[02:11] can someone here help me with my broadcom wifi card issues? [09:32] Thinking out loud, but I am beginning to wonder if we do enough to make integration into the installer easy enough for external entities (ISVs, OEMs, etc). Not sure what could be done there and to my knowledge we haven't seen requests, so something to put on the back burner, I suppose. [09:40] evand: you could ask the oem team to get some feedback from oem's see if there is anything they would like to see etc? [09:41] xivulon: how goes the python conversion? [09:41] davmor2: indeed [09:43] evand: well it might give you an idea of things they do with the installer that might be more easily done, or sanely. [09:43] ja [09:43] yikes, so it seems there is still a bug in the partitioner if you select english at the CD boot menu, but then switch to another language at ubiquity [09:43] arrr [09:43] I thought I took care of this [09:50] <_ruben> when netbooting, you cant specify multiple (remote) preseed files right? [09:52] evand: do you want me to run a quick test and see [09:52] davmor2: nope, I've already confirmed it. Thanks though [09:52] np's [09:55] _ruben: Template: preseed/include_command [09:55] Type: string [09:55] Description: for internal use; can be preseeded [09:55] Shell command to run that may output a list of preseed files to load [10:06] <_ruben> evand: ah sweet .. and that supports http files and the likes? [10:11] I believe you can just add wget into the command, though I'd probably put fetching the preseed files in early_command [10:11] <_ruben> hm.. that'd be a rather decent alternative indeed .. thanks for the tips [10:15] evand: I'm writing up all the installer testcases does kubuntu use m-a yet or is that not going to happen in the foreseeable future? [10:16] davmor2: definitely not going to happen this cycle [10:16] evand: Okay cool :) one less to write :) [10:18] evand, cjwatson: Also out of curiosity is it me or do you need to click on the ubiquity shortcut with more than one double click on the latest iso's? [10:18] on ubuntu [10:19] davmor2: I just tried the latest Ubuntu daily-live and it worked fine for me [10:25] evand: Figured it out if the cd is still active it just highlight and doesn't start ubiquity if you wait for the cd to stop then try it works first time [10:26] ah [10:54] ubiquity: cjwatson * r2987 ubiquity/ (3 files in 2 dirs): [10:54] ubiquity: Default hostname to 'oem-laptop' or 'oem-desktop' as appropriate in OEM [10:54] ubiquity: mode, to allow OEM installations to be fully preseeded (thanks, Twisted [10:54] ubiquity: Lincoln, Inc.; LP: #321341). [11:05] I'm going to track the aforementioned bug here: https://bugs.edge.launchpad.net/ubuntu/+source/ubiquity/+bug/318564 [11:05] Launchpad bug 318564 in ubiquity "partman crashed and wrong translation on partition page" [Undecided,New] [12:22] evand, cjwatson: You know when you do a manual install but don't format the drive so you keep the existing home content does this work on alt and live or just live? [12:24] theoretically on alternate too, but less well-tested [12:24] the code's there, but I'm reluctant to say for sure that it works [12:24] cjwatson: I think it was poorly tested on both :) that's why we are trying to ensure we cover everything that is possible installer wise this time :) [12:25] yeah [12:25] definitely worth testing on alternate so we can fix it up [12:26] http://testcases.qa.ubuntu.com/System is the current list if you guys can think of anything else for these 3 desktops [12:28] I suppose there is usb based too that I haven't added yet but other than that I think it's getting pretty complete now [12:28] rescue mode for alternate [12:29] Given some of the recent bugs, I'd really like to see an explicit "install in a language other than en_*" [12:29] "free software only" mode on desktops [12:30] err actually on both, IIRC [12:30] that is the ubuntudesktopfree link [12:30] none of which work yet [12:30] yes, we definitely at least need to run both automatic and manual partitioners through in a non-English language - preferably one with reasonable translation coverage [12:30] ah, ok, I thought that was "use free space" [12:31] oh crap [12:31] that'll be FreeSpace to be added :) [12:31] I wouldn't worry about that all that much, TBH [12:32] it's similar to "whole disk" in a lot of ways, and the ways in which it isn't similar will be covered by "resize and use free space" [12:32] I don't think you need to test it