[00:31] partman-md: cjwatson * r934 auto-setup/ (5 files in 4 dirs): (log message trimmed) [00:31] partman-md: Rearrange RAID configuration per the [00:31] partman-md: foundations-karmic-server-installer-improvements specification. Instead [00:31] partman-md: of requiring partitions to be set for use as RAID physical volumes [00:31] partman-md: first, we now offer all partitions that could be used as physical [00:31] partman-md: volumes, and automatically set them up that way on request. This allows [00:31] partman-md: us to offer our main menu option more or less all the time, and should [08:00] evand: I have been muking around with cleaning up the installer and I think I broke something on the kde side. After clicking next on the language screen it crashes...and syslog reports: debconf: DbDriver "targetdb": could not open /target/var/cache/debcond/config.dat ... any ideas as to what I might have done wrong? my cleanups arn't that dramatic so I am kinda lost with this debconf stuff, thanks... [08:03] shtylman: is this with the latest ubiquity? What version, specifically? [08:04] 1.13.8 [08:04] maybe I havn't updated in a bit... [08:12] evand: oh I see...a bit has changed after the oem merge [08:12] I think I will just kill my changes and start afresh from that [08:12] evand: some of the oem merge changes actually did what I was doing so yea :) [08:16] apologies. I really need to get better at communicating what's going on with ubiquity development. I emailed xivulon about the incoming merge, as I knew it directly affected what he was working on, but I didn't think to CC you. I'll send such mails to ubuntu-installer@ next time. [08:18] evand: no worries :) I haven't had as much time (with my move to nyc and all) to keep up with the dev as a would have liked...but now I am more settled in and hopefully can get a hold of it [08:20] shtylman: how is that going, by the way? Are you in Manhattan or in the dark territory of of a surrounding borough? [08:21] evand: hahah...it went quite well. Started work on Monday (eats up ALOT of time) ... yea, I am in Manhattan...I avoided the dark territories :) [08:21] haha [08:31] xivulon: out of curiosity, why do you use Python 2.3 in Wubi, rather than Python 2.6? [08:31] evand: just a heads up on what my changes will entail... beyond the cosmetic things I showed you, I am migrating the sig/slot connect calls over to the new mechanism in pyqt4.5 and also cleaning up some of te codebase... [08:32] good deal [08:32] do you have this in a published branch? [08:32] hi evand, because A) it leads to a smaller overall package B) there are issues with the distributions of some libraries in python 2.4+ [08:32] xivulon: can you elaborate on B? [08:33] evand: not yet... I did start a branch, but I am gonna delete it cause I am just gonna restart from the changes since oem-config merge... they arn't all that many yet or hard to do so hopefully I will have a new published branch soon [08:34] shtylman: okay, thanks [08:41] evand have to fetch the info, iirc newer versions are linked against msvcr*.dll, and I believe you cannot redistribute the CRT unless you are a visual studio licensee [08:51] evand, compiling python + extensions in mingw32 should also do the trick [08:53] hrmm [08:54] evand: http://article.gmane.org/gmane.comp.python.py2exe/652 [09:00] * evand reads up [09:02] surely if this was an issue, you would not be able to redistribute the Python for Windows installer, which I imagine would be a big deal. [09:02] ps in trunk I have created a new tool called pypack (out of pyinstaller), running that in wine with the target python version, it will convert a python script (entry point) into a self-contained directory with all required dependencies [09:02] nice [09:02] I think python devs have the license to redistribute msvcr*.dll, I do not own visual studio though [09:03] and I did not know about the implications for you redistributing it, so I went the safe route, I tried python 2.3 and I saved quite some space, so I ended up sticking with that [09:06] okay [09:06] I'll have to do some more research on this [09:07] Most annoying part of using 2.3 I found is lack of annotations and some modules (sets) have to be installed separately [09:09] have to go now, feel free to send me an email if you have other questions [10:47] ubiquity: cjwatson * r3319 ubiquity/ (bin/ubiquity debian/changelog): [10:47] ubiquity: Restore autologin-disabling code from oem-config, corrected