**** BEGIN LOGGING AT Thu Jan 28 03:00:03 2021 Jan 28 04:06:41 where is features/overlayfs/overlayfs.scc? Jan 28 04:12:51 do kernel features only work with the linux-yocto recipe? Jan 28 04:13:18 they work with any kernel recipe that inherits the linux-yocto bbclass. Jan 28 04:13:32 and to answer the other question: https://git.yoctoproject.org/cgit/cgit.cgi/yocto-kernel-cache/tree/features/overlayfs/overlayfs.scc Jan 28 04:13:32 I see Jan 28 04:15:01 yup just found out, and linux-ti-staging obviously doesn't inherit linux-yocto... Jan 28 04:15:11 nope. Jan 28 04:15:35 standard defconfig wrangling, or if they have their own config scheme is the way forward. Jan 28 04:16:47 it's kinda tricky. If I'm using yocto, I would've gone all yocto and using its linux-yocto recipe, but I guess there might be some (e.g. graphics?) features in linux-ti-staging that I might be missing without the manufacturer recipe... Jan 28 04:18:21 if a platform is well supported upstream, then linux-yocto would inherit that same support. But the vendor BSPs do have all the features, etc, that are often not in mainline. That's the dance with the vendor trees. Jan 28 04:19:11 and yah, that normally means graphics, some enhanced power management, drm, daughter cards, peripherals, etc. Jan 28 04:19:37 I'm playing with a beaglebone black at the moment, not sure which features I will really need yet Jan 28 04:20:15 less so now though, than say 10 years ago. Since most platforms pickup standard peripherals, etc. that leaves closed source blobs for the most part. Jan 28 04:20:38 the beaglebone is largely upstream, just probably some of the graphics you were mentioning, I haven't check recently myself. Jan 28 04:21:52 it feels weird to use the machine provided by the manufacturer (meta-ti) but not its kernel though, hum... Jan 28 04:23:09 that said I've already switched from uboot-ti-staging to poky's u-boot recipe for extlinux.conf support Jan 28 04:23:53 if it's building and booting fine with their tree, I'd stick with that, but really, it just depends if you want a different kernel version, etc. As you've seen the build system has that concept separation so you can make that choice. Jan 28 04:25:07 zeddii: yep I just needed to enable CONFIG_OVERLAYFS in the kernel, then I found the KERNEL_FEATURES thing, but this... :) Jan 28 04:26:12 :D it's good to check out all the things, learning a bit in each diversion. Jan 28 04:26:43 I'm checking out for the night, back at it tomorrow. cheers. Jan 28 04:26:55 o/ Jan 28 05:56:56 What commercial offerings are available in Europe for off-the-shelf Yocto based distributions with 5+ years security support? MontaVista and Wind River? Anything else? Jan 28 07:24:49 Hi, I'm using extended sdk and deployed my application, executable is copied, but not dependent library's. How to deploy dependent library's too? Jan 28 07:26:17 Hello, I have to 2 recipes with opencv dependency. How can i have 2 opencv "config file". Currently I have only one bbappend but I can get only one configuration... any idea ? what i have miss ? Jan 28 07:40:33 hi, tried several ways to override SERIAL_CONSOLES for imx8qxpmek machine (dunfell). I need to change ttyAMA0 to ttyLP0 in inittab. None of the methods i tried worked. What is the proper way to do this ? Jan 28 07:47:01 yo dudX Jan 28 07:47:33 did i miss the memo saying that KERNEL_IMAGETYPE can only have one value? or am i misinterpreting something. Jan 28 08:36:08 which steps should I do to create a custom image? I made a copy of recipes-core/images/core-image-minimal.bb with a different name and added a line with IMAGE_INSTALL += "openssh" but openssh don't appears in the rootfs Jan 28 08:38:14 droman: you need IMAGE_FEATURES_append = " ssh-server-openssh" Jan 28 08:42:01 mckoan: thanks, that will be cleaner for this package. But we will also need to install additional software like chrony, how we should proceed in that case? Jan 28 09:08:39 droman: use IMAGE_INSTALL_append Jan 28 09:11:33 that's what I did. So I guess I'm missing something because it didn't worked as documentation says Jan 28 09:29:17 ad__: give your attempts first so we can spot obvious mistakes before thinking too much :) Jan 28 09:30:03 (also, run bitbake -e | grep -e "^SERIAL_CONSOLE" and let us know if it is the content you expect it to be) Jan 28 09:31:32 droman: bitbake -e | grep -e "^IMAGE_INSTALL=" to the rescue Jan 28 09:32:16 if it does not have what you want in it, then run the command again but pipe it to less or a file and then look for the lines above the one starting with IMAGE_INSTALL= and read its "history" to investigate Jan 28 09:34:26 qschulz, SERIAL_CONSOLES uses a "=" in imx8qxpmek, actually i solved with a sed in a sysvinit-inittab_%.bbappend do_install_append Jan 28 09:34:49 hope is a proper solution Jan 28 09:46:38 ad__: proper way would be to create your own machine based on imx8qxpmek Jan 28 09:47:00 qschulz, thanks a lot. Jan 28 09:47:35 can try that Jan 28 09:50:19 ad__: FYI, you can "require" other machine configuration files Jan 28 09:58:10 qschulz: thanks Jan 28 10:01:17 Dear community, I do have a question regarding Dunfell's gitsm:// Jan 28 10:01:43 I've encountered a strange issue - I do have a project, which uses submodules Jan 28 10:02:03 which (as a git project) also is using submodules :-) Jan 28 10:02:36 Is gitsm fetches supporting fetching submodules for