[12:44] cjwatson, sure i'd be glad to, how do i revert the revision? [12:56] ubiquity: superm1 * r2256 ubiquity/ (4 files in 4 dirs): revert non critical changes from r2255 until after beta === avoine [n=avoine@modemcable003.251-59-74.mc.videotron.ca] has joined #ubuntu-installer === avoine [n=avoine@modemcable003.251-59-74.mc.videotron.ca] has left #ubuntu-installer [] === superm1 [n=superm1@ubuntu/member/superm1] has joined #ubuntu-installer === Starting logfile irclogs/ubuntu-installer.log === ubuntulog [i=ubuntulo@ubuntu/bot/ubuntulog] has joined #ubuntu-installer === Topic for #ubuntu-installer: Development of d-i and ubiquity in Ubuntu | http://wiki.ubuntu.com/InstallerDevelopment === Topic (#ubuntu-installer): set by cjwatson at Fri Jan 5 15:12:40 2007 === spam [n=smp@203.129.255.178] has joined #ubuntu-installer [08:57] hi there.... [09:01] i have using ubuntu installer ..and doing some study on this.Can anybody tell me how does it check internally the distribution name and version name so that if proper repository address has not given it shows error.."invalid repo link" or "release file missing' [09:02] during mirror selection stage at the installation process [09:09] is there anybody... [09:36] spam: could you tell me the exact message, please? Don't paraphrase [09:36] spam: at any rate, it looks in the Release file on the mirror [09:37] which has various metadata like that [10:33] cjwatson..:"the specified ubuntu archive mirror is either not available,or doesnot have a valid Release file on it.Please try a different mirror" [10:36] Cjwatson: r u there..? i was out of system....but i want to know what udeb does this checking during mirror selection stage.I checked one udeb is there"choose-mirror" is it the only mirror which checks the release file on the mirror? [10:48] spam: yes, choose-mirror (and choose-mirror-bin, which comes from the same source package) is the one you're looking for [10:49] spam: please don't do the "are you there" thing, by the way - I'm extremely busy this morning and it may take me a while to respond [10:52] cjwatson:actually my problem is some what different. I have created one local repository and from there i am installting my softwares..no problem i am facing but i have given my version name."my-ver".So when installer is checking for version /code name the error is coming. [10:54] so where i need to change exactly..as i changed the string name mentioned in "choose-mirror" pkg source code and after successful comilation i have included that one in the installer and compile it successfully,...no problem came but the same problem persist for choosing mirror. [10:54] do you have dists/*/Release files on your mirror? [10:55] yah...and exactly in the same order.only the string is different [10:55] exactly where is your Release file? what directory? [11:01] is it ok/ [11:02] /my-distro///dists//my-ver//Release [11:02] same what debian repository follows... [11:04] and when it fails i went to shell prompt and from there i tried to connect the specified location using wget ..and i got success [11:04] did you preseed mirror/suite=my-ver? [11:04] or otherwise change the installer's default suite to retrieve [11:05] (e.g. by changing /etc/lsb-release in the initrd) [11:05] the first few line r like.. [11:05] Architectures: i386 [11:05] Codename: anant [11:05] Components: main [11:05] Date: Wed, 29 Aug 2007 10:24:05 UTC [11:05] Description: BOSS Anant Repository [11:05] Label: anant [11:05] Origin: boss [11:05] Suite: unstable [11:05] Version: 2.0 [11:05] MD5Sum: [11:05] that Suite: is bogus [11:05] it should match the subdirectory of dists/ [11:06] that means in this case it should be anant ?? [11:06] if that file is in dists/anant/Release, yes [11:06] you also need to change the installer to look there; it can't guess that it needs to retrieve dists/anant/Release by magic [11:06] changing /etc/lsb-release in the initrd is the easiest way [11:07] I suggest reverting whatever changes you made to choose-mirror, as that shouldn't be necessary [11:07] yah..thats it.. [11:08] ok..but when i am choosing "enter manually.." option for mirror selection then also do i need to