**** BEGIN LOGGING AT Fri Mar 22 02:59:58 2013 Mar 22 06:19:45 Packaging error compiling my sdk with the latest denzil: http://pastebin.com/CF6jSfky. Any ideas? Mar 22 06:19:52 And, good morning :) Mar 22 07:57:08 Hm. Can't figure this one out. When I do "bitbake udev" it compiles systemd-v189. That somehow seems significant. Mar 22 08:08:37 morning all Mar 22 08:10:01 morning Mar 22 08:11:03 good morning Mar 22 08:11:18 good morning Mar 22 08:11:48 tasslehoff: which machine? Mar 22 08:12:18 mckoan: beagleboard w/angstrom. Mar 22 08:13:32 mckoan: my rootfs compiles just fine, but the sdk fails with the pasted udev dependency error. Mar 22 08:15:22 tasslehoff: grep -r IMAGE_DEV_MANAGER meta* may point out some problem Mar 22 08:19:19 mckoan: some udev entries in angstrom files there Mar 22 08:19:59 mckoan: but what is the solution here. I suspect I don't want to remove those Mar 22 08:21:21 tasslehoff: if is angstrom related you may try asking on their channel too ;-) Mar 22 08:22:57 mckoan: now I can. thanks :) Mar 22 08:45:08 is it possible to build a distro-agnostic sdk? Mar 22 08:45:28 * tasslehoff ponders workarounds while waiting for answers on #angstrom Mar 22 08:50:09 tasslehoff: yes, you have to use oe-core only Mar 22 08:51:07 tasslehoff: also yocto will work with a charm ;-) Mar 22 08:52:36 mckoan: I'm indeed moving to yocto. I just have to release one more rootfs+sdk from our old angstrom setup scripts regime Mar 22 08:52:48 and getting this sdk to compile is the final task Mar 22 09:04:20 mckoan: you said the grepping might "point out some problem". do you understand why there's a problem with having udev in there when compiling my sdk? Mar 22 09:07:29 tasslehoff: I don't use angstrom Mar 22 09:11:53 gm bluelightning Mar 22 09:30:40 morning woglinde, all Mar 22 09:34:41 win 12 Mar 22 09:34:50 whoops Mar 22 09:36:00 hello Mar 22 10:50:00 hello Mar 22 10:50:11 world Mar 22 10:50:14 can someone build apache2-native? Mar 22 10:51:21 tasslehoff: angstrom is using systemd as init manager so that's why you got it built when called 'bitbake udev'. Mar 22 10:51:30 tasslehoff: as systemd provides udev Mar 22 10:57:52 someone can give me some links about native-sdk stuff of OE? Mar 22 10:57:56 hrw: I understood that, but not why the sdk didn't work. koen told me to find and rebuild the offending package, so I'm trying that now. Mar 22 10:58:05 i wonder why someone would use udev and systemd in embedded devices with fixed hardware and fixed purpose ... any reason for that? Mar 22 10:58:37 BlindMan: definition of embedded hardware differ between developers Mar 22 10:58:50 Well, systemd isn't particularly tied to hotplug. Mar 22 10:58:52 I need it for Qt/X11, but I found only stuff about Qt Embedded, Mar 22 10:58:57 BlindMan: not all use cases conform to that Mar 22 10:59:01 One might like the parallel init stuff, or the service monitoring. Mar 22 10:59:09 BlindMan: also lot of uses of OE is not so embedded like you mentioned Mar 22 10:59:43 for my work sysvinit is better than systemd Mar 22 10:59:54 hrw: i know, i know, hence the limitation of fixed hardware for fixed purposes ... Mar 22 11:00:22 hi blindman Mar 22 11:00:26 hi woglinde Mar 22 11:01:13 BlindMan: for target like beaglebone I think I prefer to have system more flexible than strictly embedded Mar 22 11:01:17 actually it worrys me that i need systemd for X stuff and the uberdependency on systemd for running a system at all with treating legacy like crap Mar 22 11:01:39 BlindMan: if you use angstrom then you use systemd. dot. period. Mar 22 11:01:59 hrw hm I am using it with sysvinit Mar 22 11:02:02 * woglinde runs Mar 22 11:02:12 woglinde: so you have own hacks/changes Mar 22 11:02:27 iirc angstrom went to systemd Mar 22 11:02:41 and I do not talk about hrw: from a system architecutre perspective, why would anything rely that heaily on systemd? (answer udev, i know) ... but then only udev with systemd is kind of a shitty design decision and as "user" yo ucan't do anything about it Mar 22 11:03:03 but with some tweaks you can use sysvinit again Mar 22 11:03:06 silvio_: what's the