**** BEGIN LOGGING AT Mon Jun 06 02:59:57 2011 Jun 06 06:30:04 morning Jun 06 06:50:52 sigh, ubuntu cant find the package ubuntu-omap4-extras Jun 06 06:56:23 ah hah! http://omappedia.org/wiki/Ubuntu_PPA Jun 06 10:37:52 janimo, are you already working on lightspark ? Jun 06 10:38:02 woglinde seems to have probs building it Jun 06 10:38:36 ogra_, yes, I am looking at it on and off but did not build on ARM yet. I'll have a look Jun 06 10:38:50 I just know it takes a lot of time to build Jun 06 10:39:03 yes, and takes a lot of ram too apparently Jun 06 10:39:07 ogra_, btw I put two merge request related to the SD alignment BP Jun 06 10:39:09 janimo you need a machine with 1 gig Jun 06 10:39:18 woglinde, ok I have a Panda Jun 06 10:39:20 with 1G Jun 06 10:39:21 the amf3_parser file needs it Jun 06 10:39:28 woglinde, cant be since it built on the buildds Jun 06 10:39:42 unless thats something very new Jun 06 10:39:52 the ac100 with 512 and swap isnt working Jun 06 10:40:02 I gave up after 1 1/2 hours Jun 06 10:40:05 the babbage boards we use as buildds are Jun 06 10:40:06 compiling this file Jun 06 10:40:13 woglinde, try booting with nosmp ;) Jun 06 10:40:26 ogra its not compiled with -j Jun 06 10:40:27 2 Jun 06 10:40:38 its a boost or gcc error Jun 06 10:40:40 no, but memory management in kernel is different Jun 06 10:41:21 all i know is that it builds fine on the build machines Jun 06 10:41:26 and they only have 512M Jun 06 10:41:32 woglinde, I'll do a build on oneiric - that is our target now Jun 06 10:41:34 but they are also UP machines Jun 06 10:41:52 janimo I tried with latest boost Jun 06 10:41:56 still the same problem Jun 06 10:42:12 you need to add armel in the control file Jun 06 10:42:27 shouldnt be needed in ubuntu Jun 06 10:42:31 but I think you are clever enough to find it our yourself Jun 06 10:42:33 i thing that was already patched Jun 06 10:42:41 *think Jun 06 10:42:54 I tried latest bazaar Jun 06 10:43:04 and its really only the amf3_parser Jun 06 10:43:08 all other files compiling Jun 06 10:43:09 you should have tried the latest packaging branch Jun 06 10:43:19 I wanted 0.4.8 Jun 06 10:43:22 *g* Jun 06 10:44:10 you should have tried the latest packaging branch with a new upstream tarball :) Jun 06 10:44:51 ogra I doubt it changed something on the amf3_parser Jun 06 10:44:54 https://code.launchpad.net/ubuntu/natty/+source/lightspark is the most recent packaging branch Jun 06 10:45:16 0.4.6.1 Jun 06 10:45:17 to old Jun 06 10:45:20 woglinde, no, but you dont need to care about hacking the control file etc Jun 06 10:45:25 ?? Jun 06 10:45:38 dch -i will make it 0.4.8 Jun 06 10:46:04 woglinde, the AMF parser was just rewritten yesterday upstream :) Jun 06 10:46:10 ogra why care= Jun 06 10:46:14 janimo lol Jun 06 10:46:16 okay Jun 06 10:46:19 you get the tarball, unpack it, cd into the dir, bzr branch the debian dir, type dch -i and fire off a build Jun 06 10:46:20 I will take look Jun 06 10:46:23 woglinde, but we are only packaging git master and ,ater for oneiric Jun 06 10:46:43 why I am always to early Jun 06 10:46:50 via debian sync most likely Jun 06 10:47:55 woglinde, but good that you reminded me, I'll do a git master build on ARM to see if anything broke lately Jun 06 10:51:08 hm I will try it too Jun 06 10:51:20 looks like the rewrite dont will fail Jun 06 10:55:35 woglinde, are you trying to run latestt lightspark on natty? Jun 06 11:06:34 janimo yes Jun 06 11:07:06 woglinde, it may be the case that even x86 fails to build on natty Jun 06 11:07:10 did not check recently Jun 06 11:07:33 janimo as I said it was only the amf3 parser Jun 06 11:07:37 all other files compiled Jun 06 11:07:47 so its only linking left Jun 06 11:07:52 ok Jun 06 11:07:57 but I will see it in a few minutes Jun 06 11:08:20 I will try it against the boost libs from oneiric Jun 06 11:15:32 so now lets see whats happen Jun 06 11:18:37 lool: hi. is there any tool that i can use to parse a 'local PPA' and process the associate data e.g. list of packages (src, bin), dependencies, versions, .... ideally I think I am looking for a parse of files Packages and Sources... Jun 06 11:19:07 ndec look into apt-get code? Jun 06 11:20:20 ndec, grep-dctrl Jun 06 11:20:35 woglinde: thx ;-) i was looking at a tool more 'ready to use'... Jun 06 11:20:40 with the Packages.gz file in /var/cache/apt Jun 06 11:22:42 ndec: Yes, launchpadlib Jun 06 11:23:12 ndec: http://paste.ubuntu.com/619829/ list-ppa-dsc.py Jun 06 11:23:42 yeah, lp-lib for remote stuff ... Jun 06 11:24:44 lool: ogra_: thx! it looks like grep-dctrl might be want i need... i will dig into this. Jun 06 11:25:21 ndec: BTW I got the script from a Launchpad developer, it's not my writing Jun 06 11:25:53 you just dont want to be blamed for bugs ... :P Jun 06 11:42:03 *sigh* Jun 06 11:43:13 woglinde, FWIW I successfully built LS HEAD on ARM/Oneiric without issues Jun 06 11:43:51 janimo I am fighting with scripting/abc.cpp Jun 06 11:45:02 maybee I should switch to a usbdisk Jun 06 11:45:10 