/srv/irclogs.ubuntu.com/2010/02/26/#ubuntu-arm.txt

NCommanderpersia:Dove boards have PCIe slots :-)01:24
rcn-eeusers who have used ubiquity/oem-config before...  anyway to force it to run on both tty0 and ttyS2 on first boot?01:27
=== bjf is now known as bjf-afk
rcn-eeping for ogra, this must be an oem-config bug.. on first boot after you create your first user, oem-config exits, then restarts oem-config again from scratch...  force a reboot, then on the next boot oem-config starts again... however user does exist if you switch to tty2...01:44
bobrownIs it possible to buy an ARM laptop today and easily install the latest Ubuntu Linux on it?03:11
=== DavidBz is now known as berco
ograwoah, mounting /proc in a qemu-arm-static chroot and then installing mono explodes in a funny way10:31
loologra: So how does it fail?10:56
loologra: Could you file a bug about that?10:56
ogralool, well, it tries to enable binfmt :)10:57
ograi doubt there is any easy way around10:57
ograapart from having a fake proc, afaik persia and lifeless had some fuse based ideas for that during the sprint10:57
loolIt didn't fail for me10:57
loolupdate-binfmts: warning: /usr/share/binfmts/cli: no executable /usr/bin/cli10:58
loolfound, but continuing anyway as you request10:58
loolupdate-binfmts: warning: found manually created entry for python2.6 in10:58
lool/proc/sys/fs/binfmt_misc; leaving it alone10:58
lool(apt-get install mono-runtime)10:58
ograright and then it tries to use the existing cli handler during package install10:58
ograah10:58
ograi did apt-get install tomboy10:58
ogralool, did you ever have contact with hppfs ?11:00
ograseems that could work as a fake proc in a chroot11:00
loolNo11:00
loolI'm not sure I'm in line with the fake proc idea11:01
loolI think qemu-arm would have to emulate some of it due to the emulation it does11:01
ograwell, how else would you handle stuff that needs proc in a foreign chroot11:01
ograright, so you still need to fake something11:01
loolI'd personally prefer to use containers if we want to support this11:02
ograthats a lot of overhead11:02
loolNo11:02
loolIt's almost zero overhead11:02
ograwell, we only need proc containers do so much more11:02
ograand are quite complicated to set up for a user11:03
ogralool, http://paste.ubuntu.com/384298/11:08
ograso ita a SIGABRT and i guess because it sees the binfmt handler and tries to exec it11:09
loologra: Could you file a bug?11:09
ogragar, apport doesnt know about qemu-arm-static11:10
loolThe unsupported syscall 242 ones are probably harmless (affinity stuff)11:13
loolThe unuspported syscall 26 is just a consequence of trying to run gdb under qemu-arm, wont work because ptrace() isn't emulated (and far from trivial I'm afraid)11:14
ograyep11:15
ograbug 52837711:15
ubot4Launchpad bug 528377 in qemu-kvm (Ubuntu) "qemu-arm-static fails installing mono assemblies if /proc is mounted in the chroot (affects: 1)" [Undecided,New] https://launchpad.net/bugs/52837711:15
ograi actually only started experimenting with mono in chroot again out of desparation, i have no idea what to do with the iso-codes hang11:17
ograseems no matter what kernel or config i use, installing the ubuntu-netbook task hangs reliably at iso-codes unpacking11:18
ograi thought its caused by swapping but that was a red herring11:19
loologra: So the way you word the bug report, it sounds like of /proc is NOT mounted in the chroot, one can actually install mono; is this correct?11:42
ograno11:42
ograbut i dont get the segfaults11:42
loolWhat do you get?11:42
ograit simply doesnt execute the installation of the assmeblies11:42
ograand dpkg fails11:42
ograi'll add a log for comparison once my machine has spare cpu cycles (rootstock running atm)11:43
asachi doko ;911:58
ograoh, a rare guest :)11:59
asachehe11:59
dokoI was forced :-/12:03
