**** BEGIN LOGGING AT Tue Jan 19 03:00:19 2021 Jan 19 06:41:54 wayland recipe has GPLv3 deps in RDEPENDS_ptest, and I have blacklisted that license. I'm not interested in wayland ptest package so what is the correct way to disable it? I have ptest in DISTRO_FEATURES but I only want ptest in my selected packages that I include separately, so I don' Jan 19 06:42:13 * don't have ptest-pkgs in EXTRA_IMAGE_FEATURES Jan 19 06:47:00 so AFAIK I'm not asking bitbake to build wayland-ptest, but it still checks the rdepends-ptest and gives error about incompatible license Jan 19 07:49:40 yo dudX Jan 19 07:56:03 'morning Jan 19 08:19:56 Hi all, one of my custom machine is inherited from an other, KBUILD_DEFCONFIG in machine conf. I want to override it but if I place defconfig inside SCR_URI I get following issue :  [NOTE]: defconfig was supplied both via KBUILD_DEFCONFIG and SRC_URI. Dropping SRC_URI defconfig. So how can i override properly KBUILD_DEFCONFIG  ? Jan 19 08:52:34 v0n: kernel drivers are either built as modules or built-in, there is no such thing as builtin kernel modules. If you want drivers to be built in, you need to modify your kernel defconfig Jan 19 08:52:53 you can't have out-of-tree drivers builtin AFAIK (but I've very limited experience with oot driver) Jan 19 08:53:14 v0n: also wrt your two rootfs thingy... what about just bundling the initramfs inside the kernel directly? Jan 19 08:53:28 no need to care about wic having to handle two rootfs in two partitions Jan 19 10:03:56 Strange issu after migrating to Gatesgarth: FileNotFoundError: [Errno 2] No such file or directory: '/path2builddir/tmp/deploy/licenses/sc15-hw-image/recipeinfo' in fact reciping Jan 19 10:04:13 is located inside /path2builddir/tmp/deploy/licenses/sc15-hw-image/-TIMESTAMP Jan 19 10:25:34 to add a command to devtool, what is needed so it also gets added to the esdk? if i drop the .py in scripts/lib/devtool/whatever.py and have a register_commands method, it works nicely in a local environment. but the script isn't being picked up and packed upon esdk generation. Jan 19 11:08:09 and another one: how to remove a specific directory from the esdk? is there some form of POSTPROCESS hook for it too? Jan 19 11:22:20 rburton, abelloni: master built green: https://autobuilder.yoctoproject.org/typhoon/#/builders/113/builds/631 :/ Jan 19 11:23:04 rburton: guessing intermittent race :( Jan 19 11:42:05 yes, edk2 is the bane of my life Jan 19 11:48:29 abelloni: that pipe thing is a makefile race, its running " /bin/sh ../../bash-5.1/builtins/psize.sh > pipesize.h" twice and the two commands race Jan 19 11:48:56 ah! Jan 19 11:49:21 * RP isn't sure he wants to debug make races in bash's makefiles :/ Jan 19 11:49:47 maybe we can have a look to see if this was introduced with v5.1 Jan 19 11:57:11 abelloni: its in 5.0 but only one makefile reference to it so no race. Two references in 5.1 Jan 19 11:57:13 abelloni: looks like a makefile bug to me Jan 19 12:34:01 Hi! I have a question about the build/ directory. Does it have to be put under source control? Jan 19 12:35:13 chrysh: it shouldn't Jan 19 12:35:28 would you version your binaries or rather your sources? Jan 19 12:37:22 chrysh: but there's probably a more precise question that brought you here I feel like :) Jan 19 12:37:49 chrysh: what would you like to improve/work-around by having the build directory under source control? Jan 19 12:44:00 qschulz: I have seen management insist on that before :/ Jan 19 12:44:48 RP: what for? because of the build/conf dir? Jan 19 12:45:04 qschulz: because they want to be sure it can be reproduced/fixed later Jan 19 12:45:11 otherwise, if they don't change the directory of sstate-cache and downloads directory, that's one of the stupidest idea ever Jan 19 12:45:23 the path* directories* Jan 19 12:45:29 qschulz: I didn't say I agreed with it, I just have seen it before Jan 19 12:45:41 qschulz: entire TMPDIR checked into clearcase Jan 19 12:45:53 * qschulz screams Jan 19 12:46:17 qschulz: I apologise for the nightmares Jan 19 12:46:42 Hi quentin! :) I just don;t understand why the instructions to add a layer tell you to edit conf/bblayers.con Jan 19 12:46:45 f Jan 19 12:46:59 which would make your change volatile Jan 19 12:47:14 