**** BEGIN LOGGING AT Fri Mar 20 02:59:57 2020 Mar 20 03:01:15 New news from stackoverflow: What is difference between Yocto Release vs poky release Mar 20 04:01:25 New news from stackoverflow: Point releases vs milestone releases in Yocto Mar 20 06:25:35 mckoan|away: thanks for the hint yesterday about autoloading modules. It was not really obvious with all the different names referencing the module, that the actual filename without .ko was suppesed to go into that variable. Everything works now! Mar 20 08:27:01 Hello, I have 2 recipes trying to edit .ssh/authorized_keys to add public key. Yocto doesn't accept that, someone here have a workaround to this kind of issue ? cheers Mar 20 08:29:27 NiniC0C0, can you modify both recipes ? Mar 20 08:30:31 NiniC0C0, change them to install keys in .ssh/authorized_keys.d/, then add a postinsall script that cat's everything to .ssh/authorized_keys Mar 20 08:30:43 kroon yes yes both of them a personal recipe. One is always apply (for all board) after that i have to add 1 more key according to the image name :) Mar 20 08:32:09 kroon hooo thx for the idea ! Mar 20 08:32:19 a postinstall/postremove script that cats' those files should do the trick Mar 20 08:32:36 kroon I keep in touch :P Mar 20 08:33:04 that should work unless you're on a readonly fs Mar 20 08:37:52 erm, that should just work Mar 20 09:20:48 what should I do with my cmake recipe when it tries to find linke from my host? /lib/ld-linux-aarch64.so.1: No such file or directory Mar 20 09:21:21 LDFLAGS_append has no effect Mar 20 09:21:59 stuom1: fix the cmakelists, probably Mar 20 09:22:43 what should I put there? It doesnt have any path set Mar 20 09:26:16 @kroon/NiniC0C0: I guess you could also use a image specific .bbappend to add the image specific stuff to the generic stuff. This should also work with a read only fs. Mar 20 09:32:53 RobertBerger the problem with the bbapend is that I don't want to have IMAGE specific dependency :S Mar 20 09:41:13 @NiniC0C0: OK I see Mar 20 09:56:36 kroon confirmed ROOTFS_POSTPROCESS_COMMAND made the tricks. Thx for advice, you have save my day :) Mar 20 10:02:23 NiniC0C0: what exactly are you doing? I think what kroon suggested is a pkg_postinst_ontarget task in your package recipe? Mar 20 10:07:38 qschulz hello, yesterday i have followed your advice, put some hidden trouble left. So now I have a recipe ssh-keys.bb to add "golden keys" same for all developers. I have also ssh-keys-project1.bb ssh-keys-project2.bb, but as I have to merge authorized_keys file, yocto doesn't multiple access to the same file so I have followed kroon advice and Mar 20 10:07:39 now all recipes put public keys inside a folder and use ROOTFS_POSTPROCESS_COMMAND to merges all keys. Sounds bad idea ? Mar 20 10:10:31 NiniC0C0: not particularly, but you could do everything in your ssh-keys-project1.bb (I think you might need to put an RDEPENDS on ssh-keys-golden if that is a way to make sure this package is installed before -projectX.bb) by using pkg_postinst_ontarget_${PN} Mar 20 10:10:37 which runs on the first boot Mar 20 10:10:41 on the target Mar 20 10:10:47 (first boot after installation) Mar 20 10:12:47 qschulz in fact I have just add DEPENDS (needed at runtime ?) I have to take a look the doc because I don't know this command pkg_postinst_ontarget_${PN} :) Mar 20 10:15:02 qschulz after reading i think i will keep ROOTFS_POSTPROCESS_COMMAND because all our systems doesn't have persistent memory :S Mar 20 10:15:13 So it's always the first boot ^^ Mar 20 10:17:14 NiniC0C0: mmmmm, you actually don't need to run it ontarget since it's a simple cat Mar 20 10:17:52 