=== superm1|away is now known as superm1 [09:29] mpt: Still working on it, but how is this so far? (http://evalicious.com/tmp/usb-creator.png) [09:29] the slider will definitely get a human readable value on the right [09:31] evand, wow, nice work [09:31] The listboxes should have a border [09:32] "Other…" probably should be on the left rather than the right (if only to reduce the odd gap above the "USB disk to use:" label) [09:33] noted. Yeah, the gap is a bug in gtk, unfortunately. [09:33] http://bugzilla.gnome.org/show_bug.cgi?id=101968 [09:34] ah, yes, that famous bug [09:34] but no, I meant above the "USB disk to use:" label, not above the source listbox [09:34] ah, whoops [09:36] "Other…" should have 12px inside horizontal padding (i.e. between the left/right of the label and the left/right edge of the button) [09:36] (and Glade should do that by default, dagnammit) [09:38] heh [09:41] Ideally "How" should be aligned exactly with "Stored", but that's currently practically impossible in GTK [09:41] yeah, I tried to do that as best I could [09:41] but "How" should at least not be to the *left* of "Stored" in Ubuntu's default theme [09:42] but packing it into the option button posed two problems [09:42] one, it doesn't fully fill the remaining space to the right (which I ran into with another spec in the previous cycle) [09:42] and two, the button then becomes centered on the combined label ("Stored in reserved") and slider [09:43] cjwatson, evand, please have a look at 268123 [09:43] yeah, ugly [09:43] it's a blocker for wubi and cking is out of ideas [09:43] evand, System > Preferences > Appearance > Visual Effects has the same problem [09:45] evand, looks like "Make Startup Disk" has more margin to its right than the rest of the window does [09:46] hrm, well spotted [09:46] And finally, "Size" and "Free Space" should be the same width (though I guess that's probably also really difficult) [09:47] I can always give it a shot and make a note of it if I don't find a solution for that in time [09:48] thanks for the pointers. I'm going to take a short nap then get back to finishing these hopefully before the UI freeze cut off time, whenever that may be. [09:48] do let me know if anything else comes to mind [09:48] either here or in email [09:49] xivulon: no idea without digging deeply into it, and I unfortunately don't have the time for that right now with the impending freeze [09:49] evand, what happens when either of those lists has more than one item? [09:49] evand: I'm preseeding ubiquity, and am encountering a loop when partitioning. Would you have time to review/troubleshoot a preseed file? http://paste.ubuntu.com/45684/ [09:50] evand, does the window resize, or does the list get a teeny tiny scrollbar? [09:50] xivulon: OK, I think I've got it [09:50] cjwatson you are a * [09:50] mpt: I was going to pack a scrollbar around them. Thoughts? [09:51] I don't understand this GTK idea of a scrollbar being an independent entity [09:51] makes my brain hurt [09:51] persia: select_disk looks wrong, but don't quote me on that. Unfortunately I quite need to get to bed. [09:51] but yes, they should have vertical scrollbars :-) [09:52] heh [09:52] persia: if cjwatson doesn't reply before I come back I'll take a closer look then [09:52] evand: OK. Thanks. select_disk came from debconf-get-selections on the livecd, but that doesn't make it right :) [09:52] you should never preseed partman-auto/select_disk [09:52] start out with the installation guide rather than with debconf-get-selections [09:53] OK. [09:53] d-i partman-auto/init_automatically_partition select some_device [09:53] d-i partman-auto/disk string /dev/sdb [09:54] d-i partman-auto/method string regular [09:54] d-i partman-auto/choose_recipe select atomic [09:54] and in general everything else should have d-i in the first field rather than console-setup or ubiquity (subtle debconf ownership stuff) [09:55] Oops :) [09:56] one of these days I'll write some kind of Pure Magic heuristic tool that figures out the preseed file you actually wanted to write [09:56] oh, mpt