where do you specify then if you want to add a layer permanently? Jan 19 12:47:58 chrysh: didn't recognize your nick until now, sorry :D but hi o/ Jan 19 12:48:52 chrysh: that is up to you. kas allows you to do this I think. Otherwise, you can just install the file yourself with a script. Or this script could call bitbake-layers add-layer for example Jan 19 12:49:12 but yes, the build directory isn't meant to be shared, it's a build directory afterall Jan 19 12:50:55 (FWIW, we install local.conf and bblayers.conf from our versioned git repo) Jan 19 12:51:06 into build? Jan 19 12:52:21 yeah, well... we have our own script we source that is using oe-init-buildenv, we do some crazy stuff in there and then we cp the local.conf and bblayers.conf into the build/ directory Jan 19 12:56:40 and that's the recommended way to do things with yocto? Jan 19 12:58:02 or is there a comprehensible documentation how you as a manufacturer get a bsp from your vendor and what you have to do then? Jan 19 12:58:45 RP: you happen to know if there's a form of POSTPROCESS command, or filtering mechanism for layers being copied into the esdk? Jan 19 12:58:50 I would assume your own meta layer in order to keep your changes grouped somehow. but the descriptions I found where all more in detail, and there was no high level view on how to get started Jan 19 12:59:26 chrysh: look at the live coding video on kas. :) Jan 19 13:02:24 I have some trouble to create a recipe for 'package pynq 2.6) inherited from pip/setuptools3,RDEPENDS_${PN} += "${PYTHON_PN}-numpy" is set but i got ModuleNotFoundError: No module named 'numpy' any clue to debug that ? thx! Jan 19 13:02:32 sorry for double post Jan 19 13:03:28 LetoThe2nd: I'm not sure there is any postprocess like that. It would be a bit dangerous as the sstate hashes would be very likely to change so it wouldn't be encouraged Jan 19 13:05:39 chrysh: I wouldn't say our way of doing it is the recommended way, I'm not sure there is an officially recommended way honestly. But kas should allow you to set this all up with a config file you can then version control Jan 19 13:07:05 RP: hmm. I'm seeing the week picking up a lot of undesired stuff when a layer includes a directory with helpers and such that is meant only for system development Jan 19 13:09:29 s/week/week/ Jan 19 13:09:38 meh. *esdk Jan 19 13:19:48 only possible to change existing recipes using .bbappend or are there other methods (such as directly changing the recipe itself which is probably not the correct way I assume)? Jan 19 13:30:57 LetoThe2nd: ah, SDK_LAYERS_EXCLUDE ? Jan 19 13:32:00 RP: but that excludes a complete layer, right? Jan 19 13:32:08 LetoThe2nd: correct Jan 19 13:32:22 LetoThe2nd: code is in meta/lib/oe/copy_buildsystem.py Jan 19 13:32:23 @RP then why include it all? Jan 19 13:32:52 * LetoThe2nd will probably just hard patch populate_sdk_ext.bbclass Jan 19 13:33:27 Respectively copy_buildsystem Jan 19 13:49:37 ok got it Jan 19 14:32:33 qschulz: hi -- I meant builtin *drivers* Jan 19 14:33:46 coming from Buildroot, I prefer a separated initramfs image, it makes it easier for development and tftpboot to copy only the kernel or the rootfs image on my server depending on what I'm working Jan 19 14:35:34 but since it seems more complicated to enable builtin drivers in the kernel and having to deal with 2 rootfs images rather than building an initramfs inside the kernel and having another rootfs and wic image, I might go the yocto way... Jan 19 14:40:36 Hi, I'm struggling with how native recipes work. I have a python script "foo.py" which I want to install as a host build tool in a recipe "foo-native.py". This recipe inherits "native" and installs to ${D}${bindir}. Now I want to use this script in another recipes do_compile task. I searched the web for minimal examples but nothing worked for me. Jan 19 14:41:01 Can someone tell me how this works or where I can find an easy example of this mechanism in action? Jan 19 14:45:28 v0n: "more complicated to enable builtin drivers"? Jan 19 14:46:05 dleppich: first, the recipe should be called foo-native.bb Jan 19 14:46:32 dleppich: then just add foo-native to your script recipe's DEPENDS? Jan 19 14:47:02 then just call foo directly since it'll be in the