instead of the sdcard Jun 06 11:45:17 woglinde, most definitely Jun 06 11:45:47 but all other files compiles Jun 06 11:45:47 It did not even occur to me you use SD Jun 06 11:46:01 I have enough Jun 06 11:46:01 enough if one large file needs swapping Jun 06 11:46:13 if SD is used that is becoming very slow Jun 06 11:46:17 yes Jun 06 11:46:20 I know Jun 06 11:46:21 ;) Jun 06 11:46:40 I gave up on SD cards except for testing our images Jun 06 11:47:11 and for boot partition Jun 06 11:58:20 good move Jun 06 11:58:52 so now lets see with external harddisk Jun 06 12:41:33 its linking Jun 06 14:56:26 In 5 min we'll start the linaro public plan review Jun 06 14:56:27 https://wiki.linaro.org/Cycles/1111/PublicPlanReview Jun 06 14:56:38 today presenting the Android and Dev Platform teams Jun 06 14:56:51 Dev Platform is the one also working with Ubuntu on ARM Jun 06 14:57:05 so folks here can probably be interested in hearing it Jun 06 15:32:34 once one installs the omap4 add ons, is there a step after that to get them working ? (sound, bluetooth) Jun 06 16:39:36 rsalveti: do you know what it takes to get a linux-linaro-2.x kernel work with an ubuntu userland? Jun 06 16:40:22 ppisati: should work just fine, we have the packages for that available at the repo, but if you want to grab from sources, just use the same config used by the packages Jun 06 16:40:25 and should be fine Jun 06 16:40:45 rsalveti: uh... i'll try again, but it hanged at boot Jun 06 16:40:56 ppisati: which kernel? Jun 06 16:41:02 could be that the kernel is broken also ;-) Jun 06 16:41:06 if you got 39 Jun 06 16:41:16 i tried 39 for sure Jun 06 16:41:23 don't remember if i tried .38 too Jun 06 16:41:29 anyway, i'll do that Jun 06 16:41:44 and i'll use the .config from linux-linaro-natty Jun 06 17:01:13 does anyone here have a pandaboard with latest ubuntu, working sound, and a bluetooth menu that has more than just on/off? If so, can I discuss the install procedure with you to work out what i missed? Jun 06 17:04:51 MrCurious_: just the same conversation as minutes before, rsalveti said that those can be found as readymade packages at the repo Jun 06 17:05:37 someone with an irc-shell should make a public log Jun 06 17:06:27 i don't have enough diskspace on salaliitto.com to do that, as i already log more than one big channel Jun 06 17:06:42 MrCurious_: There is a known issue with audio and a workaround. See bug 746023 comment 45. Jun 06 17:06:43 Launchpad bug 746023 in alsa-utils "No sound on omap4" [High,In progress] https://launchpad.net/bugs/746023 Jun 06 17:07:34 I will be testing bluetooth with a BT keyboard here shortly. Just finally got one in. Jun 06 17:20:00 gruemaster: can you ping me when you do that, i cant work out how to get my bluetooth to scan for devices, or be findable by another computer. i have installed the omap4 ubuntu add-ons, but still, it doesn't work like it appears in the help pages Jun 06 17:20:50 Ok. Just imaged my new 16G class 10 SD so I can kill two birds as it were. :P Jun 06 17:21:21 i got significantly better performance by moving root to a USB drive Jun 06 17:21:35 Well, duh. :P Jun 06 17:21:47 and the SD's stopped getting corrupted Jun 06 17:21:51 i have weak sd's Jun 06 17:21:59 SD is horribly slow on every system I have seen (not just panda). Jun 06 17:22:30 my robot design depends on being able to do bluetooth serial ports. Jun 06 17:25:22 Just curious, but what are you using for a controller? Jun 06 17:27:22 you mean as the main computer, or to control the motors/servo's Jun 06 17:27:31 Motor controls. Jun 06 17:27:56 to control the motors, the plan is to use bluetooth modules + atmega328 + 754410 + some passives Jun 06 17:28:33 that way i can easily pair the subsystems with other pieces, like a lego brick approach, but wireless for all but power Jun 06 17:29:18 Do you have a link to a supplier or are you building this out of components & making your own boards? Jun 06 17:31:47 designing my own board Jun 06 17:31:54 ah. Jun 06 17:32:02 but i can send you a link for the $10 serial bluetooth modules if you want Jun 06 17:33:18 I have a ProXR 410 serial relay from http://controlanything.com that I plan on using for test automation. Jun 06 17:33:59 Having "fun" relearning how to program the serial port in C, especially since they only have examples in VB6 and .net. Jun 06 17:45:12 grue: http://www.satistronics.com/serial-port-bluetooth-module-masterslave-hc05_p2863.html 3.3v ttl serial port, any speed. very easy to configure and work with Jun 06 17:45:41 documentation is only slightly errored. just ignore everything above the command set, and you will be fine Jun 06 17:46:24 Cool Jun 06 17:49:51 Heh. FAQ on their website: Can I trust Satistronics? Jun 06 17:53:18 i tried a look alike module from elsewhere, and it was locked into 9600 8n1 Jun 06 17:53:23 and a lame name Jun 06 17:53:34 but the satistronics ones are completely unlicked Jun 06 17:55:39 unlicked? Reminds me of my test lab back at Intel. We put "Do not