**** BEGIN LOGGING AT Thu Mar 20 02:59:59 2014 Mar 20 04:14:23 Bertux_ I hope your're still around or you find this in a log. I've never done it with Yocto, but I know from other Linux experiences that you can run QT stuff directly to the framebuffer device. Some VGA devices on the PC side even have some acceleration. What hardware are you developing on? Mar 20 08:49:30 good morning Mar 20 09:43:03 morning all Mar 20 10:36:42 hi, any reason why a git repository is not refetched if erm... delete the directory manually in the downloads/git2 folder? Mar 20 10:36:49 and then run bitbake myrecipe Mar 20 10:39:54 because it didn't need to re-fetch if it already had done a build? Mar 20 10:40:39 I removed the workdir fwiw. Mar 20 10:40:45 I also tried -c cleanall Mar 20 10:40:56 something is fundamentally fubar'd. :D Mar 20 10:41:16 I do not yet have PV = "${PVBASE}.${SRCPV}" in the recipe. Mar 20 10:41:18 lpapp: if you delete the git/ directory in downloads you also need to delete the .done stamp that says its completed teh download Mar 20 10:41:29 but I thought that would not be necessary for cloning if there is no clone yet. Mar 20 10:41:47 rburton: yeah, I can double check, but I thought -c cleanall would do that. Mar 20 10:42:21 yeah, find ./sstate-cache/ -name \*foo\* returns nothing. Mar 20 10:42:34 ah, sorry, you meant download .done Mar 20 10:43:28 yeah, I deleted them, but it is still not cloning. Mar 20 10:43:50 I am not sure what more I can do for it to get refetched. Mar 20 10:44:06 I could wipe my build folder out, but it would be a lot of time to rebuild the things. :/ Mar 20 10:46:27 bitbake recipe -c fetch -f Mar 20 11:06:12 rburton: that fetched it, interesting, thanks. Mar 20 11:06:16 must be a bug I assume? Mar 20 11:06:23 (or some local setup issues) Mar 20 11:06:59 lpapp: local setup or it just didn't need to re-fetch Mar 20 11:07:14 I should add PV = "${PVBASE}.${SRCPV}" to the recipe though, right? Mar 20 11:10:54 rburton: ^ Mar 20 11:11:09 literally have no idea why you would do that, so don't ask me. Mar 20 11:12:39 ok Mar 20 11:12:56 (based on a mailing list discussion, it would be done for making sure the stuff is refetched) Mar 20 11:13:16 not in my previous case, but in general, if you need to do it while the thing is already built, for instance changing SRCREV) Mar 20 11:13:21 SRCREV.* Mar 20 11:31:51 hi, any ideas if FILESEXTRAPATHS_prepend takes a distro name at the end? i'm trying building a poky-tiny image for imx6 and hitting some issues with busybox config Mar 20 11:32:19 overrides like that work on all variables Mar 20 11:32:29 so FILESEXTRAPATHS_prepend_poky-tiny is fine Mar 20 11:32:36 great Mar 20 15:40:19 hi there Mar 20 15:41:43 hi Xz Mar 20 15:42:59 Hello Guys, how do you manage package update on the target ? I explain using a very simple example, I have an XML file version 1 on my system. The version 2 of this XML add 10 new items. When I update the package, I want the system to add the lines to the xml file, not replacing it. Is this possible to do with yocto? ( in my usage, it is more to update a database... ) Mar 20 15:44:07 weebet: is that XML file packaged as part of the package or not? Mar 20 15:45:13 yes, it is part of the package created in recipe ( XML-config-1.0.bb ) and now working on XML-config-2.0.bb. Mar 20 15:47:04 sorry must read XML-config_2.0.bb / XML-config_1.0.bb Mar 20 15:51:27 weebet: does the file get edited at all after installation or just stays static? Mar 20 15:51:47 bluelightning: yes the file get edited after installation Mar 20 15:52:35 ok, if you haven't already you'll probably need to add the file it to CONFFILES in the recipe (so it isn't just overwritten) and then use a pkg_postinst function to make the appropriate additions if the file is already present Mar 20 15:59:19 Thank you bluelightning Mar 20 16:01:01 bluelightning : one more question. I work with Danny so are using RPM, does RPM will know that it is upgrading from version 1 to 2 ? or it will install both package beside each other ? Mar 20 16:01:44 Say, I have a custom software called management_1.0 if I rpm -ivh management_2.0 it will install management_2.0 and will not upgrade management_1.0 Mar 20 16:03:18 weebet: it'll automatically upgrade since the names match Mar 20 16:07:03 Hey, I've got a problem when I first build yocto porject, can anyone help? Mar 20 16:07:36 Currently 1 running tasks (991 of 5707): 0: linux-yocto-3.8.13+gitAUTOINC+14441275c1195a1bd52a6455ecde006c4722d45f_6f3e338aa9496cf68ad03a98f66c2e98975829c7-r4.1 do_fetch (pid 2469) Mar 20 16:07:38 Shadow_Shanghai, yes. Don't ask to ask, just ask and provide enough information for people to review and respond. Mar 20 16:07:59 My problem is, as you can see Mar 20 16:08:15 the building process stuck on this task Mar 20 16:08:18 linux-yocto is a huge repository, it takes time to download Mar 20 16:08:34 really? Mar 20 16:08:50 fortunately, you only have to do the big download once so long as you don't do a clean all Mar 20 16:08:53 yes, really :-) Mar 20 16:08:55 how huge is it, then? Mar 20 16:09:25 $ du -hs linux-yocto-3.8 Mar 20 16:09:25 2.2G linux-yocto-3.8 Mar 20 16:09:44 that includes some of my personal branches/trees, but that will give you an idea Mar 20 16:10:13 just watch the download dir and as long as it's