PATH variable for this task Jan 19 14:47:33 v0n: how to build drivers in the kernel aren't specific to Yocto but the linux kernel so not sure to follow? Jan 19 14:48:06 dleppich: If you do that, you also have to add "inherit native" in the recipe: https://docs.yoctoproject.org/ref-manual/classes.html#native-bbclass Jan 19 14:48:20 qschulz: I mean the kernel recipes (linux-ti-staging for example) seems to have drivers set as modules by default so using built-in drivers require a custom defconfig. No big deal but it looks like "the yocto way" is to go with modules Jan 19 14:50:24 qschulz: I speaking in terms of changes required to achieve what you want. modules + builtin initramfs seems straight forward in yocto compared to builtin drivers + separated initramfs image (I might be wrong, hence this discussion). Jan 19 14:52:09 qschulz sp creatomg a recipe kungfu-native.bb and then adding to DEPENDS will automatically include that recipe? Jan 19 14:52:19 so creating Jan 19 14:53:06 v0n how do you enable a kernel module in yocto? Jan 19 14:54:04 qschulz: I tried it. This is my directory structure (http://ix.io/2Mzv), foo.py (http://ix.io/2Mzw), foo-native.bb (http://ix.io/2Mzy), test-foo.bb (http://ix.io/2MzB). When trying to bitbake test-foo I get the error: "foo.py: command not found" Jan 19 14:54:13 jeremyjohnsonthe: CONFIG_=m in the kernel config Jan 19 14:54:50 v0n thanks! Jan 19 14:58:17 dleppich: you are missing a DEPENDS = "python" in your foo-native (well if you're using a recent Yocto release, RDEPENDS_${PN} = "python" is better) Jan 19 14:58:18 also wondering if the path to python isn't /usr/bin/python? Jan 19 14:58:56 otherwise check in your test-foo WORKDIR/recipe-sysroot-native if foo.py is there Jan 19 15:00:50 v0n: it's not the "yocto way" it's just how ti wanted to do it by default. Yocto is a buildsystem, just modify its metadata to build whatever you want Jan 19 15:01:29 also... if you go for non0bunlded initramfs but depends on module... you better keep the kernel and initramfs partition in sync otherwise you'll be screwed at some point Jan 19 15:01:53 non-bundled* Jan 19 15:02:12 depend on modules in your initramfs image* Jan 19 15:07:28 qschulz: Damn, my example wasn't minimalistic enough. I tried the RDEPENDS but it did not work out. I changed the "foo.py" to "foo" containing an "echo" statement. Still no change. I checked the sysroot dir, it contains two executables "guards" and "quilt" but no "foo". Jan 19 15:07:52 I checked this directory: build/tmp/work/x86_64-linux/foo-native/1.0-r0/recipe-sysroot-native/usr/bin Jan 19 15:12:09 Okay, brainlag. I checked the wrong folder. The "recipe-sysroot-native" for test-foo contains some more files, but no "foo" Jan 19 15:27:40 qschulz: true, that is why I wanted to have a kernel image with builtin drivers, and a separated initramfs image. I just wanted to make sure I didn't shoot myself in the foot by trying to do this with yocto. Jan 19 15:33:02 v0n: you just need a custom defconfig, which is anyway probably something you want to do sooner or later in order to add drivers or slim down the kernel size (and compilatinon time) Jan 19 15:33:42 dleppich: that's definitely not good. Is `foo` a name used by any other recipe (be it native or not)? Jan 19 15:34:05 dleppich: check in build/tmp/work/x86_64-linux/foo-native/1.0-r0/image/usr/bin if foo.py is there Jan 19 15:34:38 dleppich: RDEPENDS for native recipes only work for a couple of releases, it was not supported before Jan 19 15:39:30 Unrelated question, what is the recommended name for image recipes? ${DISTRO}-image-${NAME}? Jan 19 15:40:14 v0n: does not matter. try to have `image` in the name so it's super obvious Jan 19 15:40:24 qschulz: you're right. I'll add my own defconfig and maybe a vanilla kernel depending on what's inside linux-ti-staging. Thanks. Jan 19 15:40:27 you can also put them in recipes-images if you want Jan 19 15:40:47 v0n: for which board are you using Yocto? Jan 19 15:40:58 qschulz: yeah this part is also not very strict, most of the time I see recipes-core/images Jan 19 15:41:12 v0n: yeah or this Jan 19 15:41:14 qschulz: a beaglebone black with a custom daughter board Jan 19 15:41:34 v0n: are you using blobs or closed source stuff from ti? Jan 