**** BEGIN LOGGING AT Wed Sep 20 03:00:02 2017 Sep 20 08:22:05 I am trying to use the "devtool deploy" subcommand but I get an error "No files to deploy - have you built the [recipename] recipe? If so, the install step has not installed any files." Looking at the source https://github.com/openembedded/openembedded-core/blob/master/scripts/lib/devtool/deploy.py I do not understand why I get this error as my "D" variable points to the correct directory Sep 20 08:34:54 hey all Sep 20 08:35:20 is there a way to expand a flagged variable from a shell script in a yocto recipe Sep 20 08:35:45 i.e get the value of VARIABLE[flag] from within a do_install ? Sep 20 08:36:36 boucman_work: tested ${VARIABLE[flag]} ? Sep 20 08:37:07 I tried, but flag is a shell variable. so I'm not sure it's possible at all... Sep 20 08:37:49 my code is for name in $VAR do stuff with $VAR2[$VAR] Sep 20 08:37:53 hmm, yeah, because the ${ } is pre-parsed before sh handoff and execution Sep 20 08:38:16 yeah, I guessed so... i'll have to find a trick around that Sep 20 08:38:59 i'm no expert, and maybe bitbake exports some sh function that you can use Sep 20 08:40:26 i'll have a look... but the BB doc is still a bit messy in my mind Sep 20 08:51:09 boucman_work: try ${@d.getVarFlag('MYVAR', 'flagname')} Sep 20 08:51:52 I just abandonned flages and exported VAR2_flag that's good enough for my usecase... Sep 20 08:52:19 bluelightning: that wouldn't really help since flagname is only known at shell execution time... the only way would be to go through python I guess Sep 20 08:52:37 boucman_work: right, yes Sep 20 08:53:03 or somehow have a primitive "export all flags of VAR2" at the beginnning of the shell script that would be expanded by bitbake before the shell is run Sep 20 08:53:21 but I got my workaround, thx guys Sep 20 09:18:05 hi! my embedded board has two rtc's: /dev/rtc0 and /dev/rtc1. i'm not sure how to select the correct one? /dev/rtc is linked to /dev/rtc0, but /dev/rtc1 is the correct one Sep 20 09:18:45 ionte: the simplest trick is to just deactivate the unwanted one, probably in the device tree Sep 20 09:19:08 LetoThe2nd guess so. ... Sep 20 09:58:54 How do I configure the Linux Kernel in/from/using Yocto? Sep 20 10:00:23 User12345678: depends a bit. either through just a defconfig file that you place in your layers kernel recipe, or through fragments, if its a little bit more elaborate Sep 20 10:01:16 If I go through the deconfig file, any step by step on how to do it? Sep 20 10:01:37 *I'm a few days new to Yocto* Sep 20 10:02:12 User12345678: it should all be nicely documented in http://www.yoctoproject.org/docs/2.3.1/kernel-dev/kernel-dev.html Sep 20 10:02:20 http://www.yoctoproject.org/docs/latest/kernel-dev/kernel-dev.html#generating-configuration-files Sep 20 10:02:23 etc Sep 20 10:03:27 rburton: btw, you coming to prague? Sep 20 10:03:54 Ok, thanks LetoThe2nd and rburton Sep 20 10:05:12 LetoThe2nd: sadly travel budget is "restricted" Sep 20 10:05:29 rburton: awww Sep 20 10:17:16 If I may ask Sep 20 10:17:50 is there any plan to have linux-yocto-custom supporting defconfig generated with savedefconfig? (and not the copy paste from .config) Sep 20 10:18:19 And one more question -> is there a rule of thumb as when shall I remove sstate directories? Sep 20 10:18:40 e.g. /sstate-cache/ Sep 20 10:18:49 or any other ones? Sep 20 10:19:25 For example, when I switch between machines it helps to keep things working...... Sep 20 10:23:22 Why do we need a deconfig file if the .config file defines how to build the kernel? Sep 20 10:24:17 User12345678: because the .config file should not be part of the git repository of the kernel sources Sep 20 10:24:47 User12345678: basically its just a convention that the defconfig file you provide gets forwarded to the kernel as its .config Sep 20 10:26:21 LetoThe2nd: And what's about re-using the default configs for architectures? Sep 20 10:27:21 lukma: never done that, so i can't comment on it. but i guess that would correlate with the fragments systme Sep 20 10:28:26 LetoThe2nd: Fragments are ok