lick" stickers on our high output powersupplies to appease the safety managers. Jun 06 17:56:33 un locked Jun 06 17:56:34 sorry Jun 06 17:56:52 heh Jun 06 17:57:12 Heh, I understood what you meant. The typo just triggered a flashback. Jun 06 17:59:54 Ok, install went semi-smooth on my class 10. oem-config restarted once but kicked through on second retry. Not sure why, but I have seen that issue sporadically. Jun 06 18:03:41 my class 10, 16g 20m/s SD corrupts every install. Jun 06 18:05:19 Very odd. Have you been able to get any logs from this? Jun 06 18:08:42 i just see write errors, then it remounts ro Jun 06 18:09:11 i believe it to be a bad sd, or that the SD timings are not compatible with the pandaboard Jun 06 18:10:01 You could try flashing the image to it then resizing the partition manually (if you have a linux desktop/laptop). Jun 06 18:12:38 Hmmm. Massive disk i/o waits just running update manager. Jun 06 18:21:35 i can live my life happy believing that one SD card is evil Jun 06 18:22:20 i am more interested in working out how to verify i have all the omap4 proprietary drivers installed correctly Jun 06 18:22:42 Getting there. :P Jun 06 19:26:06 MrCurious_: I seem to remember something about bluetooth support being added in the kernel after Natty release. I'll check with the kernel guy in the am (he's in Italy). Jun 06 19:27:06 rsalveti: Does that sound right? BT in kernel update? Jun 06 19:27:33 it is in, but userspace doesn't support it properly Jun 06 19:27:40 you still need to do some hacks to make it work Jun 06 19:27:57 from what I heard this should be fixed with 39, but probably not for natty :-( Jun 06 19:28:13 Do we have steps for natty? Jun 06 19:28:41 think so, is the same instructions I gave you while testing before the release, let me try to find it Jun 06 19:30:19 http://paste.ubuntu.com/603343/ Jun 06 19:30:24 guess that's it Jun 06 19:30:26 GrueMaster: ^ Jun 06 19:30:34 Yep, looking. Jun 06 19:32:18 gruemaster: thanks! Jun 06 19:35:32 Does anyone have debs of a newer qemu (which doesn't fail on rootstock with "qemu: uncaught target signal 11 (Segmentation fault) - core dumped" Jun 06 19:36:06 The version in natty crashes this way while trying to compute the SSH RSA keys Jun 06 19:41:53 MrCurious_: Bug 789095 has more info. Jun 06 19:41:53 Launchpad bug 789095 in linux-linaro-omap "Bluetooth does not work on PandaBoard" [Undecided,Confirmed] https://launchpad.net/bugs/789095 Jun 06 19:43:31 ty Jun 06 19:44:22 this is good info Jun 06 19:46:18 i think if i canreplicate what warmcat (last comment) did i will be happy Jun 06 19:47:55 Is it possible that bug 674146 is the same as I'm seeing in qemu/rootstock? Jun 06 19:47:57 Launchpad bug 674146 in gcc-4.5 "dpkg segfaults during debootstrap on natty armel" [High,Fix released] https://launchpad.net/bugs/674146 Jun 06 19:51:24 Anyone? Sorry for my persistence. I'm a bit in a drag here, as our build server is down due to this bug. Some pointers to what to do would be nice, thanks. Jun 06 20:51:22 I'm trying to run chroot dir /bin/bash into a armel rootfs, but I'm getting chroot: failed to run command `/bin/bash': No such file or directory Jun 06 20:51:59 This being on a server, it is possible that either the server or the FS stops this operation. How can I check that? Jun 06 21:23:43 sveinse: chrooting to armel-ubuntu from different arch? Jun 06 21:24:35 gildean: yes from i386 Jun 06 21:25:54 I'm strace'ing it, and I see the chroot() then chdir("/"), but the following execve("/bin/bash",...) fails Jun 06 21:34:03 I'm comparing the strace against a strace from my amd64 desktop where everything works, and I see minimal differences, and just the execve() call which fails Jun 06 21:56:45 Anyone knows if server apps (like apparmor) could interfer with qemu and binfmt? Jun 06 21:59:07 I don't understand why my chroot fails. The error is foo, and strace doesn't reveal anything. Jun 06 22:11:41 I think I'm going crazy. This, of course, being the day when the armel build server stops working combined with the exceptional occasion with almost no response in the forum. Tough luck... Jun 06 22:37:09 sveinse, Just to verify, you're using qemu-debootstrap to create the chroot (or some mechanism that depends upon this)? Jun 06 22:37:38 yes, i believe rootstock is using qemu-debootstrap, doesn't it? Jun 06 22:39:53 I'm suspecting that the reason rootstock/qemu crashes, and the fact that I cant run chroot into a armel rootfs, might be related Jun 06 22:40:17 I think it tries to do the same sort of thing with it's own (separate) debootstrap wrapper, but it's close enough. Jun 06 22:41:20 The weird part is that I don't trigger the crash on my desktop machine (which is amd64) Jun 06 22:42:06 Have you tried just executing a static armel binary? Jun 06 22:42:20 where can I get one? Jun 06 22:42:30 That would help differentiate whether qemu is completely not working or whether it's something more complicated. Jun 06 22:43:23 I wonder if there are any files within a normal