[09:10] StevenK: So we should drop moblin-media [09:10] StevenK: Can you give elisa a go? :-) [09:10] lool: I have just commited the seed changes, replacing it with totem and gthumb [09:10] You're in love with gthumb aren't you [09:10] persia is to blame [09:11] Let me install elisa [09:11] persia: What about eog? [09:11] I'm not a huge fan of gthumb, I just think that f-spot has too many dependencies for intrepid, and look forward to directhex's improvements for jaunty. [09:11] Um [09:11] elisa is *enormous* [09:11] lool: Doesn't have a thumbnail overview function. Can't mount cameras. [09:12] StevenK: Well, it's python, it has many plugins, many python deps, but I'd say the packages themselves are big because of some artwork [09:12] But, but, 88MB [09:12] StevenK: It's a bit like emacs, you can run it fullscreen and replace your whole OS with it! [09:13] lool: That's precisely the behaviour we *don't* want for smaller devices. [09:13] Sure. I'll also upload -default-settings that calls elisa rather than hildon-desktop? [09:13] persia: Ok, I agree f-spot is a bit heavy [09:13] :-P [09:13] (even I prefer it) [09:13] I think elisa is too big [09:13] persia: I don't see the relation [09:14] elisa is sexy, has bling, big icons, is easy to navigate in because it's meant to be possible to use it from your TV [09:14] Big icons == badness on small screens [09:14] No. Big icons == badness on low resolution screens. [09:14] I think the media center design caracteristics make it a good fit for PMP [09:15] None of the icons for Ubuntu Mobile are large enough for my 5" screen. [09:15] persia: Remember it's meant to be displayed full screen [09:15] In fact, at the same resolution, we need *bigger* icons on smaller screens. [09:15] Each icon is displayed on the full screen; it's scalable not big in term of pixmaps size [09:15] It will be big on all screen sizes :) [09:16] No, it will be small on my 3.5" 852x480 screen. [09:16] Just lots of pixels. [09:16] Comparatively to your screen, it will be big [09:16] OK. I'll accept that. [09:17] I think it's too big. But I'll try it [09:19] StevenK: Don't forget it would provide all media playback facilities, import and all, for photos, videos, music, online viewing (youtube flickr etc.) [09:19] lool: How does the import work? Would that conflict with the stuff being done by thunar? Have you tested it on recent daily images? [09:19] StevenK: Make sure you try on a host with OpenGL though; I'd say that's the biggest blocker in our ecosystem [09:20] If it requires GL, then I'd say it isn't suitable currently [09:20] persia: I don't know the work done by thunar; I think elisa will import media on an usb stick when it gets automounted [09:20] Yeah, GL is right out. [09:21] StevenK: It's purely GL [09:21] libdvdread3? Give me a break [09:21] StevenK: Eh how do you intend to play DVDs in your media center! :) [09:21] lool: I'm very leery of adding anything to the seeds that hasn't been tested. Please test it against the images. [09:22] persia: Are you suggesting I don't test stuff I add to the seeds? [09:22] StevenK: And, yes, it's quite easy to attach a DVD player to a MID (I have done so, to boot off the liveCD). [09:22] lool: Are you not suggesting elisa be added to the seeds? Am I confused? [09:22] I'm suggesting StevenK tries out elisa in the eventuality of adding it to the seeds [09:23] Oh. Then I suggest you like to convince other people to test stuff before it gets added to the seeds :) [09:25] TBH with the state of the poulsbo drivers, I don't think we can target elisa as a default [09:26] But GL can't stay a blocker for much longer with the trains of Clutter apps arriving [09:26] Oh, I don't think it's just about the psb drivers. I dislike dependency on GL. [09:26] Yeah, that's all the more reason to not depend on GL for intrepid. It's exceptionally likely that jaunty will require it. [09:29] persia: Can we discuss the upstart session job WRT openvt? [09:30] lool: Sure. What's your question? [09:30] * persia boots today's image for discussion [09:31] (I'm still syncing today's image) [09:31] persia: I'd like to revisit where we left stuff in terms of startx from upstart [09:31] That's fine. We can still talk about it. [09:32] OK. My memory is that the issue was that we needed to start *after* hal, but wanted to start *before* other services in runlevel 2. [09:32] As a result, I generated a sysvinit job that starts the