submodules ? Jan 28 10:52:25 morning all Jan 28 10:54:47 Hi all, someone here knows the simplest way to have 2 opencv  configuration file ( or other package of course) and how to switch according board or recipe ? Jan 28 10:55:13 I have the following recipe which works and compiles both the program and the library: https://pastebin.com/Hz4cXbx0 Jan 28 10:55:23 NiniC0c0: when does the opencv configuration apply? at build time for the opencv recipe? Jan 28 10:56:09 intera91: install -m 0777 ${WORKDIR}/build/lib/mylibrary.so ${D}${libdir} is incorrect Jan 28 10:56:29 qschulz currently I have set EXTRA_OECMAKE_append inside a bbappend but it's probably not the best idea Jan 28 10:56:41 NiniC0c0: that is not the question I asked :) Jan 28 10:57:13 intera91: you need to version your library Jan 28 10:58:11 qschulz So I'm not sure to understand the question :) Jan 28 10:58:37 now my problem is that the library ends up in the -dev rpm and QA complains that the executable requires the library: requires mylibrary.so()(64bit), but no providers found in RDEPENDS_myprogram? [file-rdeps] Jan 28 10:58:51 NiniC0c0: this configuration file, it applies to what? opencv or the recipe depending on opencv? Jan 28 10:59:02 intera91: version your library Jan 28 10:59:10 intera91: library.so.1.0 or whatever Jan 28 10:59:24 and not a symlink, the actual library Jan 28 10:59:41 you can create a library.so if you want, that'll land in -dev package Jan 28 10:59:48 library is versionned and both mylibrary.so.1.0.0 and a link called mylibrary.so are copied in build/lib Jan 28 11:00:00 intera91: no Jan 28 11:00:17 do_install only installs library.so Jan 28 11:00:28 in the recipe you snet Jan 28 11:00:46 if you added an install target to your cmake, just remove the do_install Jan 28 11:01:15 so remove the install step in cmake then Jan 28 11:01:17 NiniC0c0: I'm trying to understand if you want to have a different opencv per recipe that depends on it (impossible) Jan 28 11:01:33 intera91: does not matter, your do_install in the recipe overrides the one in your cmake Jan 28 11:01:54 ok removing the do_install and trying to build again Jan 28 11:01:56 thanks# Jan 28 11:02:13 currently I have only one opencv_%.bbappend to configure opencv recipe. app-test1.bb and app-test2.bb has some different opencv dependencies but opencv_%.bbappend is parsed all the time. what i have miss, seems simple :S Jan 28 11:02:51 NiniC0c0: app-test1 and app-test2 are supposed to be installed in the same image and for the same machine? Jan 28 11:04:19 intera91: there's a lot of implicit/hidden stuff going on in Yocto. When you inherit cmake, a default do_install task is created for you which calls the install target from your cmake Jan 28 11:04:29 when you set do_install, you override this implicit do_install Jan 28 11:04:41 so usually, you want to do do_install_append if you're missing something Jan 28 11:04:42 qschulz nop app-test1 is for image1.bb and app-test2 is for image2.bb Jan 28 11:04:50 NiniC0c0: ok, but same machine? Jan 28 11:05:19 qschulz nop Jan 28 11:05:41 qschulz it to different app but each need different opencv configuration Jan 28 11:05:59 NiniC0c0: that's an issue Jan 28 11:06:06 because recipes cannot modify other recipes Jan 28 11:06:12 only configruation files can Jan 28 11:06:20 so... either two machien configuration files Jan 28 11:06:32 or two distro configuration files Jan 28 11:06:41 in your case, two distros seem to make more sense Jan 28 11:07:14 NiniC0c0: however, in some cases, you could hack around stuff, I wouldn't recommend but it could be a nice work-around if it's the only thing that needs to be differnet Jan 28 11:07:30 because having two machine or two distros induces a lot of build time overhead Jan 28 11:07:47 so, the work-around would be to have two opencv recipes, named differently Jan 28 11:07:58 one with the conf file you want for app-test1, one for app-test2 Jan 28 11:08:16 in app-test1, you would depend on opencv-test1, app-test2 on opencv-test2 Jan 28 11:08:24 qschulz interesting... I will think about that . Thx you for your time Jan 28 11:08:32 this works if opencv isn't used by other recipes Jan 28 11:08:46 otherwise it starts to be a nightmare :) (having to duplicate all those recipes too) Jan 28 11:21:36 qschulz thx for all details! Appreciate that Jan 28 11:26:42 is it possible to fetch these informations json formatted? https://wiki.yoctoproject.org/wiki/Stable_branch_maintenance Jan 28 11:27:02 from a trusted and maintained source Jan 28 11:28:05 chris_ber: what do you want to do with it :) ? Jan 28 11:29:14 we have progress,  the error is now: QA Issue: myprogram: Files/directories were installed but not shipped in any package: Please set FILES such that these items are packaged. Alternatively if they are unneeded, avoid installing them or delete them within do_install. Jan 28 11:29:14 myprogram: 3 installed and not shipped files. [installed-vs-shipped] Jan 28 11:30:39 inspection of produced rpms still show no library in the main rpm Jan 28 11:31:10 the -dev contains the library still Jan 28 11:33:24 intera91: remove your FILES_${PN} Jan 28 11:34:11 so the cmake install must install in build/usr/bin and build/usr/lib? Jan 28 11:34:30 or is that ok to have build/lib and build/bin? Jan 28 11:35:52 removed FILES_${PN} line rebuilding Jan 28 11:36:34 Issue: /bin/myprogram contained in package myprogram requires