ubuntu-minimal installation which are static Jun 06 22:43:34 I usually just create a quick hello.c and gcc -static it :) Jun 06 22:43:44 hold on Jun 06 22:46:24 ok, I'm able to run the static app from my bash shell (proving the binfmt+qemu works) Jun 06 22:49:54 Interesting: "./hello" works (hello is armel static). But running (as root) "chroot . /hello" does not work! Jun 06 22:50:28 "does not work" isn't very descriptive Jun 06 22:50:37 chroot: failed to run command `/hello': No such file or directory Jun 06 22:51:35 try chroot . ./hello Jun 06 22:51:51 same result Jun 06 22:52:03 file ./hello Jun 06 22:52:18 ./hello: ELF 32-bit LSB executable, ARM, version 1 (SYSV), statically linked, for GNU/Linux 2.6.16, not stripped Jun 06 22:53:09 it was exact same error for chroot . ./hello? Jun 06 22:53:21 Aha! That sounds like /usr/bin/qemu-arm-static doesn't exist in the chroot Jun 06 22:53:42 interesting Jun 06 22:53:53 This needs to be manually copied in, unfortunately, and updates are ugly. Jun 06 22:54:02 yeah, let me try Jun 06 22:54:15 Essentially, binfmt-misc is calling /usr/bin/qemu-arm-static when attempting to run the foreign binary. Jun 06 22:54:47 voila! Jun 06 22:54:54 And if there's not an extra copy in the chroot (which wouldn't be created by debootstrap normally), then it fails with "command not found" Jun 06 22:55:29 Excellent. So, I think this probably represents a bug in rootstock: it probably needs to be ported to use qemu-debootstrap from qemu-user-static rather than whatever it's doing now. Jun 06 22:55:46 This means that a rootstock image is not purely without clutter or change from the host pc which created it? Jun 06 22:56:01 From the dependencies, it looks like it still expects the deprecated qemu-kvm-extras-static Jun 06 22:56:16 If it puts /usr/bin/qemu-arm-static in it, I mean Jun 06 22:56:29 rootstock may have some provision to delete this file, but I doubt it's clean. Jun 06 22:57:00 At least it sheds light on a large mystery for my part Jun 06 22:57:09 rootstock is a fun toy, but I wouldn't use it for any production purpose. Jun 06 22:58:19 I know you/ubuntu doesnt use it any more, but what other options are there to putting together a rootfs image? Jun 06 22:58:55 Because using it for production is exactly what it's used for today... :o Jun 06 22:59:28 Lots of folk here use it, but it was never used for the Ubuntu images. Jun 06 22:59:57 Do you have access to an armel development host? Could you build images there? Jun 06 23:00:03 Takes a list of packages and outputs a rootfs. Nice. Jun 06 23:00:07 Not cross-building is the first step to reliability. Jun 06 23:00:55 Yes, I see that, though difficult at times. Like having a large, fast, buildserver which is useless in a armel farm Jun 06 23:01:53 But, back to another qemu issue: Jun 06 23:02:40 qemu crashes (consistenly on this machine) while compuing SSH2 keys Jun 06 23:02:58 With lots of output or almost no output? Jun 06 23:03:16 "Creating SSH2 RSA key; this may take some time ...qemu: uncaught target signal 11 (Segmentation fault) - core dumped Jun 06 23:03:17 Segmentation fault" Jun 06 23:03:33 dpkg: error processing openssh-server (--configure): Jun 06 23:04:10 Get a stack trace (you can see what was being called from /var/lib/dpkg/info/openssh-server.postinst ) Jun 06 23:04:38 I need to modify the rootstock (in order for it not delete the rootfs when failing) to inspect deeper Jun 06 23:04:44 If the crash is in ssh-keygen, then it probably needs fixing, but I doubt this is the case. Jun 06 23:05:05 More likely the crash is in the kernel, in which case it's qemu not handling something in the ideal manner. Jun 06 23:05:33 FYI I see some chatter on google about uncaught target signal 11 against armel Jun 06 23:05:46 That, or step through it more slowly, perhaps using qemu-debootstrap Jun 06 23:05:55 so it's not restricted to ubuntu or rootstock Jun 06 23:06:16 signal 11 just means "Segmentation Fault". Jun 06 23:06:39 aha Jun 06 23:06:51 If qemu was catching segfaults, we'd all be complaining about that because it would make it hard to debug things :) Jun 06 23:07:29 So, quick process to reproduce: Jun 06 23:07:34 1) make a directory Jun 06 23:07:52 2) populate a cross-chroot in that directory with qemu-debootstrap Jun 06 23:08:02 3) chroot into the populated chroot Jun 06 23:08:07 4) apt-get install openssh-server Jun 06 23:08:18 5) After failure, inspect the postinst Jun 06 23:08:36 6) Install your favorite debugging tools, and find the crash. Jun 06 23:10:16 you need apt-get in there as well, dont you? Jun 06 23:10:23 Hrm. qemu-debootstrap lacks a manpage: it takes all the same arguments (with the same meanings) as debootstrap, so just use --arch=armel and otherwise follow the debootstrap model Jun 06 23:10:32 apt-get where? Jun 06 23:11:13 If you do 3) apt-get need to be inside the chroot to do 4) right? Or have I misunderstood? Jun 06 23:11:34 baaah Jun 06 23:11:36 forget it Jun 06 23:11:46 Ah good. I was getting very confused :) Jun 06 23:11:57 my bad, sorry Jun 06 23:12:01 No worries. Jun 06 23:15:51 Is there a way to create a rootfs where everything is downloaded, but not configured, and then later when the real target executes, let it do all the postinst steps? Jun 06 23:16:39 Rootstock at least fires up qemu to run the installer (or post installer, not sure on its specific task) Jun 06 23:18:54 Sure. You can separate --first-stage and --second-stage in debootstrap. Jun 06 23:19:10 Take a look at the qemu-debootstrap code for a simple walkthrough of how one does this with qemu. Jun 06 23:19:23 Replicating that with hardware ought be relatively easy. Jun 06 23:19:53 Mind you, it's not necessarily safe to install lots of other stuff before running --second-stage, as the base system isn't configured yet. Jun 06 23:20:15 I see debootstrap retrieves, extracts, unpacks, configures. It seems to be the unpacking a later is done emulated. Am I right? Jun 06 23:21:05 *The unpacking stage and the preceeding steps are done emulated. (it's getting late) Jun 06 23:21:53 I'll check qemu-debootstrap in all cases, thanks Jun 06 23:22:00 I'm not finding the source you're quoting. Jun 06 23:22:28 But debootstrap pulls the base system in a way that doesn't much care about dependencies, so it kinda handy to get a base. Jun 06 23:22:59 It does all of retrieval, extraction, unpacking, and configuration. Based on the arguments, it can perform subsets of these actions. Jun 06 23:23:31 unpacking is safe to do on a foreign system: it's just unrolling tarballs. configuration is *not* safe to do on a foreign system. Jun 06 23:23:56 Oh? How come? Jun 06 23:24:27 foreign being the target system which shall run ubuntu in the end? Jun 06 23:24:47 Because configuration scripts can run arbitrary binaries to accompish their goals. Running these in a chroot with a foreign architecture will not work. Jun 06 23:25:06 Mind you, if you do things like the qemu-static binfmt trick, it's not really that foreign anymore Jun 06 23:25:15 (although qemu support isn't quite up to full hardware) Jun 06 23:25:51 And you could have binfmt back to intel on the armel (if such exits) to bridge the oposite gap? Jun 06 23:26:11 "foreign" meaning that the ISA for the code doesn't match the ISA for the machine on which the chroot is construted. Jun 06 23:26:36 It does exist. it's supported by qemu-debootstrap. It's so incredibly buggy that nobody should ever use it. Jun 06 23:27:18 the main issue is that qemu guest support for ia32 isn't well tested on armel hosts, and there don't seem to be any engineers wanting to try to fix discovered issues. Jun 06 23:27:20 And foreign is defined by the machine running qemu-debootstrap Jun 06 23:27:35 Sorta. Jun 06 23:27:54 A chroot is "foreign" if the instruction set in the chroot is not supported by the host. Jun 06 23:28:27 So an i386 chroot on an amd64 host is not foreign, but an amd64 chroot on an i386 host is foreign. Jun 06 23:28:57 armel chroots are foreign on everything not armel Jun 06 23:29:03 Since the configure step needs to be done "native" then you're basically tied to use all of debootstrap to make an image Jun 06 23:29:14 (although it may be that in the future armel chroots will not be foreign for armhf hosts) Jun 06 23:29:53 Well, you can reimplement debootstrap, but every image building tool I know starts with debootstrap. Jun 06 23:30:34 I'm trying to figure out why rootstock is considered "not for production" while debootstrap seems to be Jun 06 23:31:03 I'm not entirely comfortable with the passive voice in that sentence :) Jun 06 23:31:22 I don't consider rootstock suitable for production for three reasons: Jun 06 23:31:54 no offence, I'm truly curious. And since we're going to use ubuntu, I should know Jun 06 23:32:00 1) It uses qemu-static binfmt hacks, such that the resulting image a) has extra files and b) is subject to any bug in qemu Jun 06 23:33:20 2) It doesn't use the typical d-i based scripts for initial system configuration, which in practice may mean nothing more than a potential for bitrot, but may also be incomplete Jun 06 23:35:01 Hrm. I remembered there being three, but the third one escapes me now Jun 06 23:35:25 Most likely, it was solved (since a lot of the issues passing through my head have corresponding memories of someone telling me they fixed it) Jun 06 23:36:09 I'll give you 3) openssh keygen & installation works fine on qemu-debootstrap, but fails on rootstock Jun 06 23:36:32 I've tried twice now Jun 06 23:37:16 OK, I've surely got a lot of very useful info here Jun 06 23:37:19 Heh. Then 3) it's buggy :) Jun 06 23:37:27 :D Jun 06 23:37:59 I'm very grateful for the conversation! Jun 06 23:38:08 Aha! rootstock still assumes versatile as an emulation host. Jun 06 23:39:17 (and apparently hardcodes a lucid kernel) Jun 06 23:40:06 Why should either affect qemu in respect of SSH keygen... Jun 06 23:40:34 rootstock is running a full VM, while deboostrap is only running chroot/binfmt? Jun 06 23:40:57 I believe that