**** BEGIN LOGGING AT Mon Apr 27 02:59:57 2020 Apr 27 04:47:47 halstead: patchwork seems to be missing patches Apr 27 04:48:18 did something change with mailing lists that its now evading patchwork ? Apr 27 05:10:31 Very likely. I'll check into it in the morning. Apr 27 06:02:26 New news from stackoverflow: COMPATIBLE_HOST vs COMPATIBLE_MACHINE in yocto Apr 27 07:32:44 New news from stackoverflow: replace Makefile to _Makefile in Makefile Apr 27 07:51:32 khem, some oe/yocto lists got a new number prefix for their list id. no idea why someone would change that, though... Apr 27 07:56:23 Hi. Iam still working on my example tool which is using an example lib. I did one huge step... I messed up the argument order from CC Apr 27 07:56:34 http://dpaste.com/041G6BS But this is the new issue :)  Apr 27 08:53:43 Good morning all! I was wondering: what is the state of PySide2 in Yocto? Apr 27 08:54:36 My employer doesn't want to use PyQt5 and would be willing to spend some of my time on a port to Yocto. Apr 27 08:59:06 problame: I guess you can direct this question to http://layers.openembedded.org/layerindex/branch/master/recipes/ Apr 27 09:00:16 Hey. Is there a way to find automatically package name of a specific binary in the final image? For instance I have `/usr/sbin/chpasswd` - Which package build it? Apr 27 09:00:52 kanavin_home: very helpful, thx :P I also found https://github.com/meta-qt5/meta-qt5/issues/253 Apr 27 09:55:31 nacknick: if you're on a live target and are using opkg, i would typically grep all *.list files for chpasswd Apr 27 09:55:45 that's /opkg on my target although i think the standard location might be different Apr 27 10:23:21 nacknick: oe-pkgdata-util can sort of do that using the 'oe-pkgdata-util find-path /path/to/file' way, but it lists which packages that provide that path and not which package in a certain image. So if there's multiple provider, but only one who end up in the image, it won't tell you which. Apr 27 10:39:50 I have an pre-build library file inside /lib dir. I want to add a symlink to it. Apr 27 10:40:09 can I do it with do_install_append()? Apr 27 10:42:07 `ln -s {D}{base_libdir}mylib.so.1 {D}{base_libdir}mylink.so` Apr 27 10:42:14 that's should work? Apr 27 10:57:30 RP: is it about 10 years since YP 1.0 release now ? Apr 27 10:57:34 RP: nice :) Apr 27 10:58:03 Marex: 10 years since 0.9 in October Apr 27 10:58:37 RP: heh Apr 27 11:01:47 RP: nice :) Apr 27 11:20:46 Hi. I created a recipe that downloads a repository from github. I then used "devtool modify " to make some changes - namely remove files (i.e. git rm ). Commited changes and "devtool update-recipe ". But everytime I try to bitbake my recipe, quit fails and complains that such patch is deteced ( Reverse or previously applied). But that is false. Did anybody faced this problem in the Apr 27 11:20:52 past? Apr 27 11:21:29 quilt* fil Apr 27 11:21:41 quilt* fails* Apr 27 11:35:45 nacknick: no ${D} in link target Apr 27 11:36:48 because $D is the path to the staging directory on your build machine, and you don't want that on the target file system Apr 27 12:50:07 rburton: So how should I do it? Apr 27 13:00:01 RP: thanks for the help :) Apr 27 13:00:49 btw, I can ssh into centos7-1, but not into centos7-4 Apr 27 13:12:48 nacknick: just don't put ${D} in the link target Apr 27 13:17:01 kanavin_home: I had to retry as the first round had a different issue from the bzip2 change Apr 27 13:17:26 kanavin_home: the mingw failure on the last net build is from your patches Apr 27 13:17:28 include/X11/meson.build:19:4: ERROR: Problem encountered: Your fd_set is too weird. Apr 27 13:17:53 amazing