**** BEGIN LOGGING AT Tue Apr 30 02:59:58 2013 Apr 30 03:00:16 if youw ant to build nano, bitbake nano. if you want to build an image with nano inside of it, that's a different question, and you can read local.conf for examples of how to add packages to an image Apr 30 03:01:28 ok. this is my first experience building linux anything. i'm getting it...maybe slowly. i've managed to put together meta-ti and meta-selinux so i can build a 'core-image-selinux' for the TI am180x-evm devkit. Apr 30 03:01:58 but now i want to add TI's syslink package. i see a recipe for it in hob as part of the meta-ti layer. Apr 30 03:48:08 I am booting up core-image-minimal on an Atom D2500HN board. the first boot works but if I reboot, syslinux hangs. any ideas? Apr 30 07:21:34 nevermind, solved the boot issue by updating syslinux from 4.03 to 4.06 Apr 30 07:37:35 good morning Apr 30 07:37:42 morining Apr 30 08:00:27 morning all Apr 30 08:21:56 good morning bluelightning Apr 30 08:22:10 hi mattnie Apr 30 08:43:24 otavio: I posted on meta-freescale ML Apr 30 08:49:58 there are lots of pty, tty, ttyS? in /dev that are not connected. how would I make it dynamic so only connected devices are shown in /dev? Apr 30 08:54:55 Net147: I'm guessing that would be a matter of configuring udev Apr 30 08:55:13 udev generally creates the devices in /dev automatically Apr 30 08:55:24 perhaps I need to suppress creation of these static device files? Apr 30 08:55:59 I have udev already and it creates /dev/sdb etc. automatically when I plug in drives Apr 30 08:57:25 do we have a pseudo benchmark? I'm wondering if adding -fno-builtin-memcmp to pseudo/sqlite would make any measurable difference. (http://gcc.gnu.org/bugzilla/show_bug.cgi?id=43052 ) Apr 30 08:57:26 Bug 43052: was not found. Apr 30 09:04:16 maybe CONFIG_LEGACY_PTY_COUNT=256 is related... Apr 30 10:19:59 otavio: some news, with your standard kernel Linux version 2.6.35.3-11.09.01+yocto+ged51f22 my problem disappered, so must be an issue in my kernel. Apr 30 11:07:49 mckoan: :-D Apr 30 11:08:04 mckoan: which machine you'll be using? Apr 30 11:11:43 I get "invalid argument" while booting core-image-minimal vmdk image when it tries to mount rootfs.img to /rootfs/ using /dev/loop0. any ideas? Apr 30 11:13:11 in particular when /init in initrd runs "mount -o rw,loop,noatime,nodiratime /media/sda2//rootfs.img /rootfs/" Apr 30 11:16:04 bluelightning: I managed to remove 512 unnecessary /dev/{pts,tty}a0..zf device nodes by adding CONFIG_LEGACY_PTY_COUNT=n to my kernel .cfg Apr 30 11:18:46 otavio: how do I specify a Makefile file? EXTRA_OEMAKE = "'-f Makefile.x11'" doesn't seem to work Apr 30 11:22:32 panda84kde: you'll need to provide your own do_compile, just call oe_runmake -f Makefile.x11 in it Apr 30 11:25:10 rburton: thanks. That's what I've done initially, but after a "bitbake -c clean package" it turned crazy and started skipping fetch, unpack, compile and so on Apr 30 11:25:15 will double check. Apr 30 11:28:15 rburton: That's what I get with the do_compile as you mentioned: http://pastebin.com/DmLXDPFu . Really strange, it used to work... Apr 30 11:30:39 panda84kde: not sure what the problem is Apr 30 11:31:07 panda84kde: it does not seems to be the error in the log Apr 30 11:31:34 panda84kde: please start the build over. bitbake -c cleansstate Apr 30 11:31:43 panda84kde: and try again Apr 30 11:32:40 panda84kde: poky/meta/recipes-support/js/js_1.7.0+1.8.0rc1.bb does 'simular' oe_runame use Apr 30 11:35:37 otavio: thanks, it worked. Really interesting that "cleanstate" command! Learning bitbake day by day! Apr 30 12:22:23 otavio: hi Apr 30 12:22:54 otavio: is a custom board derived from imx53qsb Apr 30 12:27:55 Net147: great, I'll make a note of that Apr 30 12:30:41 how would I override classes/boot-directdisk.bblass in my bsp layer? Apr 30 12:34:50 nevermind, got it. have to change BBPATH .