dokoNCommander: you might want to look at Mikael's comments in http://gcc.gnu.org/bugzilla/show_bug.cgi?id=40860 (tracked down to the very same binutils change)12:05
ubot4gcc.gnu.org bug 40860 in libgcj "[4.4/4.5 regression] regressions in libjava testsuite on arm-linux" [Normal,Unconfirmed]12:05
rcn-eehey ogra, oem-config question...   Should it run in an endless loop in lucid? ;)12:46
ograno, it shouldnt, boot with debug-oem-config on the cmdline and file a bug with /var/log/oem-config.log attached please12:47
ograagainst ubioquity12:47
rcn-eeThank so, debug-oem-config's the trick... First thing i'll do when i get to work..12:49
ograthanks :)12:50
rcn-eeanyluck with the weird iso-codecs lockup today? ;) i'm seeing it to on my machines...  Whats weird In debian it's actually locking up at I: Extracting zlib1g...12:50
ograwith debians qemu ?12:52
rcn-eesqueeze.. so (ssh's in)12:52
ograi'm trying to debug that since yesterday but given the time it takes to get to the hang its might still take more :)12:52
ograi'm just trying it in a normal VM so that i can exclude rootstock here12:53
ograrcn-ee, btw, if you file a bug for oem-config under rootstock, please subscribe ubuntu-armel12:54
ograso we get the mails in the arm team12:54
rcn-eei know what you mean.. (QEMU PC emulator version 0.11.1)... what's weird, when i was building 'google os' images in a chroot, it locked up on that same package file..12:54
ograi think its something in qemu not handling a certain amount of CPU load or disk I/O12:55
ograbut its so damned hard to track down12:55
ynezzisn't that "I: Extracting zlib1g..."12:56
ynezzoh damn keyboard12:56
ografunnily i can just install iso-codes just fine if i install it standalone12:56
ograynezz, thats not in the VM12:57
rcn-eeseems very reasonable.. do you guys have a dedicated qemu hacker at ubuntu?12:57
ograrcn-ee, well, lool is very passionate to get all qemu bugs solved and has the knowledge ...12:57
ograhe's not a "dedicated qemu hacker" per se i think though :)12:58
rcn-eeyeah he's good.... anything show up weird in the vm machine log's file?  (i could dump it too..)12:59
ograi'm only just running it in a VM13:00
ograbefore i suspected rootstock to be at fault ... thats to closed13:00
ograhmm, fun13:00
ogramy VM doesnt even boot13:01
ograwhoops typoing the command doesnt really help :P13:01
asac/tmp/ccQPoJy6.s:1847: Error: selected processor does not support `rsc r7,r7,#0'13:03
asachmm13:03
ograwhats that ?13:04
loolreverse substract with carry13:04
asacright13:04
asacis that thumb only instruction?13:04
asacor too old13:04
loolonly available on ARM, not Thumb13:04
asacin the arm reference i have open its named in the thumb section13:05
asacyeah thats what i expected13:05
loolasac: there's a regular substract though13:06
loolasac: sbc r7,#0,r7?13:07
loolHmm don't think you can pass an immediate as first arg though13:07
asacwhat i find odd is that its not even marked as arm only in the reference quick card ... strange13:08
loolasac: I certainly see it marked as such13:09
loolasac: RSC is 5th in the substract instructions, and the fifth entry in Notes is "A"13:10
loolSo not in Thumb13:10
asacoh right. i misinterpretetd the table13:12
loolthere's a blank which is hard to read13:13
persiaogra: lool: Did you want the breakdown of the fuse proc idea?  I think StevenK fiddled with some initial implenetation stuff.13:21
ograpersia, i would, seems lool doesnt ... he wants containers13:23
persiacontainers are a more correct solution.13:23
ograbut way more complex13:23
loolno13:24
ograand do they guarantee i dont see the system kernel in /proc ?13:24
loolno13:24
ograright13:24
persiaThe fuse hack was just to use fuse.pm to check the filename and if it matched one of a set of predefined files, perform some operation (either cat a fake file, or run the real file through a filter).  If it doesn't match, feel the original file (or feed through an identity filter, depending on operation)13:25
