**** BEGIN LOGGING AT Mon Jan 29 03:00:02 2018 **** BEGIN LOGGING AT Mon Jan 29 03:03:43 2018 Jan 29 05:48:38 New news from stackoverflow: Disable root login in yocto build || Yocto - Try to use Realtime Kernel Version Jan 29 08:18:26 good morning Jan 29 08:19:29 mckoan, good morning Jan 29 09:02:39 armpit: Wasn't that rocko and was due to the wrong branch of meta-mingw? Jan 29 10:44:52 Hello, I package my kernel and dtb inside kernel-image-4.9.11+gddaf07227a3e_4.9.11-r0_fod_som.ipk package, but I have a redundant duplicate of my kernel image inside kernel-image-uimage-4.9.11+gddaf07227a3e_4.9.11-r0_fod_som.ipk which is undesirable. How do I figure out what meta-data creates and includes the kernel-image-uimage package into my final rootfs? Jan 29 11:02:34 how do I properly fork a library recipe under a different name? Yocto seems to autodetect solib names and complains that there are two same libraries in different packages Jan 29 11:03:25 I didn't choose bbappend, because original was autotools-based, and the forked one uses cmake, so too much changes for the original Jan 29 11:04:44 in other recipe I've put the new recipe name in DEPENDS, but later bitbake complains about "nothing provides original_lib_name needed by my_library_client_recipe" Jan 29 11:05:57 I might need to use provider mechanism, I guess Jan 29 11:15:50 luneff: I have never tried it myself, but it sounds similar to what Yocto multilib functionality does: http://www.yoctoproject.org/docs/2.4.1/mega-manual/mega-manual.html#combining-multiple-versions-library-files-into-one-image Jan 29 11:16:31 luneff: http://www.yoctoproject.org/docs/2.4.1/mega-manual/mega-manual.html#installing-multiple-versions-of-the-same-library Jan 29 11:21:03 I think I tried the same thing. Maybe, I just need to rebuild everything from scratch Jan 29 11:21:22 so that the old library is properly removed from everywhere Jan 29 11:24:09 luneff: as far as I know opkg does not support multilib, just wanted to say that in case you did not know and are using that: https://www.yoctoproject.org/bug/bug-11729-opkg-doesnt-support-multilib Jan 29 11:25:17 multilib is a different thing, I don't need multiple architectures at once Jan 29 11:25:50 but thanks :-) Jan 29 11:29:26 luneff: Thank you for correcting me. I was somehow under the impression that the multilib feature also encompasses using different versions of the same library... not necessarily different architectures Jan 29 11:30:12 yes, that sounds similar. For me, either a full rebuild with riddance of offending library will help, or some PROVIDES magic, I guess Jan 29 12:48:00 Hi, is there a correct way to get twisted working in rocko? I had to copy the recipes from python-incremental and python-constantly from master to make it work. But it feels wrong to have to do that Jan 29 12:48:09 How can I set PV from a file in SRC_URI ? Jan 29 12:49:58 New news from stackoverflow: How to write a yocto recipe for gRPC++? Jan 29 13:14:19 can someone explain me what is the meaning of the colon in bitbake version numbers? For example, glib-2.0-native has version "1:2.50.3-r0" Jan 29 13:21:21 mardy: : is the epoch separator, it's a generally invisible version for when versions behave oddly. dpkg has a nice summary at https://manpages.debian.org/wheezy/dpkg-dev/deb-version.5.en.html Jan 29 13:28:24 Hi! Is there a way inside a recipe to programmatically determine the actual version of a dependency? The use case is a recipe for an update bundle. A manifest file contains the versions of images in the bundle. The versions shall be generated automatically. Jan 29 13:30:10 I'm still using krogoth Jan 29 14:13:04 khem: libxshmfence fails with the same memfd failures fwiw Jan 29 14:22:20 Hi, have someone ever tried to create a python3 venv with an python3 provided from the yocto SDK? Jan 29 14:24:09 basically I can create it but it is not providing pip Jan 29 14:29:54 hi, does anyone have experience with cross-building node.js packages with yocto? I'm following this guide: https://wiki.yoctoproject.org/wiki/TipsAndTricks/NPM with this layer: https://github.com/imyller/meta-nodejs, but the build stalls for hours Jan 29 14:33:21 h0ru5_: I don't use meta-nodejs, but the nodejs in meta-oe layer builds fine Jan 29 14:34:22 oh, there's nodejs in meta-oe? I've been wondering since the meta-nodejs repo seems pretty abandoned Jan 29 14:34:47 I had to use a 3rd part fork to get support for the latest node LTS releases against Rocko Jan 29 14:34:51 JaMa: thanks. the problem is then, that meta-oe's nodejs package is a very old version, 4.6.1 Jan 29 14:35:06 ah, too bad Jan 29 14:35:19 sveinse: which one did you use? Jan 29 14:35:48 otherwise I would try to edit the recipe in meta-oe Jan 29 14:36:41 h0ru5_: https://github.com/aaronovz1/meta-nodejs.git on the pyro branch, built on rocko. That works fine for 8.9.4 Jan 29 14:37:09 h0ru5_: But I havent come around to install any npm based packages yet, so I cannot answer your question thou Jan 29 14:37:19 sveinse: did you try building packages in yocto (those that have some native parts)? Jan 29 14:37:33 aah, ok, thanks Jan 29 14:37:57 h0ru5_: No, not yet. I will. But the other dev's say that node.js on target works fine Jan 29 14:38:37 I can use meta-nodejs to install 7.10, works fine on target, but the packages compiled hang if done like the guide shows up Jan 29 14:39:06 its really time they get webassembly :) Jan 29 14:40:10 rburton: looks like make from meta-gplv2 and libshmfence break with new glibc :/ Jan 29 14:41:08 yeah just fixing xshmfence now Jan 29 14:41:22 and there's a patch in the branch for the new make we can backport assuming its trivial Jan 29 14:42:43 rburton: probably kill and restart territory? Jan 29 14:46:19 let it run unless you need the AB Jan 29 14:48:14 Is it possible to set PV from a file in the repo that SRC_URI points to? Jan 29 14:49:06 sveinse: too late for PV but you can set PKGV Jan 29 14:50:29 rburton: great. Does bb have a oneliner from reading the contents from a file, or do I need to write my own? Jan 29 14:52:36 sveinse: utils.bbclass has read_file Jan 29 15:17:24 rburton: thanks! And when I put a version number in PREFERRED_VERSION_*, I should include the epoch, right? Jan 29 15:32:36 hi Jan 29 15:33:04 rburton: thank you Jan 29 15:34:24 Are there any syntatical rules to the version numbers in packages? Jan 29 15:35:17 E.g. can a version number be '2018-01-29-2' ? Or will that interfere with the $PR scheme? Jan 29 15:36:12 do I need to explicitly depend on unzip-native in order for fetch2 to handle .zip files? I fetch a .zip file and `do_unpack` fails. Jan 29 15:38:08 log.do_unpack:`/bin/sh: unzip: command not found` Jan 29 15:40:49 No rule to make target '/usr/include/c++/5/set' Jan 29 15:41:04 I have a makefile inside my local file. Any ideas? Jan 29 15:41:10 tar file* Jan 29 15:42:44 * armpit seesh the pyro-next build was really rocko sources Jan 29 15:44:11 T_UNIX: elif path.endswith('.zip') or path.endswith('.jar'): Jan 29 15:44:11 d.appendVarFlag('do_unpack', 'depends', ' unzip-native:do_populate_sysroot') Jan 29 15:44:17 T_UNIX: if its .zip then no Jan 29 15:44:34 T_UNIX: maybe the file you have is .ZIP and the test is broken as its case-sensitive/ Jan 29 15:44:58 * armpit build failures now make sense Jan 29 15:45:46 rburton: thanks for the insight :) Jan 29 15:46:37 is path overwritten by downloadfilename *prior* to this evaluation? Jan 29 15:47:05 hm Jan 29 15:47:19 no Jan 29 15:48:14 :-D Jan 29 15:49:38 so shall I append some kind of `&dummy=foo.zip` to the SRC_URI to work around this issue/match the code? Jan 29 15:49:58 T_UNIX: so you have a case where http://foo.com/bar gives you a zip? Jan 29 15:50:06 dummy args won't help as the url is parsed Jan 29 15:50:37 you could argue that the "what do i need to extract this" logic should look at downloadfilename if set Jan 29 15:50:50 and thats just a few lines to base.bbclass Jan 29 15:51:10 yeah. It's a gitlab-runner artifact that is named after the job (i.e. `bar`) Jan 29 15:53:12 yeah, downloadfilename should influence localpath, which is what unpack() should be using, if not, that's a bug, afaik Jan 29 15:53:13 * kergoth yawns Jan 29 15:53:26 oh, i see, it's not unpack(), it's the extraction dep logic.. yeah, that should be fixed Jan 29 15:53:32 that code is pretty naïve in a number of ways, actually.. Jan 29 15:53:36 hmm Jan 29 15:54:24 it might need to construct a FetchData and look at the url scheme and localpath Jan 29 15:54:33 easily fixed, patches welcome T_UNIX :) Jan 29 15:55:52 it's odd that it infers to invoke `unzip`, but does not infer the dependency Jan 29 15:56:19 contributions are handled via ML? Jan 29 15:56:22 T_UNIX: yes Jan 29 15:56:49 T_UNIX: the unpack logic is independent of the dependency tree and they're currently behaving differently, which is the bug Jan 29 16:02:08 damn... need to set up git-send-patch w/ internal exchange -.-' Jan 29 16:15:31 No rule to make target '/usr/lib/gcc/x86_64-linux-gnu/5/include/stdint.h' Jan 29 16:15:34 any ideas? Jan 29 16:32:08 eduardas_m: I've currently multiple repos but I'd prefer to use a monorepo approach Jan 29 18:11:05 if I want to clear the cache I just do rm -rf sstate-cache/* correct? Jan 29 18:22:19 noway96: bitbake -c cleansstate recipe Jan 29 18:24:11 ty Jan 29 18:34:36 noway96: generally there's no need to clean the sstate unless you're chasing a very weird bug Jan 30 00:20:57 Hi, Has anyone seen this error: Disk Requirements: At least 30MB more space needed on the //build/tmp//// filesystem ? Jan 30 00:21:31 I have more than a TB of disk space left. But, looks like bitbake is complaining about the target rootfs Jan 30 00:30:01 batman_: you sure its bitbake? Jan 30 00:35:57 It's dnf. do_rootfs is failing Jan 30 01:24:16 batman_: probably not much help, but you should try looking at the source of do_rootfs Jan 30 01:25:46 Most likely there's a line like make_ext4fs ... -l ${SYSTEM_SIZE_EXT4} ... Jan 30 01:26:01 or some other "limit" variable Jan 30 01:26:06 and you have to bump that up **** ENDING LOGGING AT Tue Jan 30 03:00:01 2018