to work with [10:47] ubiquity: new gdm (LP: #395861). [12:49] cjwatson: I just want to check with you before I start putting pieces into place. Do you have any objections to a ubiquity-slideshow-ubuntu package (in its own source package) that ubiquity-frontend-gtk depends on? It's currently set up to provide ubiquity-slideshow, but I can't see that being necessary as putting a dependency on ubiquity-slideshow in ubiquity would require every frontend to provide a slideshow and I can't think of how we'd so [12:59] err nevermind on that not being necessary. I missed the obvious case of the different slideshow packages needing to conflict. [13:12] cjwatson, whoops, didn't mean to drop a autologin-disabling section in the merge [13:14] evand: it's fine by me [13:14] mterry: no worries, just happened to notice it [13:14] good deal [14:19] ubiquity: evand * r3320 ubiquity/ (3 files in 3 dirs): Add support for ubiquity-slideshow. [14:28] okay so looks like if no slides are found, just hides the frame like before [14:29] sounds good to me [14:40] mterry: I've reassigned all the open oem-config bugs to ubiquity, and tagged them 'oem-config' [14:40] cjwatson, hah. I was just investigating how to do that via launchpadlib [14:40] cjwatson, saved me some trouble, thanks. :) [14:40] (hmm, I may have accidentally reassigned duplicates too) [14:41] mterry: http://paste.ubuntu.com/219780/ [14:42] probably want something like 'and not task.bug.duplicate_of' in there [14:42] ...829 open bugs... :-/ [14:42] cjwatson, interesting. that makes sense now that I see it, but I was wading through a bunch of docs to get there [14:43] yeah, I've just been there before is all [14:43] the bug tagging trick is one you have to know - bug 254091 [14:43] Launchpad bug 254091 in xserver-xorg-video-intel "[Intrepid] Intel Mobile GM965/GL960 Generates a corrupt screen with compiz enabled (dup-of: 245888)" [Medium,Triaged] https://launchpad.net/bugs/254091 [14:43] Launchpad bug 245888 in mesa "Intrepid, on latest updates (mesa updates - 7.1~rc1-0ubuntu1), compiz no longer works and gives white screen on login" [High,Fix released] https://launchpad.net/bugs/245888 [14:43] err, what? [14:44] bug 254901 [14:44] Launchpad bug 254901 in launchpadlib "appending tags to bug.tags is not supported properly on lp_save()" [Low,Triaged] https://launchpad.net/bugs/254901 [14:44] curious [14:44] today, launchpadlib let me do in about three or four hours what took me two full days last time I did it [14:44] (8.04.3 change summary) [14:45] so it's well worth some investment in learning [14:45] :) [14:45] evand: yeah :-/ [14:49] * mterry gets a cjwatson-inspired email storm [14:50] cjwatson, oh, btw, I don't mean to be a pest, but I didn't get an ACK yesterday, so I'm pinging you again about it: Can you take a look at https://code.launchpad.net/~mterry/ubiquity/translated-timezones/+merge/8698 when you have time and in particular, give me an OK or not on starting an MIR for python-pyicu [14:52] ok, I'll queue it up [14:52] from what you've written, I suggest starting a MIR for pyicu anyway [14:53] I can imagine using it somehow in gfxboot-theme-ubuntu, for instance; there's a long-running bug in which people complain about the collation [15:02] mterry: what's the one-liner for sorting a list of strings using pyicu then? [15:34] cjwatson, i missed your last comment about pyicu due to irc issues. the question was, 'what's a one-liner for sorting with pyicu?' let me grab that [15:36] cjwatson, well... not really a one-liner. what it does is gives us a 'collation key' for python's normal sort algorithms. So doing 'self.collator.getCollationKey('goofy string').getByteArray()' gives a good collation key [15:37] cjwatson, but you first have to instantiate a collator with a certain locale [15:37] (persia, thanks for noting the missed question) :) [15:38] * persia lurks harder [15:47] mterry: mm, I realised shortly after asking that I could have found the answer by looking through your merge request :) [15:48] right, so it doesn't really help with the language question unfortunately since there is no preferred locale there [15:48] cjwatson, even instantiating with the C locale is fine there [15:48] I mean I suppose we could just say sod-it and do Latin-alphabet sorting [15:48] mm, sort of [15:48] cjwatson, it only uses the locale for special situations [15:48] where does Čeština go? [15:48] cjwatson, there are several layers of collation [15:49] cjwatson, even in C, it will strip accents AFAIK [15:49] or Қазақ for that matter [15:49] Қазақ is post-ASCII [15:49] (i.e. sorted after Z) [15:49] that's unfortunate [15:49] (As i recall) [15:49] I tested with sorting the language list [15:50] it'd presumably mean that everything non-Latin gets punted to the end [15:50] But it was strictly better than the current list. :) [15:50] Mostly [15:50] In fact, yes [15:50] But where would you sort them? You can't really sort glyphs in the middle of Latin [15:50] Қазақ oughta go with the Ks [15:51] I dunno, it's all terribly subjective [15:51] cjwatson, the Unicode consortium has specs for this [15:51] cjwatson, and libicu tries to follow them [15:51] cjwatson, let me find the spec [15:51] cjwatson, http://unicode.org/reports/tr10/ [15:53] right, for locale-dependent collation [15:54] cjwatson, right. but in the absence of a locale, I argue that applying 70% of the collation logic is better than the arbitrary order we have now [15:54] nobody really designs for the case where you don't know the target language because (a) the best you can do is smelly heuristics and (b) it's very rare [15:54] I'm happy for somebody to drop in a replacement collation order on the condition that if we ever get bugs about it I can assign them to that person. :) [15:54] cjwatson, unless you wanted to have sexy animation logic where we resort the list after selecting a language. :) [15:54] no, that'd definitely be confusing [15:55] agreed [15:55] "the thing I clicked on just ran away, help" [15:55] cjwatson, though, we do know the locale at that point. It's 'C'. :) [15:56] yeah, I sort of meant a useful locale [15:59] cjwatson, well, I consider the current sorting as 'random'. Having a sorting that at least puts most characters that look alike near each other will help people find their language. Though there may be odd balls like your funky K [15:59] cjwatson, you can blame me when people complain [16:00] as you say, it's an inherently unsolvable problem [16:02] Exhibit A in my 'what's wrong with the current sorting' is Finnish [16:03] so let's go with it [16:04] cjwatson, one thing that may help is trying something like the LanguageOnly screen, where all languages fit on one screen [16:04] cjwatson, have we found that to be helpful? [16:05] I don't really like it personally [16:05] in the case of ubiquity I think the other information on the front page is useful to display, and aesthetically I just find the bare grid rather intimidating-looking [16:06] it has an obvious scaling problem too - we add languages from time to time, not many but usually one or two a release [16:06] Yeah.... And I would expect it to sort down first before sorting left-to-right. But that's probably locale-dependent too [16:06] /probably/certainly/ [16:07] we do it that way largely because some OEMs complained that (IIRC) Chinese wasn't on the front page, but I can't say I like the results [16:07] That's an interesting idea. A good sorting in C locale might just be based on usage [16:07] We have popcon data for that? [16:07] for which language packs are installed... [16:07] dunno [16:08] I'm not convinced Ubuntu popcon is what you might call accurate [16:08] there's some evidence of serious skew [16:08] en would probably give false positives. and languages for where the popcon UI is not translated would have trouble [16:08] cjwatson, locale data for the firefox homepage? [16:08] firefox sends it [16:08] I don't think usage is a good sorting metric though [16:09] that requires a very very weird mindset on the part of users [16:09] hmm... but we could have a sexy web 2.0 cloud of language names [16:09] "oh yeah, I think my language is about the twentieth most popular in the world" [16:09] ... not happening :) [16:09] I'd rather have a best-effort C sort with a few weird spots [16:09] yar, but it even with my sexy new sorting, they have to think 'hmm, I think english would sort me here' [16:10] Which is something they're more used to, so... [16:10] right, but at least A-Z sorting is fairly usual [16:10] so it's only Czech, Kazakh, and the non-Latin ones who have to think [16:10] which is still a lot but it's better than now [16:10] alright. Well, first step is MIR [16:10] and *definitely* better than it would be if we shuffled the order by speaker count [16:11] :) [16:11] I must admit python-pyicu seems like a lot of bytes for a smarter sort() [16:11] yeah, but icu is already on cd for