[01:34] asac: i think you are in the email loop of my ethernet ping email to fsl [01:34] i CCed you and oliver [01:35] cooloney: hmm. ok [01:35] thanks [01:40] asac: no problem. you work so late, are you still in portland TZ? [01:40] hehe [01:40] no [01:40] off now ;) [01:40] had to fiddle with specs [01:57] asac: good night, [08:23] lool: just dowloaded your zImage in my qemu, it works. [08:23] lool: but my zImage built from latest lucid tree, fails to mount root device [09:01] cooloney: Ah [09:01] cooloney: Did you build it from the lucid source package, or the lucid tip? [09:02] cooloney: Could you try http://people.canonical.com/~apw/misc/arm/linux-image-2.6.32-13-versatile_2.6.32-13.18~masterapw1_armel.deb ? [09:02] cooloney: unpack the .deb with dpkg-deb -x linux-image-2.6.32-13-versatile_2.6.32-13.18~masterapw1_armel.deb linux-versatile/ or something like that [09:02] then use linux-versatile/boot/vmlinuz* [09:03] * apw looks blearily [09:03] apw: Just an incidental strike because you host kernel packages directly [09:04] lool: i think, i messed up my build system. [09:04] cooloney: Could be [09:04] lool: i rebuilt the kernel from latest lucid [09:04] lool: it can find the rootfs now [09:04] lool: thanks [09:05] cooloney, pretty sure i've not changed the tip significantly since that build [09:26] I was thinking we should build daily rootfs-es just like we have daily UEC images [09:28] ogra was talking about having a way to script rootstock to do just that. [09:29] There's still the issue of kernel feature support, etc. but that's separable. [09:50] persia: What issue? [09:50] I was thinking of reusing their image building infrastructure actually [09:50] patching vmbuilder to create armel images [09:51] The only major difference is bootloader on hard disk versus kernel on host [09:51] I'd love to have kexec working to hide this fact completely [09:52] Image build infrastructure also works. The main issue with kernels is that some of our userspace depends on features not available in some kernels. [09:53] By using the distro kernel (or rcn-ee's kernel), one ends up with the features. Otherwise, one may not. [09:53] * persia doesn't know about other kernels) [09:58] persia: Oh yeah; in theory we should be using one in-disto kernel for each image build [09:58] persia: In practice this will only be possible with lucid [09:58] For <= karmic, perhaps we can provide backports? [09:58] I'm not interested in redoing the work on an older kernel, but it would probably be doable to provide backports for karmic in a PPA or simply to use lucid kernel (a bit ugly though) [09:59] Will it be possible with lucid? We're nearing FF, and I haven't seen that stuff land yet. [09:59] I managed to boot karmic userspace with the lucid versatile kernel BTW [09:59] persia: Well it's in the git tree [09:59] So hopefully this will be uploaded soon [09:59] Ooh! That's excellent news. [09:59] Besides, it's not like a versatile is a new feature; we're just fixing it ;-) [09:59] The biggest problem ATM is that lucid glibc breaks gtk+ apps [10:00] heh. [10:00] That certainly has some potential for regression. [10:00] I don't think I will be able to research this; I mailed doko to ask whether he had any advice on this, or on trying out backports / rebuilds of eglibc, but it's a best effort thing [10:00] I found a workaround to the glibc issue, but I don't know whether it's 100% correct [10:01] Either set G_SLICE=always-malloc, or MALLOC_CHECK_=0 [10:01] Setting G_SLICE is safe, I'm less sure about the second one [10:02] It would be nice if someone could chase that bug; it might turn out to be a toolchain, or linux or qemu or glibc bug [10:02] Doesn't that disable checking for available memory when allocating? [10:03] MALLOC_CHECK? [10:03] Yeah. [10:03] Don't think so, I think it disables the sanity checks in glibc which do things like checking around your buffer, or checks for double free etc. [10:03] It's more of a security feature as I see it [10:03] Does Qt also break, or just gtk+? [10:03] Either it's broken, or there is a really dangerous bug, I can't tell [10:04] I don't know about Qt, didn't try [10:04] That's why I would personally start with only setting G_SLICE as a workaround [10:04] Because that just