**** BEGIN LOGGING AT Wed Apr 17 02:59:57 2019 Apr 17 05:04:44 JaMa, regarding libyaml, it does not affect warrior.. it only affects the 0.2.2 update. So if I update the warrior package then I will need to also backport the src_uri changes Apr 17 07:53:59 armpit: you're right, I was on older oe-core revision and was seeing the same 0.2.1 version being used in master, warrior, thud Apr 17 08:49:19 kanavin: I'm guessing your patch series overlaps with the AUH output a bit? Apr 17 09:00:19 RP: yes, I didn't wait for AUH :) and AUH started crunching just before the master merge (it works against master, not master-next) Apr 17 09:02:20 kanavin: right, makes sense. It does feel like we're keeping a little more up to date Apr 17 09:06:12 RP: yes, but those recipes where manual patch rebase is needed aren't getting as much attention as I would hope for Apr 17 09:09:13 kanavin: we're getting to the point where we need to do something about our patch count Apr 17 09:09:30 kanavin: reducing it is shown to help us a lot Apr 17 09:14:03 A really significant portion of my rootfs is the hwdb for (e)udev and I'm not even running systemd. How do I disable it? PACKAGECONFIG_remove_udev = "hwdb"? Apr 17 09:16:54 poky/meta/recipes-core/udev/eudev_3.2.5.bb has: PACKAGECONFIG[hwdb] = "--enable-hwdb,--disable-hwdb" .. I just haven't used this PACKAGECONFIG business before.. Apr 17 09:23:24 jofr: where are you wanting to disable it from, a bbappend or a local/distro conf? Apr 17 09:26:09 I actually got rid of it by throwing out usbutils, as I didn't really need them anyway, but I'd like to learn how. Apr 17 09:26:22 Could I do it from my image-recipe? Apr 17 09:26:55 Let's say bbappend. I'm assuming you mean bbappending to eudev*.bb Apr 17 09:38:27 jofr: you can't do it from an image recipe but yes, you could do what you suggest above from a bbappen to eudev Apr 17 09:39:15 jofr: something like PACKAGECONFIG_remove_pn-eudev = "hwdb" in your distro.conf or local.conf would also work Apr 17 09:39:37 or just set PACKAGECONFIG to what you need: PACKAGECONFIG_pn-eudev = "XXX" Apr 17 10:00:38 RP: Good point. Thanks! Apr 17 11:17:20 New news from stackoverflow: How can I add some file (maybe like /home/eric/libtest0.1.so) to Yocto rootfs lib folder(/lib)? Apr 17 12:17:30 New news from stackoverflow: How to control backlight with a button connect to a gpio? Apr 17 14:37:42 Hi, any meta-java maintainer here? Apr 17 15:48:06 New news from stackoverflow: Change gdb source file path Apr 17 16:18:10 New news from stackoverflow: Yocto issue with meta-java Apr 17 17:01:10 I'm using devtool to patch a recipe, and it wants to number the patch 0001, even though I already have 0001-0003. How can I make it continue numbering from the previous patches? Apr 17 17:58:58 is it possible to force a particular libc for a single recipe? I have a static linked pgm in my initramfs so musl would make sense Apr 17 19:40:48 pebenito: you probably want multiconfig or possibly multilib could do it too Apr 17 19:45:33 Hi, I'm trying to build a multi-layered container image with Yocto. Is it possible, in an image-class, to access individual packagegroups? Apr 17 19:48:47 New news from stackoverflow: Yocto: Remove Packages From core-image-base [closed] Apr 17 19:53:00 moosnat: what do you mean by ‘access’ ? Apr 17 19:56:10 zeddii: can I write a recipe that uses each packagegroup's rootfs as an OCI image layer? Apr 17 20:01:45 hrmm. not sure you'd really want to do that. Apr 17 20:02:01 have you seen the work I did in meta-virtualization on creating oci-images ? Apr 17 20:02:39 I have not Apr 17 20:03:02 the logic to create the oci layers doesn't really belong buried in recipes and bbclasses imho, but can more easily be done on any resulting tgz's output from the build as part of an image type class. Apr 17 20:03:41 Can you link me your meta-virtualization work please? Apr 17 20:04:11 pull the meta-virt layer :D there are also some bugzilla entries