upstart job to work around the race condition. [09:32] The three issues I'm aware of: a) startx failing due to security checks on fd0 being /dev/ttyN, b) startx being racy with other stuff, starting but the Xorg vt not being visble c) Xorg starting before hal [09:33] We added openvt for a), you added some init script for c) [09:33] I'd like to drop openvt when I upload the new xorg [09:34] If the new xorg works around the problem, I'm happy to drop openvt [09:34] I don't quite know the impact this has on b) [09:34] Do people still get Xorg in a non-visible VT currently? [09:34] I don't, but I might just be lucky [09:34] Right. We're solving that with openvt -s right now. [09:35] Well it didn't solve it int he beginning [09:35] Xorg was visible for a split second, and then boom, back to console [09:35] Xorg on VT7 [09:36] I usually get Xorg on VT1, or so it seems based on Alt-arrow changes in the booted system. [09:38] At least it still works merrily in KVM in the current image [09:39] How come the keyboard supports Finland and not France, pff [09:39] Hrm. I think we ought push the new xorg, and then try changing openvt in a variety of environments to see the results. If it works well, we can drop openvt [09:39] Which keyboard? How? [09:44] lool: Whilst you're looking at xorg, do you know why UseFBDevice is hardcoded in xorg.conf for -MID, and missing in -Mobile? Is this some lpia hackery? [09:48] I don't know [09:51] Weird, when I double click the speaker applet in mobile, I get the on screen keyboard [09:52] Actually it just pops up randomly [09:53] For me, once it was opened, it just stayed open every time I wanted to enter text, even when it obscured the place where I wanted to enter the text. [09:54] ogra: I know vertical pixels are expensive, but the title is cut for all apps in the list of tasks; either we shouldn't display text and have bigger icons, or the task list should be at the bottom to have longer titles [09:54] Actually the text could also be below the icon to help somewhat [09:54] persia: What's annoying is that it covers the panel so you can't reclick it from the na [09:54] I'd be just as happy having the text only be available in a hover context hint. On my zaurus, there are only icons, no text. [09:55] lool: Just hit the desktop button in the lower right. That lets you access the menu. The keyboard will then pop up again when you launch an application. [09:56] persia: yeah, that's the solution I came too as well [09:56] I'm not sure it's best, but have no idea how to fix it. [09:58] It took five minutes to NM to get network access pfff [10:00] That's just NM. [10:02] popup windows can't stay without decorations, otherwise they are to be confused with the underlying window [10:03] ogra: Hmm we didn't setup a call yesterday [10:16] lool: I tried the ubuntu-mobile.img daily and it installed fine, but the resulting KVM image won't boot (stops at "Booting from hard disk..." before GRUB) [10:16] mdz: I just kvm started today's ubuntu-mobile.img [10:17] "kvm -m 512 -hda ubuntu-mobile.img" is how I startede [10:17] I'm on amd64 [10:18] Err I just tried again and it's hanging where you say it hangs [10:18] Yesterday's image worked great. I'll try today's again. [10:18] Did I just wait for long or is it racy hmm [10:19] lool: kvm changed in the past few days [10:19] mdz: It worked minutes ago with the same image, same packages, I didn't upgrade /anything/ [10:19] lool: so you have an image which was booting before but is no longer booting, without change? [10:19] hmm [10:19] It doesn't load grub [10:21] ogra: Ping [10:21] I booted an old -mobile image this morning, it had no keyboard/mouse due to known bug, rsynced the last image, it booted and worked fine [10:22] StevenK: It's hard to "double click" on the Q1's touch screen, perhaps kourou should allow launching by single clicking? [10:22] lool: I've noticed that. [10:23] lool: Should I allow both, or just single click? [10:23] StevenK: Perhaps it's best to allow both, I fear if you don't the double click will cause two launches [10:23] lool: It depends which callback gets called [10:24] Hrm. the ubuntu-mobile image seems to have been updated in the last 8 hours. Anyone know why? [10:24] Bah midbrowser doesn't come up [10:24] persia: There was a respin for the ubiquity upload and the langpacks [10:24] persia: I think I mentionned this 12 hours ago here [10:24] lool: There's a bug about