growing in size, it isn't stuck Mar 20 16:10:33 Shadow_Shanghai: are you sure your firewall lets you use git port? Mar 20 16:10:35 watch du -hs downloads/git2/git.yoctoproject.org.linux-yocto-3.8.git ... or similar Mar 20 16:10:45 mckoan, it would have broken well before this point Mar 20 16:11:37 I just saw a file named git.yoctoproject.org.linux-yocto-3.8.git.lock Mar 20 16:11:52 no other directory named git.yoctoproject.org.linux-yocto-3.8.git Mar 20 16:12:11 anything else in that directory? Mar 20 16:12:23 to demonstrate a successful download of other git repos? Mar 20 16:12:53 dvhart_: life is weird, it happened to me that after I installed Yocto at a customer site after some days the sysadmin suddenly blocked git access. Mar 20 16:12:55 yes, I can use git clone to download poky Mar 20 16:13:05 mckoan, valid point Mar 20 16:13:05 does it mean my git is ok? Mar 20 16:13:21 Shadow_Shanghai, see my question above Mar 20 16:14:31 $ du -hs git.yoctoproject.org.linux-yocto-3.8.git Mar 20 16:14:31 683M git.yoctoproject.org.linux-yocto-3.8.git Mar 20 16:14:41 much smaller if it's a bare clone like in the downloads dir Mar 20 16:15:42 du: cannot access ‘git.yoctoproject.org.linux-yocto-3.8.git’: No such file or directory Mar 20 16:15:55 still there is no directory Mar 20 16:17:24 Does the directory appears when all the fetching get done? Mar 20 16:17:38 It's weird Mar 20 16:18:47 Shadow_Shanghai: please try git clone git://git.yoctoproject.org/linux-yocto-3.8.git Mar 20 16:20:03 yeah, try the download manually as mckoan suggests. Mar 20 16:20:22 it's possible it's cached somewhere else first, but I thought it did it directly in the downloads dir Mar 20 16:22:11 It's cloning Mar 20 16:23:12 Hope this can help me skip this task and move on building Mar 20 16:23:18 Shadow_Shanghai: try bitbake -c fetch virtual/kernel Mar 20 16:23:26 only Mar 20 16:25:28 thanks, I guess I need to take some patients to wait for downloading Mar 20 16:25:42 Hi, im trying to create a recipe with a small web application and im running into some problems. I have assigned a local file as source and when bitbake tries to run do_unpack after it has unpacked the file it throws an error: Failed with exit code 1. When i run that command manualy the extration completes without an error. Is there any extended loggin to see more about that error? or any sugestions to what might be wrong. Thanks Mar 20 16:27:30 btw, i don't have a do_unpack in the bb file. i do have a do_install but it seems the problem occures before that Mar 20 16:28:25 can you pastebin your recipe? Mar 20 16:28:32 sure Mar 20 16:29:22 local file as in a tarball or single file? Mar 20 16:29:30 http://pastebin.com/e3mga7dy Mar 20 16:30:08 it's a zip file with the webfiles inside a folder named application-0.1 Mar 20 16:30:34 zip files should unpack fine Mar 20 16:31:01 the SRC_URI file:// is relative to the recipe dir Mar 20 16:31:39 yes, inside recipe dir -> files -> webroot -> file.zip Mar 20 16:31:47 so does your recipe layer have a recipes-foo/bar/webroot dir where the zip file sits? Mar 20 16:32:42 yes recipes-foo/webadmin/files/webroot Mar 20 16:33:21 so "bitbake webadmin -c unpack" works but "bitbake webadmin" blows up somewhere? Mar 20 16:34:27 no i havent tried the unpack command. but that failed Mar 20 16:34:30 try the second one and pastebin the error Mar 20 16:35:06 well, try -c fetch and -c unpack and see if they actually work Mar 20 16:35:57 typically you wouldn't want to commit a zipfile/tarball to your recipe repo Mar 20 16:36:36 it's fine for initial testing, but i'd stick it somewhere else for "production" Mar 20 16:37:00 ok, i tried to find some docs on how to get this right but after going trough for about a week now i still havent found the correct solution for deploying files as a recipe Mar 20 16:37:10 normally you'd want an md5sum/sha256sum in your recipe for the zipfile Mar 20 16:37:33 ah, yeah im pusing it to a git when i get the hang of this Mar 20 16:38:09 but the unpack command failed and when i run the command manualy it dose not throw an error Mar 20 16:38:25 you are talking about do_unpack. is it possible to disable the do_unpack ? I want to copy a tar.gz on my target and since I add it to SRC_URI = file://... bitbake will unpack it. So I pack it twice... to what I want Mar 20 16:39:00 yeah, do_unpack Mar 20 16:39:09 do i need to zip the zip? Mar 20 16:39:41 oh, i want it to be unpacked and deployed to a specific folder on the build Mar 20 16:40:28 weebet: there's a url parameter to disable the unpack. ;unpack=false or ;unpack=0 will disable extraction Mar 20 16:41:35 hmm when i look in core2-64-poky-linux/webadmin/0.1-r0 it extracted the folder webadmin-0.1 and that folder contains the corret files Mar 20 16:41:58 what else is do_unpack trying to accomplish Mar 20 16:42:12 Krise1: I don't understand the question Mar 20 16:42:36 compressed files in SRC_URI get extracted, it's that simple Mar 20 16:42:42 thank you kergoth ! Mar 20 16:43:05 kergoth : where did you learn that ? Mar 20 16:43:28 what's the difference between binutils-cross and binutils-crosssdk? Mar 20 16:44:06 weebet: I examined the code. do_unpack in base.bbclass calls bb.fetch2.unpack(), which is in poky/bitbake/lib/bb/fetch2/__init__.py Mar 20 16:44:27 i have created simple recipe