modify the d-i? [11:09] "enter information manually" lets you choose the hostname and directory, not usually the suite [11:09] you can avoid modifying the initrd by putting mirror/suite=anant on the kernel command line, if you prefer [11:12] will it be ok for my case if other parameters are same as in debian in my local mirror? [11:14] can i set "mirror/suite=anant" inside default file for PXE net booting installation/ [11:16] default file is in "ubuntu-installer/i386/pxelinux.cfg/default" which is the default configuration file... [11:16] or i am missing something? [11:16] parameters> that's too general a question so I can't be sure. Try it :-) [11:18] yes, you can set it on the appropriate 'append' line in pxelinux.cfg [11:18] yes .i got it. [11:19] thanks..for your help.let me try this one === xivulon [n=c2325681@207.250.49.24] has joined #ubuntu-installer [11:29] cjwatson, evand, had little time this w/e and almost no internet connection, would like to catch up briefly on any progress you have made [11:29] tonight I'll download and test the new image [11:30] aside from what I told you about Friday's work, I have almost entirely been panicking about other things [11:32] ubiquity: cjwatson * r2257 ubiquity/ (configure configure.ac): bump to 1.5.18 [11:35] cjwatson, is grub safe in loopinstallations? [11:35] I'd like to start having more people involved in the testing process, it does not matter if they do not end up with ubuntu installed, but at least the should not find their MBR overwritten [11:36] Did not go that far in the installation process myselef on friday, but noticed that "install grub in hd0" was selected when I got stack at the "ready to install" page [11:40] I haven't got that far in my testing [11:41] np will give it another go tonight [11:41] oh, no, it isn't yet [11:41] it needs a wubi change [11:41] I'll commit that shortly [11:41] I added a preseeding facility which should make it safe, but wubi needs to use it [11:42] are you goint to merge into wubi branch? I can only do that tonight [11:42] To build wubi, remove the nsis fodler, and run "make clean; make prerequisites; make all" [11:43] You only need "make all" once if plugins are cleaned/modified, to rebuild wubi itself, make is enough [11:46] The other question which we left pending was whether Ubiquity can simply skip mounted devices/partitions (except when loopinstalling) [11:46] I've committed it straight to the wubi branch [11:46] thanks [11:47] I'm not going to attempt to build it though, I'm not convinced enough that I can do that correctly [11:47] skip mounted devices/partitions> sorry, I don't know what you mean [11:47] I assume that ubiquity at some stage prepares a list of partitions (and devices)... [11:48] If some of those was already mounted, and we are not in a loopinstallation, would it be possible to skip it? [11:49] I am not sure whether this is already current behaviour [11:49] I still don't quite understand what problem you're trying to solve [11:50] The issue is to go around wubi read-only mode, but would also help when a LiveCD user had the windows partition mounted and then run Ubiquity targeting that partition [11:50] but what is the problem? [11:51] If a partition is mounted it cannot be resized online [11:51] that is, what undesirable thing happens that you want not to happen? [11:51] oh, you mean in the manual partitioning UI? [11:51] So for instance in read-only mode you have /isodevice mounted. I'd still want users to run ubiquity to be able to install on a different partition [11:52] I imagine the partitioner will simply not allow that. I don't think removing it from the manual partitioning UI is a good idea though; it would be very confusing to have only part of the disk listed [11:52] in any event, that's post-beta === cjwatson is focusing brutally on beta at the moment :) [11:52] cjwatson "simply not allow that" is more than good enough [11:53] one problem is that it might allow resizing other partitions [11:53] on the same disk [11:53] or creating/deleting partitions [11:53] none of that will work if a partition is mounted [11:54] Is modifying other partitions a problem? (change in partition