error Apr 27 13:18:01 kanavin_home: you'd need to talk to halstead about the ssh issues Apr 27 13:18:01 NO YOU'RE WEIRD Apr 27 13:18:46 reminds me of the parallel port "printer on fire" printk() Apr 27 13:19:51 kanavin_home: for context I reran -next with the bzip2 ptest and your series minus rpm and friends changes Apr 27 13:24:32 I accidentally remove deploy/images/raspberrypi3 directory. How to rebuild? Apr 27 13:24:41 removed Apr 27 13:24:53 bitbake myimage -f? Apr 27 13:25:06 OK. I will try with -f Apr 27 13:25:37 kanavin_home, I've re-run your account provisioning. Please try now. Apr 27 13:29:22 kanavin_home: the gplv3 build came from sstate so I've killed it, will run manually and force it Apr 27 13:29:56 halstead: thanks, sorry if I woke you :( Apr 27 13:30:31 halstead: neither of us can get into debian8 fwiw Apr 27 13:32:59 halstead: thanks Apr 27 13:33:07 debian 8 works now Apr 27 13:33:46 kanavin_home: non-gplv3 contains a successful rpm-native build Apr 27 13:33:54 New news from stackoverflow: How to enable logging of python script in yocto poky layer? Apr 27 13:39:16 RP, I re-ran your account provisioning on debian8-ty-1.yocto.io and it shows no changes needed. Can you try again while I'm watching the logs? Apr 27 13:40:49 rburton: does not work. it says ".../images/raspberrypi3/bcm2835-bootfiles/*: No such file or directory" Apr 27 13:40:51 halstead: works... Apr 27 13:41:13 halstead: not sure what happened, maybe I did something wrong :/ Apr 27 13:42:00 RP, How odd. I'm not sure. Maybe there was an unreported change? Glad it's sorted. Apr 27 13:42:35 halstead: yes, working now which is the main thing Apr 27 13:51:29 if I want to support multiple OE core releases with a single branch in a layer, but I need to handle new PACKAGECONFIG flags which are not defined (and thus generate a warning) in older releases, is checking for LAYERSERIES_COMPAT in a bbappand what I want to do? Apr 27 13:51:33 or is there a better way ? Apr 27 13:52:58 Marex: We set those in distro.conf, e.g. PACKAGECONFIG_pn-foo = "bar baz" Apr 27 13:53:34 JPEW: and how does that exactly handle the difference between different OE core versions ? Apr 27 13:54:59 Our distro.conf ends with a line: "require conf/distro/include/${DISTRO}-${@d.getVar('LAYERSERIES_CORENAMES').split()[0]}" Apr 27 13:55:17 RP: that build is coming from sstate unfortunately :( Apr 27 13:55:41 Marex, I use LAYERSERIES_CORENAMES in my layer.conf, and add subdirectories to BBFILES based on its value Apr 27 13:55:46 we probably need to add a PR increase to the recipe Apr 27 13:56:00 Marex: So, we have version specific distro files (e.g. poky-zeus.conf, poky-dunfell.conf) Apr 27 13:56:15 JPEW: well that already fails if you build distroless :) Apr 27 13:56:30 HELP??? I accidentally deleted deploy/images/raspberrypi3 directory. No `bitbake ` is getting stuck. How to fix it? Apr 27 13:56:40 kroon: oh that's interesting, it's like dynamic layers, but for corenames Apr 27 13:56:42 Now* Apr 27 13:57:04 nacknick: restart the build Apr 27 13:57:15 Marex, yes Apr 27 13:57:22 Marex: How? Apr 27 13:57:23 nacknick: delete tmp and restart Apr 27 13:57:34 nacknick: it will reassemble the rest of the stuff from sstate cache Apr 27 13:57:37 nacknick: Is your sstate in tmp? Apr 27 13:58:06 If it is, deleting tmp will rebuild everything from scratch, which is unnecessary :) Apr 27 13:58:09 Marex jpew sstate-control? Apr 27 13:58:32 nacknick: sstate-cache/ , should be on the same level as tmp/ Apr 27 13:58:32 nacknick: Whats SSTATE_DIR in local.conf? Apr 27 13:59:03 sstate-cache Apr 27 13:59:06 RP: is what kroon said, about using LAYERSERIES_CORENAMES