= ":${LAYERDIR}" to BBPATH := "${LAYERDIR}:${BBPATH}" Apr 30 12:36:22 Net147: best practice would be to use BBPATH =. "${LAYERDIR}:" Apr 30 12:36:59 bluelightning: yep, I just did that Apr 30 12:37:19 Net147: there are probably better ways of doing what you need than providing your own overlayed version of boot-directdisk.bbclass as well Apr 30 12:39:02 bluelightning: I am putting syslinux 4.06 in my BSP and the path to mbr.bin has changed so I need to adjust the bbclass Apr 30 12:39:26 bluelighting: with older versions of syslinux, it boots the first time but subsequent boots hang after printing SYSLINUX line Apr 30 12:42:28 Net147: ah, I see... I guess you would only need that bbclass until the new version of syslinux is merged into OE-Core Apr 30 12:43:03 bluelightning: yep. I have submitted patch to update syslinux and guile to oe-core mailing list already Apr 30 12:43:15 Net147: yep, saw that today, thanks :) Apr 30 12:43:58 bluelighting: slowly learning yocto. previously I have been using buildroot. Apr 30 12:44:21 ahh I can't spell today Apr 30 13:10:36 if I want to remove some files from a recipe, is it best to create a .bbappend and add do_install_append? Apr 30 13:11:44 or is there another method to add/remove files for a recipe in bsp layer? Apr 30 13:13:18 FILES_something -= ${sysconfdir}/blah ? Apr 30 13:13:40 Net147: there is no -= operator Apr 30 13:14:07 Net147: you could only delete them in a do_install_append Apr 30 13:14:20 Net147: out of curiosity, what files are you removing? Apr 30 13:14:57 bluelightning: lib/udev/rules.d/80-net-name-slot.rules Apr 30 13:15:27 bluelightning: I want eth0 instead of the auto-generated network interface names Apr 30 13:15:55 bluelightning: e.g. enp2s0, enp1s2, etc. Apr 30 13:15:55 Net147: that sounds like policy rather than something machine-specific; I don't think it belongs in the BSP Apr 30 13:16:21 bluelightning: what would you recommend? Apr 30 13:16:25 Net147: that sort of thing would normally be done in your own custom distro layer Apr 30 13:17:05 bluelightning: still in do_install_append or ? Apr 30 13:17:16 Net147: yes Apr 30 13:18:26 bluelightning: I haven't got up to learning distro conf yet Apr 30 13:18:50 Net147: http://www.yoctoproject.org/docs/1.4/dev-manual/dev-manual.html#creating-your-own-distribution Apr 30 13:19:16 bluelightning: thanks. i'll go through it Apr 30 13:20:14 bluelightning: should be in new layer or bsp layer? Apr 30 13:20:23 Net147: a new layer Apr 30 13:20:47 the idea is that distro policy is kept separate from machine support, so you can use the two interchangeably Apr 30 13:20:55 bluelightning: okay. so then I should have like meta-myproject and meta-myproject-bsp. currently I just have meta-myproject which is the bsp. Apr 30 13:21:25 Net147: typically yes Apr 30 13:27:29 bluelightning: then I should put syslinux and guile recipes into meta-myproject instead of the bsp layer since it isn't machine specific Apr 30 13:29:15 Net147: I suspect that might be the best course yes Apr 30 13:29:52 bluelightning: well, thanks for all the advice Apr 30 13:30:03 np Apr 30 13:34:30 any hints on how to debug when a recipe is "not installable"? Apr 30 13:35:24 i'm trying to create a recipe that *only* creates a user by inheriting from useradd Apr 30 13:36:34 when