before I depart. Is disabling rows (which I need to figure out how exactly to do in a treeview) where the CD will never fit and adding a warning icon to rows where it might sufficiently self-explanitory, or should I add (dare I suggest) a tooltip or some other aid? [09:57] cjwatson: Thank you. [09:57] xivulon: uploaded; *should* work ... [09:57] cjwatson: kickstart ;) [09:58] heh [09:58] doesnae work with ubiquity though [09:58] cjwatson thanks, cannot test now though, davmor2 can you please regenerate the initrd and see if it works? [09:58] evand, ah, I didn't make a complete suggestion for that, did I [09:59] davmor2: err, wait a couple of hours for the new busybox-initramfs to arrive [09:59] hmm [10:01] evand, I think a line of text underneath the USB disk list, describing the situation for the selected disk: [10:02] " has enough free space for ." [10:02] " contains too many files to fit alongside." [10:02] " is too small for ." [10:05] Actually, better for the second case: [10:06] " is too full to fit (N MB more needed). ( Open )" [10:06] Hi guys not going to be around much today I can probably test it tonight or tomorrow at some point [10:06] (where "Open" opens the disk in Nautilus) [10:07] #ubuntu-testing [10:07] mpt: on bug 254652, would "Installation Failed (Faulty CD)" work? [10:08] mpt: I'd like to put the "damnit it wasn't the installer's fault" up top, if possible [10:09] cjwatson, as long as the reason is prominent in the window, I don't think it needs to be wedged into the title too [10:09] hm, ok, I was just concerned that people wouldn't bother reading the more verbose text [10:10] but if you're sure, ok [10:10] In alerts I think people are more likely to read the primary text than the window title [10:10] (to the point where most alerts shouldn't bother with titles, unless they have no parent window) [10:11] I'll make the primary text a little more verbose as well; "This particular error is due to" -> "This is due to" [10:11] little LESS verbose [10:12] great [10:16] ubiquity: cjwatson * r2809 ubiquity/debian/ (81 files in 2 dirs): [10:16] ubiquity: Fix window title of bulk copy failure alerts, and remove a bit of [10:16] ubiquity: unnecessary verbosity from the primary text (LP: #254652). [10:58] I have a two-line patch for partman-partitioning/lib/disk-label.sh Where's the best place to try to push it for review? [11:01] branch of lp:~ubuntu-core-dev/partman-partitioning/ubuntu, or attached to a bug if you can't be bothered :) [11:01] I don't mind branching. [11:12] I'm also getting a repeatable crash of partman, even with the now fixed preseed. I'm not finding anything informative in /var/log/installer/debug, /var/log/partman or /var/log/syslog. When running ubiquity under strace -f, I am unable to find the reported exit value of 10. [11:13] Branch with 2-line patch for review is lp:~persia/partman-partitioning/ubuntu [11:21] partman behaves oddly if partially preseeded. Are you using ubiquity --debug? (/var/log/installer/debug isn't much use without that.) [11:22] that partman-partitioning branch is fine; I'll merge it [11:22] I was using just "-d". Does it need the entire word? [11:23] -d is fine too [11:23] can I see the debug file? [11:23] Yeah, I was afraid my problem was with preseeding :( [11:24] though cf. bug 268593, there might be a more recently-introduced bug than the one I'm thinking of [11:24] Take me a bit to pull it out: it's on a KVM instance with no functional mouse :) [11:24] scp to the host? [11:25] Actually, bug 268593 looks exactly like what I'm finding. [11:26] you might try not preseeding partman-auto/init_automatically_partition [11:26] it shouldn't be needed if you've preseeded partman-auto/disk and partman-auto/method [11:26] (I don't think ...) [11:26] partman-partitioning uploaded, though CIA is sulking and refuses to report it [11:26] I'll give that a shot when I've extracted the log. [11:27] Thanks. The next alternate CD might actually work (assuming it builds today) [11:31] /var/log/installer/debug is http://paste.ubuntu.com/45721/ [11:31] partman-partitioning: cjwatson * r683 ubuntu/debian/changelog: releasing version 59ubuntu5 [11:31] partman-partitioning: cjwatson * r682 ubuntu/ (debian/changelog lib/disk-label.sh): merge from lp:~persia/partman-partitioning/ubuntu [11:32] http://paste.ubuntu.com/45723/ is the associated /var/log/partman [11:33] oh, not quite the same thing [11:34] basically the problem is that ubiquity's state machine drove partman the wrong way and ended up backing up all the way out [11:34] did you interact with partman at all, or was it entirely automatic? [11:34] from the timestamps I'm guessing the latter unless you were super-quick [11:34] entirely automatic: X doesn't recognise the KVM keyboard or mouse on the image I'm using. [11:35] (which makes it a little tricky to "Try Again") [11:35] usually the process for debugging this kind of thing is to bring up two windows, one with partman and one with ubiquity's partman state machine, and do a tedious manual trace [11:37] Oh well. I was hoping for a silver bullet. [11:38] I'll dig through by hand then. Running `sudo partman` generates some errors that udpkg isn't found, and one that apt-install isn't found, but formats the disk as requested and mounts it to both swap and /target [11:40] oh, yeah, you can't realistically run partman from a shell [11:40] I wouldn't bother with that, it'll be too different [11:40] I was just happy it worked : it confirmed for me that the patch you recently merged was sufficient. [11:41] But yes, it doesn't really help debug ubiquity. [11:47] Hm. OK, the problem is that when ubiquity is in automatic mode it immediately responds to all debconf INPUT commands for questions that are marked as seen with 30 question skipped. [11:47] This is done at a very low level, and is across the board. [11:48] However, ubiquity's partman state machine relies on being able to go back and forward a lot, which involves it being informed about INPUT commands for questions that are marked as seen. [11:49] --automatic defeats this right now and so everything breaks. [11:49] Essentially, partman doesn't work with --automatic then? [11:50] Right. This might not be too hard to fix, though; all ubiquity should need to do is explicitly mark questions as unseen when it wants to be notified of them again. [11:50] What if I preseeded *every* internal value set during the desktop install? [11:51] You can't do that with partman [11:51] Oh, that doesn't sound so bad. Is that intrusive enough to warrant delay, or something that could be done sooner? [11:51] I was afraid of that. Perhaps the best solution is to go fix X :) [11:52] Preseeding only works with straight-through paths, and partman repeatedly asks the same questions with different parameters [11:52] That's why there's this completely separate partman-auto/{disk,method} approach for preseeding [11:52] Oh, so partman-auto is intended to inform partman without blocking the queries? [11:53] Ish [11:53] It bypasses all the interactive stuff [11:54] But ubiquity does its own scanning in order to be able to build a UI [11:54] Of course, that isn't necessary if partman is preseeded, so we could also make ubiquity check for that ... [11:54] I have to go and do some other things, but I'm happy to work on this. It's a generic bug, not lpia-specific or anything [11:55] OK. No rush, as I only encounter it trying to work around something else :) [11:55] Please let me know if I can help test, but I'll admit to only limited understanding of the ubiquity and d-i code. [11:55] All this ubiquity/partman state machine stuff is my fault anyway [11:58] Well, you're also the person to thank for heaps of it working as well :) [12:10] Hmm, of course part of the problem here is that questions are getting marked as seen even though they were purely internal and never shown to the user [12:10] that really can't be helping [12:39] OK. With X hacked to work a bit better, I can now reliably reproduce something more akin to bug 268593, at least for /var/log/installer/debug and /var/log/partman [15:07] cjwatson1: do you think it's possible to build an i386 and an