ograwhat we need is a /proc that emulates a generic arm machine13:25
ograhrm, now i cant even purge iso-codes from my VM13:26
ograjust sits there in the ldconfig triggers13:27
persiaogra: The fuse hack can give us that (at least enough to fake most stuff).13:28
persiaJust needs someone to write up the perl.13:29
ograyeah13:29
loolpersia: I find it hard to imagine that we will get cpuinfo emulation in containers in the short term13:29
persiaYour python is better than your perl, right?  How are you at python regexes?13:29
loolpersia: However I do believe we should use containers for things like binfmt13:29
ograpersia, i come from perl ... just needs a little refresh i guess13:30
persialool: So you think it makes sense to combine a container and the fuse-proc hack?13:30
loolpersia: The cpuinfo hack could either be done with fuse or in qemu; I think I find it more useful in qemu though13:30
persiaogra: StevenK suggested perl was the better language for it because it's mostly just matching regexes and then running text filters.13:30
loolI find things are complex and fragile enough that I'd prefer not to have to fiddle with fuse, personally13:30
ograyeah13:30
ogralool, to run containers you need to do massive changes to yxour chroot (which is the part i find complex)13:31
persialool: The advantage of fuse over qemu is that it's trivial to write such a hack in perl, but fixing qemu requires real thought.  If you're up for doing it, it would be loads better.13:31
loologra: No13:32
ograhuh ?13:32
ograyou need at least to create lots of stuff in /dev, need to hack mountall defaults etc13:33
persiaogra: Shouldn't it be as simple as http://blog.bodhizazen.net/linux/lxc-configure-debian-lenny-containers/ ?13:34
ograno13:34
ogralenny is way easier :)13:35
ograno upstart ... etc13:35
ograhe wrote one for lucid too13:35
ograhttp://blog.bodhizazen.net/linux/lxc-configure-ubuntu-lucid-containers/13:35
ograbut effectively /proc will stil show you the host CPU and features13:35
persiaWhy doesn't udev work?13:36
ograso its pointless to use containers imho13:36
loolAgain, containers are not going to solve the cpuinfo problem13:36
persiaContainers solve other issues.13:36
persiaThey don't solve /proc13:36
loolThey actually help a lot with /proc, but not for cpuinfo13:36
persia(well, they solve other /proc issues, but mostly of type /proc/nnnn)13:36
persiaRight.13:36
persiaWell, also not proc/[a-z]* mostly.13:37
ograwould they be able to solve binfmt ?13:37
ogra(i doubt that)13:37
ogragiven that binfmt would still need a qemu-arm-static wrapper around the interpreter calls13:38
* ogra goes for some lunch13:39
ograurgh14:15
ogramaximizin the qemu window while it does something and then restoring it to its original size reults in massively blurry fonts14:16
ograthe funny think is that its only blurry in the center of the window ... top and bottom are crisp14:18
mikeulI have a couple of patch suggestions for rootstock, where should I submit them?  On launchpad?14:24
loolmikeul: Ideally, submit them as merge proposals14:24
mikeulvia launchpad14:24
ograyes14:24
loolmikeul: bzr branch lp:project-rootstock my-feature; cd my-feature && hack && bzr commit; bzr send14:24
mikeulthey're trivial, but it'll be a good getting-to-know-launchpad exercise.14:25
mikeulbzr send?  I did bzr push"14:25
loolmikeul: Checkout "bzr help send" then  ;-)  bzr push is when you want to do things manually14:27
ogralool, send requires an email setup, no ?14:29
* ogra checks the help14:29
ograah, not a local mailserver, just a client14:30
mikeulwhy?  who gets an e-mail?14:31
loolmikeul: the email contact of the lp project14:31
loolmikeul: if you prefer, you can just submit your branch for merging via the web ui14:31