NiniC0C0: what about pkg_postinst? https://www.yoctoproject.org/docs/latest/mega-manual/mega-manual.html#new-recipe-post-installation-scripts Mar 20 10:20:41 qschulz D: Ok, pkg_postinst_ontarget_${PN} sounds better because i don't have to edit all images recipe. That's right ? Mar 20 10:22:27 Good Morning, hope everybody is fine especially these days. I'm afraid I'm pretty badly stuck on compiling a recipe for my target architecture (Allwinner H2). I'm getting a warning similar to `/_fftpack.cpython-37m-arm-linux-gnueabi.so uses VFP register arguments, libfftpack.a(cffti1.o) does not`. So, I tried enforcing FPU handling in the machine Mar 20 10:22:28 configuration setting `target_FPU=soft` (or hard or neon-vfp). But neither worked and it always fails (after about 30 minutes of heavy compiling) ;( Any hints how to proceed with this step? Mar 20 10:22:54 This is a more extensive description of the issue: https://stackoverflow.com/questions/60772241/recipe-compilation-fails-due-to-floating-point-unit-compatibility-issue-i-assum Mar 20 10:25:58 emrius: have you seen https://github.com/gpanders/oe-scipy Mar 20 10:26:58 @erbo yes that's the recipe I'm using. I asked gpanders for help but he could not help unfortunately https://github.com/gpanders/oe-scipy/issues/2 Mar 20 10:29:03 One step further: turns out that this compiles: Mar 20 10:29:05 `arm-poky-linux-gnueabi-gcc -Wall -g -march=armv7ve -mthumb -mfloat-abi=softfp -fstack-protector-strong -D_FORTIFY_SOURCE=2 -Wformat -Wformat-security -Werror=format-security --sysroot=/home/marius/mender-orangepi-warrior/build/tmp/work/armv7vet2hf-neon-poky-linux-gnueabi/python3-scipy/1.3.3-r0/recipe-sysroot -shared -Wl,-O1 -Wl,--hash-style=gnu Mar 20 10:29:06 -Wl,--as-needed -fstack-protector-strong -Wl,-z,relro,-z,now` Mar 20 10:29:34 Removed the linker arguments. Mar 20 10:32:39 New news from stackoverflow: Recipe compilation fails due to Floating Point Unit compatibility issue (I assume) Mar 20 11:15:48 NiniC0C0: sorry, fighting with a board. pkg_postinst (not _ontarget), yes, sounds better Mar 20 11:16:05 NiniC0C0: but I think you might have to do it for both the golden keys and the client keys Mar 20 11:41:40 NiniC0C0: (depends in which order they're installed in the rootfs, and that, I don't know) Mar 20 12:09:11 Hey all. I have a weird problem. I have a pre-existing recipe that hasn't changed (and has no dependencies, it's a little c program) but now fails to get installed into initrd as a result of a complaint that `no package provides /usr/bin/python3` Mar 20 12:09:27 I'm in krogoth world, and also fiddling with moving things to thud. Mar 20 12:09:48 Anything changed or anything I should know about that could cause what seems to be an environmental failure? Mar 20 12:24:03 JoeR: we did get stricter about dependencies, that probably means there is a script in there which has a python3 path at the start of it Mar 20 12:24:25 OK. That's an easy grep. Thanks! Mar 20 13:30:10 * Crofton|road grumbles about package going backward ERRORS that do not stop the build annoying jenkins Mar 20 14:15:56 Crofton|road: We made them warnings Mar 20 14:16:30 Yeah, I suppose I should also Mar 20 14:16:46 ERROR_QA_remote = "version-going-backward" Mar 20 14:16:56 WARN_QA_append = " version-going-backward" Mar 20 14:17:29 I figured there was a solution, I just like complaining Mar 20 14:17:33 and thnaks Mar 20 14:30:45 ah you mean _remove not _remote, I was scratching my head a bit :) Mar 20 14:42:12 Crofton|road: Heh ya Mar 20 14:42:28 copy-paste doesn't always work with my client, so I re-typed it :) Mar 20 15:41:23 Is there a way to access the