matter with meta-toolchain-sdk ? Mar 22 11:03:15 OE-Core will continue to support both systemd and sysvinit as options, FWIW Mar 22 11:03:39 BlindMan: I refuse to get into discussion about init systems. Mar 22 11:03:54 one prefer cherries, other like pears Mar 22 11:04:05 hrw: i was just wondering like educate myself, why would someone do that ... maybe i don't see the light? :) Mar 22 11:04:08 and speaking of pears... time for one Mar 22 11:04:17 hrw: wasn't meant as a religious war :) Mar 22 11:04:55 BlindMan: I do not use systemd on any of my machines because they are running ubuntu. But I do not care about upstart at all (even with my Canonical hat on). Mar 22 11:06:10 mckoan, I m looking for solutin, now I try the " meta-toolchain-sdk" and probably it solve my trouble, thanks ;) Mar 22 11:06:16 my fear is that systmd+udev interaction tends to be too complex for "simple" stuff Mar 22 11:06:38 BlindMan: but no one forces you to go that way Mar 22 11:06:40 and not easier, straight forward, self-explenatory Mar 22 11:11:02 BlindMan: everybody draws their line differently.... udev and/or systemd requires linux... linux is already too complex for "simple" stuff.... Mar 22 11:13:24 fatal: I know it's linux-only and yes, you can't know everything, but in certain settings it usually an evil necessity to care about everything :( (or have answers) Mar 22 11:15:20 BlindMan: that exact argument could have been made by my father who thinks computers are evil. Mar 22 11:15:57 BlindMan: maybe you're just getting too old. ;P Mar 22 11:17:06 fatal: no, it's called management :) Mar 22 11:17:42 keeping things simple have always been the best strategy for every fileld of application Mar 22 11:19:01 fatal: actually operating devices successfully, you need to know what it does when different scenarios arise Mar 22 11:19:27 fatal: at least that#s the goal, which can't be entirely reached of course :) Mar 22 11:19:49 certainly not with linux Mar 22 11:22:15 fatal: because ... linux not proofably correct or what do you mean? Mar 22 11:24:25 because linux is way too complex and with all the bugs fixed all the time I can guess you can say it's proofably in-correct. Mar 22 11:25:17 enough of this meta-discussion though... Mar 22 11:27:01 lol Mar 22 11:27:05 what a troll Mar 22 13:49:19 ok question I have Mar 22 13:49:30 coreutils - do we build it with ACL or without? Mar 22 13:49:41 cause I do not know which patch to send for merging Mar 22 13:53:52 sent Mar 22 14:23:25 did packagegroup-sdk-target replace task-native-sdk? Mar 22 14:26:50 spacecolonyone: I think so yes Mar 22 14:27:34 Know of any compatibility issues? I'm was rebuilding an image on the 2012.12-yocto1.3 branch of angstrom and based on systemd-image (angstrom specific, I know) Mar 22 14:27:54 and it failed first on do_compile for git, then on gcc Mar 22 14:28:52 git I got to build by just doing 'bitbake git' gcc I had to bitbake git -c clean then bitbake gcc Mar 22 14:29:43 My guess is there is some lurking compatibility issue in my recipe, but I don't understand why the build failed, then worked when I built the recipe directly Mar 22 14:30:18 anyway, I've got to run. I'll be back in a few Mar 22 14:32:58 without more details I'm not too sure... Mar 22 14:40:15 I need to build my own toolchain, like "meta-toolchain-sdk", someone can suggest some stuff to read that can help me? Thanks Mar 22 15:10:27 silvio_: meta-toolchain is not enough? Mar 22 15:26:44 hrw, it seams that there are not the headers and library of software that I add myself Mar 22 15:27:32 silvio_: best would be creating meta-toolchain-silvio.bb with your packages listed there Mar 22 15:28:30 silvio_: http://git.linaro.org/gitweb?p=openembedded/meta-linaro.git;a=tree;f=meta-linaro/recipes-extra/meta-toolchain-hhvm;h=bc5120d8a342aa4503281fa8c30513a8fc3a01ef;hb=HEAD - look at both meta-toolchain-hhvm.bb and Mar 22 15:28:34 packagegroup-core-standalone-hhvm-sdk-target.bb Mar 22 15:28:35 hrw, is what I want to do, but I dont know the best practise and I cannot foud stuff or example Mar 22 15:28:51 first one is using latter one. latter one