changes the memory allocator of glib with another safe one [10:05] I was just thinking of ways that one might determine whether it's a gtk+ or glibc issue. [10:09] I don't think it's a glib issue because it's only in qemu/versatile [10:09] But I could be wrong [10:09] It might be that it gets the toolchain calls wrong [10:09] Also, it only happens with a particular request size === ogra_ is now known as ogra [10:44] lool, why are you duplicating work ? [10:54] ogra: Sorry? [10:54] We discussed a lot of stuff above, so please be more specific [10:54] * persia suspects it's a rootstock vs. vm-builder things. [10:55] right [10:55] lool, you seem to reimplement exactly what will be in rootstock by A3 [10:56] Err I could argue that rootstock is a duplication of vm-builder [10:56] (Which itself duplicates part of d-i, but that's a bit far fetched) [10:56] lool, well, why not use vmbuilder as a backend then, the plan for rootstock is to actually use it for image building with L+1 [10:56] I was proposing to reuse the vm-builder daily image build infrastructure instead of coming up with a new one for rootstock [10:57] and i'd love (and was hoping) to have your help on the code ... thats why i gave you full commit access to the branch etc [10:57] if you feel vmbuilder is a better backend feel free to change it [10:57] You effectively received help from me before doing that [10:57] starting a second project from scratch feels like a huge waste to me [10:58] The backend now is debootstrap + apt-get? [10:58] yes [10:58] plus oem-config for the system setup [10:58] How is it from scratch?? vm-builder has been around for much longer than rootstock; I mentionned supporting armel even before you started rootstock [10:59] I also discussed the design issues with you at the Dublin sprint [10:59] Why reinventing "set hostname, set password" etc. was not a good reuse of our debconf stuff [10:59] thats why oem-config is used now === JamieBen1ett is now known as JamieBennett === asac_ is now known as asac [17:46] which bootloader do you generally use for ARM? [17:47] whatever bootloader the board needs :) [17:47] it really depends on the board [17:47] mbligh: hey Martin; we use either U-Boot or RedBoot [17:47] mbligh: For dove boards, we only have U-Boot support from the vendor, for imx51 boards we get both RedBoot and U-Boot support with different features [17:48] ok, any idea where the config file lives / what the flashing procedure for a new kernel is? [17:48] grep -r vmlinuz /etc turns up nothing [17:48] mbligh: That depends of the board; where is the kernel stored? [17:48] Is this qualcomm? [17:48] it's in /boot [17:48] mmm. can't say ;-) [17:48] Qualcomm has multiple stages of bootloaders, so it's a tad complex, and there are signatures at various levels [17:49] mbligh: Recent versions of U-Boot and RedBoot can load files such as kernels or boot instructions from a filesystem [17:49] I would expect that's your case, but perhaps only the kernel lives on a fs [17:49] mbligh: Check your BSP docs; there's certainly some info on writing your own kernel [17:50] and surely some doc about the used bootloader [17:51] * ogra hands asac a beer some chips and and https://edge.launchpad.net/ubuntu/+source/mono/2.4.3+dfsg-1ubuntu2/+build/1495864 [17:52] ogra: Does mono actually work properly in the recent versions? [17:52] lool, no idea, it didnt build until some mins ago, the last version we had came from karmic into lucid [17:53] lool: i think a testsuite is run in the build [17:53] and the mcs compiler also is mono ... [17:53] maybe it doesnt work fully, but it probably should work somewhat [17:53] asac: You might recall the weird banshee bugs [17:53] which are still open :) [17:53] no. that was before my time. i wanted to check if it now is better [17:53] i think we might want to disable jit on arm [17:53] --with-jit=no [17:54] i will try that in a ppa if we still have the problems [17:54] * ogra is happy as long as the apps we ship are installable at least ... [17:54] everything beyond is a bonus [17:55] ogra: will you give some love the ftbfs list and give them back for a bit ? [17:55] i have a call and would check afterwards what can be retried [17:55] dmart: get me a mwc2010 pass! [17:55] as soon as its published, yes [17:55] thx [17:55] but publisher will still take a while [17:57] ogra: http://people.canonical.com/~ogra/arm/build-arm-rootfs was an early version of rootstock, right? [17:57] lool, yep [17:57] I'll drop references to it from the wiki; I don't think it's a good idea to still point people at it [17:57] lool, btw, do you think the lucid versatile kernel could work with jaunty qemu ? [17:58] lool, where will you point to ? [17:59] there was no rootstock in jaunty and i havent tested it on jaunty since ages, build-arm-rootfs works for jaunty users [17:59] ogra: it would be nice if you could test the new versatile kernels with jaunty qemu-system-arm [17:59] I will point to rootstock for now [17:59] ok [17:59] It would also be nice to provide backports of rootstock [17:59] i will do so for karmic [18:00] not sure for jaunty since it wasnt available there [18:00] I personally would prefer supporting a single tool with the latest Ubuntu release and then perhaps investing in backports rather than supporting different tools for multiple Ubuntu versions [18:00] What wasn't available in jaunty? [18:00] rootstock [18:00] I was thinking of backports, so we could make anything available anyway [18:00] yeah, thats surely possible [18:01] e.g. backports of debootstrap, rootstock, qemu-kvm, linux [18:01] can rootstock backport? I thought it depended on features not available until karmic. [18:01] persia, the code is that bloated because it still works without qemu-arm-static [18:01] debootstrap is commonly backported, and tends to go through smoothly. linux is almost never backported: that probably needs a PPA. I'm unsure about qemu-kvm. [18:02] qem-system-arm is all that counts [18:02] and that used to work in jaunty [18:02] Concerning linux, it's probably the one we can avoid backporting in any case since it's typically not installed on the system [18:02] the question is if the new versatile will work [18:02] It's more of a companion/helper to the rootfs [18:02] i dont want to keep the different kernels in the script but unify on the archive version [18:03] So just get debootstrap backported (can be very quick if people are willing to test), and then backport rootstock. [18:03] right [18:04] but its essential to know if the kernel works with qemu [18:04] i want to reduce the bloat of the script if possible [18:04] given that all the lucid features add a lot of extra code [18:05] Well, backport the current version to karmic and jaunty, and then trim it down for lucid fowards. [18:05] i wonde if i should drop all the locale handling ... oem-config can care for that [18:07] in case of the debconf frontend i could even drop the tasksel stuff [18:07] not sure about the gtk/qt UIs === jds_ is now known as Hoonse [18:16] hi guys [18:23] Hey Hoonse [18:23] Hi [18:24] is anyone here who set up ubuntu on the beagleboard sucessfully? =) [18:24] several people i think [18:24] There's some instructions on the beagle wiki that are up to date [18:24] * persia hunts [18:24] Hoonse, http://elinux.org/BeagleBoardUbuntu [18:25] Yeah, that's it :) [18:25] there should even be test images [18:25] yeah i read this already but for this you have to compile ubuntu for your own right? [18:25] No. [18:25] no? [18:25] You have to create your own rootfs only. [18:26] But you can do that from precompiled binaries. [18:26] the rootfs is the "folder tree" ? [18:26] Yeah [18:26] roughly, yeah [18:26] the bits and bytes! [18:26] err, that wikipage even has ready made images [18:26] you dont need to compile anything :) [18:26] can i use for example the synaptic packet manager for installing things? [18:26] Some people actually believe the rootfs is the soul of your board [18:27] lol [18:27] "the thing where the magic happens" =) [18:27] Hoonse, yes, you can use synaptic/apt-get or aptitude [18:28] Most stuff should just work. IF something doesn't work, we'd appreciate help getting it fixed :) [18:28] but i think the "normal" programms that you use for the pc arent working. only a few right? [18:28] Anything that works for any other architecture *should* work. [18:28] There are a couple things that just won't, like WINE, but those are exceptions. [18:28] oh my god that would be great... [18:29] That's the point :) [18:29] i think