that: it seems to have died in the last xul changes. [10:25] persia: Because dailies were generated? [10:25] lool: Yes, but I downloaded ubuntu-mobile about 5 hours ago, and I just discovered that what I have doesn't match what is on the server, and am downloading it again. [10:25] This is more recent than the other stuff. [10:25] StevenK: Also, it's kind of weird that one can select an app as in highlight the entry [10:25] StevenK: But *only* ubuntu-mobile was updated out of the set I track (ubuntu desktop alternate lpia, ubuntu studio alternate i386, ubuntu desktop live i386, ubuntu mid, ubuntu mobile) [10:25] It's like you could highlight it or a group of apps and do something with them [10:26] But Enter, of Shift+Arrows don't do anything to the selection [10:26] You can't, but iconview allows selection [10:26] persia, lool: -mobile is generated at 2:45am BST [10:27] mdz: I suspect kvm/qemu mangles the disk image to allow booting [10:27] mdz: I'm computing the md5sum of the images after rsyncing [10:27] StevenK: There was an update today later in the day. [10:27] mdz: And it doesn't match anymore, but it matched after the rsync [10:27] I'm pretty sure I'll rsync it and it will boot [10:28] mdz: Yup, that worked [10:29] mdz: It's booting again [10:29] 20081002 was published at 0327 BST [10:29] lool: The live image, or the post-install image? [10:29] It would have hit cdimage.u.c approximately five minutes after [10:29] persia: The .img live image from cdimage in kvm [10:29] Hrm. Odd. Perhaps I updated as early as that. [10:29] mdz: And it boots again [10:30] I did launch the installer, I hope it didn't mangle sda when probing [10:30] lool: That's not the described issue. It's the post-install boot. [10:30] apt-get install devscripts debootstrap germinate bzr liburi-perl zsh openssh-client [10:30] Well I got the live image in the same state [10:30] Oops, wrong window [10:31] StevenK: Do you think you could add icons to the list of menus? [10:31] Hurrah. passwd/auto-login preseeding *is* working. [10:31] lool: Icons to the drop downs themselves? [10:31] lool: We don't even have an icon for the Home menu. [10:32] StevenK: yes [10:32] Personally, I'm afraid of the C code that drives that drop-down, given m-b-f [10:32] persia: It's quite easy to find one IMO [10:32] I suspect it neatly surpasses my knowledge of GTK and C, too [10:32] StevenK, oh, sorry, feel free to change the timeout value ... [10:33] ogra: I did. [10:33] ogra: It's now 50, and there is a message, check out the branch on antimony [10:34] i just didnt want it to st there forever [10:34] i dont care about the length [10:34] ogra: Can you please tag and push your changes to ubuntu-mid-default-settings? :-) [10:35] oh, right, probably should also create a branch for the mobile settings package [10:35] persia: the ubiquity "user identity" screen overflows on the Q1 [10:35] lool: Which image? [10:35] persia: rmid [10:35] mid [10:36] OK. Which direction? It works for me at 800x600 under vesa, so I'm a little confused. [10:37] persia: vertically, in French, "step N out of 7" is over the "this name will be used to identify the computer blah" [10:37] Ouhc, python traceback [10:37] Odd. I'll take another look. [10:38] at bootdevice=self.db.get('grub-installed/bootdev') [10:38] Yeah, that's something in grub-installer that I keep getting interrupted from investigating. Unfortunately, the alternate desktop CD has inconsistent packages today, and I'm *not* asking for any respins during beta for lpia desktop alternate. [10:39] Could you -vvv the issue, what's inconsistent? === njpatel is now known as njpatel_away [10:39] Which issue? [10:40] lool, regardnig yourcomplaint about the menu size on mobile, did you test on the Q1 ? [10:40] or on something smaller ? [10:40] * ogra still finds the menu finger friendly [10:41] ogra: I'm about to test on the Q1 [10:41] ogra: But if the size is sufficient in the menu, why not make the panel as small? [10:41] I'm writing the fixed ubuntu-mobile.img now [10:41] i use the panel with my thumb [10:42] ogra: And not the menu? [10:42] if i hold my device with oth hands i usually can reach all four corners with my thumb [10:43] if i do some action in an app or the menu i use my index finger [10:43] Right, so the menu isn't usable with thumb only [10:43] try it [10:43] I will, but there's a logic flaw here :) [10:43] i agree it *could* be bigger but not that it *must* be