mylibrary.so()(64bit), but no providers found in RDEPENDS_rinicom? [file-rdeps] Jan 28 11:38:03 qschulz: we implemented a toolchain to organize our dependencies (apps, libs). One feature is, that it can show us with graphviz dot that lib-xy-v0.2.3 dependes on lib-abc-v3.2.1. In addition it shows e.g. that for lib-abc also the version v3.2.2, v3.3.0, etc are available. Now i want to expand it to yocto. Since yocto/openembedded not using branches instead of tags i need something to rely on. Jan 28 11:39:21 difference is now the -dev rpm contains no file Jan 28 11:41:16 *yocto are using branches instead of tags Jan 28 11:42:52 with yocto, apps and libs it is easy to get lost in versions Jan 28 11:52:49 chris_ber: i don't understand what you are actually looking for Jan 28 11:54:07 simply a source where i can find the the yocto project versions and the corresponding branches taht i can rely on. Jan 28 11:54:24 which Jan 28 11:54:39 and whats wrong with the git repo? Jan 28 11:55:01 there are branches and there are tags. bot even in numerical and codename format, AFAIK Jan 28 11:56:03 LetoThe2nd: in the git repo i can see the branches, but i can't order them by the yocto version. Tags? i can't see tags, but i will check again Jan 28 11:56:49 chris_ber: https://git.yoctoproject.org/cgit/cgit.cgi/poky/refs/ shows a lot of tags. Jan 28 12:03:21 ok, now i see what went wrong. I didn't look into the git repo of ycto. Instead i looked into the git repo of openembbed and openembedded is not using tags Jan 28 12:05:37 chris_ber: https://github.com/openembedded/openembedded-core/tags Jan 28 12:05:42 want one more try? ;-) Jan 28 12:08:49 i feel a bit stupid. maybe i have to get a new identity in this channel :D But there is one exception: i started to look here: https://github.com/openembedded/meta-openembedded <- in this git repo, there are no tags. Jan 28 12:10:37 chris_ber: hehe yes. thats correct. the key point its that meta-openembedded essentially is a community-based best-effort offer, which branches match the state of the same-named branches in openembedded-core Jan 28 12:11:27 chris_ber: so as in contrast to oe-core and poky, meta-openembedded does not have coordinated states/releases. it just is there, and matches as good as possible. Jan 28 12:15:14 how do you manage this repo by yourself? Sometimes changes in meta-openembedded could lead to a big rebuild and this could/is very annoying. Therefor we changed to a commit hash, but this is ... well you know, not the very best option. Jan 28 12:16:15 chris_ber: it depends on the use case. we're not afraid of rebuilds, hence not much of a concern in the way you have it. Jan 28 12:18:00 allright, thanks for you time. I will think about. Jan 28 12:18:07 it Jan 28 12:18:13 :) have fun Jan 28 13:08:30 intera91: /usr/lib is where the libs should be Jan 28 13:10:18 chris_ber: otherwise, https://layers.openembedded.org/layerindex/branch/master/layers/ might of help too Jan 28 13:45:53 halstead: Any update on being able to add my reproduciblility summary page to the website? Jan 28 13:48:23 I just do not understand...if I do bitbake -c cleanall -f , and then do bitbake , how can a file in build/tmp/.. be missing ? Isn't that sequence of commands supposed to regenerate all files that are missing? Jan 28 13:48:34 or should I just wipe tmp completely? Jan 28 13:49:09 chrysh: You only clean the parts of tmp that belongs to the recipe. Not any other recipe. Jan 28 13:49:16 chrysh: It depends.... if the file missing is part of the image recipe then that *should* work. If the file is part of another recipes it might still be missing Jan 28 13:49:25 chrysh: What file is missing? Jan 28 13:49:35 chrysh: a) whenever you think you need -f, you're in trouble anyways. b) wiping tmp is no problem, if you don't "accidentially" wip sstate too. Jan 28 13:49:39 yea, but if I clean A and rebuild A, shouldn't that involve all the recipes? Jan 28 13:50:29 chrysh: define "all the recipes" Jan 28 13:51:01 if you clean an image, but nothing that the image consist of has changed, then only the image should be rebuilt. Jan 28 13:51:03 I might have removed it by accident, it is crtfastmath.o in sysroots-components/../libgcc/.. Jan 28 13:51:19 LetoThe2nd: -f is really useful when using externalsrc "bitbake -fc compile my-recipe" (or even better "bitbake -C compile my-recipe" ;) Jan 28 13:51:53 chrysh: You need to "-c clean" the recipe that provides that file then Jan 28 13:52:12 Then it will rebuild that recipe, and in turn the image Jan 28 13:52:26 JPEW: TBH i've never needed that. devtool modify myrecipe, then bitbake myrecipe always happily rebuilds. thats at least my experience Jan 28 13:53:32 LetoThe2nd: Ya, I don't use devtool as much as I should; I have a lot of sandboxes outside of Yocto I have a deal with anyway so it's easier just to point externalsrc at them anyway. Also, I still have to use 2.2 so devtool behaviors can be... inconsistent Jan 28 13:53:35 but what would be the way to go if I wanted to start a build form zero really. nuke tmp/? Jan 28 13:54:20 chrysh: Yes. If you have a sstate cache (outside of tmp/ !) it should rebuild fairly quickly Jan 28 13:54:39 chrysh: assembling from zero: nuke /tmp. really start from zero: nuke /tmp and sstate :) Jan 28 13:55:29 chrysh: recipes cannot impact other recipes. So when you run the cleanall task of the image recipe, it'll clean