that should deploy a web application to a folder. but it thorws an error on bitbake saying Task 0 do_unpack filed with exit code '1'. but the files get extracted to the folder Mar 20 16:44:53 Krise1: so look at the do_fetch log, and post it to a pastebin for us to look at Mar 20 16:44:58 summarizing isnt' going to solve your problem Mar 20 16:45:15 Thank you kergoth. Mar 20 16:45:17 JaMa: hi! I am a bit puzzled with the PACKAGESCONFIG of qtbase.inc. The intent is that users create a .bbappend to replace things? Mar 20 16:45:23 weebet: np Mar 20 16:45:48 Krise1: typically you would do something like this: http://pastebin.com/FmFRj7Ws Mar 20 16:46:34 unpack=no would let you manually do it Mar 20 16:46:37 I think Krise1 wants it to extract the zip, but weebet doesn't. Krise1 is getting a do_unpack failure Mar 20 16:46:41 * kergoth shrugs Mar 20 16:47:06 otherwise leave it out and let it do its thing Mar 20 16:47:26 ok, but do the files have to be hased Mar 20 16:47:39 look at the pastebin he just showed you Mar 20 16:47:43 yeah kergoth thats right Mar 20 16:47:49 theres md5 and sha checksums of the zip file, that's all Mar 20 16:47:52 and yes, those ones are required Mar 20 16:48:01 ok il add it Mar 20 16:48:03 How can I bbappend .inc file? Mar 20 16:48:08 Xz: you can't Mar 20 16:48:48 kergoth: that's sad :( Mar 20 16:50:46 extinction is sad... Mar 20 16:53:52 ERROR: Function failed: Unpack failure for URL: 'file://webroot/webadmin-0.1.zip Mar 20 16:54:15 paste the details Mar 20 16:55:06 http://pastebin.com/3a5M9Yrm Mar 20 16:56:30 I want to add one flag to GCC globally, for every package. Can I do that somehow in my machine.conf file? Mar 20 16:56:46 is there something funky about that zipfile? Mar 20 16:57:11 the only thing i see that looks weird is "shieldbox-webadmin-0.1/: ucsize 4096 <> csize 0 for STORED entry" Mar 20 16:57:12 no it's a standard zip file. when i extract i manualy it all works Mar 20 16:57:38 should i use tar instead? Mar 20 16:57:51 try it and see Mar 20 16:57:57 oops, mtg time... Mar 20 16:59:24 do I just append that flag to CLFAGS in machine.conf? Mar 20 17:01:11 Xz: use SELECTED_OPTIMIZATION Mar 20 17:01:18 append to it Mar 20 17:01:23 in local.conf Mar 20 17:01:51 BAM! that worked. thanks mr_science Mar 20 17:03:12 khem: I need to append that flag for everybody Mar 20 17:03:24 khem: not just for me, *once* Mar 20 17:04:03 Xz: so add it to distro conf Mar 20 17:04:31 khem: CFLAGS, or SELECTED_OPTIMIZATION? Mar 20 17:04:54 khem: Yocto glossary doesn't reallt say much about the second Mar 20 17:05:21 SELECTED_OPTIMIZATION? Mar 20 17:05:35 khem: yeah Mar 20 17:05:54 that should be a filed as bug in yocto docs then Mar 20 17:05:59 but is the right set here Mar 20 17:06:23 some packages like eglibc do not want to honor your options for valid reasons Mar 20 17:06:34 and if you butcher in CFLAGS ou might get more issues Mar 20 17:09:51 so SELECTED_OPTIMIZATION changes CFLAGS for choosen subset of packages, not all of them? Mar 20 17:12:08 it changes it for all Mar 20 17:14:51 then I don't understand the difference between SELECTED_OPTIMIZATION and CFLAGS. They both do exactly the same - pass flags to gcc for every package (including binutils/libc) Mar 20 17:15:06 look at the code Mar 20 17:15:14 in some cases recipes alter it Mar 20 17:15:41 since you have modified compiler to hack the option in Mar 20 17:15:57 you might very well hack the compiler further to bundle it with O2 Mar 20 17:16:07 or any O level Mar 20 17:16:21 then you dont deal with OE metadata Mar 20 17:33:43 khem: hey, I'm just preparing some fixes for the python manifests inc file and the scripts, but I've noticed some of the dependencies added in 2.7 haven't been added to the 3.3 version - does this change look OK? http://pastebin.com/GkKk7dcU Mar 20 17:40:07 when i add install -d to the bb file the log says omitting directory. why is that? Mar 20 17:41:13 can i move a directory contents with intall -d? Mar 20 17:41:21 install -m Mar 20 17:47:39 install doesn't have a "move" option afaik Mar 20 17:48:18 and which task is spitting out "omitting directory" ? Mar 20 17:49:13 yup Mar 20 17:50:19 btw adding the content to tar worked. guess the zip file was broken. but how would i go about moving all the files to a specific directory Mar 20 17:50:31 should i use some bash script Mar 20 17:54:28 you can do bash commands inside do_install Mar 20 17:55:04 ok, so cp -R is okey. cause i read on the openembedded that it should be avoided Mar 20 17:55:49 well, it should be but sometimes you kinda have to Mar 20 17:56:02 at least for initial testing, etc Mar 20 17:56:04 ok, that worked. thanks Mar 20 17:56:15 at some point you would want to clean things up Mar 20 17:56:41 yeah, but for production what would be the best route to take. moveing a tar file and running install.sh script Mar 20 17:57:33 "production" usually means installing correctly to image/usr and what not Mar 20 17:58:15 yes, but i would have to use a shell script for that or would i package it another way Mar 20 17:58:32 Krise1: there's nothing wrong with using cp to copy a subtree. install is only appropriate for individual files/dirs. sometimes you're just stuck with it Mar 20 17:58:44 eg. meta-sourcery still uses cp to extract bits from the external toolchain sysroot Mar 