separately [12:32] in fact, yes, it's really just a subset of "resize and use free space" [12:33] cool phew [12:33] cjwatson: the non-english is that on alt and live? [12:35] evand: ^ [12:35] davmor2: ideally both, yes. [12:36] evand: right so let me get it straight for the test write up. You select english on the initial language screen but then install in a non-english lang is that correct? [12:37] I think selecting a different language at the boot screen would be the most appropriate test [12:37] I know it wouldn't catch this current bug, but I think it's the most common path [12:37] and testing both paths seems like a bit too much work for humans (some day I will finish setting up automated tests) [12:38] evand: the testcase that are currently being written are for both humans and auto's so that is so much of an issue :) [12:39] so long as one set of tests does not get confused for the other :) [12:40] infact the only non-auto ones are the networkless installs it's hard to see if they worked without a network connection :) [12:41] evand: tests layout will look something like http://testcases.qa.ubuntu.com/Ubuntu/Applications/gcalc [12:41] noted; thanks [12:43] hrm, the difference between language from CD boot and language from ubiquity/localechooser that's triggering this bug is not as obvious as a diff between the debconf databases, it seems. [12:43] * evand goes back to digging === davmor2 is now known as davmor2-lunch === davmor2-lunch is now known as davmor2 [14:09] ah, got it! Somewhere there's a missing/failing call to locale-gen [14:10] evand: congrats [14:10] evand: um, careful about running locale-gen before partitioning [14:11] indeed, this might not be the right solution [14:11] it eats lots of memory and we generally try to avoid it before swap is up [14:11] ok [14:11] debconf ought not to need the locale to be generated [14:13] ok, I'm probably missing another piece of the puzzle; noted [15:19] cjwatson: do you know what 'sections' are required to have a full ubuntu mirror? [15:29] DogWater: all of them. Do you really mean sections, or do you mean components? [15:31] i guess what im wondering is what do i need to put into -s in order to get it? ;-) [15:35] in the debmirror utility [15:42] ah, debmirror uses section to mean component [15:42] cjwatson: like for example: main,restricted,universe,multiverse [15:42] DogWater: -s main,main/debian-installer is the minimum sensible [15:42] DogWater: -s main,main/debian-installer,restricted,restricted,restricted/debian-installer is what I use [15:43] thanks [15:43] DogWater: for *everything*, you want something like -s main,main/debian-installer,restricted,restricted/debian-installer,universe,multiverse [15:44] so were you able to figure out that bug that only I would find? ;-) [15:50] I did say I managed to reproduce it [15:51] haven't fixed it yet, but that's more because I've been buried in debconf half the day than anything else [15:52] I have a test case for it in my code now [15:53] basically, the problem is that the preseed command really ought to have slightly different option semantics from everything else [15:54] aside from the workaround I suggested, you could also use 'preseed -- preseed/late_command string ...' to work around it [15:54] but I'm going to try to make that unnecessary [15:56] I think this problem lies slightly lower than I understand at the moment. LC_ALL=es_ES.UTF-8 df and LC_ALL=es_ES.UTF-8 gedit both fail in jaunty when they worked fine in 8.10. [15:56] fail in what way? [15:57] df is not translated, neither is gedit. GTK reports: Locale not supported by C library. Using the fallback 'C' locale. [15:58] right, that will happen if a locale isn't generated [15:58] but that shouldn't affect debconf in any way - it does its own localisation handling [15:58] ah. Red herring aside, did we preemptively generate them in 8.10? [16:00] I hope not, but they might have been available by way of a language pack on the live CD [16:01] hrm [16:01] oh, okay so just preseed -- then the original command? [16:01] cool, i think im just going to setup and keep an updated mirror though probably will be easier and more reliable [16:04] hmm, maybe not, preseed command handling is more broken than I thought [16:05] hopefully i can get debian 4 preseeding to work as well as ubuntu's kickstart (now) works. [16:05] i'm trying to completely automate the installation of every distribution [16:05] well, every one that has a reasonable install