everything for the image recipe. But the image recipe is just a recipe Jan 28 13:55:41 so all other recipes won't be impacted Jan 28 13:58:31 thanks all! that solves a big misunderstanding about bitbake I had for a long time! Jan 28 14:05:36 that recipes can affect other recipes? ;-) Jan 28 14:11:00 hi I'm Joseph Jan 28 14:11:55 usr8392123: o/ Jan 28 14:12:16 nice to meet you q Jan 28 14:13:03 LetoThe2nd: that recipe dependency does not apply to clean Jan 28 14:13:30 just curious, how many developers are there official working for yocto? I want to know the company's size Jan 28 14:13:56 with make, any changed file is recompiled. I don't have a good understanding yet how the mechanism is for bitbake Jan 28 14:14:38 In my head it makes sense that when the dependency graph can be generated when building an image, it should also be able to do that when cleaning, but what do I know Jan 28 14:16:13 chrysh I hope bitbake uses hashes... Jan 28 14:16:30 usr8392123: you hope? Jan 28 14:16:49 yup, make doesn't uses hashes, so if you put a space it will recompile it... Jan 28 14:17:06 don't know about how bitbake detects it though, I have to still check it Jan 28 14:17:07 LetoThe2nd: What happens if i nuke only one of the two, either only sstate_cache or only tmp/? Jan 28 14:17:48 chrysh you'll get arrested, I hope Jan 28 14:17:54 :) Jan 28 14:18:28 bitbake does not look at the individual files of a package unless you are using externalsrc Jan 28 14:20:31 olani[m] depends on the type of build Jan 28 14:22:35 usr8392123: yocto is no company, hence nobody is officially working for it ;-) Jan 28 14:25:02 LetoThe2nd: well isn't Richard technically paid to work on it by the LF? Jan 28 14:25:42 he's contracting. Jan 28 14:25:42 chrysh: what if I want to clean everything for an image recipe but not the package recipes? Jan 28 14:26:07 but contracting != being part of a company, and the question was explicitly: " I want to know the company's size" Jan 28 14:26:22 LetoThe2nd: indeed :) Jan 28 14:26:35 qschulz: yea, what then? Jan 28 14:27:04 chrysh: well the cleanall of the image recipe would clean all my package recipe Jan 28 14:27:07 s Jan 28 14:27:17 ok Jan 28 14:27:36 in the end, everything is "just" task dependencies Jan 28 14:27:50 when you bitbake , IIRC something:do_build is called Jan 28 14:28:02 this task depends on others Jan 28 14:28:24 and DEPENDS for example is more or less just a task dependency from recipe A to recipe B Jan 28 14:28:44 so I have to go through the dependencies by hand? Jan 28 14:28:47 when you request cleanall task for a recipe, you will execute the whole dependency tree of that task too Jan 28 14:28:57 however, there aren't any :) Jan 28 14:29:42 chrysh: if you want to remove all from an image and the recipes it needed to build but not for any other image/machine/package recipe, then... yeah Jan 28 14:29:58 otherwise like others said, remove the sstate-cache, deploydir and tmp Jan 28 14:30:23 which is basically a different way to do everything from scratch (git clone poky and all) Jan 28 14:30:32 found this in a bb recipe, but when is it being called? python () { ... Jan 28 14:30:41 lxc: at parsing time Jan 28 14:30:47 it's an anonymous python function Jan 28 14:31:08 lxc: https://docs.yoctoproject.org/bitbake/bitbake-user-manual/bitbake-user-manual-metadata.html?highlight=anonymous#anonymous-python-functions Jan 28 14:32:08 qschulz thx Jan 28 14:32:39 chrysh: the mechanism for rebuilding on Yocto is "simple"... My understanding so far is the following: each task has a hash of its content. I'm pretty sure adding a space there or a comment would trigger a re-execution of that task Jan 28 14:32:52 anonymous functions are not found in listtasks? can they be executed with bitbake -c? Jan 28 14:33:07 any task that needs to be re-executed, will trigger a re-execution of tasks that depends on former task Jan 28 14:33:18 lxc: they are executed at parse time, so... always Jan 28 14:34:02 chrysh: this can involve other recipes as DEPENDS and RDEPENDS use cross-recipes task dependencies Jan 28 14:34:49 chrysh: Yocto also keeps a list of variables that are used in a task (after they are expanded as much as Yocto can?) so that any change will trigger a re-execution of the task Jan 28 14:35:04 this can be looked into with bitbake-dumpsig -t Jan 28 14:35:22 LetoThe2nd cool!=D Jan 28 14:35:40 for files in SRC_URI, ONLY if it starts with file:// (or for externalsrc, but different thing) will they be watched over Jan 28 14:36:00 and they are watched over by storing a map of "filename": hash of the file Jan 28 14:36:29 and will be put as "dependencies" of the do_fetch or unpack task, don';t remember Jan 28 14:38:30 LetoThe2nd Wait a minute, if it is not a company then who collects the fees/money and who pays the people that are working on it (not including the people who do it for fun)? Never mind, not my business 8)  just happy there exists yocto on this planet Jan 28 14:39:03 usr8392123: do you pay to have Yocto? Jan 28 14:39:05 qschulz: Ok, so I will add spaces to files when I want to recompile :P Jan 28 14:39:12 chrysh: noooooooooo Jan 28 14:39:38 For clarity, I meant the huge "Membership fees" https://www.yoctoproject.org/join/ Jan 28 14:40:06 qschulz Nope, and I like to keep it that way X-P Jan 28 14:40:33 :D I know it's a joke, but for completeness sake, this will actually work only if you change the