for it. Apr 17 20:04:27 I'm only half kidding, the commit in the layer has a long writeup on what it can do. Apr 17 20:04:38 so that's really the best place to look and collaborate. Apr 17 20:04:43 I'd love some help with it :D Apr 17 20:04:57 right now it is single layer by design, but could easily be extended to do multiple layers. Apr 17 20:08:12 zeddii: awesome! as my work is in the MVP phase, I might start with a single layer, then look at how to modify it for multiple layers Apr 17 20:09:17 if you follow what I put in that commit log (I should also write a README), you can spin out an oci layer than can be pushed to dockerhub, or unbundled and run, directly with no mods. If there are gaps, let me know, I have cycles to fix bugs and add features. Apr 17 20:09:31 s/oci layer/oci image/ Apr 17 20:12:18 How might it be extended to multiple layers? Apr 17 20:14:22 it only takes a single image_rootfs at the moment. it could either take a list, or could be run multiple times to add layers to the oci image. Apr 17 20:17:48 so if the `sloci-image` invocation at the bottom of your image-class passed a series of rootfses, that would produce more layers? Apr 17 20:18:27 the tool itself doesn't handle it at the moment (or at least I don't recall it being able to handle it), so some extension at that level would probably be needed. Apr 17 20:19:50 I might try to replace `sloci-image` with buildah, actually Apr 17 20:33:56 heh Apr 17 20:34:07 moosnat, I've looked at them *all* Apr 17 20:34:10 literally all the tools. Apr 17 20:34:16 they are all way to bloodly complex Apr 17 20:34:37 * zeddii disliked buildah more than most Apr 17 20:34:51 they all want to *build* things, versus assemble. Apr 17 20:35:30 you may be right, lol Apr 17 20:36:14 and try making one of them into a host tool Apr 17 20:36:24 I spent multiple days on that. Apr 17 20:36:27 come to think of it I can't find any way to use buildah to manipulate the individual layers Apr 17 20:36:32 their dependencies are literally endless Apr 17 20:36:45 so you have to make a zillion go-native, foo-native, tools. Apr 17 20:36:57 and pretty much none of them really want to be built as native. Apr 17 20:37:14 i nearly destroyed a keyboard in anger at one point ;) Apr 17 20:37:34 hmm, okay Apr 17 20:38:45 at some point, I'm sure things will outgrow the low dependency option currently in place, but definitely a lot of work to drop a replacement in Apr 17 20:38:59 gotcha Apr 17 20:40:15 I'd personally like a tool that's more actively developed, but this may work for POC purposes Apr 17 20:40:34 yah, but the spec isn't changing Apr 17 20:40:51 so as long as there aren't missing features, it's no big deal. Apr 17 20:41:10 let's just say that finding something that runs in a cross environment or another build system, is not easy. Apr 17 20:41:27 Thanks for the heads-up Apr 17 20:42:00 I thought for a while I was going to be able to use unmoci, but the dependencies were just so huge. Apr 17 20:42:51 alright, well I read through sloci-image, and it looks easy enough to maintain if the spec has a breaking change Apr 17 20:42:58 I'm going to go ahead and import meta-virt Apr 17 20:43:01 thank you! Apr 17 20:43:02 I have to head out now, but the meta-virt list, bugzilla or on here are good ways to find me. I'm happy to share my pain. :D Apr 17 20:43:11 I was about to write my own from scratch Apr 17 20:43:30 then found that, and it suited the initial need. I was just waiting to see if anyone else popped up looking for more features Apr 17 20:43:33 * zeddii really flees Apr 17 20:48:12 Hi, I made a C++ library and the install rule of the Makefile is copy the real library `xxx.so.1.1.0`, create the symlink `xxx.so` and finally call `ldconfig -n`. Now when building the yocto library recipe recipe I'm getting the error `Files/directories were installed but not shipped`. I guess it's because of symlink created previously on the image rootfs and not added to the