20 17:59:40 ok, thanks. cause i can't see any other way for that web application Mar 20 18:00:13 * kergoth nods Mar 20 18:00:35 unless there's an upstream buildsystem to call into, it makes sense to retain the copy in the recipe, but its a case by case thing Mar 20 18:01:21 one more thing. in my image i use udev, and udev add gnome-desktop-testing. and i have to remove it everytime i want to do a build in hob. and udev is still compiled into the image without the gnome-desktop-testing package Mar 20 18:01:47 why is udev importing gnome-desktop-testing Mar 20 18:02:15 no sorry, it's glib-2.0 thats improting the gnome-destop-testing Mar 20 18:02:48 stupid me, i should remove glib-2.0 Mar 20 18:04:50 Hey, just by curiosity how do you do a extract with change directory, I'm think more specificly for Krise1 case that want to move a directory. could he simply use tar xvf file.tar.gz -C ? Is there some option do to this in recipes? Mar 20 18:04:53 lots of things are build deps but not runtime deps Mar 20 18:05:14 is it actually installed in your rootfs, or just pulled in at build time? Mar 20 18:05:35 both Mar 20 18:05:54 but i think i added a recipe that pulls it in Mar 20 18:06:24 that's weird because i have a bunch of gnome/glib deps in my xaorg image but nothing has pulled in gnome-desktop-testing or whatever Mar 20 18:06:47 hmm, i don't even have X in my image Mar 20 18:07:16 ie, glib-2 is a dependency for lxdm login manager Mar 20 18:07:19 kapare: yes, subdir= in the url Mar 20 18:07:22 kapare: iirc anyway Mar 20 18:09:24 kergoth, thx Mar 20 18:12:18 zecke: you can do it from distro config without .bbappend Mar 20 18:15:40 ah thanks. I didn't know about PACKAGECONFIG yet and I am reading base.bbclass for it now Mar 20 18:16:45 zecke: of course the code is the ultimate reference, but that variable is also documented in the manual ;) Mar 20 18:21:51 bluelightning: yes, but it looks like only how to use it in your package but not how to influence it from the distro Mar 20 18:21:58 (at least with the dora documentation) Mar 20 18:22:26 JaMa: will you accept tufao and libsystemd-qt patches for meta-qt5? Mar 20 18:22:40 zecke: the _pn- override is usable for setting any variable in a recipe-specific manner Mar 20 18:23:12 zecke: i.e. you can set PACKAGECONFIG_pn-yourrecipe = "..." Mar 20 18:23:21 zecke: I haven't seen them yet, but maybe I will :) Mar 20 18:23:36 and _append_pn-foo and _remove_pn-foo. i find often my intent is more accurately reflected by using those rather than an override Mar 20 18:23:50 though admittedly, the fact that you can't "un-remove" something is occasionally irritating Mar 20 18:23:58 kergoth: :) Mar 20 18:24:17 JaMa: I haven't written them yet. But the intent is to have !QtProject libs/tools in recipes-qt? Mar 20 18:24:26 RP: thoughts on changing it so _append/_remove act in order? e.g. an _append defined after a _remove would bring it back, and another _remove after that would remove it again.. i think that would be most intuitive Mar 20 18:24:29 true, in this case you usually want just append Mar 20 18:24:35 RP: currently a single _remove anywhere will always override all _appends Mar 20 18:24:56 RP: afaict anyway Mar 20 18:27:02 kergoth: I'm not sure it is what everyone expects Mar 20 18:27:17 kergoth: a lot of people just want "remove this thing, I don't care how it got there" Mar 20 18:27:32 bluelightning: PACKAGECONFIG_pn-foo = "bar" supposed to work outside of local.conf ? Mar 20 18:27:45 bluelightning: i think file parse orer would still take care of that in a lot of cases, but its a good point Mar 20 18:27:54 i don't think being entirely unable to undo a _remove_ is kosher, though Mar 20 18:27:57 i kinda expected it to work in an image recipe but it doesn't... Mar 20 18:28:04 mr_science: it'll work from any conf file, provided that there's not another PACKAGECONFIG_pn-foo line somewhere afterwards Mar 20 18:28:06 if my distro wants to remove something today, it *must* use := with oe_filter_out, otherwise the user can't add it back Mar 20 18:28:10 even if they want it Mar 20 18:28:16 mr_science: no, that can't work Mar 20 18:28:32 mr_science: recipes cannot change eachother's configuration Mar 20 18:28:47 so you can set it in the original recipe but not another recipe? Mar 20 18:28:58 okay, makes sense i guess Mar 20 18:29:22 mr_science: right, or you can set configuration from "above" via conf files for both recipes Mar 20 18:30:07 e.g. DISTRO_FEATURES is set from above and acted upon by many recipes Mar 20 18:32:39 seems like "above" would include the image level but okay... Mar 20 18:33:09 mr_science: the image is just another recipe Mar 20 18:33:36 yeah, but it's somewhat "special" compared toa package recipe Mar 20 18:33:53 conceptually at least Mar 20 18:33:57 mr_science: really all you can do at the image level is decide what packages you want in it and then process the files in the resulting filesystem afterwards (e.g. with ROOTFS_POSTPROCESS_COMMAND) if you need to Mar 20 18:34:26 same concept as my current ml argument Mar 20 18:35:03 to me (being a gentoo guy) autogen.sh isn't "configure" per se, it's preparing the source for configure Mar 20 18:35:15 if images could influence how other recipes were built, you could never have a consistent package feed when building >1 