task differently every time otherwise you'll reuse the sstate-cache of a "space-added" task later Jan 28 14:40:37 chrysh: ^ Jan 28 14:41:06 usr8392123: hehe, but thats not what you asked. Jan 28 14:41:17 usr8392123: well, one has to pay for servers hosting multiple YP services, at least RP is a contractor, so he kinda needs to be paid :) Jan 28 14:41:20 I know :) Jan 28 14:41:41 you got me there, noted :) Jan 28 14:41:43 usr8392123: there are at least one fulltime and one part time contributor being paid. theres a webmaster/admin, pplus the infrastructure.. Jan 28 14:42:12 usr8392123: but those are technically not part of any form of "yocto company" ... that was your wording :) Jan 28 14:42:35 hahaha indeed :') Jan 28 14:42:47 I should mind my wording in the future, you guys are quick Jan 28 14:42:48 usr8392123: free software isn't (usually) "free" :) Jan 28 14:43:10 :) Jan 28 14:43:10 at one point the project is so big it cannot be maintained on just free time Jan 28 14:43:35 (and not wanting to maintain a project on free time even if it's not big is also valid :) ) Jan 28 14:43:39 usr8392123: the project is hosted by the Linux Foundation so they collect the membership dues, buy infrastructure and pay people to do things as directed by the project members Jan 28 14:45:05 usr8392123: the testing infrastructure is quite vast in itself, needs admin work along with the other IT services, I'm funded by the project as is the LTS maintainer/work. Jan 28 14:45:18 hello guys ! Jan 28 14:46:09 In my opinion the engineers in general need to get paid more than the business persons who drive luxury cars etc.... but hey just my opinion Jan 28 14:46:20 RP: like is said, "there are at least one fulltime and one part time contributor being paid." :) Jan 28 14:46:24 qschulz: hello again Jan 28 14:46:25 so removed the do_install function and the FILES_$ line and we are back to yesterday error Jan 28 14:46:25 QA Issue: /bin/myprogram contained in package MYPROGRAN requires mylibrary.so()(64bit), but no providers found in RDEPENDS_rinicom? [file-rdeps] Jan 28 14:46:38 I've one question.. in the following bbclass Jan 28 14:46:39 https://github.com/Xilinx/poky/blob/rel-v2019.1/meta/classes/kernel-fitimage.bbclass Jan 28 14:46:39 the generated fitimage/s is/are copied in the DEPLOYDIR. Wat do I need to do to add them also in ${D}/fit folder ? Jan 28 14:46:48 LetoThe2nd: just agreeing :) Jan 28 14:47:44 intera91: are you linking against the correct library? Jan 28 14:48:03 cmake should link against the versioned library Jan 28 14:48:07 linking against the one in build/lib Jan 28 14:48:10 one sec Jan 28 14:48:15 I've tried adding Jan 28 14:48:15 kernel_do_install_append() Jan 28 14:48:16 but the task is run before the fitimages are compiled Jan 28 14:49:29 thekappe: https://github.com/Xilinx/poky/blob/rel-v2019.1/meta/classes/kernel-fitimage.bbclass#L502 Jan 28 14:50:34 intera91: I've checked again and /lib/ seems to be fine too (in addition to /usr/lib), so that's not an issue Jan 28 14:50:50 thekappe: the task is added between do_bundle_initramfs and do_deploy Jan 28 14:51:38 thanks for checking now I have to explore the  CMakeLists.txt to understand how the software links against the library Jan 28 14:51:59 intera91: remove the symlink first and check if it finds the versioned library Jan 28 14:52:10 at least you'll get that out of the way Jan 28 14:52:16 at present in build/bin/ ldd myprogram results in not finding mylibrary.so Jan 28 14:52:27 (and symlink is anyway by default only in -dev packages and those aren't instaleld byu default in the rootfs Jan 28 14:52:43 intera91: are you running ldd manually? Jan 28 14:52:58 yes from the command line just now Jan 28 14:53:12 well, you don't have the sysroot set for it so it's no surprise I'd say Jan 28 14:53:24 indeed Jan 28 14:53:25 intera91: bitbake -c devshell Jan 28 14:53:55 from the shell that pops up, you can use the same command as one would do in a recipe Jan 28 14:55:34 so for gcc, you would use $CC for example Jan 28 14:56:44 bitbake -c devshell does not work Jan 28 14:56:57 qschulz, the strange thing is that from log.task_order Jan 28 14:57:04 maybe cause am working in toaster Jan 28 14:57:21 do_install is run before do_bundle_fitimage and do_deploy Jan 28 14:57:43 Is Nicolas Dechesne here? Jan 28 14:58:02 matthewcroughan: ndec Jan 28 14:58:06 sweet Jan 28 14:58:23 This is not strictly yocto related, but I've been having trouble with sphinx and #sphinx-doc is dead. Jan 28 14:58:25 https://youtu.be/5Hjd2tNd6hI Jan 28 14:58:50 The yocto docs moved from DocBook to Sphinx, and I'm very curious how you managed variables and substitutions, because I cannot figure it out. Jan 28 14:58:51 matthewcroughan: what's the issue? Jan 28 14:58:54 ndec: how much are you gonna pay me so i'll testify that you are *not* around? ;-) Jan 28 14:59:14 "this is not the ndec that you are looking for!" Jan 28 14:59:34 matthewcroughan: poky.yaml defines variables that needs to be substituted Jan 28 14:59:43 in what? The docs? Jan 28 14:59:58 LetoThe2nd: who are you talking about? ;) Jan 28 15:00:00 --> do_install (22688): log.do_install.22688 Jan 28 15:00:01 do_populate_sysroot (23115): log.do_populate_sysroot.23115 Jan 28 15:00:01 then you have the :ref:``, :term:`` etc mechanism Jan 28 15:00:01 do_package (23114): log.do_package.23114 Jan 28 15:00:02 do_packagedata (23327): log.do_packagedata.23327 