**** BEGIN LOGGING AT Tue Oct 23 03:00:00 2012 Oct 23 08:30:05 morning all Oct 23 08:37:46 hi bluelightning Oct 23 09:26:51 good morning Oct 23 09:27:37 hi mckoan, afournier Oct 23 10:39:30 quick question, opinionated answers expected, how do you manage your network connection on an embedded device. Only ethernet with options of static IP or dhcp. Current methods I'm thinking about is connman (through dbus?), or parsing of config file and restart network daemon. Oct 23 10:40:04 Changed will be made from a C application. Oct 23 10:40:16 s/changed/changes Oct 23 10:44:52 jackmitchell: connman does rely almost completely on dbus for configuration, yes Oct 23 10:45:04 works well though if that's OK for your application Oct 23 12:31:10 i am working on a mysql recipe, and i use the adduser class, i would like the homedir to be in /var/mysql, but i don't know which of those options is the best: to let add user create the homedir, or to do it in the do_install ..? Oct 23 12:38:44 afournier: I think the useradd class is supposed to take care of that... Oct 23 12:38:48 afournier: I presume you're aware that meta-oe has a mysql recipe already? Oct 23 12:39:06 yep Oct 23 12:40:58 the one in meta-oe does not use the adduser class :/ Oct 23 12:41:42 it does a "grep mysql /etc/passwd || adduser --disabled-password --home=/var/mysql --ingroup nogroup mysql" in pkg_postinst_mysql5-server Oct 23 13:06:50 afournier: ah, right Oct 23 13:07:16 afournier: the recipe probably predates useradd.bbclass and hasn't been updated to use it Oct 23 13:07:32 patches more than welcome... Oct 23 13:14:40 good Oct 23 13:15:10 i am still learning oe, i'll send patch when i'll feel more confortable with it Oct 23 13:15:47 and i still have this post to do on the ML regarding linux 2.6.32 Oct 23 13:30:41 bluelightning: is it weird that glib is put into glib-2.0 directory in the rootfs lib folder, but the headers in the glib-2.0 folder point to includes in the glib/ folder which doesn't exist Oct 23 13:31:17 jackmitchell: not sure, you'd have to compare it with how it is packaged elsewhere Oct 23 13:33:46 bluelightning: I was hoping a nice simple dbus interface with connman would be easy but it's a pain the neck with multiple dependencies - not really embedded friendly for an 'embedded network manager' Oct 23 13:34:17 bluelightning: not to blame you, just venting ;) Oct 23 13:35:42 jackmitchell: multiple dependencies for connman itself or for using a dbus interface? Oct 23 13:37:01 bluelightning: dbus interface - I wouldn't really expect to be expected to use glib in a piece of embedded software Oct 23 13:37:41 I guess it depends on how you define embedded Oct 23 13:37:59 I suppose Oct 23 13:38:39 jackmitchell: connman itself has its own C wrapper for dbus IIRC, I don't know if that's easier to use than the one provided in glib Oct 23 13:39:19 bluelightning: oh really, I will investigate Oct 23 13:39:44 not sure if it's easy to use outside of connman itself though Oct 23 13:43:41 actually connman itself needs glib, so I'm not sure you'll save much Oct 23 13:59:33 connman itself is not very embedded friendly :( its api uses variadic types which are a pain to use with C, and it does not work with nfs rootfs :( Oct 23 14:01:03 http://code.google.com/p/badvpn/wiki/NCD sounds very interesting Oct 23 14:13:15 ensc|w: I'm pretty sure you can tell connman to ignore the interface being used for nfs Oct 23 14:13:37 at the moment I think we may be disabling it completely in that case though Oct 23 14:14:39 great, yet another way to handle network interfaces... Oct 23 14:24:02 bluelightning: yes; I can do this. But then, DNS is not working because DHCP won't be used Oct 23 14:24:17 ditto for ntp Oct 23 14:26:56 all this managed stuff sucks when you want to do things which are not covered 100% by its design :( Oct 23 14:27:04 ugh another network manager Oct 23 14:29:12 ccssnet: there exists another one, written by a SUSE employee... don't remember its name atm... Oct 23 14:30:01 sigh Oct 23 14:30:04 wicd Oct 23 14:30:11 ya that ones good Oct 23 14:30:46 but from an administrative perspective the more that exist. the more headaches on script automation of systems your in charge of Oct 23 14:30:47 lol Oct 23 14:31:16 i suppose manual is best option there either way Oct 23 14:32:23 no, not wicd, but https://gitorious.org/wicked Oct 23 14:32:40 o man... another one? Oct 23 14:32:49 for a gui app, wicd is nice Oct 23 14:32:57 * ccssnet looks at wicked Oct 23 14:34:10 yet another package in the repos Oct 23 14:34:22 in otherwords downloaded and bookmarked Oct 