in layer.conf , something which is a good practice in your book ? Apr 27 13:59:41 Marex: no tmp dir Apr 27 13:59:51 JPEW: sstate-cache Apr 27 14:00:07 should I delete it? Apr 27 14:00:52 nacknick: delete tmp/ , keep sstate-cache/ Apr 27 14:00:58 nacknick: Marex is correct, deleting tmp should be sufficent. I always forget where the default is because I never use the default Apr 27 14:00:58 nacknick: then re-run bitbake Apr 27 14:01:04 JPEW: I don't have tmp dir Apr 27 14:01:39 nacknick: you should have one Apr 27 14:01:42 build-dh-stm32mp1-dhcor-avenger96$ ls Apr 27 14:01:42 bitbake-cookerdaemon.log cache conf sstate-cache tmp Apr 27 14:01:48 this is how it looks for me , (example) ^ Apr 27 14:01:59 notice that sstate-cache/ and tmp/ dir Apr 27 14:03:38 Marex: I have not. I searched tmp in whole bitbake folder Apr 27 14:04:45 JPEW: Marex: Any other solution? Apr 27 14:07:24 kanavin_home: I forced rpm-native Apr 27 14:11:51 nacknick: it's in your build directory Apr 27 14:12:04 it's not :| Apr 27 14:12:06 nacknick: you said you deleted deploy/ , deploy/ is a subdir of tmp/ Apr 27 14:12:15 nacknick: so if you deleted deploy/ , cd .. Apr 27 14:12:19 I have a Yocto version that don't has tmp dir Apr 27 14:12:37 ohhh Apr 27 14:12:55 you want me to delete the whole "tmp" dir. It's called "BUILD" here Apr 27 14:13:05 but I will take years to rebuild it Apr 27 14:14:44 nacknick: it will reassemble the whole thing from your sstate-cache Apr 27 14:15:03 the OE build usually takes an hour or a few Apr 27 14:15:37 yes but I configured to not stripped whole binaries, it takes a long time Apr 27 14:15:44 the sstate-cache contains most of the stuff you need to reassemble the content of tmp/ without compiling much Apr 27 14:15:46 not strip Apr 27 14:16:12 striping binaries of debug symbols has nothing to do with this Apr 27 14:17:36 OK. I will try Apr 27 14:41:43 RP: so is there a build I could look at? let me know Apr 27 14:47:45 Hey together, is there a way to define a `COMPATIBLE_KERNEL` or similar in a recipe? Apr 27 14:48:43 kanavin_home: yes, I said before, I hacked the non-gplv3 build and forced rpm-native to compile in it Apr 27 14:49:20 /home/pokybuild/yocto-worker/non-gpl3/build/build/tmp/work/x86_64-linux/rpm-native/ Apr 27 14:49:33 RP: is that on centos7-4 ? Apr 27 14:49:50 kanavin_home: yes Apr 27 14:52:06 RP: right, i can look now what happens Apr 27 14:53:23 RP: Apr 27 14:53:23 -bash-4.2$ ls -l gcc Apr 27 14:53:23 lrwxrwxrwx. 1 pokybuild users 12 27.4. 13:13 gcc -> /usr/bin/gcc Apr 27 14:53:23 -bash-4.2$ pwd Apr 27 14:53:23 /home/pokybuild/yocto-worker/non-gpl3/build/build/tmp/hosttools Apr 27 14:53:34 as I suspected, hosttools contains a link to the wrong gcc :( Apr 27 14:55:15 kanavin_home: hmm Apr 27 14:56:12 that would be an autobuilder bug. And hence "my" problem :/ Apr 27 14:57:06 RP: if I can help, let me know. What I did notice is that the check for gcc version runs it against gcc in PATH, which isn't entirely correct - it should be run against gcc in hosttools/ Apr 27 14:57:43 because if wrong gcc makes it into hostools/ somehow, it won't be caught or reported Apr 27 14:58:55 kanavin_home: PATH should always have hosttools first though Apr 27 14:58:55 the way I got it there was by running bitbake before installing the tarball - it does report a wrong version but *also*, crucially, creates the symlinks to the wrong versions in hosttools/. If you install the tarball, that wont recreate the symlinks. Apr 27 15:00:26 I have no idea why this is non-deterministic though. Apr 27 15:02:57 kanavin_home: I'm going to rerun this with a PR bump and extra debug in autobuilder-helper