i add it to MACHINE_ESSENTIAL_EXTRA_RDEPENDS in my machine conf i fail in do_rootfs for core-image-minimal Apr 30 13:40:15 maybe a better question would be what is the recommended way to add a specific user account to my image? Apr 30 14:06:06 can I have a .bbappend that applies to any version instead of specific version? Apr 30 14:08:23 Net147: no, it must be a specific version I'm afraid Apr 30 14:23:09 has booting core-image-minimal using vmdk been tested recently? I get "invalid argument" when /init inside initrd runs mount -o rw,loop,noatime,nodiratime /media/sda2//rootfs.img /rootfs/ Apr 30 14:59:19 I keep getting "Your version of bblayers.conf was generated from an older/newer version of bblayers.conf.sample and there have been updates made to this file"... Apr 30 14:59:41 YPTM: Saul is here Apr 30 15:00:38 YPTM: Laurentiu Palcu joined Apr 30 15:00:40 YPTM: Welcome to the technical team meeting, please let me know who's on the bridge. Thanks! Apr 30 15:00:45 YPTM: Kevin Strasser is here Apr 30 15:00:47 YPTM: Scott Rifenbark is on the call Apr 30 15:00:52 YPTM: Jeff Polk is here Apr 30 15:00:54 YPTM: Beth flanagan is here Apr 30 15:00:55 YPTM: Björn Stenberg is here Apr 30 15:01:04 Net147: delete the build directory, it was created with another poky version, thus bblayers.conf is outdated/newer Apr 30 15:01:04 Song_Liu: YPTM: Richard is on the call Apr 30 15:01:10 YPYTM: Tom here Apr 30 15:01:14 YPTM: Corneliu joined Apr 30 15:01:34 YPTM: Make here now Apr 30 15:01:36 'er.. Mark Apr 30 15:02:03 YPTM: nitin getting on the bridge Apr 30 15:02:21 Corneliu: I haven't updated poky at all... Apr 30 15:03:10 YPTM: Denys is here Apr 30 15:03:18 YPTM: jzhang on the call Apr 30 15:03:59 yptm: ross on the call Apr 30 15:04:01 YPTM: Any opens? Apr 30 15:04:16 Net147: copy meta-yocto/conf/bblayers.conf.sample over build/conf/bblayers.conf Apr 30 15:04:46 Me & Saurav joined YPTM Apr 30 15:04:47 Net147: and re-enter your information there Apr 30 15:05:21 Unrelated to the call, but it seems that an image build from sstate results in an incomplete license manifest. The generation code ignores licenses from LICENSE which didn't get licenses collected in deploy/license, and do_rootfs doesn't recrdeptask do_populate_lic, leaving a manifest with all empty LICENSE: fields Apr 30 15:05:27 hmm Apr 30 15:05:27 Corneliu: I don't see any version change Apr 30 15:05:28 Song_Liu: AlexG here too, just that I will drop in 20 minutes Apr 30 15:05:32 * kergoth investigates Apr 30 15:05:55 https://www.google.com/calendar/embed?src=3a0927ctps51u8rs2jv1uligdc%40group.calendar.google.com&ctz=America/Los_Angeles  Apr 30 15:06:11 Corneliu: both are version 6 but I add print statement in sanity check script and it says one of them is 5 Apr 30 15:09:13 Net147: are you using multiple build/poky directories/environments? Apr 30 15:09:14 YPTM: Paul Eggleton joined Apr 30 15:09:35 https://wiki.yoctoproject.org/wiki/Yocto_Project_v1.5_Status#Yocto_Project_1.5_Release_Criteria Apr 30 15:12:46 kergoth: ergh. Apr 30 15:14:27 Corneliu: no. must be a typo in local.conf. I think I fixed now. Apr 30 15:16:14 it seems when I try to set DISTRO ?