11:15] or in the underlying C code [11:34] xivulon, yes, that's kind of like what I suggested at UDS -- when you're outside the map area you see the whole map, but when you're inside the area you're auto-scrolling through a zoomed-in version [11:38] mpt I'll see if I can implement a sample [11:39] So, if the map area is X pixels wide and Y high [11:40] When you're hovering over (0,0), the area should show a zoomed-in version of the ((0,0), (X/n, Y/n)) part of the map [11:41] yep [11:42] When you're hovering over (X,Y) (the bottom-right pixel), the area should show a zoomed-in version of the ((X-X/n, Y-Y/n), (X,Y)) part of the map [11:42] etc [11:42] Where n = 2 or thereabouts [12:12] cjwatson: on the host in question, NM *was* installed, but nis didn't work at all until I uncommented the configuration in /etc/network/interfaces. [12:13] I imagine from the user's report that ypbind just sat there until it timed out and was backgrounded. Certainly logging in as root to enable /e/n/interfaces took deciseconds. [12:14] Perhaps NM simply doesn't work with NIS, or NM is activated after NIS. [12:15] it wouldn't surprise me if hardly anyone else has tried NM with NIS [12:57] Grumble [12:58] Yeah. NIS only really works if you expect to always be connected to the same network. network-manager is made for the exact opposite use case. [12:59] ...and yet NM is installed by default [12:59] I guess I must have told the users to install NIS by hand. [12:59] And nis isn't. [12:59] I can tell them to purge nm at the same time [13:09] I'm having hard time installing hardy atm. netboot fails with "Couldn't find task {minimal,standard}" [13:09] probably known already? [13:09] That sounds like a bashism [13:10] (Just guessing, really.) [13:10] there are two errors, one with "minimal", the other with "standard" :) [13:10] Oh, never mind then. [13:10] I tried to be clever.. === cjwatson_ is now known as cjwatson [13:14] which mirror? [13:15] those tasks seem to exist in the master archive [13:15] local, I'll update it again.. [13:16] check for Task: lines in the Packages file [13:16] also check syslog to ensure it did actually find your mirror successfully; I could imagine that that error might happen if apt-setup broke somehow [13:22] now it fails to find kernel modules :) [13:22] but it does use our mirror [14:17] so d-i just needs an upload to use -3? [14:19] debian-installer: cjwatson * r865 ubuntu/ (10 files in 4 dirs): * Move to 2.6.24-3 kernels. [14:19] hehe [14:19] debian-installer: cjwatson * r865 ubuntu/ (10 files in 4 dirs): * Move to 2.6.24-3 kernels. [14:19] ha, timing [14:20] the second one was because CIA said ENOSPC (at the remote end I assume) and I didn't bother checking whether the message had actually made it [14:24] debian-installer: cjwatson * r866 ubuntu/debian/changelog: releasing version 20070308ubuntu25 [15:16] regarding scrolling a zoomed tzmap, I'm actually working on that functionality now (while I work on more important things), as I have to work with the tzmap to fit ubiquity in 800x600 anyway. [15:41] evand in the scheme suggested by mpt the map does not pan and there are no scrollbars/buttons, have a look at the video [15:41] I can send you some cairo/pygtk code in the coming days if you wish [15:43] mpt: can you clarify? Are you talking about the entire map being zoomed in or just a boxed portion, like a magnifying glass? [15:43] xivulon, I'm not sure what you mean by "the map does not pan" [15:43] but I agree it shouldn't have scrollbars/buttons [15:43] magnifying glass [15:44] but the magnifying glass area being the map area itself [15:44] while the full world does not move [15:44] my plan was to make the map permanently zoomed in, and have the map move as you move toward an edge. I thought this was in line with what we discussed at UDS, someone please correct me if I'm wrong. [15:44] to further clarify, there would be no scrollbars [15:45] evand, my suggestion was to have the map zoomed out when you're not mousing over it, and zoomed in when you are [15:45] I guess I resumed the original mpt proposal after looking at the video [15:45] I probably didn't express myself very well [15:45] I could draw some sketches if it would help :-) [15:45] evand see http://www.youtube.com/watch?v=YrFz7IxXppU [15:45] 1min 10 sec [15:46] mpt: if you have the time it would be very much appreciated. We lost some of the notes from that session and I've been working off of memory. [15:48] xivulon: I already watched it. I see what you're saying, but that's fundamentally different from what I thought was the plan. But I'll wait for mpt's sketch before I assume I know exactly what that was :) [15:51] * mpt should have brought his inky pen to the office [16:12] evand, e-mailed [16:16] mpt: thanks, that definitely helps! [16:16] do you mind if I put this on the specification? [16:17] I have started a pygtk widget if it is of any help [16:18] evand, that's fine [16:19] evand, ping. [16:19] evand, are there any plans to make the preseed file language independent? [16:22] mebrown: can you elaborate? You can preseed when locale is set to something other than English, you just use the English keys in your preseed file. [16:23] evand, there are