base [16:08] cjwatson: do you have the link to the fixed amd64 initrd still or should i go fish? [16:13] I put it in the bug of which yours is a duplicate [16:13] damn, i almost got it to install in less than a minute ;-) [16:13] if this machine was faster i bet it would do it [16:13] wow [16:13] partitioning =/ [16:13] if i use a clean disk its almost instant [16:14] that's really very impressive [16:14] doesn't it take a while to unpack all the packages? [16:14] depdends on the machine [16:14] on the celerons it takes about 30-40 seconds but on some faster ones its pretty quick [16:15] if i could make it partition faster it would be great [16:15] but i understand that mkfs.ext3 is a dog at times [16:15] I'm really struggling to see how the preseed command in Kickstart ever worked for anything complex :-/ [16:15] its argument handling is all wrong [16:16] mpt: http://people.ubuntu.com/~evand/tmp/ubiquity-9.04-keyboard.png is the current UI for your suggested change to the keyboard layout page. Please do let me know if this is not suitable or if you have any comments. [16:19] could you collapse down the "standard" variant the same way the variant list box does? [16:19] evand, on what evidence is the suggested layout suggested? [16:19] so that should be USA, not USA - USA [16:22] the double dhcp pauses, the formatting take the second longest after 'select and install software', crap this one took 5:45 from pxelinux to reboot [16:26] you can shorten the pause with netcfg/dhcp_timeout [16:26] cjwatson: we just talked through it (I'm at Millbank today and tomorrow), and now plan to be going with drop down boxes with "(recommended)" next to USA. Any part of that stand out as wrong to you? [16:27] what, replacing the full list boxes with drop-downs? [16:27] With the potential to add a small portion of the keyboard selected as a bit of a visual aid [16:27] yes [16:27] I'm not all that keen on that, I have to say [16:27] it sounds clumsy to use [16:28] all attempts I've seen to display keyboard layouts visually have been worse than useless [16:28] so I wouldn't bother reserving space for that [16:28] hmm, ok [16:28] and, once you throw that out, the dialog looks awfully empty with just drop-downs? [16:28] indeed [16:29] so perhaps what we had before, just with "(recommended)" next to the default selection? [16:29] the problem with displaying an image of the keyboard is that many layouts and variants differ in rather fine details, and you end up squinting at this tiny little image trying to figure out whether it's right [16:29] mpt's suggestion was to show about 1/4th of the keyboard, the top right portion [16:30] I'm probably off on the fraction [16:30] I actually thought your mockup above was fine, with the exception of "USA - USA" -> "USA"; I'm curious as to what mpt thought was wrong with that [16:30] due respect but that's hopeless :) [16:30] heh [16:30] won't distinguish lots of layouts at all [16:30] it would be little more than decoration [16:30] mpt: can you elaborate, rather than me possibly paraphrasing you incorrectly :) [16:30] and it's lots of fragile code to generate, from what I've seen elsewhere, for pretty minimal return [16:31] the only thing I've ever seen try to display a keyboard image is GNOME's layout selector, which is the worst layout selector I've ever seen [16:32] oh wow [16:33] I see what you mean by squinting [16:33] cjwatson, my hypothesis was that you don't need to see the whole keyboard (like Gnome's, which I agree is horrible) to be able to tell whether the layout matches your keyboard [16:33] That for most layouts, the top right quarter or so is enough [16:33] I don't think that holds [16:33] so we could zoom and show that area by default, while still letting you scroll to see the rest. [16:33] and I don't think it's useful to show it visually anyway [16:34] my laptop keyboard *looks* completely different from a regular PC keyboard, but has the same layout as far as the installer is concerned [16:34] it would be worse than useless to show me a picture of a standard PC keyboard [16:35] hmm [16:35] we only care about the scancode->keycode->keysym level, not the geometry of the physical keys [16:35] Ok, so how else are people supposed to know what layout their keyboard is? [16:35] (no sarcasm) [16:35] the best way to tell whether a keyboard layout matches your keys is to type stuff with it [16:35] that's why we provide the test entry box [16:36] ideally, we should port the