call stack in with BitBake or so? Mar 20 15:42:37 RobertBerger: in which order tasks are executed within a same recipe you mean? Mar 20 15:42:52 What I am after is this do_unpack[postfuncs] += "do_unpack_extract_submodules" but if do_unpack is called from do_unpack_and_patch do something differently in do_unpack_extract_submodules or call another version of it Mar 20 15:43:09 RobertBerger: use bitbake -g -u taskexp to see dependencies Mar 20 15:43:27 well I would like it programatically Mar 20 15:43:37 in my recipe Mar 20 15:44:07 just have two different fonctions Mar 20 15:44:09 Uhm, sounds like an XY problem Mar 20 15:44:22 but yes, same as kriive :) Mar 20 15:44:23 OK let me show you the precise problem Mar 20 15:45:00 meta-java and the archiver class are incompatible Mar 20 15:45:39 If the archiver is used it calls do_unpack_and_patch Mar 20 15:46:18 between unpack and patch I need a special version of do_unpack_extract_submodules Mar 20 15:46:27 This is without the archiver class: Mar 20 15:46:56 https://github.com/RobertBerger/meta-java/blob/a97d27ea0d6701df222c79fb1ce983fc39308c0c/recipes-core/openjdk/openjdk-8-common.inc#L174 Mar 20 15:47:21 https://github.com/RobertBerger/meta-java/blob/a97d27ea0d6701df222c79fb1ce983fc39308c0c/recipes-core/openjdk/openjdk-8-common.inc#L33 Mar 20 15:47:44 This is with the archiver class (and my hack, which I would like to get rid of): Mar 20 15:48:47 I replaced unpack_and_extract with my own version: https://github.com/RobertBerger/meta-java/blob/a97d27ea0d6701df222c79fb1ce983fc39308c0c/recipes-core/openjdk/openjdk-8-common.inc#L240 Mar 20 15:49:34 which calls now a custom version of do_unpack_extract_submodules: https://github.com/RobertBerger/meta-java/blob/a97d27ea0d6701df222c79fb1ce983fc39308c0c/recipes-core/openjdk/openjdk-8-common.inc#L265 Mar 20 15:50:01 Which is this: https://github.com/RobertBerger/meta-java/blob/a97d27ea0d6701df222c79fb1ce983fc39308c0c/recipes-core/openjdk/openjdk-8-common.inc#L49 Mar 20 15:50:45 If you compare do_unpack_extract_submodules and do_unpack_extract_submodules_archiver you see that the difference is just the relative path Mar 20 15:51:06 tar xjf ${WORKDIR}/../${CORBA_FILE_LOCAL} instead of tar xjf ${WORKDIR}/${CORBA_FILE_LOCAL} Mar 20 15:53:02 If I could find out somehow that the call tree is unpack_and_extract -> unpack -> unpack_extract_submodules I guess I could come of with a better solution Mar 20 15:56:08 RobertBerger: have you reported tha meta-java does not work with the archiver class? Mar 20 15:56:57 @qschulz where? Mar 20 15:57:35 I am not sure I can put this somewhere into the yp bugzilla Mar 20 15:58:28 I only had some e-mail exchange with one of the meta-java maintainers and he reviewed my hacks. Mar 20 15:58:51 Trying to clean them up so I can push something upstream Mar 20 16:00:07 It looks like you are lucky if you get meta-java to build ;) It's highly dependent on the host gcc. Mar 20 16:00:32 So I guess the archiver class is one of the least concerns of meta-java Mar 20 16:00:36 RobertBerger: I think that's the root of your issue, can we start with that? What is happening with the incompatibility? what are the error messages? when you are doing what? etc... Then someone might have more context and give th eproper answer :) Mar 20 16:01:18 So where should I report it? Mar 20 16:02:54 I do have a pretty good idea what's the problem and have also a hack which fixes it. But my hack is ugly. Mar 20 16:03:40 Apparently no one ever built meta-java using the archiver class as well ;) Mar 20 16:03:59 it's very easy to reproduce. Mar 20 16:05:31 If you use meta-java and activate the