Jan 28 15:00:02 do_package_write_rpm (23352): log.do_package_write_rpm.23352 Jan 28 15:00:03 do_package_qa (23353): log.do_package_qa.23353 Jan 28 15:00:03 do_bundle_initramfs (23557): log.do_bundle_initramfs.23557 Jan 28 15:00:04 -->do_assemble_fitimage_initramfs (23559): log.do_assemble_fitimage_initramfs.23559 Jan 28 15:00:04 do_deploy (23600): log.do_deploy.23600 Jan 28 15:00:10 thekappe: please use pastebins Jan 28 15:00:26 qschulz: that doesn't let you do if statements. Jan 28 15:00:32 ndec: heh, just joking. somebody asked for you concerning sphinx. Jan 28 15:00:43 matthewcroughan: you have to be more specific, what is your question? Jan 28 15:00:45 Like "if arm64 is set as a var, then .. include: |VARIABLE|/thing.rst" Jan 28 15:00:51 qschulz, yap sorry, but it's blocked Jan 28 15:01:11 thekappe: pastebin is not *one* website :) Jan 28 15:01:15 qschulz: Check this out. https://github.com/NordicHPC/sphinx_ext_substitution/issues/1#issuecomment-768306419 Jan 28 15:01:18 there's probably one you can find Jan 28 15:01:44 I've defined exactly what I've done below this comment in Sphinx, but that is not seemingly possible, and I find it hard to believe you guys can live without this sort of feature. Jan 28 15:02:11 matthewcroughan: we decided not to use 'built in' variables substitutions. and implemented something fairly custom, see Jan 28 15:02:11 http://git.yoctoproject.org/cgit/cgit.cgi/yocto-docs/commit/documentation/sphinx/yocto-vars.py?id=dc5f53fae8fdfdda04285869dd1419107b920bfe Jan 28 15:02:14 I cannot `.. include:: |ARCH|-qemu-command.rst`. So RST substitution isn't enough. Jan 28 15:03:01 ndec: you know, I found these two things that might interest you. First, using Jinja inside the .rst file: https://www.ericholscher.com/blog/2016/jul/25/integrating-jinja-rst-sphinx/ Jan 28 15:03:33 Secondly, mkdocs, supports this stuff out of the box. https://squidfunk.github.io/mkdocs-material/reference/variables/ Jan 28 15:03:39 rigth. there are too many limitations with the |ARCH| notation.. Jan 28 15:04:06 we use variables a lot in code snippets, and here too it was not possible to use them. Jan 28 15:04:07 indeed, I'm reading your commit message and can relate to it so well Jan 28 15:04:16 hence the custom extension.. Jan 28 15:05:53 ndec: Your custom extension is pretty much exactly the same as this one https://pypi.org/project/sphinx-ext-substitution/ Jan 28 15:06:07 I cannot see how you could CONDITIONALLY include documents using poky.yaml Jan 28 15:06:33 I need the &ARCH variable to be different depending on what file is importing it. Jan 28 15:07:02 arm64.rst does .. include:: generic-template.template Jan 28 15:07:22 generic-template.template has inside of it "Use &ARCH for this" Jan 28 15:07:42 arm64.rst needs to redefine the &ARCH variable Jan 28 15:09:26 FYI, https://www.yoctoproject.org -> docs -> reference manual is a 404. (in the docs dropdown). https://docs.yoctoproject.org/ref-manual/ref-manual.html Jan 28 15:09:42 submit a bug? Jan 28 15:09:47 not sure how the web process works Jan 28 15:11:10 mcfrisk: first indications are about 6mins improvement of master-next on 70mins so around 10% speedup :) Jan 28 15:11:58 kergoth: bug would be good thanks Jan 28 15:12:01 ndec: Is what I described allowed for by yocto-vars.py ? "subst_vars_replace runs once per file" Jan 28 15:12:23 RP: nice Jan 28 15:14:05 Hahaha, yes! This happens even if you use Jinja and Jinja extensions in Sphinx/. "+ #FIXME: if poky.yaml changes, files are not reprocessed." Jan 28 15:14:54 It's just too bad, to me it seems like Sphinx has flaws that'll never be fixed, because nobody wants to do the work, due to its age and complexity. Jan 28 15:16:22 kergoth: thanks, it's a bug. i think we should simply remove the link. please open a bug! Jan 28 15:17:39 ndec: is it fair to say that yocto-vars.py is a pre-processor? Jan 28 15:17:48 So I could just do the same thing with sed in a makefile? Jan 28 15:22:03 matthewcroughan: https://git.yoctoproject.org/cgit/cgit.cgi/yocto-docs/commit/documentation/sphinx/yocto-vars.py?id=59537c7fa49e3ea6918f45b3201ad16d56988b9b Jan 28 15:22:07 that's what we did Jan 28 15:23:04 matthewcroughan: and I'm pissed because html_css_files and html_js_files Jan 28 15:23:09 also aren't watched over Jan 28 15:23:48 MkDocs sorted all this out. Ans it even rebuilds in milliseconds, not minutes or seconds. Jan 28 15:24:36 You've shown me that I need to find a better way <3. That said, you have a treasure trove of sphinx usage in the commits. Thanks so much. Jan 28 15:24:38 matthewcroughan: yes, it's a pre-processor. i don't think we are ready to migrate the docs again now ;) Jan 28 15:25:07 ndec: considering that we might not even have finished fixing the last bits, yes, please no migration again :D Jan 28 15:26:08 I don't want to moan. But I cannot believe how hard it is to inject data into these files in the "proper" way. It's very complex. I thought it was supposed to be a 'framework', i.e something that makes life easier and less complex. Jan 28 15:26:09 i am , for sure, not doing a new migration! Jan 28 15:27:01 You are the only Sphinx wizard I've found that is still active online. Jan 28 15:27:41 isn't the kernel using Sphinx as well? I'd expect there to be some active folks there Jan 28 15:28:40 smurray: that was one of the reasons to use Sphinx IIRC Jan 28 15:28:51 qschulz: right Jan 28 