thing from d-i that lets you type some keys and then infers the layout from that [16:36] I'm going to give interfacing with that a shot as part of the jaunty-ubiquity-usability specification. [16:37] but we should also default to language+location defaults, as they're correct a fairly large percentage of the time (though far from universally) and the layout inference tool takes a little while to walk through [16:38] I wish that USB keyboards actually provided reliable layout information, but they don't [16:38] (that would have been the ideal fix for this kind of thing) [16:38] I take it they often provide wrong information? [16:39] well, there are two problems [16:39] firstly, the spec is vague - it just has a hardcoded list of languages, rather than what you actually want which is a description of the Unicode codepoints that each key's supposed to generate, or similar [16:39] secondly, USB keyboard manufacturers are hopelessly cheap and often just leave out that circuitry so it just says it's English or doesn't say at all [16:40] but even if it does say, it's probably not enough information to infer an X keyboard layout :-/ [16:40] yikes [16:40] This reminds me of the X tell-me-your-dpi issue [16:40] mm, except you don't have to get out a ruler [16:41] a lot of people, to be honest, will get by by just hitting Next on the current dialog [16:41] but we've tried omitting that step in the past, and the proportion of people who weren't satisfied by that was non-trivial [16:42] (US keyboards are common nearly everywhere even if the majority of people use a native layout; Canadians are split between English and French layouts in a way that doesn't necessarily match the language they prefer; people move countries and take their old computers with them; etc.) [16:43] This smells like it'll need an "I don't know" radio button [16:44] which walks you through the inference tool? [16:45] I think the reason I prefer list boxes to drop-downs for this, by the way, is that the list of available layouts is rather long, and drop-downs are clumsier to scroll than list boxes [16:46] when I end up in that situation, I usually end up clicking on the drop-down and then using the keyboard to scroll, which is, well [16:46] yeah, to go through the type-to-calculate thing [16:47] one problem with that at the moment is that it has no backup [16:47] er, by which I mean "go back". It's common to hit the wrong key and then have to go through it all over again [16:48] I did try to fix that once but it's a bit labyrinthine - we should, though [16:48] GTK option menus currently are more painful than they should be -- their speed doesn't increase as you move further away from the top/bottom [16:48] they also don't get capped to a particular size, but end up filling the whole screen [16:48] (or occasionally buggily extending off-screen or something) [16:48] yes, that's kind of a corollary [16:48] I tend to prefer things that stay within the window [16:49] If variable speed wasn't useful, it wouldn't matter that they touch the top/bottom of the screen [16:49] hm, I think I would still find it visually unpleasant even if the scrolling were fixed [16:50] it has an unfinished kind of look when something splats up a piece of tape extending from the top to the bottom of your screen :) [16:50] it also makes it less convenient to switch windows in the middle of a task, although that's not so much of a problem in this particular case [16:54] IMO those are both bugs :-) [16:54] The former that the scroll arrows are the wrong height, and the latter that menus eat Alt Tab [16:55] ok - nevertheless, we need to work with the widgets we have [16:56] yeah [17:00] hm, actually there are two speeds, but that's not enough [17:00] * mpt reports a bug [17:16] evand: ping [17:16] kirkland: pong [17:17] evand: hey, i was curious about the state of the encrypted home installer option [17:17] evand: i'm downloading the desktop daily iso now [17:17] evand: curious if you needed my help with anything, or if you expect it to work? [17:17] ah indeed. Still broken, had to take a break from debugging it to work on other things, but will resume tomorrow. [17:17] kirkland: sure, help is welcome [17:17] let me find my notes [17:17] evand: can you describe the problem? [17:18] evand: i bet we're going to need a similar hack that i had to give cjwatson [17:18] for the server/alternate installers [17:18] evand: i'm about to upload a new ecryptfs-utils that adds filename