i only need a ftpd a httpd and a java runtime env... [18:29] persia, if lool ever fixes qemu to run on armel you could theoretically run wine in qemu-user *g* [18:30] ogra: That's advanced hackery though. The reason WINE doesn't work is that there's a check in the build scripts that gets mad if it's running on armel. [18:30] i think i will try this now =) the only problem is that my only monitor is plugged into my beagle... [18:30] WINE does the same thing for powerpc, which annoys me, because I used to run windows on a powerpc. [18:30] persia, i men the x86 version of wine :) [18:31] *mean [18:31] ogra: And that's the advanced hackery bit :) [18:31] can i do this via ssh plugged the mmc card into my server?!? [18:31] just some binfmt magic :) [18:32] Hoonse: Most people create the rootfs on their host workstation. If you're just doing a server, you might try installing openssh server, and then logging in over ssh and using DHCP to set the IP, rather than connecting a monitor. [18:32] ogra: The binfmt magic isn't the hard part. The hard part is getting the binary installed in the first place. [18:32] details :P [18:33] heh [18:33] * persia watches the publisher impatiently [18:34] * ogra is eager for the first blog entries from people running IE under wine on armel netbooks *g* [18:34] probably in 12.04 or so :) [18:38] ogra, did you catch the blog/youtube where someone ran max os (powerpc) thru pear on a n900? [18:38] max/ mac... [18:38] no, cool :) [18:39] http://www.engadget.com/2010/01/31/mac-os-x-10-3-installed-on-nokia-n900-via-pearpc-barely-usable/ [18:44] haha, i should try that on mine :) [18:44] lool, you missed to remove a sudo qemu call :) [18:44] * ogra fixes [18:50] OMG ... [18:50] i just ran rootstock -s gdm ... [18:50] And it worked? [18:50] that installs half the desktop [18:50] Well, yeah. [18:50] so i just did the formatting stuff for the mmc card... can i take the "normal" MLO uImage and u-bin files from the angström image or do i have to take "special" files? [18:50] why the heck does gdm pull in nautilus? === jds_ is now known as Hoonse [18:51] u-boot not u-bin sorry [18:52] arent they linked from the wiki ? [18:52] * ogra thought they are [18:53] ogra: gnome-session is the link. [18:53] persia, i thought there was a shrunk down gnome-session [18:53] * persia checks [18:53] gnome-session-core or -bin or so [18:53] Right. gnome-sessoin-bin [18:53] * persia looks again [18:54] I don't see it immediately then. [18:55] i hope my image size is enough ... [18:55] * ogra doesnt want to do that again ... its just to test the logic for installing oem-config-gtk if gdm is installed [18:56] * persia invokes the heavy artillery [18:57] i think i'll forbid ll useradd stuff in rootstock in favor of oem-config [18:57] up to now you can still use --user --password ... [18:57] can anyone tell me a reason why to keep it ? [18:57] (gievn its really insecure) [18:58] ogra, what would oem-config bring up on first boot then? would it require a gui? i'm thinking console installs... [18:58] rcn-ee: There's a debconf front-end [18:58] it brings up a debconf ui [18:59] looks like debian-installer [18:59] but only asks for username, passwd, language and timezone [18:59] okay, that works.... wasn't sure.. ;) [18:59] ogra: I think it's an issue with dependency resolution. I'm *definitely* getting it through gnome-panel->gnome-session->nautilus [18:59] and then removes itself [18:59] * persia blames the lack of versioned provides [18:59] in the wiki there are "usefull packages" listed what should i do with dem?!? [19:00] Hoonse: Install them on your workstation :) [19:00] Hoonse, you would install them thru the '--seed ' [19:00] or afterwards in the booted system [19:00] yeah but i need at least the wifi drivers do get this thing to the www [19:01] the key ones are linux-firmware and wireless-tools for wifi devices... since ubuntu's repo gives us the firmware... [19:01] please excuse my strange english skills... normally i speak german =) [19:03] your skills are fine ... [19:04] lol thanks [19:04] when i spt-get install the xfce4 has this any affect on my PC UBUNTU system?!? i am a little afraid right now =) [19:05] no, it wont touch your host system [19:11] hmmm the rootstock script is