table are not well digested maybe)? [11:54] it is not possible to make any changes to the partition table when the disk is locked, which will be the case if anything is mounted from that disk [11:55] well, you can change the partition table, but the kernel will refuse to reread it until you reboot [11:56] Makes sense, even in this case though, Ubiquity has still a use though if the user pre-partitions the drives or already has a (non-mounted) partition to be overwritten. [11:56] Or if the target partition is on a different device [11:56] ubiquity: cjwatson * r2258 ubiquity/ (debian/changelog ubiquity/frontend/gtk_ui.py): [11:56] ubiquity: * GTK frontend: [11:56] ubiquity: - Make release notes button work again (LP: #138365). [11:56] right, it isn't useless, but it does need to be made clear in the UI [12:00] I'd do so that if a device is mounted the user can only pick free/space or an existing partition to install into [12:00] Anyway don't what to distract you too much with that [12:02] free space won't work either [12:02] unless it's on another disk [12:12] Just a list of available (non-mounted) partitions then (but I guess that splitting root and swap would be a problem). [12:13] What about doing the partitioning and then forcing a reboot? [12:20] too horrible for words [12:25] ahah [12:25] I agree [12:45] cjwatson, last try: partprobe [12:45] ? [12:46] http://www.cyberciti.biz/tips/re-read-the-partition-table-without-rebooting-linux-system.html [12:53] xivulon: partprobe isn't magic; it's just a userspace program (and the same libparted calls are already made by the installer). That article is incomplete [12:54] xivulon: if the kernel refuses to reread the partition table because the disk is locked, no userspace program can possibly help [12:59] makes sense [01:08] http://lkml.org/lkml/2006/9/8/46 and http://www.gelato.unsw.edu.au/lxr/source/include/linux/blkpg.h [01:08] could it be that there is some in-kernel support which is simply tapped by userspace toos? [01:09] not as far as I can see [01:09] I would rather you tried it than that you pasted URLs here :-) [01:10] I wish, but have to use a windows machine during the day [01:10] it may be true that you can get away with BLKPG_ADD_PARTITION (which libparted already uses) if that specific partition isn't in use [01:11] it's risky for a partitioner to rely on the exact details though [01:11] particularly if any kind of resizing or insertion of partitions is involved === mpt [n=mpt@121-72-130-17.dsl.telstraclear.net] has joined #ubuntu-installer [03:12] weird, I could have sworn that the linkbutton issue was a GTK+ bug. [03:13] but I suppose I was just not setting it up correctly. [03:13] I think the code fell out during the automation rearrangements; it was there originally [03:13] yikes, sorry [03:13] no worries, caught it === cr3 [n=cr3@modemcable178.77-70-69.static.videotron.ca] has joined #ubuntu-installer === avoine [n=avoine@69.70.0.36] has joined #ubuntu-installer === xivulon [n=c2325681@207.250.49.24] has joined #ubuntu-installer === cr3 [n=cr3@modemcable178.77-70-69.static.videotron.ca] has joined #ubuntu-installer [05:18] yeah i had thought it was a GTK+ bug too, i had another pyGTK app that it randomly showed up in too. i guess i'll have to take a look at your fix for it [05:19] you just need to do gtk.link_button_set_uri_hook(some_function) [05:19] gtk.LinkButton doesn't know how to launch a browser itself because that's pretty system-specific [05:19] so it doesn't by default call /usr/bin/sensible-browser then? [05:19] no [05:20] there's some argument that it should but then your application would still break on other systems ... === superm1 [n=superm1@ubuntu/member/superm1] has joined #ubuntu-installer [05:27] ubiquity: evand * r2259 ubiquity/ (debian/changelog ubiquity/osextras.py): * Fix typo in realpath_recurse (LP: #144460). [05:28] Launchpad Janitor angers me. [05:29] Though I suppose this can be fixed with a capital D [05:45] cjwatson: now that we have ubiquity-dm, should we make that an option in isolinux in Hardy or document it as a kernel cmdline option? [05:46] something like that in hardy, yeah, though in that case it