Ubuntu is currently shipping Linaro's QEMU, which I thought was no longer targeting versatile as primary. Jun 06 23:41:28 Rootstock works in three stages Jun 06 23:42:25 1) debootstrap to populate a chroot Jun 06 23:42:37 2) debootstrap under qemu-static to configure the chroot Jun 06 23:42:47 3) fully-emulated environment to install random stuff Jun 06 23:43:27 Mind you, I'm not that familiar with the code, but that's how it appears to me from a quick look. Jun 06 23:43:47 sounds like I should move away from rootstock and integrate debootstrap instead Jun 06 23:43:48 debootstrap itself doesn't do anything architecture related. Jun 06 23:44:04 qemu-debootstrap (also not something I'd use for production) does the chroot/binfmt stuff. Jun 06 23:44:23 * persia has fairly strong feelings about doing things natively, if this isn't already obvious Jun 06 23:44:41 I understand that Jun 06 23:45:36 What's the most powerful armel machine available? Because that is what you need to do such things natively, IMHO Jun 06 23:46:31 Available? Possibly TI omap4 panda or nVidia Tegra. Jun 06 23:46:42 All the hardware I've used it IO-bound. Jun 06 23:46:57 Doing debootstrap on armel for armel is ok, right Jun 06 23:47:02 Find something (anything) with non-USB SATA, or other reasonable IO path, and you'll win. Jun 06 23:47:17 Even an OMAP3 should keep up. Jun 06 23:47:31 For I/O, Marvell is the best, but not available to the general public afaik. Jun 06 23:47:37 Native debootstrap is the basis for all official images. Jun 06 23:48:03 Is this just for image builds or for package compiling? Jun 06 23:48:12 I'm certainly happy with my Marvell IO performance, but it's not ARMv7a, so can't run native debootstrap for Ubuntu. Jun 06 23:48:17 Image builds. Jun 06 23:48:19 * GrueMaster too lazy/busy to read backscroll. Jun 06 23:48:25 Ah. Jun 06 23:48:44 Backscroll summary: rootstrap is buggy and was being used on a high-speed build server for production images. Jun 06 23:48:49 An dual core ARMv7a with SATA is what I want I think Jun 06 23:49:01 hahahah Jun 06 23:49:17 sveinse, If you can find such a beast, then yes, that's precisely what you want. Jun 06 23:49:19 Well, for cost/speed, definately omap3/4 based atm. Although the imx51 isn't bad in the Efika nettop. Jun 06 23:49:44 rootstock? ouch. Jun 06 23:49:48 Just be sure to check the internal wiring: some devices claim SATA but really just have a USB<->SATA bridge, and you could do that yourself. Jun 06 23:49:52 persia, You work at reuters or something? That's the most compact summary I've ever read Jun 06 23:50:10 sveinse, I haven't worked with reuters is well over a decade, but thanks :) Jun 06 23:50:11 The efika is true sata. Freescale fixed it post babbage3. Jun 06 23:50:24 GrueMaster, Really? Nice! Jun 06 23:50:27 But only Cortex A8 (single core). Jun 06 23:50:38 debootstrap is single-threaded anyway. Jun 06 23:50:42 That's what I was told at UDS. Jun 06 23:51:20 Yea, but having one core focus on one thing while the other core handles normal tasks does provide some improvement. Jun 06 23:51:29 I heard the mx51 Efika's were currently at reduced prices, to clear inventory for the mx53s, so if single-core is enough, now is probably a good time. Jun 06 23:51:44 $125 for the nettop. Jun 06 23:51:53 Ah, you mean like handling the IO requests. Heh. Yeah. Jun 06 23:52:05 Yep. That's about half what they were. Jun 06 23:52:09 I thought I heard a rumor about a armel server blade (for energy efficiency). I have no idea from where, though Jun 06 23:52:29 I've heard of that too. Jun 06 23:52:59 sveinse, I've been hearing all sorts of rumors like that for the past couple years, often from folk who really ought to know. I have yet to see any retail products like that which support ARMv7a Jun 06 23:53:13 (retail servers with ARMv5te are plentiful though) Jun 06 23:53:49 https://www.genesi-usa.com/store/ Jun 06 23:53:59 Oops. $129. My bad. Jun 06 23:54:04 That's like USB OTG. Have you guys ever seen a product with USB UTG? Which /uses/ the OTG part? Jun 06 23:54:18 Yes. Jun 06 23:54:22 UTG? Jun 06 23:54:28 sorry, typo Jun 06 23:54:38 its 2am here soon Jun 06 23:54:43 Ah. Jun 06 23:54:50 Time for more coffee. :P Jun 06 23:55:41 Point is OTG is well defined by the USB standards, but it's very hard to come by any product and/or cables that are wired correctly Jun 06 23:56:06 You're not shopping in the right places :) Cables are plentiful in some shops. Jun 06 23:56:07 It seems the industry and the standards dont line up on it Jun 06 23:56:37 The main issue is really that it's hard to figure out what it ought do. Jun 06 23:57:05 The product we're taking forward does include OTG. But all cables I've had in my hands are not correct Jun 06 23:57:12 I have found a few OTG cables that do work. mini to usb host. I've had a keyboard plugged into my beagle that way and it worked fine. Jun 06 23:57:18 In the limited context of using OTG on an Ubuntu system, the main issue is figuring out how to provide access