running... everything seems awesome =) [19:12] it will take a while, get a beer and some dinner :) [19:12] =) [19:14] when everything is done i want to put the beagle in my roomba vacuumcleaner with a webcam and wifi. prog a java tool or something you can enter via a web browser to controll the roomba... i think i will create a domain like vacmyhouse.com or something so that people can clean my home =) [19:14] heh [19:16] ooh, rcn-ee [19:16] I has kernel config options for you! [19:16] hey cwillu_at_work [19:16] if only my server wasn't wedged right now :p [19:17] ouch... i know that feeling, it's my day off and i'm rotating new harddrives on the beagle farm.. .;) [19:17] hmmm my girlfriend is screaming right now... i think i should go to her before she is making a scene AGAIN... -.- brb when script is finished^^ [19:48] * ogra wonders what hapens in ubuntu-system-service's unpack phase that reliably locks up his VM [19:52] strace it :) [19:52] i cant [19:52] its inside a VM [19:52] And you can't install strace in the vm, and strace dpkg while unpacking? [19:52] i just want to get that codepath tested damned [19:53] that would require any kind of option to get *inside* the VM [19:53] the VM is spawned by rootstock as backend [19:53] i cant easily get to it [19:54] i suspect its the bug that lool just reasearched (and i think fixed too) with libc [19:54] it simply takes 40min to get to that point, that the annoying part ... but i think i'll kill the build now [19:55] Ah. Yeah. It would be painful to untangle that sufficiently. [19:55] holy sh this takes really a while... [19:56] * ogra just goes back to the other kernel and tries again [19:56] Hoonse, yes, under which host release do you run it ? [19:56] i think 9.04 [19:56] yeah, that was really slow [19:57] karmic (9.10) wa a lot faster [19:57] seems so =) [20:09] hmmm i got some warnings... is this normal? [20:11] Hoonse: What sort of warning? [20:11] Setting up wdiff (0.5-18) ... [20:11] install-info: warning: maintainer scripts should not call install-info anymore, [20:11] install-info: warning: this is handled now by a dpkg trigger provided by the [20:11] install-info: warning: install-info package; package wdiff should be updated. [20:11] depends on the warnings :) [20:11] this sort of... [20:12] Hoonse: You can safely ignore those. Those are bugs in the packages being installed. [20:12] great its cleaning up =) [20:12] aaaand done yeah [20:12] If you're feeling especially adventerous, you can try to fix the bugs, but I'd suggest leaving them alone unless they are causing you issues. [20:13] ogra: how is the mono give back going? [20:13] someone needs to take over? whats the status? [20:13] asac, i simply gave back everything [20:14] lest see what sticks to the wall and what falls off [20:14] *lets [20:14] right :-P [20:14] i'll do a second run tomorrow [20:14] do you get emails for give backs? [20:14] hmm. i think so [20:14] yep [20:14] for failures at least [20:14] coool. then be accurate ;) [20:14] talk to you tomorrow [20:14] i'll check the ftbfs list tomorrow morning and we'll see [20:14] One only gets give-back fail emails if one was an uploader. [20:14] ejoy your evening [20:15] persia, well, the ftbfs list will have them [20:15] Indeed. [20:16] most of it still sits in the queue [20:16] gnome-sharp2 is currently building [20:17] ogra: I wasn't done with the wiki page, just went for dinner [20:17] ah [20:31] does anybody know of any ARM CPUs with compression offload or accelleration? [20:34] are these boot fixup for 9.10 to do or not (i am talking from the wiki page =)) [20:35] Hoonse, just 9.10 for the moment... (unless you've installed a backup battery on your c3/c4..) [20:35] lucid (10.04) when released, hopefully won't require the fixup.. ;) [20:35] k then i really should do this... =) [20:35] therealgalen: Lots of them have DSPs available, but you'd have to check carefully to see if any of those vendors have kernels that do that sort of offloading. [20:36] rcn-ee, who doesn't install the battery on a c3/c4? [20:37] me.. ;) [20:37] and i have 3 other's beagles couple bx's and c2 with out the battery hookup.. [20:38] rcn-ee, whats that fixup ? anything we can pull into the distro to help ? [20:38] it's the util-linux... missing