= "mydistro" it gives that error Apr 30 15:17:15 For 1.4 "Distribution Testing was performed on: Ubuntu 12.04, Ubuntu 12.10, Fedora 17, Fedora 18, OpenSUse 12.1, OpenSuse 12.2, CentOS 6.3 " Apr 30 15:24:43 Corneliu thanks.. ya the CentOS was left off the completion criteria, which is why it was brought up as a future issue Apr 30 15:31:12 looking for comments on "How to generate rootfs from package feed" , https://lists.yoctoproject.org/pipermail/yocto/2013-April/015730.html Apr 30 15:32:12 Saurav: what you want sounds a lot like what happens already - if you rebuild an image and no packages need to be re-build, it uses the package feeds in deploy/ Apr 30 15:32:39 Saurav: unless you mean from an arbitrary feed, and not the one that it built Apr 30 15:33:45 I meant passing a prebuilt (from yocto) to poky , to generate rootfs Apr 30 15:34:18 prebuilt package feed(from yocto) Apr 30 15:34:42 Corneliu: anything special I need to do in distro conf to avoid bblayers.conf sanity check error? Apr 30 15:35:30 Net147: if you are not working with git it should not generate issues Apr 30 15:36:45 Corneliu: I am using git Apr 30 15:37:37 Net147: changing between commits can cause the bblayers.conf file to be outdated/newer compared to the repository's state date Apr 30 15:38:19 Corneliu: well I was using poky distribution and it compiled ok. then i'm following http://www.yoctoproject.org/docs/1.4/dev-manual/dev-manual.html#creating-your-own-distribution Apr 30 15:38:23 Net147: I think you'll have to drop LCONF_VERSION to 5 in your conf/bblayers.conf Apr 30 15:38:49 bluelightning: any reason why I have to do so? Apr 30 15:39:12 Net147: unfortunately, LAYER_CONF_VERSION which this value is compared to is different between OE-Core and Poky Apr 30 15:39:44 rburton: pass a prebuilt package feed to a new build directory so that, rootfs will be generated by installing packages, not by compiling again. Apr 30 15:39:48 bluelightning: it's not mentioned anywhere in documentation that I need to drop LCONF_VERSION when creating distribution... Apr 30 15:40:08 Net147: no it isn't, sorry about that; I hadn't appreciated this problem would occur Apr 30 15:40:33 Saurav: i think that is something in the planning under "improve deployment", but i'm curious as to why your main build and image pokys are separate. Apr 30 15:42:24 rburton: We have a scenario where binaries(package feed) are built and used by different users to make required rootfs. i am curious. if any thing related is already implemented. Apr 30 15:43:53 yptm: thanks for joining the meeting. Have a nice day/evening Apr 30 15:48:08 Song_Liu: goodbye. See you in 2 weeks :) Apr 30 16:00:11 bluelightning: can I make distro that extends poky distro? Apr 30 16:00:51 Net147: sure, you can; you don't have to do that just to resolve this particular issue though Apr 30 16:01:02 bluelightning: generally recommended to start from scratch though? Apr 30 16:01:25 I think so yes Apr 30 16:03:38 then do I need meta-yocto and meta-yocto-bsp in bblayers.conf still? Apr 30 16:05:19 Net147: if you don't use anything from them, no they're not needed Apr 30 16:33:40 if I want to force .ipk packaging in distro config can I just do PACKAGE_CLASSES = "package_ipk" ? Apr 30 16:36:01 Net147: yep Apr 30 16:38:28 bluelightning: great Apr 30 16:40:06 where does the oe in TARGET_SYS = "i586-oe-linux" come from? Apr 30 16:40:50 a setting somewhere? Apr 30 16:42:37 Net147: it's composed of TARGET_ARCH, TARGET_VENDOR and TARGET_OS Apr 30 16:43:02 TARGET_ARCH would come in via the machine configuration (usually set by an inc file that it would be including) Apr 30 16:43:24 TARGET_VENDOR ("-oe" by default) could be set by the distro config Apr 30 16:44:10 ignoring the gnueabi fun, TARGET_OS is always "linux" unless you want to do the work to support some other target operating system ;) Apr 30 18:39:08 hello, is it possible in .bbappend to override all of .bb file? Apr 30 18:47:51 alex_kag: if you want to do that, don't use a .bbappend, just use a .bb