to a useful filesystem when in gadget mode. Jun 06 23:57:37 Mini USB seems to be better supported, but micro is not Jun 06 23:57:44 Gadget mode is an entirely different thing. Jun 06 23:58:07 There's some micro cables around, but yeah, they're lots harder to get. Jun 06 23:58:08 I also found a micro to mini adapter. Just got it Saturday, but haven't teted it yet. Jun 06 23:58:27 are you using the beagle xM? Jun 06 23:59:15 I have one, yes. Jun 06 23:59:16 sveinse, Do you *need* to be micro, or would mini work? Jun 06 23:59:56 Too late in the design of the product. The tools for the enclosure are being built now Jun 07 00:00:35 Oh well. Maybe find some online source of cables and point customers there. Jun 07 00:00:55 But mini are deprecated by USB, that's what puzzles me Jun 07 00:01:21 It is? That's annoying. I have *lots* of mini cables and ports around. Jun 07 00:01:44 True. They are taller than micro, and the EU wanted a standard deployed across all cell phone markets. Jun 07 00:02:04 That's why the sudden conversion to micro. Jun 07 00:02:20 http://www.penguin.cz/~utx/hardware/USB_Mini/ Jun 07 00:02:32 Gadget cables are plentiful. It is the Host cables that are hard to find. Jun 07 00:03:21 Host cables being usbmicro to USB B (square) or USB A female receptacle? Jun 07 00:03:47 Right. Jun 07 00:04:32 * sveinse want 10k cables of usbmicro to USB A female receptacle Jun 07 00:05:05 Would you like fries with that? :P Jun 07 00:05:08 That quantity ought make it easier. Jun 07 00:06:30 Hmmm. Just looked closely at my micro<>mini adapter. It is for plugging a micro cable into a mini device, not the other way. Useless. Jun 07 00:06:49 sveinse: persia: rootstock uses qemu and debootstrap Jun 07 00:07:04 but it copies the qemu-arm-static before doing the second phase of debootstrap Jun 07 00:07:08 it can also run inside a full vm Jun 07 00:07:17 and that's why it's using vexpress kernel Jun 07 00:07:20 yup. I just hoped finding it off the shelf somewhere. Yet the 5 cable samples I've tested all wires the ID pin incorrectly, so it's not trivial Jun 07 00:07:28 but if you run with root, it just uses user mode emulation Jun 07 00:07:50 Which for some reason is crashing while computing ssh host keys Jun 07 00:07:59 and the reason why rootstock is not suitable for production is that it's not the official tool to generate images Jun 07 00:08:19 and not doing all the specific hacks used by live-rootfs to create the rootfs Jun 07 00:09:05 rsalveti, Do you think it ought migrate to use the omap3 kernel? Jun 07 00:09:09 it creates the user with it's own way, and set the images using it's on setup Jun 07 00:09:20 I have two things when I start: A debian repo and a list of packages. They are to be installed into a image which shall be put into an sdcard in production. Jun 07 00:10:29 And if qemu-bootstrap is also dodgy because of qemu, then the only solution is to let an armel machine build the image for production Jun 07 00:10:58 persia: last time I checked omap3 support you needed to create a whole bin image Jun 07 00:11:03 with x-loader/u-boot and stuff Jun 07 00:11:25 the only real problem with versatile is the ram restriction Jun 07 00:11:32 besides that it works better with qemu Jun 07 00:12:03 Oh. I thought the kernel team dropped versatile because omap3 support had improved enough to make it more suitable with qemu. Jun 07 00:12:22 persia: well, it's hard to push omap 3 specific patches upstream Jun 07 00:12:40 because upstream wants to emulate it exactly as the hardware should behave Jun 07 00:13:02 How far do we have to push? I thought Ubuntu was shipping a special linaro-patched qemu Jun 07 00:13:05 and versatile is kind of a virtual machine for qemu, no one actually check if the kernel really works with real hardware Jun 07 00:13:26 yes, but last time I checked it was still missing some bits, would be nice to check it again Jun 07 00:13:41 rsalveti: My rootstock crashes when install openssh-server. It did not prior to it being upgraded from maverick to natty. Nor does my desktop machine (amd64) crash it. Jun 07 00:13:42 Some people have real versatile hardware, but not many, and it'&s known that the real hardware *can't* run Ubuntu (not ARMv7a compliant) Jun 07 00:13:55 sveinse: unfortunately I'd recommend you to run natively on an ARM board if you want a stable tool Jun 07 00:14:08 rootstock supports running on arm currently Jun 07 00:15:08 versatile express should be fine Jun 07 00:15:15 don't know how well that's support in qemu Jun 07 00:15:24 sveinse, What's your timeframe for getting the SD images complete? If it's in July sometime, we should have clear instructions available for you beforehand. If it's sooner, you may have some muddling to do. (And no, you don't need to tell me the answer, just apply your answer to my comments) Jun 07 00:15:50 rsalveti: interestingly, this is the first time since the start of our project that it has crashed. It has been faithful since like nov -10. Daily builds, 10-20 builds per