Sep 20 10:28:39 Just want to say that it is a bit misleading Sep 20 10:29:22 # To use "in-tree" defconfig -> avoid code duplication Sep 20 10:29:22 COMPATIBLE_MACHINE_XXX = "XXX" Sep 20 10:29:22 KBUILD_DEFCONFIG_pn-linux-yocto-custom_XXX = "imx6q_XXX_defconfig" Sep 20 10:29:24 like i said, i cant really comment on it. in my usecases, using a default configuration provided from somewhere else is just not desirable. Sep 20 10:29:36 and then you can avoid code duplication Sep 20 10:30:02 but imx6q_XXX_defconfig is really the .config Sep 20 10:30:21 and not the "excerpted" version which is created with savedefconfig Sep 20 10:34:30 hello guys !! I have a zip file wich contains a folder called "driver" which, in turn, contains another zip file that represents the driver source code Sep 20 10:34:58 I have written a recipe to fetch the main zip file and extract it to the proper tmp folder Sep 20 10:35:04 fberg: let me guess. your question is how to pack it a thrid time? :-) Sep 20 10:35:12 yep Sep 20 10:35:27 cool! Sep 20 10:36:01 I have the ../tmp/.../4.3.19-r0/main_folder/driver/driver_source.zip Sep 20 10:36:52 how can I mange to unzip the soruce code and compile it with the same recipe Sep 20 10:37:35 Hello, I am trying to create a bitbake recipe that makes an sdcard image using wic. However, I get wrong permissions on my rootfs folders and systemd fails to boot: https://pastebin.com/QxW9CPFs Sep 20 10:37:46 fberg: to me it sounds liek you need to create a custom unpack stage for that recipe Sep 20 10:38:08 as you can see bin, boot folders, etc. belong to group pulse instead of root for some reason Sep 20 10:39:19 ok ! I thought there was some "ready to go" goodness Sep 20 10:39:27 thank you leto Sep 20 10:39:36 here is my bitbake recipe: https://pastebin.com/jJa20T1Q Sep 20 10:40:25 and here is the kickstart file: https://pastebin.com/9L5YQv0c Sep 20 10:40:31 fberg: probably not for such specific things. there is tricks for unpacking debs, handling uncommonly named archives, but i've never heard of a nested partial unpacking thing Sep 20 10:48:54 where do I look for a proper example on how to do a bitbake recipe of an image using IMAGE_FSTYPES = "wic"? Sep 20 10:49:14 I only found wic-image-minimal.bb Sep 20 10:49:31 but that does not provide a solution to my problem Sep 20 11:04:21 eduardas_m: what do you want to do? We use wic, and I can probably share the recipe Sep 20 11:05:00 We do that because we need a specific partition table and fs label, which wic can make Sep 20 11:05:09 sveinse: here is my bitbake recipe: https://pastebin.com/jJa20T1Q I just want my systemd-based rootfs to boot Sep 20 11:05:30 it is not booting because of bad permissions on system directories Sep 20 11:06:06 I need a recovery partition with initrd for recovery image Sep 20 11:06:14 so that is why I am using wic Sep 20 11:07:06 sveinse: here is the kickstart file: https://pastebin.com/9L5YQv0c Sep 20 11:08:56 sveinse: how do you ensure correct system folder group and user ownership? Sep 20 11:09:13 mine become wrong: https://pastebin.com/QxW9CPFs Sep 20 11:09:57 which is strange because I did not have problems using wic separately (not in bitbake recipe) Sep 20 11:12:37 I'm not familiar with the IMAGE_BOOT_FILES option, but at at first glance this seems ok Sep 20 11:13:24 I use IMAGE_BOOT_FILES to put initrd and other recovery image related stuff to recovery partition Sep 20 11:13:54 it is unrelated to the problem, the problem is how the user image rootfs is treated Sep 20 11:15:28 sveinse: why I get "drwxr-xr-x 2 1000 pulse 4096 Sep 19 2017 bin" is what baffles me Sep 20 11:16:39 it should be "root root" instead of "1000 pulse" Sep 20 11:18:12 and pulse is strangely enough a group used by pulseaudio as far as I understand Sep 20 11:19:42 something isn't running under pseudo and its picking up the host UIDs Sep 20 11:20:23 why can't you just make the recovery image exactly like a normal image? Sep 20 11:20:39 rburton: it really looks that way, but I can not even start imagining what am I supposed to do Sep 20 11:22:02 rburton: I need