bigger :) [10:44] ogra: The world wont fall under two small icons; I first noticed because it was visually weird [10:44] i fully agree about the other two points [10:45] but at lest the tasklist one will require patching.... so nothing for intrepid anymore i fear [10:46] mdz: How did you call KVM for the install, and how did you call it for the boot of the result? [10:46] framing the popup windows is just a config change i think [10:46] persia: ah, good question [10:46] in order to get it to boot, I had to specify the .img as -hda [10:46] RIght, and grub got installed on hda. [10:46] and install to a disk image on -hdb [10:47] then swap them [10:47] I thought GRUB was supposed to handle that case correctly these days, though [10:47] should that matter in the times of UUIDs ? [10:47] persia: how do you do it? [10:47] If you want to install for KVM, I think you need to run something like -hda disk.img -hdb ubuntu-mobile.img -boot b [10:48] mdz: I repeatly replicate your results under KVM, but reboot with the same definitions of -hda and -hdb, so have no issues (grub points to hdb, which boots). [10:48] ogra: the menu entries are too far to the right for my thumb [10:48] mdz: On real HW, I remove the install USB key, and grub did the right thing. [10:48] persia: -boot b means boot from floppy [10:48] I thought that was -boot a [10:48] lool, but you can reach the menu icon with it to open the menu [10:48] ogra: So I'd say it doesn't matter much to have big icons in the top level panel; the only useful big icon there is the close one [10:49] ogra: What's the point if I'm going to use my index anyway [10:49] And I need it for the browser or thunderbird too [10:49] Reading the manpage, I have no idea how to boot from the second hard drive. If this was a CD image, there'd be no such issue. [10:49] persia: so is it actually clobbering the boot sector on ubuntu-mobile.img? [10:49] Looks that way to me. Let me try something. [10:49] persia: I think we need a standard recipe for this in the wiki [10:49] lool, well, i'll look if i can do anything about menu size, but i fear that will be a heavy patch [10:50] mdz: Yes, that's precisely what it's doing. If you later try to boot with only -hda to the modified image, it doesn't work. [10:50] mdz: I think you're right. [10:50] ogra: Well as I was saying, there are many apps / menus too [10:50] ogra: Not sure it's a good idea to make the menu entries much bigger [10:50] persia: ideally we would boot from it as a USB device, but I don't know if qemu supports that [10:50] StevenK: What do you think about creating .iso images, and pointing people to usb-creator if they don't have a USB CD drive? [10:51] ogra: Perhaps menu should be slightly bigger and panel smaller? [10:51] persia: -usbdevice disk:ubuntu-mobile.img probably works, but I don't think it has support for booting from USB [10:51] mdz: No, that's currently broken [10:51] mdz: hardy qemu doesn't. I don't have kvm running on intrepid today. [10:51] mdz: Ah you mean fake an USB hard disk, sorry, I thought passing throuhg of USB real devices [10:51] lool, 36px panel ? [10:51] lool: yes. though, pass-through was working for me earlier in intrepid [10:52] * ogra finds that a bit to small [10:52] persia: The same ubiquity screen works fine in mobile; I think we need to fiddle with gtk engines and font size [10:52] Yeah, that's *really* small. [10:52] mdz: In qemu or kvm? [10:52] lool: So you think that it's not something that needs changing in ubiquity? That's good news. [10:53] lool: kvm [10:53] I think it's fixed in qemu, and last time I checked it was fixed in debian-experimental's kvm, but this didn't build in intrepid for me, at this point I stopped trying to use that [10:53] Should be the same really. [10:53] Nah, they're not in sync [10:54] kvm regularly pulls newer qemus, but there's a delay [10:55] Ah, that makes sense. Too bad we're seeing that for release. Still, the fact that one can't boot off USB is likely worse than anything else. [10:55] given that kvm falls back to qemu-style emulation if kernel support isn't available, it seems a bit odd that they need to be separate [10:55] I still think using CD images is the easiest workaround. [10:57] probably so [10:57] but it would be very valuable to be able to test USB images more easily with kvm [10:57] mdz, btw you shuould be able to test it in vbox if you dont get kvm to work [10:58] ogra: vbox can boot