Apr 27 15:04:15 New news from stackoverflow: Check kernel version in recipe Apr 27 15:06:48 khem: sorry, had to stop that meta-oe build as we're trying to debug something on that worker Apr 27 15:11:09 khem: that master-next is bust anyway Apr 27 15:16:00 kanavin_home: I don't understand. https://autobuilder.yoctoproject.org/typhoon/#/builders/75/builds/1817 the stdio log says its using buildtools Apr 27 15:17:45 RP: what could be happening is that hosttools/ somehow gets populated before the extended-tarball is set up and taken into use. I don;t know where or how though :-/ Apr 27 15:23:17 just ran my prune sstate script to strip out sstate i haven't accessed for a few months Apr 27 15:23:22 Before: Apr 27 15:23:22 942G /home/ross/Yocto/sstate Apr 27 15:23:23 After: Apr 27 15:23:23 36G /home/ross/Yocto/sstate Apr 27 15:23:35 i think my disk is breathing a sigh of relief Apr 27 15:26:39 kanavin_home: right, but that shouldn't happen :/ Apr 27 15:26:48 rburton: nice :) Apr 27 15:31:02 RP: hmm ok, isnt there a way to bring nodes offline so scheduler does not see them when being worked on Apr 27 15:31:57 rburton: thats great but it never reduces so much for me. are you building similar arch most of time ? Apr 27 15:32:10 khem: yes, and i haven't pruned for about two years Apr 27 15:32:39 RP: or somehow the hosttools creation code doesn't see the tools from the tarball and falls through to the host ones :-/ Again, hard to say without adding lots of logging everywhere. Apr 27 15:35:11 rburton: 1TB in 2 years, mine fills in a week Apr 27 15:35:41 khem: i never build meta-oe :) Apr 27 15:36:14 yeah I guessed Apr 27 15:36:41 I have an archive file that contains a binary file and a symlink to the binary file. I extract them both to ${D}. In the final image I can find the binary file, but not the symlink. WHY? Apr 27 15:37:50 khem: I needed to actually run builds :/ Apr 27 15:38:40 kanavin_home: if I run this on the console it works. If run from the autobuilder it does not Apr 27 15:40:02 kanavin_home: I think I might understand Apr 27 15:41:08 nacknick: try 'oe-pkgdata-util list-pkg-files -p [recipename]' to see if the files are packaged where you expect Apr 27 15:41:42 rburton: thanks Apr 27 15:41:48 nacknick: if you're using a new enough release, oe-pkgdata-browser is nicer Apr 27 15:41:56 ah heck, to fix this I have to change parameters to one of the scripts Apr 27 15:41:58 @rburton: Is this prune sstate script of yours somewhere lurking around to give it a try? Apr 27 15:42:12 Another question please. If I want to add a line to a file that resides inside /etc. What is the correct way to do it? Apr 27 15:42:20 RobertBerger: mine is dumb, it's just a find script using -atime and -delete Apr 27 15:42:32 nacknick: find the recipe that writes that file, change it Apr 27 15:42:53 @rburton: so something like that? find ${sstate_dir} -name 'sstate*' -atime +3 -delete Apr 27 15:43:22 RobertBerger: find $SSTATE -type f -atime +$DAYS -delete Apr 27 15:43:25 it's a "general" file. Several packages could use it Apr 27 15:43:30 rburton: Apr 27 15:43:46 nacknick: but only one recipe wrote the file into the image Apr 27 15:43:55 @rburton thanks Apr 27 15:44:49 rburton: OK Apr 27 15:55:46 kanavin_home: I understand what broke so you can look at how we fix buildtools... Apr 27 15:56:53 RP: yeah, I think adding nativesdk-libgomp-dev to the tarball should be enough, but need to run tests on centos7 first Apr 27 15:57:54 RP: then there's the question of what to do with ubuntu 16.04, as its compiler too produces binaries that segfault :-/ Apr 27 15:58:42 RP: btw, rpm 4.16 deprecates berkeley db in