a separate partition for recovery initrd, dtb and kernel and another partition for user image... as far as I know only wic allows me to easily do that Sep 20 11:22:34 sure but can't you construct that partition using "normal" image stuff, IMAGE_INSTALL etc Sep 20 11:24:55 how can one put image into non-rootfs partition using IMAGE_INSTALL? Sep 20 11:25:20 I mean initrd, not image Sep 20 11:25:23 sorry Sep 20 11:26:23 if I am not using wic, how else can I define all of the partitions I need? Sep 20 11:26:53 eduardas_m: are you talking about partitions or about initrd images? Sep 20 11:27:19 I need 4 partitions total: recovery partition, 1st user partition, 2nd user partition (fallback), user data partition Sep 20 11:27:51 initrd is in recovery partition Sep 20 11:28:11 it contains a minimal recovery image that uses sysvinit Sep 20 11:28:15 this is not my area but i thought the point was you can build multiple images for each partition and then wic glues them together into a disk image Sep 20 11:28:17 For initrd, IMAGE_FSTYPES="cpio.gz.u-boot" (for our type) and then PACKAGE_INSTALL with all packages you'd want into the image Sep 20 11:28:30 ah, right, separate partition for that Sep 20 11:28:43 rburton: that is exactly what I am doing Sep 20 11:29:10 hello Sep 20 11:29:46 is there a difference between cortexa9hf-neon-mx6qdl/ and 'cortexa9hf-vfp-neon'. ? Sep 20 11:30:03 pagios: depends on the context ;-) Sep 20 11:30:31 Yeah, I wondered about somewhat the same: Sep 20 11:30:56 in a general one, as "neon-mx6qdl" != "vfp-neon", there is at least a lexical difference Sep 20 11:31:19 We've changed yocto from krogoth to pyro and I've noticed that my qt app, which is built to tune, has changed directory from 'cortexa9hf-neon-poky-linux-gnueabi' where most other packages are to 'cortexa9hf-neon-mx6qdl-poky-linux-gnueabi'. Sep 20 11:31:35 well i am trying to install a mass ipk of softwares on my imx6 machine, i did create a webserver on my local machine did point my opkg.conf to my local machine and did opkg update and then opkg install python-pip for instance , i get nothing to install Sep 20 11:31:43 sveinse: iirc qt ends up being machine arch via mesa Sep 20 11:31:51 We don't touch PACKAGE_ARCH, so something has changed behind the scenes Sep 20 11:32:16 rburton: do you know who is the wic maintainer (expert)? how can I contact them? Sep 20 11:32:44 rburton: not quite, because neither of these two are MACHINE. That's a third type Sep 20 11:34:03 And if my recipe does not contain "include qmake5" then the target build directory changes Sep 20 11:35:12 eduardas_m: ed is, apparently not here right now Sep 20 11:36:07 anyone? Sep 20 11:36:48 * opkg_prepare_url_for_install: Couldn't find anything to satisfy 'python-pip'. Sep 20 11:37:24 rburton: thank you Sep 20 11:38:34 LetoThe2nd, i am trying to install the ipk on my yocto Sep 20 11:39:00 pagios: could be several things. if you set PACKAGE_FEED_URLS in local.conf to the base URL of the feeds then it will write the correct opkg.conf for you in the image. if you didn't run bitbake package-index then its possible the indexes are stale. Sep 20 11:39:14 (ie if you built pip after you build the image, the index will be stale) Sep 20 11:44:07 rburton, so you mean instead of writing directly to opkg.conf SRC/GZ etc , i write PACKAGE_FEED_URLS http://localmachine/ipks/ in local.conf and do a bitbake update? Sep 20 11:49:39 hi, i'm writing a bb script that deploys my qt application. The 2 executables are deployed and installed. But the lib is not installed on the target(and is not in the content of the ipkg package). on the other hand in the sub dir image/usr/bin/lib contains my .so file Sep 20 11:50:19 *image/usr/lib Sep 20 11:56:15 rburton, after modifying local.conf do i need to issue any command for the changes to take effect? Sep 20 11:56:19 i did bitbake update Sep 20 11:58:32 can I do IMAGE_INSTALL_append for different image names? i.e. IMAGE_INSTALL_foo-image_append and IMAGE_INSTALL_bar-image_append? Sep 20 11:59:25 - package python-pip-dev-9.0.1-r0.cortexa9hf-neon does