from USB? [10:58] mdz: Well they work decently until you try to install them [10:58] mdz, no but vditool can convert .img to .vdi [10:58] lool: They even install properly, it's just that one needs to manually reinstall grub on the resulting drive. [10:59] ogra: How does that help the grub target issue? [10:59] vditool DD ubuntu-mobile.vdi ubuntu-mobile.img [10:59] and [10:59] vditool SHRINK ubuntu-mobile.vdi [10:59] lool: well, they boot anyway :-) [10:59] lool: the web browser doesn't start up [10:59] persia, just testing that [11:00] mdz: Just raised that minutes ago (/lastlog midbrowser) it's caused by recent xulrunner changes it seems [11:00] ogra: ubiquity is looking for the first HD to install grub, which is the sensible behaviour. [11:00] lool: I don't know if it is caused by those, but the timing is similar. [11:00] persia, ah [11:01] ogra: So basically, for the virtual install experience, we need to recreate the "boot off USB" model, where the USB stick is *not* hd(0,0,0). [11:01] GRRR /usr/lib/jvm/java-6-openjdk/bin/apt is an alternative for "apt", I don't want an apt command launching java [11:01] yeah, that wont work in vbox [11:01] at least not in the free edition [11:02] Yeah. Let's make it a CD. It would be nice to be able to test USB images in virtual environments, but the virtual environments don't support that. We could use d-i directly, and have alternate disks, but I like the live images. [11:03] evand's work on usb-creator makes it nearly painless to convert a liveCD into a live USB stick. [11:03] And the virtualisation environments will boot off CD optionally, and then boot off the install HD later. [11:04] Otherwise users are just going to make a mess, and end up with images that boot to Error 21 [11:06] bah, vbow panics anyway [11:06] *vbox [11:06] i thought the kernel tem fixed that [11:06] panics? That's not good. [11:07] There were reports of working vbox with both 32 and 64 bit guests in #ubuntu-testing recently. [11:07] if you add a kernel option to drop virtualization it works [11:08] but apparently the old oops is still there for me otherwise [11:08] I wonder whether we should provide some magic hda image for qemu/kvm which allows chosing the boot device [11:08] or a cdrom (or a floppy :) image [11:08] I'd rather do a CD image. [11:09] Well netbooks don't always have CD [11:09] and UMPC even less so [11:09] There are plenty of people (like me) that have CD drives that can connect to a MID, or have Mobile devices with built-in optical drives. [11:09] So it makes sense to provide people first with USB images [11:09] It's also trivial to convert a CD image into a bootable USB key with usb-craetor. [11:09] Still an additional step [11:10] To help us with virtual testing which is only a development use case [11:10] For real HW, I agree that USB is the way to go, but mdz brings up a useful point that people may want to play with virtualisation, and doing so corrupts the existing images. [11:10] I wouldn't care so much if the problem was just that the result didn't boot, but that it corrupts the local image is a *big* issue to me. [11:10] Yeah, and I propose a bootstrap generic cdrom image to use our usb images [11:10] rather than switching our usb images [11:10] or we could have a cdrom-creator :) [11:11] We don't have cdrom-creator, it's past UI Freeze. [11:11] In July, I'd be inclined to agree with you. Today, I'm not. [11:11] You didn't expect we'd switch from USB to CDROM images before release, did you? [11:12] Actually, yes, I do. It's about 20 lines of code. [11:12] It's trivial to test, doesn't affect translations, etc. [11:12] oh, it didnt panic thist time [11:12] I want to release USB images, that's what people want to use to install on their devices [11:13] persia, and in vbox i actually see what you mean with the panel applets [11:13] Even though it's *trivial* to corrupt an image just because you wanted to try it before using it? How do you think that will affect users' confidence in the results. [11:13] usb-creator works *really* nicely, and does the job. Lots of us who have lots of little devices have a CD drive around. [11:14] * persia has bugs against usb-creator, but minor ones, really. [11:14] usb-creator is an additional step and not everybody has a cd drive [11:14] Yes, but it means that it's difficult to corrupt the image, rather than being trivial. [11:14] It also allows people to play in a KVM environment