favor of something called 'ndb backend', or sqlite Apr 27 15:59:26 fray: ^^^ Apr 27 16:00:26 rburton: Regrading my previous question. Is ".so.1" convention extension is applied to symlinks as well? I can't create symlink with ".so" extension? Apr 27 16:00:46 nacknick: you can name a symlink anything you want Apr 27 16:01:01 nacknick: but the default packaging rules put .so and .so.1 in different packages Apr 27 16:01:22 we went through this last week and the tips and tricks article i posted - which i wrote - explains all this Apr 27 16:01:43 so If I want it to be inside "release" version and not "dev" I have to add version Apr 27 16:02:00 a version Apr 27 16:02:18 well either 1) version the libraries or 2) change the FILES so the unversioned library goes where you want it. the wiki page explains this. Apr 27 16:21:08 I am having trouble trying to do something that seems pretty basic. I am attempting to remove DISTRO_FEATURES such as bluetooth from Poky. After building, I get the following error: https://pastebin.com/QCjWKKv1 Apr 27 16:59:30 There is a recipe that has 2 tasks: `pkg_postinst_${PNLIBSEGFAULT}` and `pkg_prerm_${PNLIBSEGFAULT}` - What does it mean? Apr 27 17:04:15 they're not tasks, they're package postinst and pre-remove hooks Apr 27 17:15:57 I am having trouble trying to do something that seems pretty basic. I am attempting to remove DISTRO_FEATURES such as bluetooth from Poky. After building, I get the following error: https://pastebin.com/QCjWKKv1 Apr 27 17:41:34 khem: thanks for the merge, but I didn't have time to test with the `libedit` suggesetion. Apr 27 18:04:09 kanavin_home: autobuilder buildtools issue is fixed Apr 27 18:05:21 RP: I am about to send the patch for omp.h issue as well :) tested on centos7 - seems to work (rpm-native builds, and 'bitbake rpm' succeeds in packaging everything, with multithreading working as expected) Apr 27 18:05:57 RP: it does require producing a new tarball though Apr 27 18:13:39 kanavin_home: thanks. We'll figure out the tarball Apr 27 18:14:00 kanavin_home: we can probably convince sakoman to take this for the next point release :) Apr 27 18:15:33 kanavin_home: yes please tag the patch with [dunfell] so I don't miss it! Apr 27 18:23:03 RP: patch sent :) Apr 27 18:23:29 sakoman: too late, I sent it just before looking here :) Apr 27 18:25:18 RP: the next issue is most likely segfaults on ubuntu 16.04 - I guess we will have to bump minimum gcc to 6.x and install the tarball on ubuntu 16.04 as well. Maybe let's respin first to confirm the centos and debian are sorted at least. Apr 27 18:31:11 kanavin_home: right, will take a while since I'll have to spin a new buildtools first Apr 27 18:31:20 * RP also needs to eat, need to remember that Apr 27 18:33:28 kanavin_home: no worries, I'll look for it Apr 27 21:34:47 is there a way to generate a list of certain package variables and put that as a file into an image? Apr 27 21:35:53 You can write a bbclass, set it as INHERIT += in local.conf, have that bbclass write that variable to a file somewhere in a staging dir and append a task to the image recipe that collects the various values from the staging area Apr 27 21:38:57 there already is image-buildinfo.bbclass take a look maybe thats a good starting point Apr 27 22:33:29 khem: autobuilder is ok to use now btw, found the problem earlier and now fixed Apr 27 23:39:34 RP: ok thanks Apr 28 01:06:14 New news from stackoverflow: How do I generate a multi-boot Grub EFI WIC image from Yocto Apr 28 02:36:35 New news from stackoverflow: how to change the button size on title bar on xwayland? **** ENDING LOGGING AT Tue Apr 28 02:59:59 2020