not have a compatible architecture Sep 20 11:59:31 what could be the issue Sep 20 12:00:37 phako[m]: why don't you just set in IMAGE_INSTALL in the respective image recipes? i mean, thats what they're for Sep 20 12:01:01 because $REASON Sep 20 12:01:06 historically grown Sep 20 12:05:09 pagios: PACKAGE_FEED_URLS controls a file in your image, so rebuild the image. if you can't do that then just check the urls are the right ones. Sep 20 12:05:16 hey all Sep 20 12:05:24 phako[m]: yes, pn-[recipe name] is the override Sep 20 12:05:50 what is the proper way to have an image depend on another image... I tried EXTRA_IMAGEDEPENDS, but it doesn't seem to work... Sep 20 12:08:08 ah ha, EXTRA_IMAGEDEPENDS adds a dependency to do_populate_sysroot, an image probably doesn't have that target. Sep 20 12:08:20 i'll have to add the dependancy myself, I guess Sep 20 12:22:23 Anyone around that might know why WIC doesn't generate a startup.nsh file? Sep 20 12:22:30 for EFI things Sep 20 12:29:19 kergoth: ping when you're around, something seems to have gone wrong since http://lists.openembedded.org/pipermail/openembedded-core/2016-May/122235.html Sep 20 12:29:38 bitbake core-image-minimal fails due to 9 failed download tasks. For some reason it can't find dependencies in any mirrors. I just ran bitbake with -k option to continue as much as possible and see what happens, although it feels strange that it doesn't continue by itself. Sep 20 12:32:36 mflow: what files? everything in oe-core should be on the yocto source mirror Sep 20 12:35:42 rburton: https://pastebin.com/FYguxUm1 Sep 20 12:37:01 mflow: is that running in a container or something else that isn't a normal linux system? Sep 20 12:37:09 note how its moaning that files it created don't exist Sep 20 12:37:16 ERROR: m4-native-1.4.18-r0 do_fetch: Fetcher failure for URL: 'http://ftp.gnu.org/gnu/m4/m4-1.4.18.tar.gz'. The fetch command returned success for url http://ftp.gnu.org/gnu/m4/m4-1.4.18.tar.gz but ~/Software/Yocto/downloads/poky/m4-1.4.18.tar.gz doesn't exist?! Sep 20 12:37:21 oh Sep 20 12:37:26 dont use ~ in local.conf Sep 20 12:37:32 it won't get expanded Sep 20 12:37:54 neither env variables? like $HOME? Sep 20 12:39:15 environment is pruned before build. not sure if home gets removed. you could try ${HOME} or just use absolute paths Sep 20 12:43:38 Hi all, I am looking at the boost package and see it generates one package per library. There is a FILES_boost-X-Y for each library, but they get packaged into libboost-X-Y--.rpm so it looks like the lib part is being added automatically. Is there a way to change the name chosen? Changing the FILES_boost name does not seem to be reflected in the final rpm file names. Sep 20 12:44:50 i.e. the final rpm name seems to match the name of the actual library, but I would like to change this to include some additional info Sep 20 12:45:04 the name is indeed based on the library version. Sep 20 12:45:15 rburton: ${HOME} works, but not $HOME Sep 20 12:45:34 mflow: yes, bitbake won't expand $HOME at all. bitbake only expands ${FOO} Sep 20 12:45:44 (thats why i said try ${HOME}) Sep 20 12:48:20 rburton: is it possible to change it to libboost-NNN-X-Y where NNN is some additional string? I tried setting FILES_boost-NNN-X-Y but it seems to be ignored and based only on the final library file name, which I don't want to change (reason for this is I want to have two diff configs of boost available to chose from) Sep 20 12:49:07 for a start turn off debian-renaming, then you can use PKG to set the name Sep 20 12:49:12 boost is a tricky one as it splits so much Sep 20 12:50:23 thanks, where/how do I turn off the debian renaming? Any other packages do this as an example? Sep 20 12:58:47 ah, looks like setting PKG_ is enough Sep 20 13:59:35 i am stuck with this error: https://pastebin.com/3MY21R5e Sep 20 14:00:14 pagios: your opkg isn't configured to accept packages with the cortexa9hf-neon architecture. Sep 20 14:00:39 that's either because it isn't the correct architecture, or your opkg configuration isn't correct Sep 20 14:01:15 