Apr 30 20:09:08 hi everyone, i had a question regarding moving from YP 1.3 to YP 1.4 and was wondering if someone could help me out. Apr 30 20:09:25 i am having an issue when trying to fetch from git.yoctoproject.org Apr 30 20:10:22 i noticed in the refrence documentation (section 4.2.3: proxies and fetching source) that some chagnes where made when going through a proxy Apr 30 20:11:17 I looked at site.conf.sample and tried specifing my proxy in my own build/conf/site.conf but still getting a failure when trying to fetch Apr 30 20:12:02 is there something that I missed? Apr 30 22:15:10 why is temp directory build/tmp-eglibc instead of build/tmp when I create custom distribution? Apr 30 22:17:36 I've got three shared objects a yocto-built binary links against which are present in my image, but which the RPM database is not aware are installed, which means the Yocto-produced RPM containing my binary won't install due to missing dependencies. Any suggestions on how to hack around that? (I don't really care how ugly they are if they're easy.) Apr 30 22:17:57 however you did that, something is invlaid.. Apr 30 22:18:03 Net147: meta/conf/distro/defaultsetup.conf:TCLIBCAPPEND ?= "-${TCLIBC}" Apr 30 22:18:03 meta/conf/distro/defaultsetup.conf:TMPDIR .= "${TCLIBCAPPEND}" Apr 30 22:18:12 this sounds like the items were installed into the sysroot, but never into a -dev package.. Apr 30 22:18:17 (or final package) Apr 30 22:18:28 or you have a host-contamination problem where you are picking up host junk.. Apr 30 22:18:47 if you are using the OE-core compiler, you should get a warning message when host contamination occurs.. (Warning: blah is unsafe when cross compiling" Apr 30 22:20:24 fray: talking to me? Apr 30 22:21:44 yes Apr 30 22:23:59 my root problem is that I have a non-yocto RPM with some binaries in it, but whoever constructed this RPM did it in a way that it doesn't claim to provide the shared objects in question Apr 30 22:24:22 that would be a problem.. you may have to extract it and repackage that RPM Apr 30 22:26:07 Unfortunately, that won't work. This RPM is provided by a third party, and may or may not be obtained and installed by the end user. So the end user wouldn't be installing the fixed RPM, but rather the original broken one. Apr 30 22:26:24 no I'm saying YOU need to do it, or provide a script that does it or.... Apr 30 22:26:36 or don't use RPM to install that.. use rpm2cpio and cpio to extract it Apr 30 22:26:46 or rpm -Uhv --nodeps .... Apr 30 22:26:51 lots of ways to install a broken package Apr 30 22:27:08 just be aware all of them turn off dependency scanning.... Apr 30 22:27:48 JaMa: thanks Apr 30 22:28:05 I can certainly do that. If I were writing my package's .spec file by hand, I could instruct RPM not to emit 'requires' entries for the three shared objects provided by this broken RPM. Apr 30 22:28:23 if there are requires entries, they are most likely valid... Apr 30 22:28:30 it's the provide entries I thought were missing Apr 30 22:29:31 if ther equires entries are not libfoo.so style, but a package that doesn't exist in the YP.. then it might be easier to find the package providing the functionality and adding an RPROVIDES.. Apr 30 22:29:44 the third-party RPM provides three shared objects but does not have 'provides' entries for them. my yocto-built RPM obviously has 'requires' entries for these three libraries. Apr 30 22:29:45 alternatively you can just tell RPM that whatever it is "exists".. eve if it's not in the RPM data.. Apr 30 22:30:14 If I were writing my RPM's spec file by hand, I could