the FILE_${PN}. How can I fix this problem ? **** BEGIN LOGGING AT Wed Apr 17 22:49:39 2019 Apr 18 01:38:14 Hi, Is someone can help me with the error `QA Issue: myapp rdepends on mylib-dev [dev-deps]` ? In the myapp.bb I already have `DEPENDS="mylib-dev"` Apr 18 01:43:13 get rid of the "-dev" Apr 18 01:43:57 @tz Sorry, typo. I currently have `DEPENDS="mylib"``. Apr 18 01:44:27 that... should be enough Apr 18 01:46:05 @tz, can I show you a paste bin of the two recipes ? Maybe I made a mistake... Apr 18 01:46:26 can't hurt! Apr 18 01:46:49 I don't proclaim to be any sort of expert, though Apr 18 01:47:04 just a friendly dev who's avoiding their normal dayjob ;) Apr 18 01:47:14 procrastination is a wonderful thing Apr 18 01:54:08 The library: https://pastebin.com/7RbBh9Jq, the application https://pastebin.com/L0e5aWpL Apr 18 01:55:39 @tz sorry for the delay, I'm working remotely from home and I'm not really well installed. Apr 18 01:55:46 all good Apr 18 01:56:56 the summary on the gdd lib doesn't have a leading double quote " -- copypaste issue? Apr 18 01:59:20 Yeah, copy paste issue. Sorry. Probably when I erase the HOMEPAGE and other link to my company. Sorry. Apr 18 01:59:43 yep, figured as much Apr 18 02:04:22 honestly not sure, it looks fine -- perhaps check the logs in tmp/work/(lib|app)//temp Apr 18 02:04:52 do_install's I'd guess Apr 18 02:06:17 I tend to use CMake for my projects and it just seems to work out exactly what it needs and how to deploy it, relatively painless in that respect Apr 18 02:06:26 Could it be related to the Makefile of the gdd ? Because I have basic knowledge with Makefile. I'm not 100% sur of the way I produce the output of the library and the install rules. Apr 18 02:07:18 Here is the Makefile of the library https://pastebin.com/FSE3jzJc Apr 18 02:07:52 perhaps, I don't think I've worked with a makefile in earnest for several years if I'm completely honest Apr 18 02:08:20 >> perhaps check the logs. The log contains onlly this error and no other clue. Apr 18 02:09:35 wonder if you run bitbake with verbose (-v) flag it gets any extra output Apr 18 02:09:46 for the logs Apr 18 02:10:23 Sure, I try now Apr 18 02:16:29 Here is something that look strange https://pastebin.com/2GfMxj8t Apr 18 02:17:37 Line 342 and 343. Could it be that objdump detect that netclient application depends on gdd library automatically ? Apr 18 02:20:11 seems like it, what's in the log for do_package and do_package_qa? Apr 18 02:22:50 log.do_package_qa --> https://pastebin.com/YNg5G6eM Apr 18 02:24:24 log.do_package -> https://pastebin.com/zmMEBRmB Apr 18 02:24:35 sounds a bit like it, try a: bitbake -c devshell intelia-netclient, then run the failing objdump command manually, see if it spits out anything abnormal Apr 18 02:24:56 shame it doesn't include the output Apr 18 02:25:01 of the objdump Apr 18 02:27:46 objdump -> https://pastebin.com/q14pz5YP Apr 18 02:29:01 was that the .debug one? what about the other invocation? Apr 18 02:29:47 What do you means by .debug ? I don't understand. Apr 18 02:30:01 ah, there were two objdump invocations in the logfile Apr 18 02:30:08 was curious what the output of the other was Apr 18 02:30:32 Oh sorry I did not notice that. Apr 18 02:30:33 lines 8 and 14 Apr 18 02:34:05 For the debug -> https://pastebin.com/nY7wRQDX Apr 18 02:35:45 Could this be because my makefile has to produce 2 libraries, including a debug and a release? And that bitbake expects the release to be in the FILES _ $ {PN} and the other in FILES _ $ {PN} -dev? Apr 18 02:36:10 honestly not sure Apr 18 02:40:41 nothing I can see really stands out I'm afraid Apr 18 02:40:46 looks... fine Apr 18 02:41:02 sorry mate :( Apr 18 02:43:57 Thank you so much @tz. I will continue tomorrow and If I did not found the problem I will ask again here when there will be more active members. Apr 18 02:44:43 sounds like a plan Apr 18 02:44:47 good luck :D **** ENDING LOGGING AT Thu Apr 18 02:59:57 2019