neverpanic, how can i make opkg accept the neon architecture? Sep 20 14:02:33 edit the configuration in /etc/opkg Sep 20 14:04:30 neverpanic, i edited /etc/opkg/opkg.conf and added the line: arch all 100 rerun opkg install and got same error Sep 20 14:05:05 hi ... I'm kind of doing my own sdk by create an image recipe and inherit the populate_sdk class ... but I'm not getting the license.manifest file anymore ... any clue on how to get it back? Sep 20 14:05:55 pagios: what makes you think adding the 'all' arch will make opkg accept the 'cortexa9hf-neon' arch? Sep 20 14:06:32 neverpanic, i added now arch cortexa9hf-neon 100 Sep 20 14:23:03 i cant find Packages.gz in any directory how can i create it using bitbake? Sep 20 14:23:44 bitbake package-index ? Sep 20 14:30:39 Thanks neverpanic it works now Sep 20 14:37:29 pagios: yep, bitbake package-index, or you can use opkg tooling to do it Sep 20 15:09:28 Tartarus: not aware of any issues, last time i was working with efi hardware my minnowmax was booting wic images fine. *shrug* Sep 20 15:09:42 kergoth: when was that? Sep 20 15:09:43 haven't messed with that board in a while, maybe something recent changed something? Sep 20 15:09:59 kergoth: The problem I see atm is that it expects to copy in a startup.nsh Sep 20 15:10:06 But that's not going to have been created if you only have wic Sep 20 15:10:23 (Even then it's going to not be where the code is looking for it, at least today) Sep 20 15:10:39 I've got a patch atm that generates the startup.nsh, if one isn't found to copy in Sep 20 15:11:14 sounds reasonable Sep 20 15:12:18 OK, thanks. I'll post and CC you shortly Sep 20 15:12:44 (well, relatively, now I need to spit out an image vs master and test that, jsut to be safe, not pyro where AGL is atm) Sep 20 15:13:03 it adapts to the bootloader in question, i'm assuming? or did you add a new env var for the efi file path to boot? Sep 20 15:13:54 I'm re-using the existing efi or systemd or throw an expcetion logic to set a variable I use later on Sep 20 15:14:37 Current patch https://www.irccloud.com/pastebin/DfufwgWO/startup.nsh.patch Sep 20 15:14:49 Hey all, I am having some trouble generating source code patches. It seems when I edit code in build/tmp/[work, work-shared] (notably u-boot-fslc or kernel-source) then commit and generate patches (git format-patch) quilt rejects those patches as non-applicable. I got around this with u-boot by cloning, modifying and generating patches outside yocto. Sep 20 15:16:30 Any questions or comments appreciated. Sep 20 15:17:27 https://pastebin.com/Z0vhkM55 Sep 20 15:18:43 kergoth release Sep 20 15:18:50 kroogth Sep 20 15:18:52 damnit Sep 20 15:19:00 krogoth release Sep 20 15:20:54 majuk: you may find it useful to look at 'devtool' Sep 20 15:21:09 lukma: Thanks. Sep 20 15:21:17 as well as bitbake -c do_populate_sdk_ext core-image-XXX Sep 20 15:21:37 this will give you the "sandbox" for developing your application Sep 20 15:38:46 Yeaaa, I wish I understood programming well enough to sort through this. Sep 20 15:54:25 majuk: https://stackoverflow.com/questions/38841096/can-a-yocto-recipe-be-built-using-the-sdk Sep 20 15:55:14 lukma: Thanks again. Sep 20 16:29:09 Can't build core-image-minimal image due to gcc error while compiling cross-localedef-native recipe from build-qemux86/tmp/work/x86_64-linux/cross-localedef-native/2.25-r0/build/glibc/ which includes struct __locale_struct two times: https://pastebin.com/axpxZAZi Any ideas what can I do to fix this issue? Sep 20 16:33:19 The system locale.h header is included because inside build-qemux86/tmp/work/x86_64-linux/cross-localedef-native/2.25-r0/git/localedef/include/libintl.h #ifdef HAVE_LIBINTL_H #include_next #endif Maybe I can try to undef HAVE_LIBINTL_H Sep 20 16:47:44 mflow: is it master ? Sep 20 16:50:02 mflow: and what OS are you running on build host Sep 20 16:53:16 hmmm it seems that you are building glibc 2.25 but your build host has newer glibc e.g. glibc 2.26 so in this case you need to backport a fix to localedef Sep 20 16:53:59 update SRCREV_localedef to