openoffice [16:11] I wonder if we could do a reasonable job in less space given the limited input data [16:11] python-pyicu itself is a small addition [16:11] it's >200KB - not huge but not that trivial either [16:12] if it'd be really hard to do a decent job independently, I'll defer [16:12] cjwatson, we could hardcode it. we could strip accents, but unless we hardcoded list of accents, we'd need to get that from *some* library [16:12] cjwatson, if language list doesn't change unless we hear about it, hardcode might not be the worst [16:12] but... i want icu for sorting country list anyway [16:13] so it's already there [16:13] and that should take into account the locale, which we definitely don't' want to reimplement [16:13] ok [16:13] I don't really want to hardcode as such [16:14] we hear about changes, but we have to sort the list in several places, so my gut feel is that hardcoding would get out of date [16:15] Is there an issue with adding python-pyicu to the CD in terms of size? [16:16] there's always a size issue with adding things to the CD :-/ [16:16] it won't be the worst offender, but it'll crowd out part of a language pack or so [16:16] always tradeoffs [16:53] hi all, i'm trying to build an iso with ubuntu-cdimage. i need to find out if $(BASEDIR)/tasks/auto/$(IMAGE_TYPE)/$(PROJECT)/$(DIST)/MASTER something creates this or should i create this file manually [16:58] rbelem: update-tasks creates that [16:59] rbelem: well, update-tasks copies it into place - the file itself is written by make-master-task, called by germinate-to-tasks [17:02] cjwatson, nice! Are there other commands should i run before build-image-set? [17:03] build-image-set should already call those things for you ... [17:03] and it should be fine on its own although we generally run it via one of the cron.* wrappers [17:04] cjwatson, hum... i will check what i'm doing wrong [17:04] cjwatson, thanks for your help :-) [17:35] mpt: Regarding your most recent comment on bug 154506, the text as of today's live CD is "Install Ubuntu 9.10". Would you suggest I change this to "Install Ubuntu 9.10 Permanently"? [17:35] Launchpad bug 154506 in hundredpapercuts "Ubuntu LiveCD "Install" icon confusing" [Low,Confirmed] https://launchpad.net/bugs/154506 [17:36] evand, wellllll, it occurred to me a couple of years later that "permanently" is a strong word :-) [17:36] heh [17:36] so I thought maybe "Install Ubuntu on this computer" [17:37] the tricky thing about that is that A) it's really long and B) you might be installing to a USB disk [17:37] you can do that? [17:37] yarp [17:37] ((B)) [17:37] Why wouldn't you be using usb-creator to do that? [17:38] they don't do exactly the same thing [17:39] usb-creator puts the live media on a usb disk. You can have persistent storage, but as a file that serves as a copy-on-write layer. [17:40] installing via ubiquity means that all writes go straight to the filesystem, there is no pristine copy of ubuntu with a layer of changes on top of it like there is with the live media [17:41] What's "a copy-on-write layer"? [17:42] persistent storage (i.e. doesn't go away on reboot). It's OK for a while and for quick demonstrations and the like, but it has some weird properties so you wouldn't want to use it long-term [17:43] in particular if you upgrade the system you will run out of space on the stick eventually - when upgrading system files it never frees the storage used for them at the start of the stick [17:43] huh [17:43] I had no idea that using usb-creator was not the recommended way of setting up a USB installation you plan to work from. [17:44] usb-creator is best thought of as taking a live CD and putting it on a USB stick [17:44] ubiquity: superm1 * r3321 ubiquity/debian/ (oem-config-gtk.postrm oem-config-gtk.preinst changelog): [17:44] ubiquity: Divert the ubiquity-gtkui.desktop file when oem-config is installed as [17:44] ubiquity: oem-config now depends on ubiquity. [17:44] name change ideas welcome :) === dpm is now known as dpm-afk [17:44] I'm not convinced people care about treating a USB disk like a CD where the data goes away at the end [17:45] writing to a USB disk is pretty simple. If they mess it up, they're probably willing to run the utility to write an image to it again [17:45] evand, so would it be possible to make usb-creator set up Ubuntu on the USB device in the same way that Ubiquity does? [17:45] superm1: eww. I wonder if we can do better than that. [17:45] superm1: (and can we have consistent formatting in scripts please?) [17:45] Or is that effectively equivalent to just using Ubiquity? [17:46] that [17:46] cjwatson, oh my mistake on the formatting. bad copy paste job. will clean up [17:46] superm1: I assume you're trying to arrange for ubiquity not to be visible in the applications menu at the configuration stage [17:46] evand, ok, usb-creator perhaps should include a one-sentence disclaimer about that under the radio button for storing your data on the USB disk [17:47] superm1: maybe the answer is to get round to having oem-config remove itself [17:47] cjwatson, no i was actually referring to the installed system [17:47] superm1: though I agree the diversions are ok for the moment [17:47] mpt: though we could preseed some assumptions, like the choices available for formatting the disk (just blow it away, people don't dual boot on usb disks, for example). [17:47] cjwatson, but that is the proper solution i agree [17:47] evand, I can't suggest a wording for that right now, partly because I don't fully understand the issue, and partly because I have a headache [17:47] superm1: what's the distinction between installed system and configuration stage? [17:48] I meant configuration stage as in when the end user gets oem-config presented to them [17:48] cjwatson, oh i suppose they are identical, you're right. at first i thought you meant configuration stage as when the system was getting installed [17:48] superm1: hmm, I'm not sure the semantics of that diversion are quite right - should it remove the diversion on upgrade too? [17:48] * cjwatson squints at policy [17:49] that would be version dependent I suppose [17:49] mpt, cjwatson: "In time this space will run out and further changes will no longer be saved." ? [17:49] IE you want the diversion to go away at version 2.0.0 or so, then you'll remove it on upgrade with that version [17:49] well, no, it's better for the previous version's postrm to do it [17:49] err will not be saved [17:50] superm1: any reason not to do the diversions unconditionally in preinst/postrm? [17:50] evand, I have the niggling feeling that the sheer variety of ways of installing Ubuntu makes the average process more difficult than it could be [17:50] but anyway [17:50] Ian told me once that the maintainer scripts were designed such that common cases could usually be run unconditionally [17:50] mpt: that's been an increasing concern of mine [17:50] fortunately colinux didn't happen, but the list is still too long [17:50] cjwatson, no particular reason I can think of. just thinking back and i've always seen it ran only in case statements [17:51] evand, so is it that the your-stuff area is write-only, space used by deleted files isn't reallocated? [17:51] (on USB sticks with usb-creator) [17:51] mpt: that's my understanding. [17:51] mm, a lot of people overconditionalise maintscripts, and hardly anyone gets the rollback cases right [17:51] let's see, preinst install/upgrade you clearly want to add the diversion [17:52] preinst abort-upgrade is error unwind from postrm upgrade so needs to undo whatever postrm upgrade does [17:53] or abort-install in that case too i suppose [17:53] no preinst abort-install [17:53] postrm upgrade goes after preinst upgrade, so ok that shouldn't remove the diversion [17:54] oh i was thinking of postrm abort-install. where's that cheat sheet at... oh yeah http://women.debian.org/wiki/English/MaintainerScripts [17:55] hmm. ok. I think you're right that it has to be conditional. [17:55] unfortunate, that does indeed mean it'll need a version guard to remove it later [17:55] which is sort of a shame because it never goes away [17:55] oh well [17:56] ubiquity: superm1 * r3322 ubiquity/debian/oem-config-gtk.postrm: clean up formatting from previous commit [17:56] evand, so maybe something like "Deleting files from this area will not increase space available." [17:56] er, "space available" -> "available space" [17:57] (reinsert headache disclaimer here) [17:57] deleting files from the persistent area will increase available space [17:58] Then what doesn't? [17:58] replacing files in the non-persistent area with files in the persistent area will decrease available space more than intuition would predict [17:58] deleting files in the non-persistent area will not increase available space [17:58] ah, so this is mainly about updates and installing new packages? [17:58] mostly, yeah [17:59] Whereas if you