archiver and build openjdk-8 it will not be able to extract the sources and apply the patches Mar 20 16:05:59 because the path is wrong for the archiver class Mar 20 16:08:47 do_unpack_and_patch() sets WORKDIR itself to ${ARCHIVER_WORKDIR}, which defaults to ${WORKDIR}/archiver-work/ Mar 20 16:09:25 and so the relative paths need to be /../ Mar 20 16:09:50 I would happily report it somewhere just don't know where ;) Mar 20 16:13:23 RobertBerger: http://git.yoctoproject.org/cgit/cgit.cgi/meta-java/tree/README#n107 Mar 20 16:14:06 I already tried on openembedded-devel@lists.openembedded.org but I can try again Mar 20 16:18:12 RobertBerger: if you've done it, then good enough :) Mar 20 16:18:23 you can ping gently if there's some ongoign discussion or something Mar 20 16:19:27 It would really make things easier for people if you would point to the whole thread or maybe even give the actual error (people are lazy, they won't try to build meta-java with the archiver class :) ) If you give a log, people are usually more enclined to help Mar 20 16:19:34 BUT I suspect I won't be able to help much Mar 20 16:19:40 I'll try with [meta-java] formatting, maybe this helps ;) Mar 20 16:22:43 Guys, how can I tell bitbake to use a pair of Git credentials in order to let it download from private repos? Mar 20 16:23:51 In a CROPS Docker container can I do this: $ git config --global user.name 'your user name' Mar 20 16:23:55 $ git config --global user.password 'your password' Mar 20 16:24:35 And instead of my username and password pass and env variable with the proper auth credentials? Mar 20 16:24:41 an env variable* Mar 20 16:25:09 kriive: docker takes -e arguments AFAIR for environment variables Mar 20 16:25:36 but why not use SSH keys instead? Mar 20 16:26:03 well, then you have to pass the private key to the container, so... not extremely safe if it's not your container :) Mar 20 16:27:45 But if I put an ssh key into .ssh/idblahblah in the container, does bitbake in the CROPS container use that key or should I put that key somewhere else so that bitbake recognizes it? Mar 20 16:29:21 I use CROPS container + gitlab ci, so I would inject the key before starting bitbake, I hope that is sufficiently secure Mar 20 16:29:55 bitbake just calls ssh. whether ssh uses that key depends on its filename, your ssh config, and whether an agent is involved (probably not given crops) Mar 20 16:32:01 Alpright, thank you guys! Mar 20 16:32:05 kergoth: mmmm, then password asked for every repo requesting the ssh key :/ Mar 20 16:32:33 password or the passphrase of the key? see man ssh_config for how to configure which keys are used by default lacking an agent Mar 20 16:33:06 Yeah, I guess I would not encrypt the key Mar 20 16:33:55 New news from stackoverflow: Activating High Resolution timer on ARM V7 -Yocto Mar 20 16:38:26 kergoth: I meant passphrase Mar 20 16:39:02 go passphrase-less or arrange to use an agent on the host and map SSH_AUTH_SOCK into the container, i'd say Mar 20 17:04:02 New news from stackoverflow: compiling hardware-accelerated VLC into YOCTO for Raspberry Pi 3 Mar 20 17:29:06 qschulz: Its solved for yoe distro here https://github.com/YoeDistro/yoe-distro/blob/master/envsetup.sh#L487-L514 Mar 20 17:34:26 khem: ? Mar 20 17:37:56 khem, that's super useful! Mar 20 18:34:56 qschulz: that was for your ssh passphrase issue for docker use Mar 20 19:04:25 New news from stackoverflow: yocto offline build fails cause fetches from recipes Mar 20 19:34:31 New news from stackoverflow: yocto: how do I install header files along with kernel module in sdk **** ENDING LOGGING AT Sat Mar 21 02:59:57 2020