safely. [11:14] Come on, we're speaking of people running in kvm, not the target *users* [11:15] well, if we could get vditool on antimony converting them for vbox would be trivial [11:15] Plus, we should fix the corruption [11:15] And no, playing in KVM from the live environment *isn't* suitable, because you can't reboot unless you create a special casper-rw filesystem. [11:15] its two additional commands [11:15] lool: How do we fix the corruption. Change the standard behaviour of grub-install? === njpatel_away is now known as njpatel [11:15] Is that *really* safer than changing to CD images? [11:16] ogra: Yes, but I don't want to have multiple images. Trying to explain the multiplicity of images for Hardy was very annoying. [11:17] indeed [11:17] and users using vbox usually have vditool installed locally [11:17] ogra: Also, we want to encourage *installs*, not live environments. There are too many little annoyances in live environments. I *really* like having them, but only for demo purposes. [11:18] mdz: At least the install of mobile worked fine with a fixed image on the Q1 [11:20] * persia goes back to lpia, still entirely certain that CDs are far superior to USB keys [11:21] * lool still believes we shouldn't change the format of the images just before release [11:21] lool: The current format is broken... [11:21] I actually have a good argument for ISO images, but I'll hold it until we release [11:21] persia: Is not [11:22] What's your argument for ISO images. Don't hold back. [11:22] I booted our two images today without running any conversion tool whatsoever and I installed one of them successfully on a device, still no conversion tool whatsoever [11:22] I already gave you the argument in another discussion, but I'm glad you missed it [11:22] No, but you needed to run dd. usb-creator provides a GUI. [11:23] Please share again. digging through logs just takes me longer and distracts me more. [11:23] You have no chance :) [11:24] the install seems to work fine if i boot vbox from secondary master [11:24] ogra: I think it would work fine in kvm/qemu as well if I managed to boot from hdb [11:24] ogra: vbox supports that! Excellent. Any idea how to do it in KVM? [11:24] But this doesn't seem to be possible with -boot [11:24] It's only hard disk [11:24] persia, no kvm user :) [11:24] ogra: How about qemu ? [11:25] lool: No: using the images in KVM is actively dangerous. That's why I want CDs. [11:25] It's not [11:25] they both draw massive ressources on my laptop ... [11:25] ogra: I can't figure out how to do it, and otherwise running the install corrupts the install image. [11:26] lool: It's running one program. Not an uncommon action. [11:26] Especially if someone wants to try different applications in a development environment before committing to their hardware. [11:26] It's running what? [11:26] As a long-time pocket computer user, I'll say that it takes a *lot* for me to want to change anything on my device, for fear of breaking something. [11:27] Running the installer. [11:27] So if any corruption goes on, we should fix it [11:28] * ogra watches the installer in vbox ... [11:28] OK. That happens as a result of the grub-installer code to select the right boot disk for a given environment. Except under virtualisation, hdb is the wrong selection. [11:29] soren has been having lots of trouble trying to determine when something is running under KVM for some X issues, and that's with the full Ubuntu environment. Doing it in busybox is *even harder*. Making changes to grub-installer that don't work in busybox is a bad idea. [11:30] Oh, and it's even the wrong selection under virtualisation *except* when installing off a USB image. [11:30] Sorry, didn't get your comment on grub-installer [11:30] What's the issue with it? [11:30] lool, btw, there is a nice howto on the maemo wiki how to put your system on the internal flashcard intead of the internal rom, i use a 4G sdhc card on it now and its nearly twice as fast [11:30] lool: Which part don't you understand? [11:30] ogra: But you're using the special bootloader? [11:30] yep [11:30] you have to [11:30] Yeah [11:30] but 4G is a very good size :) [11:30] ogra: Didn't think it could make the OS faster, I thought it was the same type of hardware, just soldered on the mobo [11:31] persia: I don't understand what you refer to at all [11:31] lool: OK. Do you understand the problem at all? [11:31] sdhc