**** BEGIN LOGGING AT Tue Jul 26 02:59:56 2011 Jul 26 06:29:33 someone noticed that openembedded-commits@openembedded.org no longer works? is it because git is not on melo now? Jul 26 07:36:20 good morning Jul 26 07:38:57 morning Jul 26 07:39:12 (russian) http://www.ak-systems.ru/jml/ru/production/item/58-ip-plug Jul 26 07:39:26 new plug-style computer made in russia Jul 26 07:39:35 based on Kirkwood 88F6283 Jul 26 07:42:24 Jay7: is it your company? Jul 26 07:43:32 Jay7: unfortunately I'm unable to find english pages Jul 26 07:44:50 mckoan: use google translator Jul 26 07:45:06 it's not my company :( Jul 26 07:45:32 price is expected about $100-150 Jul 26 08:04:45 Hi guys Jul 26 08:06:12 I need to use eglibc 2.13 in my build but I'm still using yocto 1.0 (that has eglibc 2.12) and I cannot update to yocto 1.0.1. Can I simply download the new eglibc recipes from 1.0.1 and put them in the yocto 1.0 meta tree? Jul 26 08:06:25 Do you think it will work? Jul 26 08:22:15 Jay7: seems quite nice, pity they did not export the sata interface too Jul 26 08:22:52 eFfeM_work: do Kirkwood 88F6283 have sata? Jul 26 08:23:06 i thought so, let me check Jul 26 08:23:20 I'm not very familiar with platforms/SoCs :) Jul 26 08:25:43 Jay7: it has: http://www.marvell.com/processors/embedded/armada_300/assets/armada_310.pdf Jul 26 08:25:56 hm.. interesting Jul 26 08:26:05 pity they did not go for the 6282 Jul 26 08:26:23 the openrd boards also have the sata and some of the plugs too Jul 26 08:26:33 we are waiting for second line (independent from ak-systems) in november Jul 26 08:27:25 I used to do some prototyping with openrd, but unfortunately that did not result in a followup project Jul 26 08:27:33 other company will produce plug-style computers based on some Marvells too Jul 26 08:27:59 have no tech. details.. Jul 26 08:28:05 what I like about the ak one compared to the original sheeva is the dual usb and the wifi+bt Jul 26 08:28:38 2xGbLAN is nice too :) Jul 26 08:28:42 when marvell announced the sheeva they said they were eventually aiming for a $ 50 system, but unfortunately that was never made Jul 26 08:29:12 dual lan is nice if you want to do router functionality, if you want to turn it into a nas it is not too useful Jul 26 08:30:05 NAS with USB only is bad idea :) Jul 26 08:31:14 anyway, I'll try to create print-server firmware for my customers Jul 26 08:31:21 and usb-server probably Jul 26 08:31:29 s/usb/ups/ Jul 26 08:31:51 yeah that is why I miss the sata, (guess nas with usb3 would also be possible, but this one is only usb2) Jul 26 08:31:52 i.e. convert non-network UPS to network-capable UPS :) Jul 26 08:33:40 sounds nice Jul 26 08:38:59 Jay7: i browsed the globalscale web site, dream plug is also only $ 149 and supports no nand though Jul 26 08:39:06 http://www.globalscaletechnologies.com/c-5-dreamplugs.aspx Jul 26 08:39:44 this one is nice too: http://www.globalscaletechnologies.com/c-8-d2plug.aspx for playback as it has hdmi out but is 249 Jul 26 08:41:40 and iirc the armada5 series is armv7 or so instead of armv5 Jul 26 08:51:26 where is the right place to place a PREFERRED_PROVIDER_qt4-x11-free ="qt4-x11-free-gles" ? Jul 26 08:51:46 your distro config or local.conf Jul 26 08:51:56 I found a NAK on a patch where one wished to place it in a mymachine.conf Jul 26 08:51:59 JaMa, thanks Jul 26 08:52:00 :)= Jul 26 08:53:49 eFfeM_work: I got first prices :) Jul 26 08:54:23 version with 2xGbLAN w/o usb/wifi is now $130 Jul 26 08:54:24 Jay7: and? how much ? Jul 26 08:54:38 ah ok Jul 26 08:54:40 base model with all ifaces will be $150 Jul 26 08:54:58 that is roughly the same as the dreamplug Jul 26 08:55:16 I'm sure they was refereced to dreamplug :) Jul 26 08:55:26 *referenced Jul 26 08:56:02 well.. anyway, good that we got first cheap enough local devices Jul 26 09:01:06 what a waste.. 