dfb4afe551c6c6e94f9cc85417bd1f582168c843 in meta/recipes-core/glibc/cross-localedef-native_2.25.bb Sep 20 16:54:02 that should fix it Sep 20 16:54:25 JaMa: you were mentioning about a build issue around localedef too IIRC Sep 20 16:54:39 JaMa: can you describe the problem a bit here Sep 20 16:56:24 does yocto daisy include the corss compiler? Sep 20 16:56:34 sure Sep 20 16:57:02 khem: was that response to me? Sep 20 16:58:46 khem: pyro latest commit, ArchLinux x86_64 Sep 20 16:59:09 mflow: ok, try what I suggested Sep 20 16:59:26 zarzar: yes, cross compiler is fundamental block of any yocto release Sep 20 16:59:51 ok cool, i am trying to recreate a validated config Sep 20 16:59:54 mflow: bane of using rolling distros to build time based released software Sep 20 17:00:44 was there a daiss or dizzy that included arm-linux-gnueabihf cross toolchain version 4.8.1? Sep 20 17:01:06 daisy*** Sep 20 17:04:22 zarzar: https://www.yoctoproject.org/downloads/core/daisy16 Sep 20 17:04:38 I would read throuh release notes like that Sep 20 17:05:16 4.8.2 Sep 20 17:09:06 thanks khem i really need to figure out if there is a yocto daisy or dizzy with 4.8.1 Sep 20 17:12:57 dizzy had 4.9 Sep 20 17:13:10 where did you get your Yocto release from ? Sep 20 17:13:20 it seems your supplier has made changes Sep 20 17:13:34 http://csrc.nist.gov/groups/STM/cmvp/documents/140-1/140val-all.htm#2473 Sep 20 17:13:57 trying to match a fips security validated configuration Sep 20 17:14:50 yocto kernel version 3.10 (daisy or dizzy) and gcc 4.8.1 Sep 20 17:15:03 just trying to get as close as possible Sep 20 17:15:22 seems like what was used there was not of the shelf maybe Sep 20 17:15:30 off*** the shelf Sep 20 17:15:40 yes Sep 20 17:16:39 that's a bummer Sep 20 17:16:44 more complicated Sep 20 17:16:52 or may be you are on dora release Sep 20 17:16:54 see https://www.yoctoproject.org/downloads/core/dora15 Sep 20 17:17:04 maybe, does that use kernel 3.10? Sep 20 17:17:06 what version of libc does it show Sep 20 17:17:13 khem: yep, that was it :-) , now I get ../git/localedef/include/bits/getopt_core.h:1:10: fatal error: glibc/posix/bits/getopt_core.h: No such file or directory https://pastebin.com/Q1ZPjBJY Sep 20 17:18:00 mflow: yeah unfortunately it depends on glibc internal gut too Sep 20 17:18:15 so you have to backport some glibc patches as well. Sep 20 17:18:59 khem: that is it, it must have been dora Sep 20 17:19:05 khem: thanks! Sep 20 17:22:23 np Sep 20 17:22:43 mflow ok i think a whole upgrade might be too much Sep 20 17:23:11 mflow: revert back to where you were and then apply this patch alone https://github.com/kraj/localedef/commit/dfb4afe551c6c6e94f9cc85417bd1f582168c843.patch Sep 20 17:23:23 to localedef recipe Sep 20 17:34:10 how do i abort a build? Sep 20 17:34:26 khem: applying only that patch fails https://pastebin.com/c5HQA0Qp Sep 20 17:34:39 I'm not sure which is the next patch, as I've grepped all the patches and none add that header Sep 20 17:35:19 or 0024-eglibc-Forward-port-cross-locale-generation-support.patch Sep 20 17:35:36 30: headers = locale.h bits/locale.h langinfo.h xlocale.h Sep 20 17:36:10 patches are applied in the order specified? Sep 20 18:24:37 khem: What I don't understand is: why do I need to have on the host the same version of glibc as the one used to build the image? Why doesn't it use the glibc baked in the recipe? Sep 20 18:50:59 khem: yes it looks like the same issue I was seeing Sep 20 18:52:04 mflow: as work-around you can use LOCALE_GENERATION_WITH_CROSS-LOCALEDEF_forcevariable = "0" Sep 20 18:52:33 to prevent using cross-localedef-native completely Sep 20 18:54:12 14:29 < JaMa> khem: have you ever tried cross-localedef-native/2.25-r0 on host system with glibc-2.26? Sep 20 18:54:15 14:31 < JaMa> khem: it's choking on it's own locale_t https://pastebin.com/udCzFA9L but more difficult to resolve here when it mixes host's headers from glibc-2.26 with glibc-2.25 code Sep 20 18:54:19 14:34 < JaMa> I'll probably use LOCALE_GENERATION_WITH_CROSS-LOCALEDEF to bypass this, but it there is a plan to support Pyro