defines all dependencies Mar 22 15:29:19 hrw, ok thanks so much Mar 22 15:29:23 you welcome Mar 22 15:32:56 hrw, so if I understand I put in the my second one the same packages that I hve in my image? Please correct me if I m wrong Mar 22 15:45:16 silvio_: yes Mar 22 15:45:32 silvio_: you can then use packages generated by recipe in image Mar 22 15:53:28 hrw, there is a smart way to include all the dev packages of my image without rewrite them one by one? Mar 22 15:55:21 hrw, and why use RDEPENDS_${PN} insted of RDEPENDS_${PN}-dev ? Mar 22 15:55:55 silvio_: because I do not want -dev package Mar 22 15:57:08 silvio_: for dev/staticdev etc in image: IMAGE_FEATURES += "dev-pkgs staticdev-pkgs tools-debug tools-sdk" Mar 22 15:57:19 skip last two if not need Mar 22 16:04:58 hrw, now I try thanks Mar 22 16:14:09 hrw, in your example "4 inherit packagegroup", what refers to? Mar 22 16:17:41 "inherit packagegroup"? Mar 22 16:18:01 it inherits packagegroup.bbclass which sets PACKAGES to have PN only, arch to all etc Mar 22 16:19:24 ok. time to end a day Mar 22 16:19:30 have a nice weekend guys Mar 22 16:24:34 hrw, thanks good we Mar 22 16:25:20 silvio_: there is a create SDK from image function FYI Mar 22 16:25:34 bitbake -c populate_sdk your-image-name Mar 22 16:27:52 bluelightning, thanks, I see that, but is not working on my buildsystem, seams it works only in poky version after 1.3 Mar 22 16:28:08 right, yes I believe that is correct Mar 22 16:37:00 bluelightning, so my hope is create a meta-toolchain-silvio.bb that make it, do you know some guide, or doc, that explain more about it? Mar 22 16:39:24 silvio_: ok, in that case have a look at meta/recipes-qt/meta/meta-toolchain-qt* Mar 22 16:39:57 bluelightning, tahnks Mar 22 16:41:52 I added connman to my image recipe, but it still starts /etc/init.d/networking at startup. I want to disable that, and let connman handle the network. what is the recommended way to do that? I could add a .bbappend file to the netbase recipe that simply deletes /etc/init.d/networking , but that does not sound very clean Mar 22 16:43:19 dv_, do u try to use NetworkManager? Mar 22 16:43:26 no, connman Mar 22 16:43:33 dv_: I think what you need to do is bbappend netbase to just install a basic ${sysconfdir}/network/interfaces that only defines lo Mar 22 16:43:50 bluelightning: but doesnt connman read ${sysconfdir}/network/interfaces ? Mar 22 16:45:17 huh, doesnt seem so Mar 22 16:45:19 nice idea! Mar 22 16:45:22 dv_: don't think that it relies on interfaces being listed there no Mar 22 16:45:22 thanks Mar 22 16:45:25 np Mar 22 16:45:30 we should probably document that somewhere Mar 22 16:45:54 dv_: what were you finding exactly without the bbappend? Mar 22 16:46:10 that udhcpc starts before connman Mar 22 16:46:18 ah right, of course Mar 22 16:46:31 which is quite pointless, and causes unnecessary waits if no cable is connected Mar 22 16:50:52 bb Mar 22 16:53:25 I'm trying to build a recipe that adds my software to an image. I've got it all in a git repp, most of it consists of files that just need to be copied to a directory, so I gather for them I just use install in a do_install section, but I'm a bit confused as to where the tools chain is going to put all the files it grams from the src_uri. Mar 22 16:53:37 I think it sticks them in {S} Mar 22 16:54:22 spacecolonyone: correct Mar 22 16:54:47 but in the examples I see I'm not seeing things like install -m 0644 ${S}/foo ${D}$/mydir/foo Mar 22 16:54:59 is do_install already in {S} Mar 22 16:55:16 spacecolonyone: usually yes Mar 22 16:55:34 spacecolonyone: FYI if nothing actually needs to be compiled you can use INHIBIT_DEFAULT_DEPS = "1" Mar 22 16:55:43 ok Mar 22 16:55:53 that just avoids it depending on things like eglibc Mar 22 16:55:54 yea these are all python scripts Mar 22 16:56:06 ok, right Mar 22 16:56:08 I'd like to make it also add users to the image Mar 22 16:56:19 in which case ignore than and "inherit allarch" instead which does that and more Mar 22 16:56:48 and there are some config files it needs to copy (e.g udev rules and samba configs) Mar 22 16:56:49 there