2MB SPI NOR Flash for uboot Jul 26 09:04:35 Jay7, which board ? Jul 26 09:10:29 alx_: (russian) http://www.ak-systems.ru/jml/ru/production/item/58-ip-plug Jul 26 09:12:15 spasiba Jul 26 09:12:28 alx_: :) Jul 26 09:13:01 wondering what may be the4 use of that object :) Jul 26 09:15:12 alx_: router/access point/home NAS Jul 26 09:15:30 original usage is data sensors as I understand Jul 26 09:16:23 hi all Jul 26 09:16:27 there are some relations between russian magistral provider and ak-systems Jul 26 09:16:44 hi bluelightning Jul 26 09:17:38 well.. I should go now Jul 26 09:44:34 Guys...is there a "standard way" to compile a package in debug mode? I mean with the -g flag set ? Jul 26 09:44:52 iirc DEBUG_BUILD in local.conf Jul 26 09:45:23 tnx!!! Jul 26 09:45:37 but better check if it's the same in oe-core Jul 26 09:53:01 khem`: ping? Jul 26 10:37:51 Does anyone can hel me with remote debugging using gdbserver? I've compiled my app in debug mode but I still can't see symbols...it seems that the symbols table is missing Jul 26 10:42:34 salut , il y a-t-il quelqu'un qui parle francais?? Jul 26 10:47:42 mrAlmond: very vague question, does gdb have access to the debug symbols? Jul 26 10:51:15 zecke : I'm trying to do a remote debugging with gdbserver on an ARM machine so I'm using gdbserver on the target and the toolchain gdb on the host Jul 26 10:51:51 the gdb on the host reads the symbols from my application as I pass it as parameter to gdb Jul 26 10:52:11 but it will probably not read symbols of all the shared libraries Jul 26 10:52:11 mrAlmond: I understood, but when you do arm-..-gdb binary on the host, does it have access to the debug symbols Jul 26 10:52:24 mrAlmond: e.g. we are in the context of OE, do you have -dbg packages installed? Jul 26 10:52:41 mrAlmond: installed == installed in the rootfs used on the host. Jul 26 10:53:19 no..on the target there are no dbg package installed...I've just copied the binary compiled with debug symbols Jul 26 10:54:30 I'm debugging the bluez daemon bluetoothd Jul 26 10:55:03 as it crashes on my ARM platform with a "glibc detected" free() error Jul 26 10:55:05 12:46 < zecke> mrAlmond: installed == installed in the rootfs used on the host., talking about host here Jul 26 10:55:51 in general I either use gdb on the target, or let it dump core and then use gdb on the host with a debug rootfs (-dbg packages installed) Jul 26 10:55:55 yes, no debug packages installed on the host...just the single bluetoothd binary with debug symbols launched manually Jul 26 10:56:47 zecke : understood...so I should create a debug rootfs Jul 26 10:57:36 mrAlmond: well, I have not really used GDB 7.x for ARM (so there might be bugs). I would try to install -dbg packages and use that rootfs with gdb on the host. Jul 26 10:57:57 mrAlmond: alternatively try to read the coredump, or use catchsegv and see how well that goes Jul 26 10:58:45 in general having a backtrace on ARM can be a bit difficult, e.g. O2 disables frame-pointers, I don't know if GDB will always try to use the DWARF2 symbols or will still scan the prologue/epilogue of a function Jul 26 10:58:51 at the moment the core is not dumped Jul 26 10:59:16 I don't remember how to enable it...I must check ;-) Jul 26 10:59:17 ulimit -c unlimited Jul 26 10:59:20 wow Jul 26 10:59:30 that was easy! :-D Jul 26 11:00:17 hi all Jul 26 11:02:44 *** glibc detected *** bluetoothd: free(): invalid next size (fast): 0x2a08e660 *** Jul 26 11:02:45 Aborted Jul 26 11:02:51 but no core was dumped Jul 26 11:02:54 hi zecke Jul 26 11:10:15 mrAlmond: hmm, the above should enable it, did you first type it and then start program in the same shell? Jul 26 11:21:32 pb_: my journey in syscall's wonderland was not productiive... _syscallX have been removed (broken on many archs) and now it looks like klibc avoids by design the indirect system calls done with syscall() Jul 26 11:22:46 while kexec_load(2) states " Currently, there is no glibc support for kexec_load(). Call it using syscall(2)." Jul 26 11:23:06 lol Jul 26 11:23:10 stupid glibc Jul 26 11:24:03 but in SYSCALLS.def you can define your own stub it seems Jul 26 11:31:57 ^'' This is a list of system calls we invoke "directly". These are generated into syscall stubs in their own files, so the linker can do its job properly.'' Jul 26 11:32:34 apparently adding the prototype here is the correct way Jul 26 11:50:47 zecke: yes..I run that command before my app on the same shell Jul 26 11:51:25 there is also /proc/sys/kernel/core_pattern, but ulimit -c unlimited should work. Jul 26 12:10:47 zecke : The program exited with "abort" and not with a seg-fault...is the core dumped only when segfaulting? Jul 26 12:11:18 "Aborted", not "abort". Jul 26 12:12:21 i think it should dump core as well, is something printed on the console? Not dumping core or such? Jul 26 12:14:15 what could i do about the tmp/sysroots/ direcotrys to get some space removed or cleand up ? Jul 26 12:14:27 rob_w: do you use rm_work? Jul 26 12:15:43 yes but sometimes i turn it off to get a view into tmp/work Jul 26 12:41:58 zecke: unfortunately nothing is dumped Jul 26 12:42:20 hi to compile some code i'm looking for tcl.h is there something like tcl-dev? Jul 26 12:43:43 mrAlmond: next proposal... -fno-omit-frame-pointer, recompile everything, use catchsegv Jul 26 12:44:45 for "everything" you mean really everything or just the application I'm debugging? Jul 26 12:47:35 everything Jul 26 12:47:44 ok tnx Jul 26 12:48:39 or alternatively (if you have enough storage), install -dbg packages on the target and use the gdb on the target Jul 26 13:18:31 right, finished my initial version of the oe-core summary doc: http://www.openembedded.org/index.php/OpenEmbedded-Core Jul 26 13:18:40 comments welcome Jul 26 13:25:08 bluelightning: looks good Jul 26 13:25:52 hmm, we still need a link to the layers list, and an OE-core-based distros list page Jul 26 13:26:47 bluelightning: maybe "at varying states of repair" makes oe-dev to look like old unmaintained garbage and I don't think it is so bad Jul 26 13:29:43 bluelightning: and IMHO scope of meta-openembedded/meta-oe can be defined as "recipes which are needed by more than 1 meta-foo layer or by more then 1 meta-distro layer to avoid duplication" Jul 26 13:30:55 JaMa: yes it is a bit harsh I shall remove it Jul 26 13:32:53 for the second thing I will adjust the language slightly Jul 26 13:34:35 till later Jul 26 13:36:47 done Jul 26 13:36:59 JaMa: ^ Jul 26 13:37:40 FWIW I think that meta-openembedded is a bit more than that right now... the problem is how things are (or aren't) moved from there to oe-core Jul 26 13:38:56 bluelightning: btw layer index seems a bit broken now Jul 26 13:39:18 bluelightning: setup link for angstrom is missing Jul 26 13:39:47 JaMa: is there a specific page about oe-core based Angstrom now? Jul 26 13:40:18 JaMa: what else is wrong with the LayerIndex? Jul 26 13:41:52 bluelightning: this is better "setup" page for SHR http://shr-project.org/trac/wiki/Building%20SHR Jul 26 13:42:49 JaMa: all I can find is http://www.angstrom-distribution.org/building-angstrom which hasn't been updated since March 2010 Jul 26 13:43:08 JaMa: OK i will change to that link for SHR Jul 26 13:44:10 bluelightning: there is howto