ograright14:31
mikeulOK, so having pushed it to a personal branch, can I do that with "Propose for merging"?14:32
ograyep14:32
mikeulOr should I push it to the project-rootstock page instead and propose-for-merge there?14:33
persiamikeul: I usually push to some personal lp page (e.g. lp:~mikeul/project-rootstock/make-everything-better ) and then do the merge proposal on LP.14:34
ogramikeul, hmm, why do you add rmdir $BUILDDIR to the usage() function ?14:40
ogra$BUILDDIR wont exist if usage() is executed14:41
ogramikeul, i think that requires more restructuring, i'm not really comfortable with creating $BUILDDIR just because non root users call --help, could you move the variables around a bit instead of adding the rmdir ?14:45
ograits a good approach though14:46
mikeul$BUILDDIR _did_ exist when usage() is executed.  I called "sudo rootstock --help" a few times and ended up with several empty /tmp/tmp.xxxxxx dirs.14:54
ograright, thats the actual bug :)14:54
mikeulI'd be happy to move some things around to fix it differently.14:54
mikeule.g. to avoid creating /tmp/tmp.xxxxxx in the first place.14:55
ograyes, BUILDDIR= and all variables that use ${BUILDDIR} need to move down a bit ... best is below the uid 0 check14:55
ogra(the script started as 50 lines ... and grew functions in several directions)14:56
mikeulogra, I'm familiar with that.14:56
ogra:)14:57
mikeulOK, then move all the variable defs to after the cmdline parsing.  I'll move the qemu-system-arm and debootstrap checks, too, since they're also not needed e.g. for --help15:00
ograbe careful though, some need to be initialized before the cmdline parsing15:02
ograbest to only move everything with $BUILDDIR in it15:02
mikeulOh yeah, that's why I went the quick-and-dirty rmdir route.15:02
mikeulOK, can you give me an overview of how to handle that with bazaar?  I can start from scratch and just post a new branch, but I'm guessing there's a different way.15:04
ograjust delete the rmedir with your next change ... the moving of the uid 0 check is fine15:05
ogra*rmdir15:05
mikeulah, OK, so just dump more commits on top of my existing ones?15:06
ograright "revert last change and .... "15:06
mikeulfyi, I'm a git'ter.  Do you mean a) I can change the history in bzr to remove mention of 'rmdir' or b) I should make the changes "on top of" the existing ones.15:08
ograi would just do it on top15:09
mikeulOK15:09
mikeuleasy peezy15:09
ynezzwhois mikeul15:10
ynezzoops15:10
ynezz:)15:11
mikeulFYI, I'm not going to get around to doing even those small changes today, will get to it soon.  Signing out.15:27
ogralool, so i am testing the same iso-codes issue in a real VM now instead of using rootstock, i'm at the point where apt stops moving in "unpacking iso-codes" ... i see nothing in the logs, apt-get and dpkg dont seem to consume to much CPU, one intresting thing i see is that dpkg.log seems to be behind and getting slowly filled ... dou you have any idea where else to look at ?15:29
ogramikeul, dont worry :)15:29
ogramikeul, thanks for doing the work :)15:29
mikeulKein Problem, ogra, have a nice weekend.15:30
ogradu auch :)15:30
ogragrmbl ... i should have installed iotop first15:31
ynezzogra: build karmic image today, using rootstock r82 and it boots :)15:37
ogracool !15:38
ograsigh ... no strace either in ubuntu-minimal15:41
loologra: Are you swapping?15:42
loolynezz: We figured out the mountall issues with mikeul BTW15:42
ograi created a swapfile on the virtual disk after it hung already for 20min15:42
loolynezz: Missing rw on the kernel cmdline15:42
ynezzogra: can't login tho, maybe it's related to useradd http://paste.ubuntu.com/384439/15:43
ograyeah, looks like a bug15:43
ynezzlool: yep, saw it in bzr log, thanks15:43
ograi switched the default to install oem-config instead ... i havent tested with username/pw for a while ... its on my list15:44