on more recent hosts than this should be resolved as well Sep 20 18:54:24 14:34 < JaMa> toghether with backport of icu and qemu-native fix from master Sep 20 18:55:28 for whatever reason RP was seeing other glibc-2.26 related issues in pyro build on autobuilder but not this cross-localedef-native Sep 20 18:55:50 my only explanation is that autobuilder doesn't enable binary locale generation for whatever reason Sep 20 19:04:46 JaMa: You mean to set LOCALE_GENERATION_WITH_CROSS-LOCALEDEF = "0" in poky/meta/recipes-core/glibc/glibc-locale.inc ? Sep 20 19:09:13 JaMa: Doing this, the baking of the rpi-basic-image has moved forward from cross-localedef-native to glibc-locale-2.25-r0 , it only gives some warnings: https://pastebin.com/B66N9fEp Sep 20 19:30:04 i am getting errors running dora core-image-sato on ubuntu 16lts Sep 20 19:35:00 mflow: I meant to use it in local.conf, but in glibc-locale it will work as well Sep 20 19:35:09 mflow: yes I've seen the same warnings Sep 20 19:35:57 Does anyone know if Yocto supports hardware encoding out of the box? Sep 20 19:37:29 is there a way to test if it does? Sep 20 19:38:10 tcpdump: hardware encoding for what? Sep 20 19:38:11 JaMa: ok you and mflow both are facing same problem Sep 20 19:38:26 tcpdump: you would have to include whatever drivers and libraries are required in your image Sep 20 19:38:42 rburton: its hardware video encoding. Sep 20 19:38:50 JaMa: I have suggested a backport to mflow Sep 20 19:38:50 and what exactly is required depends on your exact hardware Sep 20 19:38:55 JEEB: its Qualcomm/Venus Sep 20 19:38:57 I gotcha Sep 20 19:39:01 tcpdump: the it depends on the BSP. meta-intel includes libva and gstreamer-va. Sep 20 19:39:39 tcpdump: ARM stuff tends to be rather bad unless they implement an API that is not something vendor specific Sep 20 19:40:06 there's vaapi, vdpau and the thing that raspberry pi's sysroot uses Sep 20 19:40:25 (on the background it's omx but it had some other name IIRC) Sep 20 19:42:05 right, MMAL I think Sep 20 19:42:07 JEEB: Im using WebRTC. I was using VP8, which is software encoded, but that was killing the ARM processor cores. Im now using Qualcomm/Venus hardware encoding. So Im dumping the raw frames in NV12 into a shared buffer and then WebRTC is supposed to open the Venus encoder and start crunching them into H.264. Sep 20 19:42:27 The problem I have is that WebRTC isn't able to see the Venus encoder. Sep 20 19:42:38 So Im trying to see if its a lower level issue. Sep 20 19:42:51 you'd have to know what interface the drivers provide Sep 20 19:42:56 and hope it's not specific to that vendor Sep 20 19:43:26 Im trying to see if theres some way to detemine if anything can see venus, or if its just WEbRTC Sep 20 19:43:35 Yea? When you say drivers, you mean the Venus? Sep 20 19:43:39 yes Sep 20 19:44:33 oh, I wonder if it's one of those v4l2 based things Sep 20 19:44:36 looks like venus does v4l Sep 20 19:44:40 that was recently posted on the ffmpeg-devel mailing list Sep 20 19:44:58 yea, so you would have to verify that the v4l2 device is found and usable Sep 20 19:45:09 ok, any idea on how I can do that? Sep 20 19:45:37 the first part is by checking for v4l2 devices in lshw or something Sep 20 19:45:38 Sorry, im in the deep end here? Sep 20 19:45:48 OK Sep 20 19:45:50 the second part you can try applying the v4l2 dec-enc patch to ffmpeg Sep 20 19:45:51 let me check that. Sep 20 19:46:19 I'd link you the patch but it seems like patchwork still hasn't been updated from a startssl certificate :/ Sep 20 19:46:27 so anything chromium based will just nope at it Sep 20 19:46:57 relevant thread @ https://ffmpeg.org/pipermail/ffmpeg-devel/2017-September/215649.html Sep 20 19:47:31 right, the next mail in thread had the patch https://ffmpeg.org/pipermail/ffmpeg-devel/2017-September/215650.html Sep 20 19:47:56 seems to have been validated on venus Sep 20 19:48:15 thanks! Sep 20 19:48:18 * tcpdump reads Sep 20 19:48:54 oh, geez Sep 20 19:49:03 I linked you v7 and there's already a v8 on the ML :P Sep 20 19:49:11 the funzies of mailing lists