is a class for handling addition of users - useradd.bbclass Mar 22 16:56:55 ok Mar 22 16:57:13 "git grep useradd" and you'll find a ton of examples of recipes using it Mar 22 16:57:25 I saw something saying that for installing things to /etc I should use a ${} location Mar 22 16:57:33 yes Mar 22 16:57:43 ${sysconfdir} should be used instead of /etc Mar 22 16:58:03 you can see the default definitions of such variables near the top of meta/conf/bitbake.conf Mar 22 16:58:04 I've been looking for a listing of these targets but I've only been seeing them piecemeal in existing recipes Mar 22 16:58:10 ok Mar 22 16:58:21 regarding the bbappend, this should be enough to make OE use my interfaces file, correct? http://pastie.org/7078146 Mar 22 16:58:37 finally, what about order? so my image installs samba Mar 22 16:58:47 which installs a default samba.conf Mar 22 16:59:06 I'd rater overwrite that with my own instead of modifying the stock samba recipe Mar 22 16:59:10 dv_: the SRC_URI line is not necessary since the file is already in there, otherwise yes that should work as long as the interfaces file resides in a directory next to the recipe called "netbase" Mar 22 16:59:36 dv_: (i.e. ${PN} is netbase) Mar 22 17:00:15 alright Mar 22 17:00:32 spacecolonyone: you should do that by bbappending the samba recipe in a similar manner to dv_ Mar 22 17:01:07 spacecolonyone: if you try to do it with a separate recipe the package manager will error out with a conflict Mar 22 17:01:13 oh, one more thing, then I'm done: I noticed that images by default do not include the uImage in the image's /boot/ directory. also, I want to autogenerate a u-boot boot.scr from a script I have stored somewhere (for example, in conf/ ). Mar 22 17:01:14 so my config file is part of the repo for this other image Mar 22 17:01:18 any recommendations? Mar 22 17:02:04 dv_: I'm a little hazy on the u-boot details I'm afraid... Mar 22 17:02:24 bluelightning: essentially I just need to run the mkimage command to produce a boot.scr, and then copy this boot.scr to /boot/ Mar 22 17:02:47 I thought of writing a recipe for that. that could work. but as for copying the uImage to /boot/ , I have no idea. Mar 22 17:02:54 So I could move it around to a different repo, but is there a way to add additional source uris with the bbappend and then append the do_install? Just a yes will work, I'll go google it Mar 22 17:04:11 dv_: it could be done using ROOTFS_POSTPROCESS_COMMAND ; or alternatively a postinstall script Mar 22 17:04:26 dv_: postinstall script is probably best if you're using on-target packaging Mar 22 17:04:35 s/packaging/package management/ Mar 22 17:05:05 hmm Mar 22 17:05:21 copying the uImage does sound like a postprocess command Mar 22 17:05:36 postprocess means after the image is done but before it is bundled to a tarball etc.? Mar 22 17:05:46 spacecolonyone: sure, you can add the repo to SRC_URI in the samba recipe... the file won't end up being in ${S} but if you know where it will be extracted to (e.g. ${WORKDIR}/git/some/path) then that's not a problem Mar 22 17:07:51 dv_: that's correct Mar 22 17:12:23 okay, I'd copy "${DEPLOY_DIR_IMAGE}/${IMAGE_LINK_NAME}" to "${IMAGE_ROOTFS}/boot/" Mar 22 17:12:36 uh no Mar 22 17:12:41 What is the purpose of FILES_${PN} i see some recipes defining? Mar 22 17:12:43 "${DEPLOY_DIR_IMAGE}/uImage" to "${IMAGE_ROOTFS}/boot/" Mar 22 17:13:21 spacecolonyone: defines the files the package named ${PN} contains Mar 22 17:13:44 spacecolonyone: FILES_foobar defines the files package foobar will include Mar 22 17:14:04 and ${PN} is a variable with the name of your package, defined by the filename of the .bb file Mar 22 17:14:21 I thought that would be done by whatever was fetched by src_URI and then operations in do_install? Mar 22 17:14:42 well OE usually tries to do that automatically Mar 22 17:14:47 PN includes the PV& PR appended, right Mar 22 17:14:50 but sometimes you want to define these lists manually Mar 22 17:15:03 for the packager? Mar 22 17:15:08 yes Mar 22 17:15:27 also, PN does not include PV and PR, I think Mar 22 17:15:52 so is there a guide