ograynezz, i fear you need to mount your image and chroot into it to create the user15:44
ynezzdid it already, no problem15:44
ograah, k15:45
ynezzjust letting you know15:45
ograyeah, thanks15:45
ograseems the options are wrong, i'll look into it on the weekend15:45
ogralool, so swapon after it was already stuck didnt help ...15:45
ograeven though i see 204 bytes of the swap being used now after 1h15:46
ograerr, 24015:46
ograapt-get now jumps between 80 and 95% CPU15:46
ograbut nothing happens beyond that15:47
=== bjf-afk is now known as bjf
ynezzogra: it's missing " around one param http://paste.ubuntu.com/384443/15:53
ograoh, right, it used to live directly in the installer script ... when i moved it i didnt add proper quoting15:53
ograynezz, thanks !15:53
ynezznp, that was easy :p15:54
ogra(in fact i didnt really care about it because oem-config should become the default)15:54
ograbut i'll add your fix with the next commit15:54
asacbug 50750316:14
ubot4Launchpad bug 507503 in linux-mvl-dove (Ubuntu Lucid) (and 3 other projects) "VFP/NEON state is not preserved around signal handlers, causing state corruption between user processes (affects: 1)" [High,Fix committed] https://launchpad.net/bugs/50750316:15
NCommanderramana: morning (or afternoon)16:16
asacbug 52852416:18
ubot4Launchpad bug 528524 in pulseaudio (Ubuntu Lucid) (and 1 other project) "Sound not working in all apps on dove (affects: 1)" [High,New] https://launchpad.net/bugs/52852416:18
alecrimI got this problem with ubuntu karmic http://paste.ubuntu.com/384586/ . I fixed it at Mamona with this patch (http://gitorious.org/mamona/openembedded/commit/4d4933fd97b85a52e48a5fa1a97bd34532864e32). Anyone with this problem?19:24
NCommanderARGH($*)(@*#$)(*@!$#@$20:39
* NCommander hates OOo20:39
loolalecrim_: Could you file a bug with your patch?21:18
loolalecrim_: I'm pretty sure I saw this in the past, I think it was on the sheevaplugs21:18
loolalecrim_: http://bugs.debian.org/cgi-bin/bugreport.cgi?bug=31433421:21
ubot4Debian bug 314334 in apt "apt: APT doesn't work on filesystems without shared writable mmap(), like JFFS2." [Wishlist,Open]21:21
alecrim_lool, ok!21:43
|nfecteDrcn-ee: you around?22:05
NCommanderramana: ping?22:20
|nfecteDanyone want to give me whatever boot.scr they are using for lucid alpha 3?22:21
NCommander|nfecteD: what are you trying to do?22:25
* NCommander is the boot.scr writer guy22:25
|nfecteDgetting it to boot :/22:25
NCommander|nfecteD: which SoC?22:25
|nfecteDbeagleboard22:26
|nfecteDinit: plymouth-log main process (227) terminated with status 11122:26
NCommander|nfecteD: Beagleboard isn't officially supported by Ubuntu22:26
|nfecteDthats the last thing i get during boot before it seems to hang22:27
NCommander|nfecteD: http://elinux.org/BeagleBoardUbuntu - but here's the best set of directions I can give22:27
NCommanderoh wow22:27
* NCommander feels honored22:27
NCommanderBeagle adopted my boot.scr mechanism22:27
NCommanderor something similar22:27
NCommander|nfecteD: sounds like your making it to the initramfs and then its crashing.22:28
|nfecteDyup22:28
|nfecteDseems so...22:28
NCommanderI'd recommend trying #beagle and if no response, come back here22:28
|nfecteDplymouth needs kms/framebuffer and initramfs i believe...22:29
|nfecteDearly patch, i'm messing with:22:29
|nfecteDhttp://bazaar.launchpad.net/~robertcnelson/project-rootstock/initramf...22:29
|nfecteDand boot.scr needs to be updated (simple) to also load the initramfs22:29
|nfecteDinto ram..22:29
|nfecteDfound that on the beagleboard group now22:29
|nfecteDi would think rcn built the newest rootfs with that patch so what im missing is what goes in the boot.scr22:31
* |nfecteD pokes rcn-ee22:31

Generated by irclog2html.py 2.7 by Marius Gedminas - find it at mg.pov.lt!