several places in the preseed where the string value is equal to what the GUI displays [16:23] evand, and if you change language those parts of the preseed brak [16:23] break [16:23] https://wiki.ubuntu.com/HardyUbiquityUsability has been updated. Thanks again mpt. [16:23] specifically partitioning and X [16:25] mebrown: Are you sure these are not broken elsewhere? I was able to preseed partitioner questions when using debian-installer/locale=zh_CN [16:25] evand, I had a working preseed. Exactly one change to language broke it. (fails at partitioning) [16:25] I saw something somewhere about them being locale specific [16:26] hrmm [16:26] http://pastebin.com/m758c0d2e [16:26] the preseed in question [16:27] when I change locale to zh_CN, partitioning fails (7.10, here) [16:27] It will be a couple weeks before I can start testing hardy === cjwatson_ is now known as cjwatson [16:29] partman is special, in every sense of the word [16:29] haha [16:29] but you should be able to work around it by preseeding the Chinese equivalent [16:29] which is obviously annoyingly unreliable, but should be a workaround [16:30] cjwatson, something like a feature request: [16:30] specify preseed language independent of installer langauge [16:30] don't you need to preseed partman-auto/disk, or does it assume you mean the first one it finds? [16:30] independent of final system locale [16:30] mebrown: yes, I am definitely familiar with the issue, it's just complex to fix [16:30] it is ONLY the partitioner [16:31] and only when you're asking it to partition free space as opposed to the whole disk [16:31] cjwatson, any way to just call 'gettext' on any of the string values that need localization? [16:31] if you're partitioning the whole disk, as Evan says, you use partman-auto/disk which doesn't require i18n [16:31] cjwatson, now thats funny. [16:31] mebrown: the installer doesn't use gettext, and not really [16:31] I can try to find you a better workaround [16:31] ah, which would explain why it worked for me. [16:32] cjwatson, I think we discussed this before, but I dont have a bug open: would like to be able to specify partitioning better in preseed. [16:32] right now, I have an early_script that does the partitioning [16:32] which is just wonky [16:32] and: we have had a couple things pop up on our mailing list [16:32] the right fix is to make it possible to preseed it as biggest_free (the underlying identifier) rather than as "Guided - use the largest continuous free space" [16:32] cjwatson, agreed on that part, except: [16:33] for our factory install, I know that I want to keep /dev/sda1 and /dev/sda2 and blow away everything else [16:33] and on the mailing lists, some people want to install to things like LVM or even encrypted lvm [16:33] and would be nice if they could specify that in the preseed. [16:34] right, I'm sorry but it'll be a while before we have that (not enough people are working on the partitioner as yet); it certainly isn't just you asking for that kind of thing :-/ [16:34] cjwatson, well, we have a working solution right now, so it is a feature request wishlist item [16:34] cjwatson, and it would be nice if the (I count 4) places in the preseed where it is locale-sensitive could be fixed [16:34] so I can just specify locale in the preseed. [16:35] and then... [16:35] fix up firstboot so I dont have to separately specify language/keyboard there. [16:35] I see 2 in yours, one of which is unnecessary [16:35] ? [16:35] 2 partitioning and 2 xorg [16:35] partman-auto/init_automatically_partition (necessary) and partman-auto/choose_recipe (you're just preseeding the default, so why bother?) [16:36] are you *sure* those xorg questions are locale-sensitive? that very much surprises me [16:36] so it should work ok if I remove the choose_recipe? I dont have a test system handy now, will test later. [16:36] cjwatson, an assumption on my part [16:37] they *look* similar to the partitioning stuff [16:37] they do, but it's the underlying code that matters. partitioning only suffers from this because the underlying code is very very weird and nonstandard. [16:38] (because it's trying to solve complex problems and doing stuff that the debconf system wasn't really designed to do) [16:38] you can definitely remove partman-auto/choose_recipe [16:38] ok [16:39] on oem-config, yes, though that will happen first. I asked evand to take care of that one. [16:40] I left a comment in /lib/partman/lib/base.sh a while back about the problem you're running into [16:40] just have it come up by default in the system language? (/etc/defaults/locale) [16:40] # TODO: This can be preseeded without having to use translated [16:40] # values (which are often inappropriate for preseeding across many [16:40] # machines due to including e.g. disk capacities) but it's nasty; [16:40] # you have to use runes like [16:40] # "20some_device__________/var/lib/partman/devices/=dev=sda". [16:40] # We could do with an abbreviated syntax. [16:40] mebrown: something needs to take care of copying the question in; once that happens it will work [16:41] the problem is that nothing copies preseeding across for the benefit of oem-config [16:41] but, question: [16:41] I was on an installed, running