15:40:10 I've added a new task to be sure that is run: after do_assemble_fitimage_initramfs before do_deploy Jan 28 15:40:58 Now I can see the fitimage in kernel-build-dir/image/fit/ Jan 28 15:41:32 by the way the fit/ folder is not added to the "rootfs" folder of the image to be built Jan 28 15:42:08 weird Jan 28 15:42:44 also, I have rebuilt the image after a cleansstate, so I am sure that the rootfs has been refreshed Jan 28 15:44:47 thekappe: because files have to be added between do_install and do_packages_split Jan 28 15:45:52 do_package sorry Jan 28 15:46:43 try to replace do_deploy with do_package in your task dependency Jan 28 15:46:54 and pray it's not going to introduce a circular dependency Jan 28 15:49:46 qschulz, uff.. Jan 28 15:51:56 what's make me smiling is that the files are currently installed also in ${D}/boot, and I need them in ${D}/fit Jan 28 16:06:44 thekappe: that might be easy to do in do_install_append Jan 28 16:07:36 but... probably need to create a /fit directory and I don't really know if it's easy to have a new directory at root of the rootfs Jan 28 16:07:48 thekappe: but... why in /fit specifically? Jan 28 16:07:53 qschulz, nope because I need also the initramfs fitimage, but that one is built after do_bundle_initramfs Jan 28 16:08:04 /boot is pretty much standard? Jan 28 16:08:15 qschulz, don't ask me Jan 28 16:08:17 thekappe: well, you can move everything except the one you don't have Jan 28 16:08:43 someone seems to like /fit much more Jan 28 16:09:14 the point is that a need to build a .wic image with multiple partitions Jan 28 16:09:27 in the /fit partition I need to put the fitimages Jan 28 16:09:35 thekappe: why didn't you say that in the first place? Jan 28 16:09:51 I'm pretty sure you don't need to have the fitimage in the rootfs to have a partition in wic for that Jan 28 16:10:06 don't quote me on that but please investigate before spending too much time on it :) Jan 28 16:10:06 I'm also trying to do that Jan 28 16:10:22 specifically, that's probably done for raspberrypi? Jan 28 16:10:36 or any system using u-boot and wic for that matter? Jan 28 16:10:37 part /fit --source rootfs --ondisk mmcblk --fstype=ext4 --label opt --align 4 --fixed-size 500 --rootfs-dir=${WKS_FIT} Jan 28 16:11:00 (I haven't used wic in ages so can't help with that one) Jan 28 16:11:03 also I'm setting WKS_FIT in my conf file Jan 28 16:11:28 Check this out ndec :D Jan 28 16:11:29 https://github.com/fralau/mkdocs_macros_plugin#defining-variables Jan 28 16:12:27 WKS_FIT = "${DEPLOYDIR}/my_fitimage Jan 28 16:14:22 but it seems that DEPLOYDIR can't be resolved Jan 28 16:14:56 "Couldn't get bitbake variable from /my_fitimage.env" Jan 28 16:17:37 JPEW: can I merge the patch in meta-mingw master-next btw? Jan 28 16:21:10 thekappe: DEPLOY_DIR_IMAGE https://docs.yoctoproject.org/ref-manual/variables.html#term-DEPLOY_DIR_IMAGE Jan 28 16:21:17 thekappe: DEPLOYDIR is for within a recipe Jan 28 16:26:07 qschulz, that's good to know Jan 28 16:26:11 thanks man Jan 28 16:26:35 RP: seems fine to merge Jan 28 16:28:49 JPEW: thanks, it fixes master-next builds Jan 28 16:29:05 Real data on the performance change of maser-next: https://autobuilder.yocto.io/pub/non-release/20210128-9/testresults/buildperf-ubuntu1604/perf-ubuntu1604_master-next_20210128130346_0ec9fb3f98.html Jan 28 16:29:38 3-4mins on a 70 min build on that machine Jan 28 16:30:31 not quite as good as the other numbers I saw but I'll take it :) Jan 28 16:31:01 7% disk usage reduction too Jan 28 16:32:01 mcfrisk: ^^^ Jan 28 16:35:40 and https://autobuilder.yocto.io/pub/non-release/20210128-10/testresults/buildperf-centos7/perf-centos7.yoctoproject.org_master-next_20210128130350_0ec9fb3f98.html says 7% speedup, 5mins on 53mins Jan 28 16:38:38 vmeson: not quite the 10% I was seeing from rough numbers but I'll take 7% :) Jan 28 16:44:35 matthewcroughan: pretty cool! jinja templates in docs! we need that for sphinx :) Jan 28 16:45:01 ndec: you can do it in sphinx, but it's impossible! I can't figure out how to get pagename into it, if you know what I mean? Jan 28 16:49:57 ndec: https://www.sphinx-doc.org/en/master/templating.html#pagename Jan 28 16:50:46 I can't get it to be like {% if pagename == arm64.rst %}, because it'll always fail with something like Handler for event 'source-read' threw an exception (exception: 'pathto' is undefined) Jan 28 16:50:59 any ideas? How does it even know to call that `pathto` function? Jan 28 17:11:57 matthewcroughan: never looked into that! I will check it out. Jan 28 17:20:16 rsalveti: o/ Jan 28 17:57:59 JPEW: https://autobuilder.yoctoproject.org/typhoon/#/builders/89/builds/2994 is going to be coming on the radar with autoconf 2.70 Jan 28 18:15:35 RP: Noted: error: AM_GNU_GETTEXT is used, but not AM_GNU_GETTEXT_VERSION or AM_GNU_GETTEXT_REQUIRE_VERSION Jan 28 18:15:47 Hopefully an easy fix Jan 28 18:41:43 I've run bitbake linux-ti-staging -c cleanall; and now it's stuck at: Jan 28 18:41:46 NOTE: recipe linux-ti-staging-5.4.74+gitAUTOINC+9574bba32a-r22a: task do_fetch: Started Jan 28 18:50:22 JPEW, You should have just received a Wordpress account invitation. Jan 28 18:54:26 halstead: Got it. Thanks Jan 28 18:55:47 JPEW, Can you make use of https://developer.wordpress.org/rest-api/reference/pages/#retrieve-a-page and https://developer.wordpress.org/rest-api/reference/pages/#update-a-page from you script? Jan 