**** BEGIN LOGGING AT Thu Oct 11 02:59:59 2018 Oct 11 03:21:47 cramalho: you should just track release branches and match them with oe-core release branches Oct 11 03:21:50 makes it easy Oct 11 03:28:52 khem: ok. Thanks! Oct 11 08:24:16 How get ipk's copied from the work dir to the deploy dir? Oct 11 08:25:11 And in case sstate is in use, I guess sstate package_write_ipk gets written somewhere before it gets deployed to deploy dir? Oct 11 08:49:23 is there any way to install the extensible SDK with internal/non-public git repos used in recipes? Oct 11 09:08:22 Hi! I am building a yocto image for Raspberry Pi. I want to include support for CAN bus. I think I need to add these parameters to the kernel configuration: CONFIG_CAN=y CONFIG_CAN_GS_USB=y. But I am not sure how to do it. Can anyone help me with this, please? Oct 11 09:09:46 New news from stackoverflow: If else statement using external variable in a bitbake file Oct 11 09:39:51 New news from stackoverflow: best approach of image versioning in yocto || gcc7 with older kernel 2.6.35 Oct 11 09:40:18 jostor: https://www.yoctoproject.org/docs/latest/kernel-dev/kernel-dev.html Oct 11 09:41:27 sven^: eSDK recipes are the same as normal Yocto ones, they can use local repos Oct 11 10:07:36 Hi, I just noticed that the report-error tool is not handling Errors generated by the cookerdata.py, e.g: around parseConfigurationFiles: Oct 11 10:07:36 ERROR: Layer gstreamer1.0 is not compatible with the core layer which only supports these series: thud (layer is compatible with sumo) Oct 11 10:07:36 Is it possible to fire events for such errors and catch them in the report-error tool? Oct 11 10:41:36 mckoan: Thanks! I have created a layer, and added /recipes-kernel/linux/linux-raspberrypi%.bbappend with FILESEXTRAPATHS_prepend := "${THISDIR}/files:" SRC_URI += "file://defconfig" in defconfig I have added CONFIG_CAN=y CONFIG_CAN_GS_USB=y. Does this look correct? Oct 11 10:42:11 I get this warning: WARNING: linux-raspberrypi-1_4.14.68+gitAUTOINC+8c8666ff6c-r0 do_kernel_metadata: defconfig detected in WORKDIR. bcm2709_defconfig skipped. Is this OK, or is it replacing the original config with my config instead of appending it? Oct 11 11:36:20 Hi guys. What is a right way to change a content of, e.g. the /etc/X11/xinit/xinitrc file ? Oct 11 11:36:48 e.g. I need to add there the start of my app instead of xclock and so on Oct 11 12:54:37 mckoan|away: yes, but I cannot install the SDK because the eSDK installer cannot access the private repos without authentication and ignores ssh-agent or keys I try to provide Oct 11 13:08:38 I am having trouble building my kernel via devtool. If I just do devtool modify my-kernel && devtool build my-kernel I get an error that silentconfig failed and I should run make oldconfig. When I go to workspace/sources/my-kernel and copy my defconfig there (as defconfig or .config) I get the same error. (same for workspace/sources/my-kernel/oe-workdir/my-kernel+git999) Oct 11 13:11:00 I also tried building it manually with "make O=oe-workdir/my-kernel+git999", but get the same error Oct 11 13:11:07 so what's the correct way to do this? Oct 11 14:07:58 Hiya! Shouldn't "$ bitbake-layers show-layers" give similar output to what's in bblayers.conf? Oct 11 14:36:07 Has anyone tried building two linux kernels as part of an image. One to boot and another to kexec into for crashkernel? Oct 11 14:40:52 New news from stackoverflow: populate_sdk fails when multilib enabled Oct 11 14:52:40 if i have a layer that only contains bbclasses i get a warning about BBFILE_PATTERN_layer not matching any files, should BBFILE_COLLECTIONS and BBFILE_PRIORITY_layer also not be set if the layer only contains classes? Oct 11 14:53:19 matman1: latest release can do multiple kernels Oct 11 14:54:26 rburton, Thanks! i'll take a look at how its done Oct 11 14:56:29 Do you just add the second kernel as a package? Oct 11 14:57:49 Hi, quick question: using opkg, upgrading busybox makes many of the upgrades after it fail. This is because, int the prerm, busybox removes all it's links, that are required in the control scripts of other packages. After all packages were upgraded, the links are created again. Is there a way to make all packages install sequentially? Or is there some other workaround? I could still do it in a for-loop, but I Oct 11 14:57:52 rather have some built in config in opkg or busybox, so people are less prone to break the system because