amd64 cd with extra repository where packages are in the same directory ? must tweak apt-ftparchive no ? [15:22] cjwatson1: you use dak or reprepro for building archives ? [15:25] should be perfectly possible === cjwatson1 is now known as cjwatson [15:25] Ubuntu as a whole uses Soyuz; cdimage uses apt-ftparchive [15:30] cjwatson: ok, thanks, will search how... :) === superm1 is now known as superm1|away === superm1|away is now known as superm1 [16:34] cjwatson: right I'm back home now is it possible to test wubi now? [16:35] you'd need a CD for an actual wubi test, but it ought to be possible to upgrade busybox-initramfs, run 'sudo update-initramfs -u', and see if it's possible to mount an ntfs partition from the initramfs [16:35] I can't help you with the procedure right now though [16:35] if you need a CD, that should happen tomorrow [16:37] cjwatson: cool does that mean the landscape issue is resolved now? [16:43] by means of removing landscape-client from ubuntu-desktop for the moment, yes [16:51] cjwatson: Ah cool. I'll leave for in the morning then I think as I believe it will dl the iso if the cd isn't available anway [17:44] evand: you have a pile of new bugs from me; 269044, 269048, and 269050 should be low-hanging fruit [17:44] (at least) [17:48] cjwatson: much appreciated. Gave a quick look over and I'll implement/integrate fixes once I'm done working in mpt's suggestions. [17:48] I'm just trying to get to the point where successful runs are actually possible [17:48] weird, it's been working fine for me. [17:48] but that's probably to be expected [17:49] I don't see how install.py could possibly work [17:50] indeed, I'm quite surprised given your bug report, but I've seen it happen. [17:50] you aren't being fooled due to bug 269035? [17:51] hrm, possibly. Still, I've watched it fill the disk. [17:52] oh, it fills the disk OK [17:52] the failures in install.py are after bulk copying [17:53] ahhh [18:51] hey guys, what's the traditional way of getting important early boot messages to either a) be splash-printed, or b) kill splash so that the console shows? [18:51] cf: https://bugs.launchpad.net/ubuntu/+source/initramfs-tools/+bug/268873 [18:52] just a pointer to some other code would suffice..... [18:54] usplash has a TEXT-URGENT message type; see usplash_write(8) [18:55] cjwatson: thanks [19:13] cjwatson: it's not immediately obvious to me as to how I might "usplash_read()", in that I have a question which asks, "Do you wish to start the degraded RAID?" which times out after 15 seconds [19:13] cjwatson: if usplash is enabled, that's not shown, and the timeout happens [19:14] cjwatson: i see how I could use usplash_write()... but not how I might grab that y/n input [19:14] cjwatson: would it be reasonable for me to just kill usplash if running before displaying that text and prompting? [19:17] there's an INPUTTIMEOUT command [19:18] scripts/casper-functions in casper has an example of how to use the INPUT* family [19:18] please don't kill usplash [19:19] if [ -x /sbin/usplash_write ]; then [19:19] /sbin/usplash_write "INPUTENTER $@" [19:19] read nunya < /dev/.initramfs/usplash_outfifo [19:19] fi [19:19] though with a different command and actually doing something with the input [19:27] cjwatson: thanks! [19:34] * evand thinks he's going to go blind staring at d-feet's supposedly pretty printed output. [21:38] persia: I'm fixing the lpia CD build failure; it's because the kernel team yet again forgot to tell me about an ABI bump. [21:38] I have complained about this before [22:02] cjwatson, perhaps make it the the responsibility of whomever uploads an ABI bump to update platform.intrepid in the future? [22:03] cjwatson: solved the usplash problem, with a pointer from kees... there was a missing chvt(1) call in initramfs-tools when calling the failure hooks [22:04] cjwatson: one-line patch attached to https://bugs.launchpad.net/ubuntu/+source/initramfs-tools/+bug/268873 ... kees may sponsor for me [23:57] cjwatson, re 268123 partial progress: [23:57] mount -t ntfs -> no such device [23:57] mount.ntfs -> OK