28 18:58:20 Oh, Hmm.... Were you thinking the autobuilder would generate a page with results and update the wordpress page? Jan 28 19:07:57 halstead: I'm very familiar with wordpress.... how do I get the ID of a page? Jan 28 19:08:41 JPEW, Yes I was thinking whatever produced the results could automatically update the page. Jan 28 19:09:31 JPEW, I made a new page at https://www.yoctoproject.org/wp-admin/post.php?post=22223&action=edit and I think the ID is 22223. Jan 28 19:09:32 halstead: Ah, I was thinking I could pull the latest testresult.json file and parse it in javascript, so the page would pretty much be static Jan 28 19:09:46 JPEW, That works for me. Jan 28 19:13:20 JPEW, In that case we need to put the javascript in a hook that can be called on page load. I don't want to add the ability to embed javascript dynamically. Do you already have some JS written you can provide? Jan 28 19:13:38 as usual, rm -rf build/tmp seemed to fix my issue. Jan 28 19:39:43 is there not any recipe for https://github.com/bookwar/python-gpiodev ? Jan 28 19:55:54 halstead: Yes... let me post it somewhere and I'll get back to you Jan 28 19:56:13 JPEW, Okay. I need to travel some today so I will be slow with responses. Jan 28 20:25:13 is it possible to remove a packagegroup coming from inherited image ? Jan 28 20:26:06 Sure, just VARIABLE_remove = "packagegroup" whatever variable adds it to your image. Jan 28 20:33:25 oh thanks. it seems to be in MACHINE_EXTRA_RRECOMMENDS Jan 28 20:35:35 Yeah, then MACHINE_EXTRA_RRECOMMENDS_remove = "packagegroup". Jan 28 20:35:51 Or outright overwrite MACHINE_EXTRA_RRECOMMENDS Jan 28 20:35:56 neverpanic, thanks, setting it in mymachine.conf, ok ? Jan 28 20:37:14 Err, that likely won't work. I was assuming you are editing the image recipe directly, or at least bbappending it in your layer. Jan 28 20:37:52 neverpanic, it seems to have worked, i am in do_rootfs now Jan 28 20:38:09 mymachine.conf could work if you use one of the values in $OVERRIDES, e.g. MACHINE_EXTRA_RRECOMMENDS_yourmachine = "new-value", or MACHINE_EXTRA_RRECOMMENDS_remove_yourmachine = "packgegroup" Jan 28 20:38:21 let see if it get angry Jan 28 20:38:25 Well, yes, it'll work, but it will affect all images; I assumed you only wanted it to affect a specific image? Jan 28 20:39:11 well, packagegroup is related to wifi, and the machine/board doesn't have it, so maybe machine conf is approrpiate Jan 28 20:39:29 yes. btw it worked. Jan 28 20:39:49 thanks a lot ! Jan 28 21:32:54 I'm adding a .cfg file to the SRC_URI of my kernel recipe, stored in a folder prepended to FILESEXTRAPATHS, but it isn't applied. Any clue? Jan 28 21:34:36 the file contains CONFIG_OVERLAY_FS=y but my kernel doesn't show overlay in /proc/filesystems Jan 28 21:35:04 if I enter the menuconfig and set the option manually, it works. What's wrong? Jan 28 21:35:25 depends on if your kernel recipe inherit's linux-yocto. Jan 28 21:35:51 otherwise, the config is defconfig based, or whatever scheme the provider has used Jan 28 21:36:48 zeddii: it's linux-ti-staging, and I think they are using a custom thing (it contains use-tisdk-config=ti_sdk_am3x_release) Jan 28 21:38:25 I can look quickly at their recipe. which version ? I see a few in their master branch. Jan 28 21:42:30 ahah. well, I see they are doing their own merge_config, so you'd be best off asking on one of the ti lists (I don't know which ones they are), but my quick glance says look into the variable KERNEL_CONFIG_FRAGMENTS Jan 28 21:43:30 and a note in their function: Jan 28 21:43:33 # Fourth, handle config fragments specified in the recipe Jan 28 21:43:33 # The assumption is that the config fragment will be specified with the absolute path. Jan 28 21:43:33 # E.g. ${WORKDIR}/config1.cfg or ${S}/config2.cfg Jan 28 21:43:47 I have to go, but that should get you moving. Jan 28 21:44:27 thank you zeddii! Jan 28 21:57:59 Yocto best practices question: say I am using someone else's layer to build a BSP for their platform (e.g. Xilinx's zynqmp-zcu102).  Now I want to customize it (add software, drivers, etc).  Seems ideal to make a new layer with my changes, and then still build exactly the same way (e.g. MACHINE=zcu102-zynqmp bitbake petalinux-image-minimal) Jan 28 21:58:11 My issue is, how do I now build the original zynqmp-zcu102 BSP without my changes?  Do I have to remove the layer, possibly clean everything, and rebuild?  Is there any better way to allow switching back and forth? Jan 28 22:03:14 Ooops...  Lost my connection...  If anyone had any suggestions please repost as I didn't see it.  (probably not, just wanted to make sure) Jan 28 22:24:06 whata's the website with docs and courses about embebed systems? Jan 28 22:24:20 someone like bootin? Jan 28 22:24:39 ooh, it's bootlin 😁 Jan 28 22:47:07 indeed, bootlin Jan 28 22:47:43 * abelloni puts his sales guy cap on Jan 28 22:48:04 wyre: are you looking for a specific course ? :) Jan 28 22:57:09 RP: rburton https://autobuilder.yoctoproject.org/typhoon/#/builders/88/builds/890/steps/12/logs/stdio Jan 28 22:57:21 this is with autotools 2.70 Jan 28 23:02:39 khem: I was just looking at that, I'm not really surprised. oe-core isn't clean yet either but closer Jan 28 23:03:35 khem: fixes for apr, lrzsz, at, ruby, msmtp are in -next now Jan 28 23:12:44 khem: heh, 2.7.1 is out now Jan 28 23:12:57 er, 2.71 **** ENDING LOGGING AT Fri Jan 29 03:02:12 2021