busybox happens to be one of the packages to be upgraded. Oct 11 15:24:59 rburton, I think I found it. This patch seems to be how it's implemented. https://patchwork.openembedded.org/patch/141561/ Oct 11 15:25:17 yeah Oct 11 15:25:23 should be in the docs tbh Oct 11 15:55:54 Filing a bug to document that would be good Oct 11 16:08:36 All: I am porting from Morty -> Pyro. In my do_rootfs for (core-image-myproj-1.0-r0), it succeeds in my RHEL7 Host OS. When compiling in docker on a server, it fails on -> Could not invoke dnf. Command '/home/jenkins/workspace/project/bb-layers/poky-pyro-build/build/tmp/work/myproj-poky-linux/core-image-myproj/1.0-r0/recipe-sysroot-native/usr/bin/dnf -y -c ..... Oct 11 16:08:58 The dnf binary is in the directory. Any ideas? I'm stumped. Oct 11 16:47:44 What happens when you try to run the command manually? Oct 11 16:48:24 matman1: In my native RHEL7? Oct 11 16:51:03 For the docker builder - I have just modified our jenkins docker builder to not delete workspace. After it fails, I can ssh and check if dnf really there and runs without errors. Oct 11 16:51:56 I found this (https://lists.yoctoproject.org/pipermail/yocto/2018-March/040374.html) - Trying another build with "inherit setuptools" Oct 11 16:53:44 If this is the issue, then why would it fail in docker but not native RHEL7 OS. Must be environment (if so). Oct 11 17:00:29 It could be a race condition? Perhaps more cores in docker env? Oct 11 17:07:27 Hmmm. Checked docker virtualization. Has 4 cores. Could run with BB_NUMBER_THREADS=1 to verify. I suppose. Oct 11 19:40:09 msg dreyna_ ping Oct 11 19:40:15 oops Oct 11 19:43:26 halstead: out of curiosity, when you setup build machines, do you use ext4? Oct 11 19:46:18 tlwoerner, Yes. The build directory is on an ext4 filesystem. When CentOS changed to xfs by default we used that for a year or so but experienced reoccurring errors that were hard to reproduce. Oct 11 19:46:37 do you enable journaling? Oct 11 19:47:32 halstead: I'm still debugging. It's type overlay. Oct 11 19:48:52 tlwoerner, I do on the AB workers but raise the commit time. Oct 11 19:49:06 halstead: Could overlay filesystem be of concern for our docker build machines? Oct 11 19:49:18 tlwoerner, Makes recovery easier if it crashes. Oct 11 19:49:53 halstead: do you have notes somewhere? because my next question is: to what do you set the commit time? 6000? Oct 11 19:50:34 mrk377, We haven't tested overlay much. I've setup training using docker and overlay2 and it worked for the day. Oct 11 19:51:31 i've generally disabled journaling on my oe build disk, but have had yucky post-power-glitch surprises, so i've grown a tendency to have config/dl/layers outside the build disk, and symlinked in Oct 11 19:51:40 mrk377, I wouldn't expect overlay to work well long term unless some care was done during setup to mitigate problems. Oct 11 19:52:18 halstead: Thanks for info. Oct 11 19:52:38 mrk377, But I am curious to know if it works well for you and how you are using it. Oct 11 19:52:58 halstead: you say "the build directory is on an ext4 filesystem", what other areas/directories do you define? Oct 11 19:53:47 tlwoerner, no recent notes. https://wiki.yoctoproject.org/wiki/Build_Performance was written when we used 6000. I think it is set to 300 now. Oct 11 19:54:50 tlwoerner, We use whatever the distro defaults to for everything except the build filesystem. ext4 on debian, xfs on CentOS, btrfs on opensuse, etc. Oct 11 19:56:11 halstead: excellent, thanks. i'll leave journaling off for now then and continue on with my current system of symlinks Oct 11 20:00:48 halstead: Top level: We have a jenkins master w/ 10 swarm slaves (RHEL 7) configured/managed using Ovirt using 6 Hosts for generating custom RHEL7 OS deliveries. The 6 hosts also have local SSDs for caching, but get stored off hosts once complete. All swarm slaves have 4 cores each. That is it in a nutshell. Oct 11 20:01:31 mrk377, And how are you using overlayfs? Oct 11 20:26:03 The entire filesystem is overlaid with the overlayfs (from ram disk), except for the builder workspace. It is now completely stateless so Ovirt never creates or tries to recover state. Reboot time is a few seconds. When you kickoff a new swarm slave, it automatically registers with the jenkins master. Sorry if I mentioned that the build is being done on the overlayfs. It is the opposite. It is ext4 for the workspace. **** ENDING LOGGING AT Fri Oct 12 03:00:00 2018