19 15:41:45 (e.g. prob vpu, gpu?) Jan 19 15:42:18 if not, just ditch their layer and go for full upstream. The BBB is well supported in both linux and u-boot Jan 19 15:42:19 not for the moment, but I might Jan 19 15:42:47 otherwise, you might want to check if those usually painful drivers are supported upstream decently or not Jan 19 15:43:11 but custom daughter board means custom dts to and probably new drivers Jan 19 15:43:54 then if not too far from ti's kernel, you can add dts and defconfig manually, otherwise just clone their repo and use your own git repo for fetching the sources Jan 19 15:44:08 qschulz: hum it makes me think about this, if I do need prob vpu and ti stuffs, do I need u-boot-ti-staging as well on only linux-ti-staging? (I already switched to u-boot for the extlinux support before thinking about this) Jan 19 15:45:33 v0n: nah, if ti did their job correctly, linux kernel and bootloaders aren't setting things up for the other Jan 19 15:46:07 (it's a principle of Linux kernel development upstream, do not trust any state of the HW before the Linux kernel is booted, so you need to initialize everything to known default values) Jan 19 15:47:28 as always for vendor BSPs, they might have patches they haven't sent upstream yet. Quality is variable and sometimes they break stuff while fixing other stuff Jan 19 15:47:30 qschulz: I know the linux philosophy but some nasty products rely on the bootloader passing modified dtb or blobs sometimes ;-) Jan 19 15:54:52 v0n: u-boot (even upstream) patches the dtb before sending it to the kernel so yeah Jan 19 15:55:04 v0n: very well aware of RPi and the GPU used a bootloader thingy too :) Jan 19 15:55:41 v0n: and it's not the product relying on the bootloader passing stuff, it's the vendor BSP doing that, up to you to reimplement it better if you want Jan 19 15:59:59 qschulz: yeah I prefer using vanilla bootloaders and kernel if I can, but I don't want to spend time on it if that isn't necessary. Jan 19 16:00:06 * v0n checks linux-ti-staging Jan 19 16:03:33 Hello guys, is the team technical meeting open for anyone interested in Yocto or for a certain group? I am asking because I found that the zoom meeting has a password Jan 19 16:04:07 Guest14991: https://zoom.us/j/990892712?pwd=cHU1MjhoM2x6ck81bkcrYjRrcmJsUT09 Jan 19 16:04:17 The link should include the password as it's a public meeting Jan 19 16:04:41 thanks! Jan 19 16:07:19 great to see Bootlin stepping up! :-D Jan 19 16:07:26 abelloni: ^^ awesome Jan 19 16:24:41 https://reproducible-builds.org/citests/ Jan 19 16:26:37 Guest14991: only passworded to stop zoombombing, anyone can join Jan 19 18:22:33 Hi, i'm trying to add re2 package from github.com/google/re2.git Jan 19 18:22:36 But i got the following ERROR: re2-2020.11.01-r0 do_package_qa: QA Issue: -dev package contains non-symlink .so: re2-dev path '/work/aarch64-poky-linux/re2/2020.11.01-r0/packages-split/re2-dev/usr/lib/libre2.so' [dev-elf] Jan 19 18:22:52 is is because the recipe is called re2 and the so libre2.so ? Jan 19 18:50:25 here is my libre2_2020.11.01.bb https://pastebin.com/1isJ6bxR Jan 19 19:04:06 hello. where is the essential point where bitbake tells the compiler where the target system headers are (like stdint.h) ? Jan 19 19:04:42 I am having a problem where a custom 3rd party makefile fails to find these Jan 19 20:40:41 RP: The data is all in the test results. I have an example HTML page that displays them; just a matter of making it look pretty and publishing it Jan 19 20:40:57 Is there a URL that always points to the "latest" reproducible results? Jan 19 20:45:32 JPEW: in theory the yocto-testresults repo is where they go. I'm having trouble accessing that though :( Jan 19 20:49:23 is this the issue in PDX ? Jan 19 20:49:56 I'm not sure how much infra YP shares with kernel.org Jan 19 22:40:37 Happy new year.  Trying to understand PACKAGECONFIG[...]. The docs are excellent - thank you. Jan 19 22:41:07 Then I came across this (in a QT5 recipe): Jan 19 22:41:29 `PACKAGECONFIG[xcb] = "-xcb -xcb-xlib -system-xcb,-no-xcb,libxcb xcb-util-wm xcb-util-image xcb-util-keysyms xcb-util-renderutil"` Jan 19 22:41:57 abelloni: its more that our testresults repo is huge :/ Jan 19 22:42:31 JPEW: can you use http://git.yoctoproject.org/cgit/cgit.cgi/yocto-testresults/plain/oeselftest/testresults.json ? Jan 19 22:42:55 Not at all sure how to interpret that given the documentation description of PACKAGECONFIG. Jan 19 22:45:56 First the use of space and comma delimiters is a puzzle.  