image Mar 20 18:35:29 mr_science: an image is just another recipe. Mar 20 18:35:37 it's not magic, it doesn't have any special control Mar 20 18:35:38 so conceptually a src_prepare task makes more sense than autogen in do_configure Mar 20 18:35:44 consider bitbake oneimage twoimage threeimage, what would happen if those try to alter configuration of something else? Mar 20 18:35:48 would it rebuild something else three times? Mar 20 18:35:54 there's ar eason it's the way it is :) Mar 20 18:37:26 that typo is dangerously close to talk-like-a-pirate day... Mar 20 18:37:50 :) Mar 20 19:25:18 khem: I've sent the patch anyway Mar 20 19:37:09 what causes a package to be (skipped) in the bitbake-layers show-recipes output? Mar 20 19:38:11 generally it means that it raised the SkipPackage ecxeption, which means it considers itself unbuildable in the current context Mar 20 19:38:25 which usually means COMPATIBLE_HOST/COMPATIBLE_MACHINE indicates its incompatible with your current configuration Mar 20 19:38:31 iirc, anyway Mar 20 19:38:41 thanks I'll look in that direction Mar 20 19:48:26 This is cool, but we shoudl make one on yp.org and use the oe-core toolchain Mar 20 19:48:27 http://gcc.godbolt.org/ Mar 20 19:51:29 Crofton|work: :) Mar 20 19:52:59 Crofton: that *is* cool... Mar 20 19:53:19 and feel free to fork/deploy it Mar 20 19:53:47 Crofton|work: you should've looked into my g+ posts earlier ;) Mar 20 19:54:23 it seems familiar Mar 20 20:01:35 rburton: sent an email about cogl Mar 20 20:11:43 is there anything known about mesa breaking on arm? it seems to compile fine here, but break building the rootfs http://paste.ubuntu.com/7127035 Mar 20 20:23:18 this should be the corresponding install log http://paste.ubuntu.com/7127115/ Mar 20 20:24:43 robertyang: thought about using bb.build.addtask() to add the inter-task deps in archiver.bbclass rather than using ${PN}:? Mar 20 20:25:00 robertyang: might be cleaner, since thats specifically for inter-task deps within a recipe Mar 20 20:26:03 halstead: We're going to need to check the filesystem on opensuse131-dev. I'm getting input/output errors trying to rm. Mar 20 20:58:25 robertyang: any objection to adding emission of patch series files next to patches when using 'original' archiver mode, to gain feature parity with copyleft_compliance? Mar 20 21:06:31 From what I'm seeing the hostname is set by meta/recipes-core/bases-files and this will result in default openembedded then is MACHINE exist it became it. So what will be the best way to change this to something diff to MACHINE? I try local.conf without success... Mar 20 21:06:48 There's the this comment that doesn't help me yet: set standard hostname, might be a candidate for a DISTRO variable? :M: Mar 20 21:09:12 hmm https://lists.yoctoproject.org/pipermail/yocto/2012-December/011285.html Mar 20 21:10:05 LetoThe2nd: there's no error in that install log Mar 20 21:10:13 i hate those... Mar 20 21:10:20 mr_science: yeah, thats what i thought too. Mar 20 21:10:53 usually there's an actual error in there somewhere Mar 20 21:11:17 did you look at what mesa packages are in the feed? Mar 20 21:11:38 what do you mean exactly? Mar 20 21:11:42 maybe it didn't package the mesa_blah ipk for some reason Mar 20 21:11:53 ah Mar 20 21:12:08 do an ls -l on tmp/deploy/ipks/arm-blah/mesa* Mar 20 21:12:33 -rw-r--r-- 2 jd jd 9451418 Mär 20 20:54 tmp/deploy/ipk/armv5te/mesa-dbg_9.1.6-r0_armv5te.ipk Mar 20 21:12:36 -rw-r--r-- 2 jd jd 1522 Mär 20 20:54 tmp/deploy/ipk/armv5te/mesa-dev_9.1.6-r0_armv5te.ipk Mar 20 21:12:39 -rw-r--r-- 2 jd jd 9118 Mär 20 20:54 tmp/deploy/ipk/armv5te/mesa-driver-swrast_9.1.6-r0_armv5te.ipk Mar 20 21:12:49 it didn't leave unpackaged files since that would be an error Mar 20 21:13:14 yup, i don't see mesa_9.1.6-r0_armv5te.ipk Mar 20 21:13:31 hm ok? or rather, not ok? Mar 20 21:13:33 so that would be the reason for the do_rootfs install error Mar 20 21:13:48 trying to install a package that doesn't exist Mar 20 21:14:14 so your mesa build is somewhat fubared Mar 20 21:14:53 i've had lots of weirdeness lately with changing upstream behavior Mar 20 21:14:54 hehe, this is a totally untinkered dora-10.0.1 for qemuarm Mar 20 21:15:35 meaning several things that worked fine in November don't work now Mar 20 21:15:42 on master anyway Mar 20 21:15:53 http://paste.ubuntu.com/7127377/ Mar 20 21:16:09 this is the log of a bitbake mesa -f -c package Mar 20 21:16:20 i can also try for head, if you like Mar 20 21:17:49 kapare: best to bbappend base-files, most likely Mar 20 21:18:03 LetoThe2nd: i do see a mesa package in that log Mar 20 21:18:33 so you get a cookie, since i have no idea what's wrong... Mar 20 21:18:52 mr_science: cookie? ;) Mar 20 21:19:08 the edible chocolate variety Mar 20 21:19:10 mr_science: anyways, what might help you? trying on HEAD? trying on qemux86? Mar 20 21:20:07 i guess first you might scan/grep the other build logs and see if you can figure out what happened to the mesa package Mar 20 21:20:21 hmhm Mar 20 21:20:41 if it was actually packaged (which is what it looks like) then i should end up in the feed Mar 20 21:20:47 *it Mar 20 21:22:25 What would be the best way to trace why a recipe is listed as (skipped) in the output of bitbake-layers show-recipes, it appears to have appropriate settings for