system [16:41] and was running oem-config manually to test [16:41] and when I ran echo ... | debconf-set-selections [16:41] it had no effect [16:42] you said you were preseeding languagechooser/language-name. I really recommend not doing that. The variable exposed for preseeding is debian-installer/locale. [16:42] I suspect some slight difference around there was your problem [16:42] the suggestion of languagechooser/language-name was my fault [16:42] languagechooser/language-name gets set internally by localechooser depending on the exact code path you hit and is not guaranteed to respect preseeding [16:43] evand, no it wasnt. [16:43] it wasn't? [16:43] evand, I was reading the code and that is what I came up with [16:43] before you made your suggestion on sat morning, I had been playing around for hours on friday with that variable [16:43] ah, fair enough [16:43] I just never got it to work until your suggestion to do it in post [16:44] I agree with doing it in a post-install script for now [16:44] but we'll fix the bug that requires that [16:44] cjwatson, the code already went to the factory, as we were over our deadline. we are using languagechooser/language-name for now. [16:44] ok [16:45] cjwatson, so: the final solution to this (implemented for hardy?) is that I can set debian-installer/locale and get everything to Just Work(TM)? [16:46] Do I need to open up any launchpads on these? [16:48] I've filed bug 181291 [16:48] Launchpad bug 181291 in oem-config "preseeded locale etc. not propagated to oem-config" [Undecided,New] https://launchpad.net/bugs/181291 [16:49] and tentatively assigned to evand since he said yesterday that was ok [16:49] yup, already working on it [16:49] mebrown: though I have not yet tested it, it's my belief that setting debian-installer/locale in a post-install script will work fine even for gutsy. Yes, the intent for hardy is that just setting debian-installer/locale in the preseed file should work without further post-install fiddling. [16:50] (regardless of exactly how that is achieved) [16:50] so: another bug about the localized partitioning to make it complete? [16:51] thanks for the help. (especially saturday help from evand) [16:51] anytime [16:52] next thing I noticed while working on this: [16:52] the list of keyboard layouts is not localized. [16:52] which was great for me testing, but would suck if you *only* speak mandarin [16:54] and also: I did a *quick* test, so I may have gotten it wrong, but I tried to enter chinese chars in the last screen (and only screen that accepts keyboard input, thankfully) for username and it didnt work. [16:56] yes, a bug on partman-base about the general mess that is partman preseeding would be good [16:57] keyboard layout localisation is known and indeed is a blocker to Debian switching to the same system [16:57] bug 127528 [16:57] Launchpad bug 127528 in console-setup "keyboard layout chooser not localized" [Wishlist,Confirmed] https://launchpad.net/bugs/127528 [16:58] what went wrong with Chinese entry on the username page? was the problem that it didn't have a working input method (my suspicion) or did it let you enter Chinese characters but then break? [16:59] I could not find a way to even enter chinese chars [17:00] I dont know chinese, but iirc, -space normally turns into chinese mode [17:00] and then you can enter gobbledy-gook [17:01] question: where to enter bugs in partman-base? [17:01] "partman-base does not use Launchpad as its bug tracker." [17:02] /ubuntu/+source/partman-base rather than /partman-base [17:02] or start at bugs.launchpad.net/ubuntu and enter partman-base as the source package [17:03] ctrl-space requires an input method to be active. I strongly suspect that oem-config isn't activating one. (That's a bug, and one I hadn't heard of before now, so yes, please do file it ...) [17:09] https://bugs.launchpad.net/dell/+bug/181296 [17:09] Launchpad bug 181296 in partman-base "partman preseeding should not require localized strings" [Undecided,New] [17:09] Launchpad bug 181296 in partman-base "partman preseeding should not require localized strings" [Undecided,New] [21:33] cjwatson, mpt, evand, have done a preliminary widget for the location selector [21:33] http://paste.ubuntu-nl.org/51260/ [21:36] You need the pixmap http://codebrowse.launchpad.net/~ubuntu-installer/ubiquity/trunk/download/supermario%40portablemario-20071225071604-qg5aa8kt510412xl/world_map960.png-20060222230952-12b3a9101a91fc15/world_map-960.png [21:36] within the same directory [21:48] cool; it's for evand to deal with though :) [21:59] heh, indeed and thanks. I'll take a look at it after I finish up an email to ubuntu-devel-discuss. [22:10] http://paste.ubuntu-nl.org/51266/ [22:10] added click event [22:11] evand, you owe me a beer then [22:11] I'll settle if you merge in the wubi tagged bugs though ;) [22:20] draw_map is a sink of resource (don't think there is a need to re-render the background at each refresh) if that is optimized responsiveness will improve [22:26] http://paste.ubuntu-nl.org/51267/ [22:26] last [23:58] evand, cjwaton have moved the map widget code to https://code.launchpad.net/~ubuntu-installer/ubiquity/zoom-map-widget [23:58] cjwatson