rtc date, you had mentioned a fix was heading for lucid... [20:39] ah, k [20:40] my fingers are crossed that debian picks that up the same fix for squeeze too... then no one would have noticed it was an issue... ;) [20:41] hmmm there is no ./tmp/etc/e2fsck.conf file on my image ?!? [20:46] Hoonse, that's correct.. by default the file doesn't exist... [20:47] by default everything works perfectly.. ;) [20:47] i have to make it right? [20:47] correct... [20:47] thanks [20:53] when i try to umount the ext partition if the mmc card it says the device is busy but the "writing led" isnt blinking... i am afraid to all again when i just stick it out... [20:54] -.- i was ON the partition... [20:54] sorry [20:55] so guys its finished [20:55] please cross your fingers =) [20:56] plugged in... [20:56] nothing to cross ... it will just work :) [20:56] woooohoooo i see tux on the screen *muahahahaaa* [20:57] holy sh... it works!!! [20:57] :) [20:57] thanks guys!!! [21:02] hmmm i just got a terminal on a blue backscreen?!? [21:02] but the login was at a gui... [21:03] Hoonse: Well, what desktop environment did you install? [21:03] i.e. what did you specify for the -s option [21:03] xfce4?!? [21:04] ahh Hoonse, by default it goes to 'xterm' you need to change it in the drop down on the bottom... [21:04] (at the login screen) [21:04] k *reboot* [21:04] 'exit' should do it.. [21:05] yeah thats the other option =) [21:06] that would be a nother good fixup to list on the wiki.. just remove/move the xsession file... [21:06] there is no drag down?!? [21:06] there is login restart and shutdown... [21:07] you need to click on your login name... or type it.. then the theme changes... [21:07] oh there it is... sorry [21:07] i know.. it's weird... if you don't have a 'theme' installed it's impossible... ;) [21:11] is there a tool to setup the wifi device? [21:12] network manager, once installed.... but you'll have to do... "sudo ifconfig -a", "sudo iwconfig wlanX essid "xxx"", "sudo dhclient wlanX" [21:13] iwconfig not found... [21:14] that was "wireless-tools" one of the usefull packages... you have to download, copy, dpkg it.. [21:15] http://ports.ubuntu.com/pool/main/w/wireless-tools/ it shouldn't have any dependices, dpkg will warn you... [21:15] under karmic qnd lucid you can just chroot into the SD card from your x86 host :) [21:15] an just install NM [21:15] qemu-arm-static ftw ! [21:16] yeah, that would be easier.. and while there, add the nework manager gui for xfce4, not sure on the name.. ;) [21:16] yeah [21:16] everything is so much easier in karmic and lucid [21:17] imagine lucid+1 .... it will just work on its own, we wont even need to do anything anymore :) [21:19] whats the difference between wireless-tools 29-1ubuntu... and wireless-tools 29-1.1ubuntu... [21:19] wich should i download? [21:25] Cool, linux uploaded [21:26] Hopefully a working versatile in lucid soon [21:26] lool, doe sthat have the fix for the eglibc issues ? [21:26] or do we need another libc upload for them ? [21:36] i have downloaded the files to my beagle desktop but i cant find any option in apt-get to install from disk [21:36] ogra: I don't have any fix for the regression caused by libc6 update [21:37] meh [21:37] Hoonse, from disk, use "sudo dpkg -i " [21:37] then we wont have working versatile [21:39] opps, thats why mkfs.ext4 took so long on my server.. usb 1.1 port... [21:39] hmmm i have to take the armel.deb files right? [21:41] 1.1 ? ouch [21:41] Hoonse, that's correct... [21:41] no i took 2ubuntu... [21:42] wireless-tools_29-2ubuntu6_armel.deb [21:42] looks ok [21:42] it says "packet damaged" [21:43] i translatet this... [21:43] sounds like it didn't copy all the way.... [21:44] k i will do it again... [21:45] Hoonse, https://launchpad.net/~ogra/+archive/ppa/+files/qemu-arm-eabi_0.10.5-0ubuntu1_i386.deb install that on your x86 machine [21:46] Hoonse, then you can mount your SD to /mnt and do: sudo chroot /mnt apt-get install network-manager [21:46] via ssh too? [21:46] on your intel machine [21:46] you can do chroot via ssh too indeed, yes [21:49] i have istalled this [21:49] now? [21:49] now you can execute stuff on the SD [21:49] through the chroot command [21:50] mount it to /mnt and run sudo chroot /mnt [21:50]