COMPATIBLE_MACHINE. Mar 20 21:23:15 there is another recipe with a newer version, I'm specifically trying to see why the older version is listed (skipped) in case that's also relevant. Mar 20 21:23:21 mr_science: unfortunately, i have built inside a folder with "mesa" in the path which makes grepping hard. let me fix that and then see. Mar 20 21:58:21 mr_science: does http://paste.ubuntu.com/7127584 maybe give a hint why the mesa package is not built? Mar 20 22:00:59 mh, "NOTE: Not creating empty RPM package for mesa" Mar 20 22:16:35 LetoThe2nd: sorry, was running some tests on an RC build... Mar 20 22:17:20 if all you see is the "NOTE" about not creating an empty package then it sounds like the install was image was missing some stuff Mar 20 22:17:28 s/was// Mar 20 22:17:58 so try maybe a "bitbake mesa -c install" and look at the install image dirs/files Mar 20 22:18:23 compare to what the recipe has in the FILES_${PN} bits Mar 20 22:33:10 Curious- if I run menuconfig on linux-yocto I'm able to create a config, I can compile/deploy that target and looking at the deploy folder it appears that the modules/packages are setup correctly Mar 20 22:33:34 but images like core-image-sato aren't valid targets for menuconfig Mar 20 22:34:10 and if I compile linux-yocto, then try to bitbake core-image-sato, my changes are erased and it uses a default config. Is this expected behavior? any way around it? Mar 20 22:43:28 menuconfig would be meaningless for core-image-sato. core-image-sato has no configuration Mar 20 22:43:36 core-image-sato is just an image, not a kernel Mar 20 22:44:05 if you want your menuconfig changes to persist, then you need to put your defconfig you create with it in an appropriate place so the recipe uses it Mar 20 22:44:05 asdf__, have you read the kernel-dev manual? Mar 20 22:44:13 menuconfig was never intended to make permanent changes Mar 20 22:44:21 asdf__, that process is meticulously documented there Mar 20 22:44:34 http://www.yoctoproject.org/docs/current/kernel-dev/kernel-dev.html Mar 20 22:44:49 the author is a bit of a know it all, but it's still informative ;-) Mar 20 22:45:00 hehe Mar 20 22:46:12 * mranostay yawns in Mar 20 22:46:46 asdf__, please do let me know if you encounter issues with the manual Mar 20 22:46:47 gah, I'm not sure if archiver will really work as a replacement for copyleft_compliance for us Mar 20 22:47:00 I've read that, was hoping there was a way to avoid creating a custom layer JUST to add bluetooth. Mar 20 22:47:04 the archive original sources function is rather heavyweight. emits sstate archives duplicating DL_DIR contents, etc Mar 20 22:47:19 asdf__, you can do it in local.conf as well Mar 20 22:47:36 that mdoe of operation makes sense for many of the more complex configurations, but the archive original sources only case can be quicker than that, just re-run the task instead of using sstate, and symlink from DL_DIR rather than copying Mar 20 22:47:47 which is how copyleft_compliance works, so there's less impact to enabling it Mar 20 22:47:47 KERNEL_FEATURES_pn-linux-yocto += "features/bluetooth.scc" or similar Mar 20 22:47:49 * kergoth ponders Mar 20 22:48:17 I can add kernel config changes to it? I thought it was mostly just for machine declaration and build environment setup Mar 20 22:48:26 thank you for the information, that clears it up a bit! Mar 20 22:48:34 asdf__, you do most anything in local.conf Mar 20 22:48:41 but you have to be aware of scope Mar 20 22:48:47 thus the _pn-... Mar 20 22:49:36 The other thing you could do is ask the recipe maintainer (another know it all) to add the fragment you care about, they just might to it... Mar 20 22:50:48 Ahh okay, see I wasn't sure of the scope of local.conf and didn't want to add things that didn't belong there Mar 20 22:51:17 asdf__, I would say it doesn't belong there, it belongs in a layer Mar 20 22:51:27 but if you don't want to add a layer, you can do local changes there Mar 20 22:51:30 sure, but for testing purposes it'd be okay there? Mar 20 22:51:32 just hard to share with others Mar 20 22:51:34 yup Mar 20 22:51:37 I do that sometimes Mar 20 22:51:45 http://openembedded.org/wiki/OEDAM Mar 20 22:51:58 Don't forget, get travel apporved, add your names Mar 20 22:52:26 kergoth: the new archiver when in original mode simply invokes fetch and "unpacks" anything that is a directory beforing taring it up (thus handling git clones) Mar 20 22:52:51 so for beagleboard or the like, it'd be something along the lines of: KERNEL_FEATURES_beagleboard-linux-yocto += "features/bluetooth.scc" Mar 20 22:52:58 rburton: Yeah, I know. but it uses sstate, which means it then tars up those tarballs into an sstate archive, duplicating the DL_DIR contents, to use to avoid re-running the task later Mar 20 22:53:09 or if I just added it to KERNEL_FEATURES it'd apply it to all boards? Mar 20 22:53:19 kergoth: never noticed it hitting sstate Mar 20 22:53:37 I'm talking about the refactored one, no idea on the old one Mar 20 22:53:40 can that be avoided? Mar 20 22:53:53 but copyleft_compliance doesn't tar anything, it symlinks, so re-running the task is extremely quick, so there's no need for sstate Mar 20 22:54:26 possibly, by special casing the original mode, but there'd be lots of little hacks, since there are other combination