needs to not use --automatic [05:46] indeed [06:01] Is there some magic I can do to get past the partman crash with the latest daily cd and wubi? [06:02] which partman crash? [06:03] ? [06:03] reproducing... [06:15] cjwatson, evand, if it may help with development, at this stage you can copy the ubuntu folder into a VM (ext3) drive and set the VM bootloader use the configfile in ubuntu/install [06:19] personally I don't have time to do anything other than get the standard installation flow working [06:21] "The filesystem ntfs cannot be mounted on /, because it is not a fully-functional Unix file system." [06:21] Clearly I must be doing something wrong [06:21] that only started to appear after I canceled and ran ubiquity manually using -d and --automatic. [06:21] hmm [06:22] what build? [06:22] wubi: the one from friday, ubiquity 1.5.17 [06:22] hmm, that was supposed to be fixed in .17 [06:22] why is it trying to mount an ntfs filesystem on / ? [06:23] oh, unless ... [06:23] evand: can you get ps ax output from when that happens and see if autopartition-loop is there? [06:25] will do [06:27] doesn't appear to be [06:27] hm [06:27] ok, not what I thought then [06:28] can you see what the innermost running partman script is? [06:28] ps axf may help [06:30] check.d/08mountpoint_fat [06:31] evand: ls /var/lib/partman/devices/*/*/mountpoint [06:31] err, maybe not quite that [06:31] grep -H '^/$' /var/lib/partman/devices/*/*/mountpoint [06:32] / [06:32] I'm looking for filenames under /var/lib/partman/devices [06:32] well, /var/lib/partman/devices/=dev=sda/32256-17873533439/mountpoint;/ [06:33] no others? [06:33] nope [06:33] are you following what I'm looking for here? it may be easier for you to take it from here [06:34] no, sorry I don't follow [06:34] partman-auto-loop 0ubuntu9 (included in ubiquity 1.5.17) added 'rm -f $partition_id/mountpoint' because the host NTFS filesystem shouldn't really be mounted on / [06:35] partman-auto-loop temporarily tells partman that that filesystem is mounted on / in order that it can use the mount.d scripts to mount it [06:35] but it's supposed to clean up after itself [06:35] the real / in a wubi install ought to be a loop-mounted filesystem [06:35] I think this is my fault. [06:35] i.e. =dev=loop2 or something not =dev=sda [06:35] oh, it could be that you didn't clean up all the way after cancelling, maybe? [06:35] I ran the install, partman crashed, and I hit cancel instead of continue. This brought me back to the desktop. [06:35] Exactly [06:36] (remember my comments from Friday where I was doing losetup and stuff) [06:36] indeed, I just realized I should have ran that [06:36] taking care of that now [06:37] sorry for the false alarm [06:37] it's certainly harder to clean up than it ought to be [06:37] I think that is a bug, it's just awkward to fix [06:42] cjwatson: have you gotten through a full install yet? [06:50] no :-( [06:50] ubiquity: cjwatson * r2260 ubiquity/ubiquity/osextras.py: fix realpath_root a bit more so that it really returns something prefixed with root [06:50] I got close [06:51] at least to the formatting stage [06:51] but then it decided it had enough of life and hard locked [06:51] sigh [06:51] memory? [06:51] are you asking how much I was giving it? about 768MB [06:52] huh [06:52] I'm going to try again with htop running in the background [06:52] I guess it's unionfs then [06:52] ah [06:52] any updates from pkl? [06:52] he's off ill [06:52] I've asked Ben to take over [06:52] uh oh [06:52] oh, ok [06:53] I was unaware that ben knew the unionfs internals, but I suppose that doesn't matter when you're king kernel hacker [06:53] I don't think he does much [06:53] but pkl gave a relatively detailed description of what he was doing the other day [06:54] ah, good [07:09] ugh, I think you're right about the hard lock being unionfs === avoine [n=avoine@193.231-ppp.3menatwork.com] has joined #ubuntu-installer [08:05] meh, wubi was broken by our CD designation switching from Alpha to Beta [08:05] how fragile [08:10] evand: when xivulon returns, would you ask him to rebuild wubi urgently, and when that's done rebuild the live CDs? [08:10] I've committed