day Jun 07 00:15:52 sveinse: this could be easy to fix, but then you'll end up having issues with mono Jun 07 00:16:02 Sure, we're still in development, so I can wait Jun 07 00:16:17 Hrm. My usb micro host cables were from Green House Japan, but they don't seem to be offering any now. Jun 07 00:16:20 if you don't use mono, or not plan to use, then rootstock should work in most of the cases Jun 07 00:16:54 It's not just mono: it's gotten buggier again because fewer folk are heavily using the qemu-static stuff. Jun 07 00:17:10 no mono, just this little regression :) Jun 07 00:17:28 sveinse: do you have any specific change at your rootstock? Jun 07 00:17:36 And there's something else wonky: the openssh-server case works with qemu-debootstrap and not with rootstrap, which implies something rootstrap-specific as a bug. Jun 07 00:17:41 sveinse: if not, can you paste me the logs? or cmd arguments you're using? Jun 07 00:18:25 s/rootstrap/rootstock/ Jun 07 00:18:32 sure, hold on (takes a while) Jun 07 00:24:04 persia, does this mean you are working on docs for bootstrapping or a tool? (Since referring to july) Jun 07 00:25:18 sveinse, Yes. Clear documentation of the Ubuntu image building infrastructure is something that's supposed to happen soon. Jun 07 00:25:26 excellent Jun 07 00:25:33 Presumably you could use this for your purposes. Jun 07 00:25:52 Seems so. Jun 07 00:25:56 Note that we're *not* attempting to generate documentation for full factory-automated installation at this point. Jun 07 00:26:28 No, I wouldn't expect that, nor demand such a thing Jun 07 00:26:47 I'd like it if we could provide that, but it's a hard problem still. Jun 07 00:27:21 And there are many holes to fall into I've noticed. Jun 07 00:27:43 I believe there's some work in progress to handle recovery partition creation, and presumably that could also be used to generate recovery media. Jun 07 00:28:07 E.g. rootstock as discussed. Previously (well haven't resolved it yet), the missing --sysroot from ubuntu/linary armel gcc Jun 07 00:28:19 But multicast distribution in a production environment is still several steps away (mostly because very few hobbyists have the infrastructure to test such things, and most people who build them like to sell their solutions) Jun 07 00:29:30 Our strategy is to make an sd-image which is duplicated across all units. And then we run a firstboot to make the units unique based on information from other sources Jun 07 00:30:13 I'd strongly recommend using oem-config as that "firstboot". If it doesn't work for you, let's fix that, rather than have lots of implementations. Jun 07 00:30:33 Thanks, I'll look into it Jun 07 00:31:17 oem-config basically wraps d-i components for personalisation of a preconfigured system. Does stuff like handle user creation, hostname assignment, timezone & locale information, etc. Jun 07 00:31:42 nice Jun 07 00:32:39 Since you're installing openssh-server, you may also be interested in looking at vm-builder: I'm not sure if all the work done there to ensure post-boot systems has migrated into oem-config, but you will want to have that set of hacks as well (stuff like regenerating SSH keys on first boot, etc.) Jun 07 00:33:13 openssh-server is a development tool for now Jun 07 00:33:43 Heh, OK. If you're doing something like a desktop, handheld, phone, etc. then oem-config should have all the right bits already included. Jun 07 00:34:00 If you're doing a server, then I'm not sure it's as ready. Jun 07 00:34:44 while I'm at it: Our guys are reporting stability problems between Qt (QWS) and Natty. Is this something which is familiar? Jun 07 00:35:01 I haven't had time to dig into the issue, so I don't know first hand Jun 07 00:35:21 I've had a few things crash on my Kubuntu netbook, but most stuff seems to work. Jun 07 00:35:34 (that being natty/armel) Jun 07 00:35:58 The key thing is to report the bugs: if you can make something crash, it ought be fixed. Jun 07 00:36:19 Perhaps I should diff the qt4-x11 sources and the stock qt. Because we are on stock qt (due to support from Nokia) Jun 07 00:37:14 Yes, I'll report any bugs I come across. Just wanted to hear if its known to be unstable Jun 07 00:37:38 Works for me, but that's only one data point, and I don't claim to use that system that much. Jun 07 00:38:13 but thanks, it does help Jun 07 00:40:39 rsalveti: I have to sleep now, but I'll get hold of you if I'm able to reproduce the crash Jun 07 00:40:41 Just checked: there's a heap of patches against Qt. You don't need to diff though: just `apt-get source qt4-x11` and look in debian/patches Jun 07 00:40:46 Thanks Jun 07 00:41:00 sveinse: sure Jun 07 00:41:05 kubuntu_23_arm_memory_barriers.patch is one of the things you may require :) Jun 07 00:41:12 persia, yes, there's lots Jun 07 00:42:46 persia, thanks you for your insights and discussions. I've learned a lot. In fact I shall bookmark the irc logs from this evening Jun 07 00:45:05 good night **** ENDING LOGGING AT Tue Jun 07 02:59:57 2011