somewhere that describes do_install, the packager, etc. Mar 22 17:16:03 I've got the really, REALLY old manual Mar 22 17:16:13 but that thing isn't much use any more Mar 22 17:16:20 spacecolonyone: I'd strongly recommend the yocto project documentation Mar 22 17:16:33 where does yocto fit into all this Mar 22 17:16:43 I've started hereing about it more Mar 22 17:16:48 hear* Mar 22 17:17:18 spacecolonyone: the Yocto Project is an umbrella project that helps to maintain a number of embedded bits and pieces, the foremost of which being BitBake and OE-Core Mar 22 17:17:29 ok Mar 22 17:17:38 and also aims at making it more robust, stable, and mature Mar 22 17:17:43 indeed Mar 22 17:17:58 one good example is how much faster bitbake parses recipes Mar 22 17:18:07 yes, I've noticed the debugging messages improving during builds Mar 22 17:18:08 in the OE-classic days, that was a nightmare Mar 22 17:19:27 also, the ncurses UI is nice :) Mar 22 17:19:57 UI? Mar 22 17:20:05 oh well, the output Mar 22 17:20:21 Ah, yes, that is what I've noticed improving Mar 22 18:21:58 suppose i want to use FILESEXTRAPATHS_append in the ".bbappend" file for a particular recipe in two different layers Mar 22 18:22:06 is there "proper" way to do this? Mar 22 18:24:11 for example in this case I am using FILESEXTRAPATHS_append := "${THISDIR}/files" in the .bbappend file for both layers and what i end up with is the path from the higher higher priority layer appended directly to the path of the lower priority layer Mar 22 18:24:56 erh, i mean "prepend" Mar 22 18:26:03 * waynr reads docs Mar 22 18:26:16 ah, maybe i should try using a colon separator? Mar 22 18:26:49 wooo Mar 22 18:27:09 thanks for the help whoever contributed to the documentation Mar 22 19:46:28 Have someone used OE / Yocto to build a sub distro to run inside a regular distro? Mar 22 19:48:14 We have a product and we'd like to use a controlled environment and we'd like to provide a specific Xorg version for it. So it can be distro independent Mar 22 19:48:35 but to accomplish it we need to have a micro distro wich provides Xorg and its drivers only Mar 22 19:48:49 and to have it to be build in /opt or similar Mar 22 19:48:59 sounds a little like what we do with our SDKs Mar 22 19:49:09 bluelightning: yes Mar 22 19:49:17 bluelightning: similar Mar 22 19:49:30 bluelightning: but it'd for a target arch Mar 22 19:49:38 bluelightning: in my case x86 Mar 22 19:49:42 (for now) Mar 22 19:50:12 it could be implemented in a similar way though, i.e. by creating a separate variant a la nativesdk with a different prefix Mar 22 19:50:33 althernatively you'd just create a separate distro with the prefix set Mar 22 19:50:44 bluelightning: I did the second Mar 22 19:50:49 bluelightning: and I setted Mar 22 19:51:00 bluelightning: prefix, exec_prefix and base_prefix Mar 22 19:52:21 | /home/otavio/hacking/ossystems-products-platform/build/tmp/sysroots/ossystems-x86/opt/ossystems/osmultiseat/lib/libusb-1.0.so: undefined reference to `clock_gettime' Mar 22 19:52:24 | collect2: error: ld returned 1 exit status Mar 22 19:52:27 and this happens ;) Mar 22 19:54:56 hmm... not sure, but clearly it's not finding the right libs Mar 22 19:55:19 bbl Mar 22 19:55:21 bluelightning: it seems to be provided by rt lib Mar 22 19:55:27 oh ok Mar 22 20:04:19 hm gcc 4.8 out Mar 22 20:09:23 geez, i just updated to 4.7.1 on this machine Mar 22 20:09:56 ah, 4.7.2 Mar 22 20:10:04 so now i'm behind again... Mar 22 20:10:26 hehe Mar 22 20:23:00 anyone has a clue about the link failure? Mar 22 20:39:17 -lrt Mar 22 21:25:10 I'm confused. there is no xterm in oe-core? Mar 22 21:27:02 (at least I cant find it) Mar 22 21:29:31 sounds like one of those days when i can't find my pants... Mar 22 21:30:58 inb4 "use urxvt" Mar 22 21:32:36 * mr_science likes eterm Mar 22 21:34:36 but i've been using the gnome and lxde native terminals lately Mar 22 21:35:11 the latter being less of memory hog than gnome terminal with lmost the same feature set Mar 22 21:35:18 *almost even Mar 23 01:12:29 yo Mar 23 02:51:11 * kergoth yawns **** ENDING LOGGING AT Sat Mar 23 02:59:58 2013