has up to 10M transferrates, not sure what the internal one has ... thats actually a good task for the weekend, i'll run some hdparm and bonnie tests :) [11:32] /* Allowed boot devices are: [11:32] * a b : floppy disk drives [11:32] * c ... f : IDE disk drives [11:32] Pff manpage out of date [11:32] Indeed :/ [11:32] persia: I'm afraid I don't even know which problem you're talking about [11:32] Sorry :-/ [11:33] lool: The problem is that when one runs the installer afte starting KVM with `kvm -hda ubuntu-mobile.img -hdb disk.img -m 512M` it boots into an environment where the primary hard drive is the USB image we provide (read-write), and the secondary hard drive is the virtual hard drive. [11:34] Yes [11:34] When the installer runs, the grub-installer component attempts to install grub on the default boot drive for the environment. [11:34] grrrr [11:34] In this environment, that happens to be our image. [11:34] * ogra starts over [11:34] persia: grub-installer does that? [11:34] that's awfully broken :-/ [11:34] Yes. [11:34] No. [11:34] 2G wasnt enough on the target device ... and i thought "grow dynamically" means something pfft [11:34] It's *only* broken when running an install from a USB key when claiming it's a hard drive in a virtual environment. [11:35] Oh it does know how to exclude it when it's an USB key, just not when it appears as a hard disk? [11:36] For real environments, yes. In our special KVM environment no, because we told KVM is wasn't a USB drive. [11:36] It would be nice to emulate an USB drive, but I also think the exclusion could use more sensible data such as the uuid of the live system or something to exclude this device [11:36] Really, give usb-creator a try: I think you'll find it a better experience than dd for most users, and it solves the issues. [11:37] It doesn't solve the issue, it just makes us use something else [11:37] That means updating a UUID list in grub-installer for *every* CD spin. Since grub-installer is *never* used directly, but only as part of ubiquity or debian-installer, that way lies madness. [11:38] Right. It makes it difficult to cause the problem, and we have time to maybe fix it in jaunty. [11:38] Currently, it's trivial to cause the problem, which is bad. [11:38] The problem could also very well be avoided by booting on hdb with the image in it [11:38] It's especially bad if we want to get more developers, as the number of people who can play with virtual environments is much larger than the number of people with devices that they are willing to sacrifice for development. [11:39] That's true. I just think it's easier to change how we build images than it is to fix KVM at this point. [11:39] I also think that usb-creator is significantly more user-friendly than dd, and we should recommend it's use. [11:41] Like I said before, in July, I'd agree with you. In December, I'll agree with you. Today, I want CD images. [11:42] Sorry but how can you make a point about easy of development to change the way which will only affect the released images? [11:42] It's not like developers can help us right now [11:42] Everything is frozen [11:43] No, but we can get more developers in November, and having released an image that is easy to use, hard to break, etc. is a good thing. [11:43] This image exists, it can be used from live envs [11:43] We have *no* GUI to convert a USB image into a USB drive (yes, we have usb-imagewriter, but neither of the people who have dug at the code did so enough for intrepid release). [11:43] And we have the KVM issue. [11:44] It just can't be installed to a virtual disk easily because we lack proper detection or boot switches [11:44] And a live environment isn't suitable: a couple of apt-get update; apt-get upgrades breaks it. [11:44] (this is true for any live environment: it has to do with the way casper sets up the environment, and the available space) [11:45] * lool lunch & [11:45] Have fun. [11:46] persia, i will PPA it the next days [11:47] so people dont struggel with dd at least [11:48] just on a sidenote, screenrefresh in vbox with vesa officially sucks [11:48] ogra: Sure, but we have usb-creator in the archives. As much as I can see arguments for usb images, and as much as I like usb-imagewriter, I don't think USB images are the right format for the intrepid release. [11:48] (and that's not a virtualbox issue : I can replicate on real HW) [11:49] persia, ?? [11:49]