encryption too [17:18] evand: i was hoping to do that after having verified the installer bits [17:19] evand: fwiw, i'm flying on Wednesday to .de, so I wanted to try and get the desktop installer working before i left [17:19] kirkland: we can always pair up in Berlin if you have some time to hack on it, assuming it's not fixed by then [17:19] but that's a week a way, so I imagine it wouldn't have to come to that [17:20] evand: that's fine too, i was curious how berlin week was going to work [17:20] evand: ie, coop between teams, plus alpha4 freeze [17:20] ah right, now I remember [17:20] evand: i was thinking the desktop bit is probably something simple and can be fixed ahead of then [17:21] there seems to be some wonky interaction between ecryptfs-setup-private and the ecryptfs kernel module [17:21] evand: what I really want to accomplish in berlin is some symblance of an encrypted swap solution [17:21] it doesn't like user=foobar [17:21] oh, right, i fixed that in this upload [17:21] wonderful! [17:21] evand: that should be a warning, though [17:21] I just need to upload a fix for the blacklisting problem, and it should be golden [17:22] cool, let me get this new ecryptfs-utils uploaded too, then. [17:22] I'll review that and upload it tomorrow (ubiquity change, that is) [17:34] DogWater: I think I've decided that the right fix for you is in fact a documentation fix [17:35] DogWater: and that you should put the whole value to be preseeded in double quotes [17:35] DogWater: so: preseed preseed/late_command string "sed -i ...; in-target apt-get update" [17:35] or whatever it was [17:38] hrm, porting a cdebconf plugin to ubiquity without rewriting it looks like it might involve a large saw and a welding torch. [17:39] yes, debconf doesn't have plugins and in any case there's no GTK version of that one; "port" might be a euphemism for "write new code that uses the same backend data" [17:39] indeed, I'll try to wrap my head around the latter tomorrow [17:40] DogWater: (I did try changing this in kickseed, but the result was that the "preseed" command ended up being treated radically differently from every other command, and in the end I decided it would be far too confusing) [17:41] installation-guide: cjwatson * r442 ubuntu/ (debian/changelog en/install-methods/automatic-install.xml): [17:41] installation-guide: Document that values for Kickstart's preseed command containing spaces [17:41] installation-guide: must be quoted. [17:56] ubiquity: evand * r2988 ubiquity/ubiquity/frontend/gtk_ui.py: Only show the keyboard variant selection in the suggested option label. [19:23] oem-config: cjwatson * r585 trunk/ (d-i/manifest debian/changelog): Automatic update of included source packages: user-setup 1.23ubuntu7. [19:55] oem-config: cjwatson * r586 trunk/debian/changelog: releasing version 1.54.3 [23:49] cjwatson: okay, i dug a bit deeper into the encrypted home problem on the installer... perhaps this info might help ... [23:50] cjwatson: the libecryptfs0 gets installed, however, one of the files it owns, doesn't [23:50] cjwatson: namely, /usr/lib/libecryptfs.so.0.0.0 [23:50] kirkland: that's the blacklisting issue that evand mentioned, no? [23:50] cjwatson: however, there are symlinks pointing to ie, ie /usr/lib/libecryptfs.so.0.0 and /usr/lib/libecryptfs.so.0 [23:50] cjwatson: oh, is it? [23:50] cjwatson: sorry, "blacklisting" was cryptic to me [23:51] it's curious that the symlinks are there, but I thought the cause of the missing files was established [23:51] cjwatson: okey doke, i'll wait until tomorrow, hopefully it'll sort itself out [23:51] so, as you know, ubiquity copies the live filesystem to the hard disk, and then removes some packages afterwards [23:51] to save time, it doesn't bother copying the files of packages it knows it's going to remove anyway [23:52] problem is, if it makes a mistake, it doesn't know how to put those files back [23:52] this happens in the event that a package is in the live filesystem but not in the standard desktop install, but some code that runs after the copying process asks for it to be installed [23:52] cjwatson: smart! [23:52] I believe that *a* fix is to move the conditional apt-install in user-setup-apply back to user-setup-ask [23:53] cjwatson: smart = the optimization (when it works) anyway :-) [23:53] but evand was testing this, and ran into trouble due (I think) to the problem you mentioned earlier [23:53] so it hasn't been committed anywhere yet