the relevant fix [08:12] I'm off for a few hours now [08:13] cjwatson: will do [08:13] thanks [08:13] no problem [08:28] cjwatson: I' [08:28] whoops [08:29] cjwatson: I'm not a RM by any means, but at what point does the beta release get pushed back on account of these unionfs bugs [08:29] or can we go down to the 11th hour on that one === jetsaredi1 [n=jgreenwa@pool-141-149-173-208.bos.east.verizon.net] has joined #ubuntu-installer === jetsaredi1 [n=jgreenwa@pool-141-149-173-208.bos.east.verizon.net] has left #ubuntu-installer [] === superm1_ [n=malimonc@ubuntu/member/superm1] has joined #ubuntu-installer [09:45] cjwatson: I have a fix for checking to see if we're running in automatic mode. Basically, m-a.py sets a flag when it gets handed a question and if that flag is not set when it gets to the point where it would normally dive into the gtk loop, then it doesn't have any questions to ask, so it returns self.succeeded. [09:45] It's pretty much the same thing as checking for automatic, but arguably slightly cleaner === xivulon [n=ago@87-194-85-156.bethere.co.uk] has joined #ubuntu-installer [09:51] though that's basically duplicating what filteredcommand is doing, arr [09:51] xivulon: [09:51] 14:10:31 < cjwatson> evand: when xivulon returns, would you ask him to rebuild wubi urgently, and when that's done rebuild the live CDs? [09:52] 14:10:38 < cjwatson> I've committed the relevant fix [09:53] evand, have you been seeing any weird oddities of the installer stopping at random places on the latest dailies (after the unionfs fix)? I've had random stops at apt configuration, configuring hardware, and configuring packages that have gotten installed. Nothing tangible in the logs though, so I was going to hold off reporting it until I found something in /var/log/syslog or /var/log/installer or dmesg [09:53] superm1_: unionfs is still an issue [09:53] evand, okay so that is likely where its sitting then [09:54] most likely [09:54] for some reason I wasn't experiencing it for a brief period of time [09:54] but it's definitely back [09:54] the kernel team is on it [09:54] i'm almost wondering if they are better off rolling back to the older (functional) unionfs release rather than debug these issues with the beta looming so close [09:54] they can't, though I do not recall why [09:55] ah i see [09:55] evand wubi rev 301 should be up [09:55] xivulon: thanks! [09:56] I'll explain how to rebuild yourself [09:56] It's fairly straightforward [09:56] 1 remove any nsis folder you may have [09:56] 2 make clean [09:56] 3 make prerequisites [09:56] 4 make all [09:57] <- 64-bit, no WINE atm. [09:57] hmm [09:57] but continue, I'll save it for reference [09:57] that's it, if you then edit nsis you only need to run "make" [09:57] neat, thanks [09:57] make all also updates grub4dos, the plugins, and translations [09:58] Which do not need to be rebuilt every time unless you change them or run make clean [10:00] ok [10:00] downloading the iso [10:00] I haven't built a new one yet [10:01] waiting to see if any other changes need to go in from elsewhere in the distro [10:04] Hmm I am only interested in wubi/lupin relevant changes [10:04] Are they in? === avoine [n=avoine@193.231-ppp.3menatwork.com] has left #ubuntu-installer [] [10:21] xivulon: if I understand the release manager correctly, he wants to see ubiquity 1.5.18 on the next CD, so I'm uploading that, letting it build, then triggering a new CD build [10:21] should be a few hours [10:21] What is/isn't in there? [10:23] It looks that rev 2258 are not in, anything relevant there? [10:24] as far as wubi is concerned, no [10:25] ok I am gonna test the iso I have. logging off now [10:25] best of luck -- unionfs is still an issue, so beware [10:25] I'll report back any issue [10:25] thanks === xivulon [n=ago@87-194-85-156.bethere.co.uk] has left #ubuntu-installer [] === superm1__ [n=malimonc@waluigi.student.iastate.edu] has joined #ubuntu-installer === michael_e_brown [n=michael_@adsl-67-65-14-121.dsl.austtx.swbell.net] has joined #ubuntu-installer [11:21] cjwatson, ping === xivulon [n=ubuntu@87-194-85-156.bethere.co.uk] has joined #ubuntu-installer [11:48] cjwatson, I