23 14:34:50 to bad its xml though Oct 23 14:35:00 they should have tried json... Oct 23 14:35:28 or anything thats easier to type Oct 23 14:35:31 lol Oct 23 14:58:57 otavio, msm: is there a best contact point for issues with meta-fsl-arm? Oct 23 15:21:54 ensc|w: it's another dbus implementation though Oct 23 15:22:16 ensc|w: they say it's going under a re-write so I wonder how it would be different from connman Oct 23 15:24:45 jackmitchell: I do not have anything against dbus. Indeed, dbus is a good choice for this task. But the connman api is horrible because everything is hidden behind a variant type. This makes configuration by 'dbus-send' impossible and requires lot of ugly C code. And I really do not want to use Python on embedded devices Oct 23 15:27:20 ensc|w: interesting I will have to investigate - first though to get dbus-dev into my SDK so I can actually start playing with it! Oct 23 15:27:48 I do not know, how much different wicked is... I am developing on lower levels only but have to recommend something to our customers and heard lot of complaints about connman Oct 23 15:27:58 ccssnet: I agree, I much prefer json to xml, xml is so... longwinded Oct 23 15:29:33 ensc|w: variants are totally possible with dbus-send... Oct 23 15:30:22 bluelightning: no; not when they within a dict or dict is a variant Oct 23 15:31:05 bluelightning: else, please show me a way to configure connman to use dhcp or set a static ip Oct 23 15:31:25 ensc|w: I'm not a connman expert by any means Oct 23 15:54:02 morning Oct 23 15:56:50 rburton: I'm leaving work now so can pick this up again tomorrow if you need anymore info Oct 23 15:57:21 jackmitchell: my advice is clean those hits and fingers crossed. no idea how your sysroot got messed up. Oct 23 15:57:25 jackmitchell: just cleansstate those found references like xrender etc Oct 23 15:57:47 i propose pruning .la files from sysroot Oct 23 15:58:25 iirc pb_ sent patch to do it globally Oct 23 15:58:30 ensc|w: upgrade to latest connman, it has a cli tool Oct 23 15:58:44 ensc|w: (and then submit the patch to oe-core :) Oct 23 15:58:50 kergoth: the mailing list Oct 23 15:59:08 kergoth: or if it is something sensitive, mail me privately Oct 23 15:59:23 kergoth: what is it about? Oct 23 15:59:48 kergoth: https://groups.google.com/group/meta-fsl-arm <= btw Oct 23 16:00:25 rburton: it's weird because this is my production box, so not even any dodgy development goes on here Oct 23 16:01:05 jackmitchell: agreed, weird. there was some cleanup to the x deps but nothing that shoud have produced that problem Oct 23 16:01:10 (and that problem is why everyone hates libtool) Oct 23 16:01:41 bitbake -c cleansstate libxrandr libxrender libxext Oct 23 16:01:51 bitbake -c cleansstate nativesdk-libxrandr nativesdk-libxrender nativesdk-libxext Oct 23 16:02:02 rburton, jama: still fails with same error message Oct 23 16:02:20 will pick back up tomorrow and do some more investigating! Oct 23 16:02:23 rburton: good point re the new connman client :) Oct 23 16:02:29 jackmitchell: did you use cleansstate with -f? Oct 23 16:02:44 jackmitchell: how annoying Oct 23 16:02:48 no Oct 23 16:02:50 why would -f be needed with cleansstate? Oct 23 16:02:57 it's nostamp Oct 23 16:03:01 but it ran the tasks as far as I could tell Oct 23 16:03:15 otavio: ah, right, somehow i missed that list. thanks. imx6qsabrelite is failing to build due to imx-base sucking in imx-lib, yet imx-lib is listed as compatible only with mx5. similar issue with firmware-imx. this only bit me after an update, so it likely changed in the past couple weeks Oct 23 16:03:18 bluelightning: to force it rebuilt? or what taints that checksum? Oct 23 16:04:12 right, tara all, rburton: i may nag you a bit about connman tomorrow if that's ok - I see you have an article written on the front page (even if from 2004) :P Oct 23 16:04:26 haha Oct 23 16:04:49 rburton: or maybe it was a dbus articlem either way... Oct 23 16:04:50 JaMa: that would only help if you were forcing a task that's in the normal dependency chain Oct 23 16:05:00 jackmitchell: yeah, dbus. old api, ignore it :) Oct 23 16:05:19 rburton: everything I read about dbus is old, sigh... Oct 23 16:05:36 bluelightning: can you explain? Oct 23 16:06:21 rburton: my dev machine is showing the same .la traits - so it's fairly reproduceable Oct 23 16:06:53 jackmitchell: what recipe is failing? nativesdk-libx11? Oct 23 16:06:58 kergoth: I am sure I fixed it ... Oct 23 16:07:03 yes Oct 23 16:07:13 jackmitchell: a copy of your sysroot libdir .la files might be useful Oct 23 16:07:15 kergoth: are