Sep 20 19:49:38 scratch that, latest set was v12 Sep 20 19:49:38 https://ffmpeg.org/pipermail/ffmpeg-devel/2017-September/216200.html Sep 20 19:49:39 lal Sep 20 19:49:42 sorry Sep 20 19:51:57 Haha wow Sep 20 19:53:37 Im using, for better or worse, the dragonboard 410c Sep 20 19:53:45 so that fits right in wth what Im working with. Sep 20 19:54:08 yea, it was noted as tested Sep 20 19:54:23 so if the v4l2 device is visible you could try building master FFmpeg with that patch Sep 20 19:55:47 hmmm looks like lshw isnt implimented. Sep 20 19:56:05 I may have to install it. Sep 20 19:56:17 yea, I'm too used to desktop sysroots :P Sep 20 19:56:30 yea - this is a new adventure. Sep 20 20:06:53 JEEB: is there any other way to determine if v4l2 is showing correctly, or do I need to go bake in lshw real quick? Sep 20 20:06:53 how do i diagnose errors with building dora core-image-sato on ubuntu 16lts Sep 20 20:07:26 tcpdump: check if under /dev you have the device Sep 20 20:07:38 and/or if the driver mentioned @ https://lwn.net/Articles/697956/ pops up Sep 20 20:31:46 using dora i am getting the build error pseudo_util.c:101:29: error: initializer element is not constant how do I resolve? Sep 20 20:41:59 * tcpdump sighs Sep 20 20:42:07 This is going to be ugly... Sep 20 20:43:59 ERROR: Layer 'meta-python' depends on layer 'openembedded-layer', but this layer is not enabled in your configuration ; Where can I find the openembedded-layer layer? It doesn't seem to be in openembedded-core or meta-oe layers Sep 20 20:45:02 nor meta-openembedded Sep 20 20:47:23 nevermind, seems to be in meta-openembedded/meta-oe Sep 20 20:48:34 mflow: correct, sorry for the confusing naming Sep 20 20:49:20 FYI we will in the near future have tools that will set up layers and take care of all dependencies automatically **** BEGIN LOGGING AT Wed Sep 20 20:56:15 2017 Sep 20 21:13:07 bluelightning: awesome! :-) Sep 20 22:27:54 I'm trying to patch /etc/init.d/rcS to source some extra files, but rcS isn't part of the sysvinit source, so the normal patch process doesn't seem to work. I'm now trying to patch it in do_install_append but it seems that environment variables come from the original bitbake file, not the bbappend, so I'm not certain how to apply a patch which comes from the same directory as bbappend... Sep 20 22:28:30 Any thoughts? Am I approaching this completely wrong? I'm admittedly extremely new to bitbake, OE, and yocto Sep 20 22:49:06 That appears to work, at least, a patched copy of rcS is floating around, though there seem to be multiple copies and I'm not sure which one will make it into the image ultimately... Sep 20 23:41:23 bluelightning, I am waiting for some fuzzy logic so bitbake know what I meant to do Sep 21 00:13:06 Hi, Im looking into ptest and I was wondering how this chart was populated? https://wiki.yoctoproject.org/wiki/Ptest_6964efddd31c479386d1643c1025bc102710392f Sep 21 00:13:30 if there was already framework available? Sep 21 00:14:20 How can one not exclude a recipe from world? Sep 21 00:14:43 I'm trying to compile gdb-cross but I see that it is excluded from world Sep 21 00:18:15 Not sure if it is related, but I get this in the end: ERROR: Nothing PROVIDES 'gdb-cross' Sep 21 00:25:28 ahh, gdb-cross-aarch64, got it Sep 21 00:49:57 Guest25170: https://wiki.yoctoproject.org/wiki/Ptest Sep 21 00:50:21 User12345678: look into EXCLUDE_FROM_WORLD = "1" Sep 21 00:50:31 add it to given recipe Sep 21 00:50:48 or EXCLUDE_FROM_WORLD_pn- = "1" in local.conf Sep 21 01:04:06 Guest25170, I think the ptest on the wiki is like a design spec for the framework that was added. its not what is currently implemented Sep 21 01:04:56 we have more support these days Sep 21 01:18:56 armpit: is there some place which reflects latest state of ptest Sep 21 01:29:36 khem, I don't think so. I can ask what the intent of the ptest wiki is Sep 21 01:29:54 it will be good if it reflected latest Sep 21 01:30:01 there is ptest support in other layers Sep 21 01:32:17 it could be a consolidated list **** ENDING LOGGING AT Thu Sep 21 03:00:02 2017