noticed that autopartition-loop does not include changes to make it skip mount.d scripts for mounted partitions [11:49] You mentioned that a few days ago, so I am not really surprised, just pointing out === superm1__ is now known as superm1_ [11:54] anybody here who can answer questions about ubiquity and preseed? [11:54] for gutsy? [11:55] what question? [11:57] so... I am trying to automate gutsy install. [11:57] using live cd as the base. [11:57] and I dont see much documentation on it. [11:57] basically what I have is /dev/sda1 is a Dell Utility partition [11:58] /dev/sda2 has a copy of the live CD copied to it [11:58] and has grub installed to it is bootable [11:58] and that boots and works just fine. [11:58] I can go through the graphical installer and repartition the rest of the disk and everything is wonderful [11:58] but now, I need to totally automate that. [11:58] michael_e_brown: yes [11:59] I tried starting with a preseed, but cant seem to make it do anything. [11:59] the reason that it's not documented very well yet is that it was just developed in this release cycle. [11:59] michael_e_brown: you want to run ubiquity with the --automatic flag [11:59] evand: unionfs> not sure yet, hoping not to have to decide [11:59] ok, how to pass that? [11:59] evand: m-a.py/automatic> let's stick with what we have for beta if it works, and roll that in afterwards [11:59] I have preseed=... on the kernel command line [12:00] superm1_: rolling back unionfs involves backing out lots of apparmor stuff too [12:00] cjwatson: agreed, that's why it's not checked in [12:00] cjwatson, ah that's a bit unfortunate [12:00] michael_e_brown: booting with automatic-ubiquity (and a very current image) should suffice [12:00] I just bzr shelved it [12:00] (as in, 'automatic-ubiquity' on the command line) [12:00] I have daily gutsy from today [12:01] so, just add that to the kernel command line? [12:01] cjwatson: the wubi.exe bug was not reproduceable. Do you want me to trigger a CD build now or wait for ubiquity to make its way into the archive? [12:02] michael_e_brown: yeah [12:02] booting now.. [12:02] michael_e_brown if the preseed is on hard disk you can use find_preseed=/path/to/preseed [12:02] have that. [12:02] evand: which image were you using? [12:02] cjwatson: rev231 [12:02] I have preseed=/cdrom/preseed/dell.seed [12:02] michael_e_brown: use file= [12:02] then you do not find_preseed [12:03] (or preseed/file=, equivalent) [12:03] sweet. Ok. 'automatic-ubiquity' worked. now? [12:03] cjwatson, first time I run wubi I had a crush, but was not able to reproduce that again [12:03] xivulon: I meant which CD image version, not which wubi version [12:03] xivulon: I committed a fix to wubi to make it expect Beta rather than Alpha [12:03] Right now, I have 'preseed=/cdrom/preseed/dell.seed', I need "preseed/file=..." ? [12:03] that was the rebuild I wanted [12:03] that went in fine [12:04] xivulon: you've told us several times how to rebuild wubi ourselves, which is great, but for beta I want to stick with the current build process [12:04] I used the beta CD with wubi looking for beta [12:04] three days before beta is no time to be switching which machine is doing the build [12:04] sure [12:04] cjwatson: that went in fine, but pitti was in the process of building the CD and xivulon found that crash in wubi.exe so I held off on running the build, while at the same time pushing a new ubiquity out as it seemed like pitti wanted that asap [12:05] cjwatson all other times I tried wubi went well, unfortunately I did not have --debug tracing the first time around [12:06] I am writing from latest wubi boot indeed [12:06] 1.5.18 is stuck waiting for approval [12:06] should I just trigger a new cd build instead of waiting? [12:07] evand can you quickly test wubi? [12:07] I do not see any code change that could have triggered that though [12:08] xivulon: yes, what am I looking for [12:10] hmm any time I do tab completion in /var/log/ the shell freezes [12:10] ok, so ubiquity is coming up automatically, but doesnt appear to have read the preseed. [12:10] just a sec... [12:12] xivulon: you seem to have broken CD rom detection