modes of operation it supports. e.g. original sources + srpm probably still wants to use sstate to avoid re-creating the srpm :) Mar 20 22:55:35 hmm Mar 20 22:55:38 also, is there an easy way to list what features/*.scc's are available to me on the linux-yocto kernel? Mar 20 22:55:52 (sorry for the newb questions, just starting off with a beagleboard) Mar 20 22:57:04 asdf__, hrm... I believe there is... toms or zedd might know - I usually just search the meta branch myself Mar 20 22:58:56 Thank you VERY much dvhart. Got me started. I suppose you're an author on the project. :) Mar 20 22:59:15 ;-) I'm the Intel kernel lead Mar 20 22:59:25 hooray dvhart! Mar 20 22:59:26 (for yocto) Mar 20 22:59:26 well, that would explain it! cheers! Mar 20 22:59:37 do you guys have any idea why a kernel option set in my defconfig doesn't get set when the .config is generated? Mar 20 23:00:01 side question: will Galileo be added to the BSP's on the yocto page? or is that a non-official support type of thing? Mar 20 23:00:28 Galileo is more of an independent project that uses Yocto, among other things. Mar 20 23:00:50 neabax, usually because a dependency is missing Mar 20 23:00:52 asdf__: the gailileo site has their bsp to download now Mar 20 23:01:31 dvhart: does it usually throw a warning about that? I don't see anything in my log.do_configure Mar 20 23:02:46 it's in the logs in the build dir... something like "tmp/work/MACHINE*/linux-yocto*/*/linux/.meta/standard/* Mar 20 23:02:56 something like that... just dumped that from memory Mar 20 23:03:14 (that's admittedly a totally crap location for this stuff ;-) Mar 20 23:13:35 dvhart: that seems a bit hard to find :) Mar 20 23:21:19 Alright it seems the configure step overwrites the defconfig that ends up being patched Mar 20 23:21:33 so instead of patching that option in I had to just change the defconfig in the linux-mainline folder Mar 20 23:23:42 neabax: sounds like you'd be better off using a config fragment Mar 20 23:25:48 kergoth: thanks, that seems a lot cleaner, I'll look into using that Mar 20 23:26:31 neabax, config fragments are also a lot easier to maintain Mar 20 23:26:39 you define only what you really care about Mar 20 23:27:45 yeah, meta-mentor *still* has an old busybox defconfig, from so long ago int eh mists of time that I have no idea which options we had to enable Mar 20 23:27:55 so trying to get that into a fragment now is a pain :) Mar 20 23:28:08 might just ahve to abandon it and let QA open new issues if we'r emissing something Mar 20 23:29:02 it's not super clear in the yocto manual, can you use config fragments in any recipe or does it have to be in the kernel recipe? Mar 20 23:29:20 not all recipes even have a config to have fragments of :P Mar 20 23:29:42 neabax, busybox and linux-yocto* Mar 20 23:29:50 busybox and kernel-yocto recipes support it out of the box today for sure. certain others, but theres' no generic mechanism for non-linux-yocto kernels (yet) Mar 20 23:30:22 Although you can use them on non linux-yocto sources using the linux-yocto-custom template in meta-skeleton Mar 20 23:30:49 so can I use that for linux-mainline or does it only work for kernel-yocto? Mar 20 23:30:50 I thoguth the cml class supports fragments now.. so as long as the thing being configured uses the cml class you are good.. (which of course also means it uses the kernel-like method of config) Mar 20 23:31:25 neabax, see the linux-yocto-custom recipe and the kernel-dev manual, they document this Mar 20 23:31:28 nope, doesn't yet Mar 20 23:31:48 cml1 provides a diffconfig task to let you generate fragments, but it can't use them, since cml1 doesn't provide a generic mechanism for the defconfig -> .config process Mar 20 23:32:10 every cml1 recipe doe sthat themselves, so there's no standard DEFCONFIG variable, or place in the build to inject hte fragments Mar 20 23:32:14 I thought it did Mar 20 23:34:46 I guess not.. Mar 20 23:34:51 it's apparently pretty simple though Mar 20 23:34:59 do_configure () { Mar 20 23:35:04 merge_config.sh -m .config ${@" ".join(find_cfgs(d))} Mar 20 23:35:04 cml1_do_configure Mar 20 23:35:04 } Mar 20 23:35:18 (see busybox.inc for the definition of find_cfgs) Mar 20 23:35:58 indeed Mar 20 23:37:02 a simple little cml1-config class which does that in a task between do_patch and do_configure, using DEFCONFIG, which we could adapt cml1-based recipes to use, would be nice. i created one along those lines for meta-mentor, but it can be simplified now. it had a fragment filtering pipeline to let you remove bits, but now that we can use _remove, its likely overkill Mar 20 23:37:19 https://github.com/MentorEmbedded/meta-mentor/blob/master/meta-mel/classes/cml1-config.bbclass is that one, definitely overkill :) Mar 20 23:37:44 I suspect it's not there by default because the kernel and busybox find and combine their fragments differently.. Mar 20 23:38:01 but the busybox 'way' is what I'd suggest for the default.. (minus the busybox specific line which I didn't paste above) Mar 20 23:38:35 well, like i said, there's no standard mechanism to even write .config. since its inside of do_configure, there's no injection point, so we can't modify cml1 to handle it without having to change all the recipes anyway Mar 20 23:38:39 has to be something new which