from khem http://sakrah.homelinux.org/blog/2011/03/using-openembedded-core-to-build-angstrom-for-qemu/ Jul 26 13:44:21 but not sure if it can be used as official howto Jul 26 13:44:41 JaMa: yes, it's not really right for an official link Jul 26 13:44:57 bluelightning: but point was that there is missing cell in table where the angstrom setup link should be Jul 26 13:45:03 yes Jul 26 13:45:10 kick the lead dev ;) Jul 26 13:45:10 hmm Jul 26 13:51:44 next time I go to the wifi cafee I'll try not to forget to take my external hdds to build shr-core Jul 26 13:52:04 hehe Jul 26 14:15:25 hmm Jul 26 14:15:43 khem`? ka6sox? Looks like the commits ML has been dead since the git move Jul 26 14:17:42 acked-by Jul 26 14:18:13 08:23:14 < JaMa> someone noticed that openembedded-commits@openembedded.org no longer works? is it because git is not on melo now? Jul 26 14:19:51 JaMa, is the oe.dev git ok? Jul 26 14:20:01 or does it have issues such as melo down Jul 26 14:20:04 or similar issues Jul 26 14:20:44 you need to use git from openembedded.org not .net Jul 26 14:20:53 I've updated SHR Makefile few days ago Jul 26 14:21:10 but that won't update repo url in existing checkouts Jul 26 14:27:29 ok Jul 26 14:27:32 I'll add remote Jul 26 14:27:53 GNUtoo|laptop: git config --replace-all remote.origin.url git://git.openembedded.org/meta-openembedded Jul 26 14:27:57 and similar Jul 26 14:28:08 and use github mirror where possible Jul 26 14:28:18 so it's still down? Jul 26 14:28:36 .net probably yes, .org still works Jul 26 14:29:38 ok Jul 26 14:29:49 changed trough nano .git/config Jul 26 14:34:57 for 2011.03-maintenance, which recipe should I bbappend to add new users to an image? Jul 26 14:49:45 why is oe-core using fnmatch patterns for PACKAGES_DYNAMIC although bitbake is interpreting it as a regular expression? Jul 26 14:49:57 or is there some layer which replaces the '*' with '.*'? Jul 26 14:50:54 I get some weird problems with intltool (while building shared-mime-info-native). It looks like intltool-update is being executed as sh script, instead of being caught by perl Jul 26 14:51:12 Has anybody seen that? Jul 26 14:52:30 lumag: do you have log? Jul 26 14:54:37 relevant part at http://pastebin.de/17957 Jul 26 14:56:42 I'm still using inherit perlnative to force it to use perl-native instead of perl from hosts.. but this looks different Jul 26 14:57:21 looks like it uses perl-native. Jul 26 14:58:20 Aha! Seems I got the problem. strace -f -e execve sh -c /home/users/mitya/OE/setup-scripts/build/tmp-angstrom_2010_x-eglibc/sysroots/x86_64-linux/usr/bin/intltool-update Jul 26 14:58:20 execve("/bin/sh", ["sh", "-c", "/home/users/mitya/OE/setup-scrip"...], [/* 23 vars */]) = 0 Jul 26 14:58:20 execve("/home/users/mitya/OE/setup-scripts/build/tmp-angstrom_2010_x-eglibc/sysroots/x86_64-linux/usr/bin/intltool-update", ["/home/users/mitya/OE/setup-scrip"...], [/* 22 vars */]) = -1 ENOEXEC (Exec format error) Jul 26 14:58:42 And then sh will use it as a simple shell script. Jul 26 15:04:14 lumag: i hit that yesterday, the bug is in the intltool recipe Jul 26 15:04:23 it does export PERL_virtclass-native = "/usr/bin/env perl" Jul 26 15:04:25 what it should be: Jul 26 15:04:31 PERL_virtclass-native = "/usr/bin/env perl" Jul 26 15:04:32 export PERL Jul 26 15:04:52 flags don't get pulled across for overrides, so "PERL_virtclass-native" gets exported, but "PERL" does not Jul 26 15:05:11 kergoth_: thanks Jul 26 15:05:21 haven't had a chance to send it to the list yet, but i verified that fixes it for me Jul 26 15:05:39 kergoth_: what about without this export? (works here) Jul 26 15:06:38 if you don't export it, it'll use the hardcoded full path to the perl binary, breaking use of sstate, and potentially just exploding horribly if you're in a deep path (the #! line can only be a certain length) Jul 26 15:06:54 ah, ok.. #!