you using an update repository? Oct 23 16:07:21 | /home/jack/Projects/poky-r0005/R0005/tmp/sysroots/x86_64-linux//bin/sed: can't read /home/jack/Projects/poky-r0005/R0005/tmp/sysroots/x86_64-nativesdk-pokysdk-linux/opt/poky/1.2+snapshot/sysroots/x86_64-pokysdk-linux/usr/lib/libXau.la: No such file or directory Oct 23 16:07:23 jackmitchell: before you start removing stuff on the dev machine Oct 23 16:07:24 JaMa: if your aim is to force dependent recipes to re-build the next time they are called for, you have to force-build (i.e. taint) a task that they actually depend upon; clean / cleansstate are not depended upon as part of the normal task execution for building a recipe Oct 23 16:08:11 kergoth: let me confirm it Oct 23 16:08:22 bluelightning: but without -f I sometimes see the same broken sstate package being redownloaded from SSTATE_MIRROR (or at least it seems like that) Oct 23 16:08:34 rburton: the only cleansstate I have done on dev is some earlier packages which it now passes on and a single run of -c cleansstate nativesdk-libx11 Oct 23 16:08:54 rburton: so it may already be too late :( Oct 23 16:08:57 kergoth: indeed Oct 23 16:09:19 JaMa: bitbake can't help stale packages on a mirror it can't write to... Oct 23 16:09:22 kergoth: I will provide you a patch ... hold sec Oct 23 16:10:26 rburton: 552 bitbake meta-toolchain-r0005 Oct 23 16:10:26 553 bitbake -c cleansstate binutils-crosssdk Oct 23 16:10:26 554 bitbake meta-toolchain-r0005 Oct 23 16:10:26 555 bitbake -c cleansstate nativesdk-gmp Oct 23 16:10:26 556 bitbake meta-toolchain-r0005 Oct 23 16:10:26 bluelightning: then I didn't understand tainging properly (I was expecting it to force ignoring sstate package even when it's on mirror) Oct 23 16:10:27 557 bitbake -c cleansstate alsa-lib Oct 23 16:10:27 558 bitbake meta-toolchain-r0005 Oct 23 16:10:28 559 bitbake -c cleansstate nativesdk-alsa-lib Oct 23 16:10:28 560 bitbake meta-toolchain-r0005 Oct 23 16:10:29 561 bitbake -c cleansstate nativesdk-gdbm Oct 23 16:10:29 562 bitbake meta-toolchain-r0005 Oct 23 16:10:30 563 bitbake -c cleansstate nativesdk-glib Oct 23 16:10:30 564 bitbake -c cleansstate nativesdk-glib-2.0 Oct 23 16:10:31 565 bitbake meta-toolchain-r0005 Oct 23 16:11:26 JaMa: if you taint the right task, sure, it will result in the checksum being different Oct 23 16:11:55 JaMa: worth noting however is that cleansstate will delete stamp files for all tasks for the recipe and that will include the taint file Oct 23 16:12:03 this time i'm definitely off, will talk tomorrow if the fog doesn't engulf me ;) ! Oct 23 16:14:09 jackmitchell: it's positively lifted from earlier Oct 23 16:17:16 kergoth: You're building master right? Oct 23 16:18:20 kergoth: http://paste.debian.net/203018/ should work just fine; gst-fsl-plugin will trigger imx-lib inclusion in mx5 Oct 23 16:18:33 kergoth: can you give it a try Oct 23 16:29:59 otavio: will do, thanks Oct 23 16:42:44 Hi. Is there a way to get a reverse dependencies graph for a given recipe? Oct 23 16:42:53 (including indirect dependencies) Oct 23 16:46:26 mario-goulart: you'd probably have to post-process bitbake -g output unless you want to get dirty with bitbake code Oct 23 16:59:36 bluelightning: If I understand correctly, what bitbake -g does is "given a recipe, generate a graph of the recipes it depends on". I actually want reverse dependencies (given a recipe, which recipes depend on it). Is -g able to do that (assuming I won't run -g e for all recipes). Oct 23 17:00:12 mario-goulart: you could try world as a target Oct 23 17:03:02 bluelightning: hmmm! Interesting. Thanks for the tip. Oct 23 17:04:14 some things will be excluded from that though as it's intended to be a buildable target - you could use universe if that's an issue Oct 23 17:04:39 it also won't catch recipes that are skipped for your configuration, but you may not care about those Oct 23 17:05:07 Ah, ok. Good points. Oct 23 17:07:38 otavio: that took care of it, thanks Oct 23 17:11:05 otavio: ping Oct 23 17:12:20 kergoth: I will send it to the mailing list Oct 23 17:13:01 anyone know where I can find a recent linphone recipe? Oct 23 17:15:30 otavio: thanks, appreciate it Oct 23 17:20:49 chouimat: nothing newer than in oe-classic that I'm aware of Oct 23 17:23:21 bluelightning: ok thanks I found some patches that bring 3.4.3 but they don't seems to work .. Oct 23 21:58:34 does someone have simple script to search sysroot if there are any files not tracked by sstate manifest? e.g. if someone touched sysroot from do_install directly? **** ENDING LOGGING AT Wed Oct 24 03:00:01 2012