is opt-in Mar 20 23:38:43 * kergoth shrugs Mar 20 23:41:52 simply documenting what I posted above wouldn't be a bad idea.. Mar 20 23:42:08 the only thing busybox does "special" is the setup of the initial .config based on the distro feature flags.. Mar 20 23:42:18 that could probably be do in a do_configure_prepend actually.. Mar 20 23:42:24 then the do_configure could be "more standard" Mar 20 23:43:36 I can say from a newbie's perspective, a step-by-step guide on how to, say, add the bluetooth module specifically (or any module specifically) would be useful :D Mar 20 23:44:23 Umm.. I have an example of how to enable selinux.... Mar 20 23:44:57 http://git.yoctoproject.org/cgit/cgit.cgi/meta-selinux/tree/recipes-kernel/linux Mar 20 23:45:01 add a .bbappend for the kernel.. Mar 20 23:45:08 then add a custom '.cfg' file.. Mar 20 23:45:29 the contents of the bbappend simply adds the '.cfg' to the SRC_URI.. and add the location of the .cfg to the FILESEXTRAPATHS.. Mar 20 23:46:23 so.. something like: linux-yocto_3.16.bbappend: Mar 20 23:46:35 FILESEXTRAPATHS_prepend := "${THISDIR}/${PN}:" Mar 20 23:46:44 SRC_URI += "file://myconfigfrag.cfg" Mar 20 23:46:49 asdf__, please do send me email feedback on the manual Mar 20 23:47:00 then linux-yocto/myconfigfrag.cfg contains just the items to enable/disable for your config Mar 20 23:47:07 (but I agree.. the manual should have this info) Mar 20 23:47:14 it does Mar 20 23:47:23 especially since it is fairly easy to do it.. once you know how Mar 20 23:47:46 http://www.yoctoproject.org/docs/current/kernel-dev/kernel-dev.html#changing-the-configuration Mar 20 23:48:25 yup that's it.. Mar 20 23:48:25 asdf__, there is an example there specifically how to add 8250 support Mar 20 23:48:32 both the full defconfig, and a "fragment" Mar 20 23:48:44 I can't very well document how to add all 5000 possible kernel configurations.... Mar 20 23:48:49 ;) Mar 20 23:48:56 so some level of abstraction has to be allowed for :-) Mar 20 23:50:42 I'm curious - did you read this section and it didn't make sense? Did you not find this section in your search? Sometimes creating the right structure can be a challenge. The manual is intended to be organized by commonly performed tasks - specifically the ones you have been describing Mar 20 23:50:50 So I'm concerned it isn't meeting your needs.... Mar 20 23:51:08 I guess that's a good point, it does mention that, I just fell back to the 'well does bluetooth.scc even exist by default?, how do I know which modules are supported?" Mar 20 23:51:20 the documentation is VERY good, it's just a lot for a newbie to absorb in less than 24 hours Mar 20 23:51:30 that's fair enough Mar 20 23:51:39 I was at it two years before I wrote it after all ;-) Mar 20 23:52:01 ya.. a link for beginners to 'start here'.. may be useful.. I know the few times I've had to adjust the kernel the items in the section dvhart pointed you at was enough.. Mar 20 23:52:02 So that is one good bit of feedback - how do I know which fragments already exist Mar 20 23:52:16 the scc stuff wasn't needed until you get to more advanced features.. Mar 20 23:52:17 I actually have to say, of most 'side linux projects' ive seen, the documentation for Yocto is absolutely incredible Mar 20 23:52:33 log that and print it Mar 20 23:52:34 quick Mar 20 23:52:38 before the internet dies Mar 20 23:53:53 And see, I'm uncertain what something like: KERNEL_FEATURES_beagleboard-linux-yocto += "features/bluetooth.scc" would do as opposed to adding bluetooth enabling in a config fragment. Mar 20 23:54:14 "adding bluetooth enabling" - I'm rocking english tonight. Heh. Mar 20 23:54:52 asdf__, they would do the exact same thing provided they contain the same CONFIG_* lines Mar 20 23:55:22 gotcha, and the .SCC includes all of that CONFIG_ line data Mar 20 23:55:32 (I need to go look for .scc's) Mar 20 23:55:38 the .scc is a description file that contains links to .cfg files Mar 20 23:55:50 acc is also documented in the kernel-dev manual under advanced meta data Mar 20 23:55:55 scc even Mar 20 23:56:00 There we go, that makes sense. Mar 20 23:56:17 again, just a lot of info I'm still sifting through, documentation is amazing, just overwhelming simply based on how much there is of it Mar 20 23:56:52 I don't think anything is necessarily lacking in the documentation, for the record. Mar 20 23:57:37 asdf__, https://bugzilla.yoctoproject.org/show_bug.cgi?id=6022 Mar 20 23:57:38 Bug 6022: normal, Undecided, ---, scott.m.rifenbark, NEW , kernel-dev: How do I know which features are available for KERNEL_FEATURES ? Mar 20 23:58:10 Awesome! Mar 20 23:58:25 asdf__, if you could Cc yourself, you can help review the solution. Mar 20 23:58:57 will do, thanks. Mar 21 00:24:59 hmm, any hints on what that is: do_package_write_ipk_setscene) failed with exit code '1' - real task will be run instead Mar 21 00:40:15 Jin^eLD: you'd need to look at the log for that task Mar 21 00:42:35 I think it was somehow related to a set PE Mar 21 00:42:45 it was a recipe which I wanted to rebuild automatically each time it was being built Mar 21 00:43:14 I imported it from a 1.5 year old OE core configuration where it worked OK Mar 21 00:43:25 now I removed PE it the problem disappeared Mar 21 00:48:28 nite **** ENDING LOGGING AT Fri Mar 21 02:59:58 2014