/OE/shr-core/tmp/sysroots/x86_64-linux/usr/bin/perl-native/perl Jul 26 15:07:09 yep Jul 26 15:13:22 JaMa: BTW: could you please check your meta-smartphone repo. Looks like your last push didn't fully succeed. Jul 26 15:13:45 JaMa: compare http://git.shr-project.org/repo/meta-smartphone.git/info/refs?service=git-upload-pack with http://git.shr-project.org/repo/meta-smartphone.git/refs/heads/master Jul 26 15:57:34 are there any user and group manipulation tools for creating images in the 2011.03-maintenance branch? Jul 26 16:37:24 bbl, bye Jul 26 16:46:16 Hi Jul 26 16:46:32 sakoman: ping Jul 26 16:49:40 Tartarus: yes git is currently in transition Jul 26 16:50:02 Tartarus: once it reaches its final destination I will fix it Jul 26 16:55:22 k Jul 26 16:56:09 Was were any significant changes between .34 and .39 kernel regarding wireless or specifically libertas driver? Jul 26 16:57:35 davidlt: pong Jul 26 16:59:10 sakoman: I have restricted ad-hoc network with WEP encryption on .34 with libertas, but it doesn't work on .39 kernel, any hints? According to software on .39 ad-hoc network has no encryption and is open. Jul 26 16:59:17 6 Jul 26 16:59:29 oops Jul 26 16:59:45 sakoman: One more thing, any roadmap for 3.0 final kernel showing up on your rep? Jul 26 17:00:34 davidlt: no idea! I don't often work with ad hoc networks, so no direct experience Jul 26 17:01:15 davidlt: re: 3.0 final -- no client has asked for that so it is at the bottom of the priority list behind paying work :-) Jul 26 17:01:34 Ach :) Jul 26 17:01:40 davidlt: unless someone asks for it I may end up skipping it Jul 26 17:01:59 sakoman: I think .39 has memory leak with UBIFS Jul 26 17:02:18 davidlt: any patches to fix the issue? Jul 26 17:02:37 I say patch on 3.0 changelog Jul 26 17:03:11 sakoman: that's why I am interesting, because during last test with .39 I am loosing very little memory, but still loosing Jul 26 17:05:02 sakoman: UBIFS: fix memory leak on error path Jul 26 17:05:15 UBIFS leaks memory on error path in 'ubifs_jnl_update()' in case of write Jul 26 17:05:15 failure because it forgets to free the 'struct ubifs_dent_node *dent' object. Jul 26 17:05:15 Although the object is small, the alignment can make it large - e.g., 2KiB Jul 26 17:05:15 if the min. I/O unit is 2KiB. Jul 26 17:05:21 Ups, a bit too many lines Jul 26 17:05:35 davidlt: I'll look at adding that patch Jul 26 17:05:50 sakoman: commit: 812eb258311f89bcd664a34a620f249d54a2cd83 Jul 26 17:07:28 sakoman: If you do rebase again, ping Neil to update recipe on oe-gumstix as there was problems before Jul 26 17:07:45 use kmemleak to find out the leaks I would say Jul 26 17:09:53 khem: I will look into that. Thanks! Jul 26 17:11:03 you need CONFIG_DEBUG_KMEMLEAK and debugfs support turned on in kernel Jul 26 17:12:00 khem: already reading docs :) Jul 26 17:12:45 k Jul 26 17:24:44 tartarus, what monitors that? Jul 26 17:24:59 that might be something khem needs to look @ Jul 26 17:25:59 hi khem and ka6sox Jul 26 17:26:13 ka6sox: yeah, khem is aware Jul 26 17:31:41 hi woglinde Jul 26 17:31:53 Tartarus, good...if he needs my help he will holler Jul 26 17:33:33 fray: ping Jul 26 17:40:19 fray: I don't recall what the exact config is, but "classic" 32bit userland for ppc is the 32bit multilib for 64bit Jul 26 20:16:30 Hi guys. Is a `PR` bump needed when a meta data field like `HOMEPAGE` is updated? Jul 26 20:19:49 PaulePanter: yes, you want the new HOMEPAGE to be in packages. Jul 26 20:21:31 zecke: Thanks! **** ENDING LOGGING AT Wed Jul 27 02:59:57 2011