The doc suggest, but don't state delimiter is a comma.  What does the space delimiter indicate? Jan 19 22:46:36 spaces are used to delimiante lists of items.. Jan 19 22:46:48 commas are used to delimate parts of a list.. Jan 19 22:47:02 i.e. PACKAGECONFIG = "foo bar foobar" has three elements.. foo, bar and foobar Jan 19 22:47:48 PACKAGECONFIG[foo] = "param1,param2,param3,param4" is a PACKAGECONFIG vairable that explains what 'foo' means, and it's specific parameters are 'comma' deliminated, because some parameters might have spaces in them (list of things) Jan 19 22:49:42 Thanks Frey. Okay, so `-xcb -xcb-xlib -system-xcb` is passed through to QT5 (in this case), and is not meaningful to yocto/bitbake? Jan 19 22:50:32 context matters in this case.. it's probably a lsit of arguments to something in that case Jan 19 22:51:33 Agreed, but its *not* a list of arguments to bitbake? Jan 19 22:51:51 no, I wouldn't assume so. Jan 19 22:52:20 aaargh ;) Jan 19 22:52:22 bitbake is similar to make, cmake, etc. WHere it simply handles global and 'local' namespace for variables, dependencies for tasks and executes tasks Jan 19 22:52:55 variables specific in local.conf (and others) are globally defined. variables from a recipe (.bb / .bbappend) and local to that recipe.. Jan 19 22:58:31 is meta-initramfs-image from the meta/openembedded/meta-initramfs a good start or is there better choice? Jan 19 23:00:06 Okay, so in that example is it still true that bitbake will only add to `DEPENDS` the list of items `libxcb xcb-util-wm xcb-util-image xcb-util-keysyms xcb-util-renderutil`, and in this case these are recipe names? Jan 19 23:08:37 Final puzzle is (also from QT5 recipe): `PACKAGECONFIG[freetype] = "-system-freetype,-qt-freetype,freetype"` Jan 19 23:10:32 This, and the previous example, together suggests the use of `--with` and `--without` are conventions and the meaning is positional?  I'm pretty sure that is correct, but its a big oversight in the docs so want to be sure. Jan 19 23:14:00 Also it seems that there is a short form for `--with-foo, --without-foo`: `-foo,-no-foo`. But if meaning is positional then something like this would be legal: Jan 19 23:15:16 `PACKAGECONFIG[foo] = ",,freetype,,,conflictfoo"` Jan 19 23:21:55 Just to be clear the last example would be used when trying to add a dependency for the `--without`/`-no` case. Jan 19 23:22:28 So more precisely: Jan 19 23:23:37 `PACKAGECONFIG[nofoo] = ",,nonefoo-pkg,,,conflict-with-nonefoo-pkg"` Jan 19 23:25:14 `PACKAGECONFIG[foo] = "-foo,-no-foo,foo-dep-pkg,,,conflict-with-foo-pkg"` Jan 19 23:26:38 * jonesv[m] sent a long message: < https://matrix.org/_matrix/media/r0/download/matrix.org/afSMMGzWGTxhLmsoFzqRkUcQ/message.txt > Jan 19 23:27:34 * jonesv[m] sent a long message: < https://matrix.org/_matrix/media/r0/download/matrix.org/mirrxLhgMLmCiNnVIFHTeUzT/message.txt > Jan 19 23:27:59 My device is running an apq8009, which I believe is an armv7 processor (hence my guess above) Jan 19 23:28:34 Any suggestion on what could be wrong? Even just keywords would help, I've been reading docs and DuckDuckGoing for a while, and I'm stuck Jan 19 23:30:10 https://docs.yoctoproject.org/ref-manual/variables.html#term-PACKAGECONFIG Jan 19 23:32:58 kanavin_home: I've tweaked your bash upgrade not to need the bbclass change, hope that is ok Jan 20 01:48:17 Hmm, I've hit an error with `base_do_unpack` when consuming a git repo.  Specifically it does not pass the `--origin origin` option to clone and then it assumes the remote name is origin - it is `clone.defaultRemoteName`.  Somehow setting that git config option to origin, with `--global` or `--system` does not workaround the issue. Jan 20 01:49:15 How best to override the method/function `base_do_unpack` in such a recipe? Jan 20 01:58:24 Okay its not that simple.... can anyone tell me where the git commands are prepared for `base_do_unpack` maybe I can hack in the `--origin origin` fix. **** ENDING LOGGING AT Wed Jan 20 02:59:57 2021