suppress that and create an RPM the end user wouldn't need to install with --nodeps Apr 30 22:30:40 the -best- way to do it is repackage the upstream RPM.. write a recipe that uses rpmcpio to extract it.. a do_configure / do_compile that does nothing.. anda do_install that copies the files into the right place.. Apr 30 22:30:52 YP/OE-core already knows how to extract an RPM BTW.. Apr 30 22:31:17 then the build system can repackage and add whatever missing information when it regenerates the package.. you certainly wouldn't be th first to do this.. Apr 30 22:31:26 unfortunately, I can't repackage because I can't convince my end user to install my repackackaged version instead of the broken one. Apr 30 22:31:30 and as long as you aren't redistributing something you don't have a license to you should be fine.. Apr 30 22:31:58 then they need to take it up with the original vendor.. or you need to change the way your custom code is being packaged Apr 30 22:32:42 on the "change the way your custom code is packaged" front, what do you suggest? Apr 30 22:33:16 how are you packaging it? are you dumping that binary mess into the YP sysroot and letting the recipe system do it.. or are you manually generating an RPM or? Apr 30 22:36:31 if this is happening on the OE-core side, where you've just dumped the binaries into the sysroot.. you will want to diabled the automatic dependency generation.. you can do that by defining (in your recipe) Apr 30 22:36:52 PACKAGEFUNCS = "package_fixsymlinks package_name_hook package_depchains emit_pkgdata" Apr 30 22:37:11 (see meta/classes/package.bbclass for the default.) Apr 30 22:45:04 ...if my recipe used a _prepend to add code that rewrote the .shlibsdeps file after filtering out the three libraries in question to read_shlibdeps, would that work? Not quite as heavy handed.... Apr 30 22:45:12 if I want to remove/modify the user@host information from the built kernel packages, how would I go about doing that? Apr 30 22:45:22 thats a lot more difficult.. Apr 30 22:45:44 better to just avoid calling the various items.. the do_shlib one for sure.. Apr 30 22:46:00 and then use the RDEPENDS_... and RPROVIDES_... set to whatever you want them to be Apr 30 22:47:17 okay. thanks fray, that was very helpful. Apr 30 22:48:15 for example when I do cat /proc/version in the image, it shows user@host from the build host Apr 30 22:51:13 Net147: look in the kernel's makefile for how exactly that works. It's probably a make variable assigned to some short of $(shell) expression, which means you can probabaly override it on make's command line--which you can do from your recipe Apr 30 23:18:07 evanp: thanks. i'll look into it. May 01 02:09:10 when building xf86-video-intel 2.21.3, I get intel.h:66:18: fatal error: dri2.h: No such file or directory... missing dependency? May 01 02:19:23 I guess I need opengl in DISTRO_FEATURES May 01 02:26:22 how do i find out what images i can build beyond core-image-minimal? May 01 02:53:13 cpow: you can try 'ls meta*/recipes*/images/*.bb' May 01 02:54:17 http://elinux.org/Yocto_Project_Introduction#list_available_images May 01 02:55:39 ah thanks stac May 01 02:56:12 np May 01 02:56:20 i'm pretty new to yocto... May 01 02:56:46 but i've managed to get meta-ti and meta-selinux layers together and build a core-image-selinux for a TI devkit. May 01 02:56:55 but now i'm trying to figure out what else i can do. May 01 02:57:10 i'm going to look into the eclipse plugin. May 01 02:57:25 how does that plugin communicate with the target? Ethernet? **** ENDING LOGGING AT Wed May 01 02:59:58 2013