choose "Discarded on shutdown..." you can't install updates or new packages at all? [17:59] or, you can, but they get discarded on shutdown too? [18:01] If so, that makes this much simpler [18:04] evand? [18:05] right, the latter [18:05] ok! [18:05] turns out people do want to upgrade the live USB stick quite often, we found out about this when a casper bug broke kernel upgrades a bit ... [18:06] evand, so I suggest changing "documents and settings will be:" to "documents, settings, and new or updated software will be:" [18:06] And now I'm going home, I'll read scrollback tomorrow if you have questions [18:07] hm, actually, no, I'm taking this notebook home with me, but I'll be online in a couple of hours or so [18:08] sorry, was on the phone [18:11] cjwatson, i was debugging to discover where the problem was and i found out that the seeds list are not being populated [18:12] cjwatson, i made this change http://paste.ubuntu.com/219926/ to exit when this is the case [18:13] cjwatson, do you have any clues about why this is happening === robbiew is now known as robbiew-afk [19:05] rbelem: dunno, run list-seeds under sh -x maybe? [19:27] cjwatson, in fact the problem is not in list-seeds but in germinate. it is generating the structure file without entries on the installer. [19:28] might depend on what seeds you point it at ... [19:28] that stuff has been stable for us for a long time [19:29] i'm using ubuntu.karmic and platform.karmic [19:30] can you pastebin the entire contents of the structure file it generates? or is it literally empty? [19:36] cjwatson, http://paste.ubuntu.com/219980/ [19:36] well that looks fine [19:36] installer isn't supposed to have any dependencies [19:37] what does 'list-seeds /path/to/that/structure/file installer' say? [19:38] cjwatson, it is returning nothing [19:38] oh, I know [19:38] you aren't using one of those cron.* wrappers as I suggested, and nor are you setting any of the environment variables they set [19:39] if you want to produce an install CD (text installer), you need to set CDIMAGE_INSTALL=1 in the environment [19:40] i'm running the following line [19:40] LOCAL_SEEDS=file:///home/rodrigo/devel/ubuntu/seeds/ CDIMAGE_ROOT=`pwd` PROJECT=ubuntu CAPPROJECT=Ubuntu DIST=karmic ARCHES=i386 CDIMAGE_NOSYNC=1 IMAGE_TYPE=daily build-image-set daily [19:40] you should use 'for-project ubuntu build-image-set daily' rather than setting PROJECT and CAPPROJECT; and add CDIMAGE_INSTALL=1 to that [19:41] oh and you don't need to set IMAGE_TYPE=daily, the fact that you're running 'build-image-set daily' implies that [19:41] you could just use cron.daily, that'd be easier [19:42] LOCAL_SEEDS=file:///home/rodrigo/devel/ubuntu/seeds/ CDIMAGE_ROOT=`pwd` DIST=karmic ARCHES=i386 CDIMAGE_NOSYNC=1 for-project ubuntu cron.daily [19:42] cjwatson, nice! :-) i will try this right now [19:42] :-) [19:45] cjwatson, it is working! \o/ [19:47] good [19:49] cjwatson, i'm writing a script using debmirror to reduce the disk usage. Do you think this might be interesting to add to the mainline? [19:52] sure, potentially [19:54] cjwatson, neat! i will finish it and put it in my branch === robbiew-afk is now known as robbiew [20:57] partman-base: cjwatson * r162 ubuntu/ (7 files in 3 dirs): merge from Debian 132 [20:57] Error: Could not parse data returned by Debian: HTTP Error 404: No such bug (http://bugs.debian.org/cgi-bin/bugreport.cgi?bug=132;mbox=yes) [20:59] partman-base: cjwatson * r163 ubuntu/debian/changelog: releasing version 132ubuntu1 [21:17] partman-target: cjwatson * r769 ubuntu/ (16 files in 6 dirs): merge from Debian 61 [21:38] partman-target: cjwatson * r770 ubuntu/debian/changelog: merge from Debian 62 [21:41] partman-target: cjwatson * r771 ubuntu/debian/changelog: releasing version 62ubuntu1 [22:06] Hey there, can you guys tell me how "pluggable" Ubiquity is as it stands? [22:09] As in, can I write a plugin to support different installation profiles? Does it support tours already? [23:10] svenstaro, It is but not as "pluggable" as plain debian-installer [23:34] partman-lvm: cjwatson * r1221 ubuntu/lib/lvm-base.sh: honour partman locking when checking freeness [23:35] partman-md: cjwatson * r935 auto-setup/lib/md-base.sh: honour partman locking when checking freeness [23:38] partman-lvm: cjwatson * r1222 ubuntu/lib/lvm-base.sh: mkdir -p in pv_prepare, just to be safe [23:38] partman-md: cjwatson * r936 auto-setup/lib/md-base.sh: mkdir -p in md_prepare, just to be safe