**** BEGIN LOGGING AT Wed Apr 22 02:59:57 2020
Apr 22 07:18:07 <mcfrisk> whoaa, 3.1 dunfell LTS is out! Great! Need to get my projects updated to it and chip in with the maintenance...
Apr 22 07:19:00 <interruptguy> Morning. Is there a way to re-run the Yocto `oe-selftest` without deleting the `build-st` directory? I checked the docs but there is nothing bout this. I didn't check the code yet...I thought that asking here is the first easy choice. Any help appreciated!
Apr 22 08:09:29 <yocti> New news from stackoverflow: How to add pre-built package to image in Yocto? <https://stackoverflow.com/questions/61359734/how-to-add-pre-built-package-to-image-in-yocto>
Apr 22 11:55:57 <riz46> I am noticing that whenever I run bitbake -k <image>, I am not getting a new image built in my "/build/tmp-glibc/deploy/images/intel-corei7-64/" directory. I first noticed it when I would make changes and copy the changed image to my target and no changes would occur. So I deleted the old image to see if a new one was built and sure enough nothing
Apr 22 11:55:58 <riz46> new was placed in the directory.
Apr 22 11:58:42 <kroon> riz46, why -k ?
Apr 22 12:01:34 <JaMa> riz46: are you sure that the changes you did really influence the content of the image? if it doesn't need to generate new one, then it doesn't
Apr 22 12:17:33 <riz46> JaMa Yes, I removed splash screens and did noticeable stuff
Apr 22 12:17:53 <riz46> kroon Just so it continues to build in case there are errors
Apr 22 12:21:55 <smurray> riz46: note that if there did happen to be an error, you won't get an image even with -k
Apr 22 12:22:17 <riz46> smurray There was no error
Apr 22 12:22:37 <riz46> It is just odd. I have built and deployed many images before. It just started doing this a week ago
Apr 22 12:27:41 <paulbarker> riz46: Could you post the bitbake command and the full output via a pastebin?
Apr 22 12:31:17 <riz46> https://pastebin.com/qgYEEvjB
Apr 22 12:31:59 <riz46> I actually originally noticed it because I went from nodistro to a custom distro and the terminal still displayed nodistro while booting
Apr 22 12:33:31 <riz46> That bitbake was wit no changes to the image. I just deleted the old image to see if a new one would be placed int he deploy folder and nothing was actually made
Apr 22 12:35:01 <paulbarker> riz46: If you delete contents from the deploy directory bitbake won't notice that
Apr 22 12:35:25 <paulbarker> You can try `bitbake <image> -c clean && bitbake <image>`
Apr 22 12:35:54 <riz46> OK, let me try that
Apr 22 12:37:17 <paulbarker> riz46: Also, the bitbake version should be 1.44.0 with zeus, make sure you're not mixing stable branches of layers with the master branch of bitbake
Apr 22 12:37:50 <riz46> What command can I use to check the bitbake version?
Apr 22 12:38:15 <JaMa> it's shown in your pastebin
Apr 22 12:38:24 <paulbarker> riz46: It's shown as `BB_VERSION` in the bitbake output
Apr 22 12:38:49 <riz46> Ahh, it is 1.46.0
Apr 22 12:39:01 <JaMa> "noticeable stuff" bitbake doesn't agree with that statement
Apr 22 12:39:16 <riz46> And I did the cleanall and build and it didn't provide an updated image
Apr 22 12:40:14 <riz46> And am I right by looking at the tmp-glibc folder in build as opposed to just tmp?
Apr 22 12:41:20 <JaMa> depends on your TMPDIR variable
Apr 22 12:45:08 <riz46> @JaMa That is set to just "tmp", but I noticed that I still got an output in tmp-glibc. In fact the one in tmp was 1/3rd the size and didnt work but the one in tmp-glibc seemed to boot fine. I'm not really sure of the difference.
Apr 22 12:45:17 <riz46> JaMa
Apr 22 12:46:41 <JaMa> "I went from nodistro to a custom distro" which possibly changed the TMPDIR and probably other things as well
Apr 22 12:47:54 <JaMa> and as we don't know what your custom distro sets, it's hard to guess why your image doesn't work anymore
Apr 22 12:48:49 <riz46> JaMa But I thought the TMPDIR is set in local.conf...
Apr 22 12:50:48 <JaMa> meta/conf/distro/defaultsetup.conf:TMPDIR .= "${TCLIBCAPPEND}"
Apr 22 12:51:25 <riz46> oh OK. Thanks!
Apr 22 12:55:04 <rburton> JaMa: i'd argue for removing that line because its just confusing and serves no purpose
Apr 22 12:56:10 <JaMa> except for people/CI which test multiple LIBCs in the same build directory, but otherwise I agree
Apr 22 12:56:35 <riz46> Where is TCLIBCAPPEND set?
Apr 22 12:56:53 <JaMa> in oe-core
Apr 22 13:38:22 <rburton> JaMa: they can easily add it back if needed
Apr 22 13:56:02 <darwi> Hi. What's tha best way to detect the kernel version from a recipe that does **not** inherit from kernel.bbclass?
Apr 22 14:06:30 <darwi> LINUX_VERSION works only on kernel recipes
Apr 22 14:07:06 <smurray> darwi: what's the context that your recipe needs to know it for?
Apr 22 14:08:15 <darwi> smurray, remove wireless-regdb-static package from packagegroup-base-wifi RDEPENDS if linux version <= 4.14
Apr 22 14:09:44 <smurray> darwi: there's not really a way to do it that comes to mind.  You can't get at per-recipe variables like that from another recipe
Apr 22 14:10:44 <yocti> New news from stackoverflow: Core image minimal fails <https://stackoverflow.com/questions/61367173/core-image-minimal-fails>
Apr 22 14:12:35 <smurray> darwi: I've seen things like that statically managed in machine conf files with the MACHINE_ESSENTIAL variables, but if you want to make it more automatic I don't have a good idea atm
Apr 22 14:19:41 <nacknick> Hey. I created a recipe that create a library under `/lib`, added its package to local.conf (IMAGE_INSTALL_append), the package was built successfully, but in the final image, I can't find the library under `/lib`. Any idea why?
Apr 22 14:20:57 <JaMa> is it static .a or dynamic, but without version .so ?
Apr 22 14:21:23 <nacknick> I can see the library file under "work/.../image/lib"
Apr 22 14:21:27 <nacknick> .so file
Apr 22 14:23:29 <JaMa> then it's probably in PN-dev package, not PN which you probably installed
Apr 22 14:24:12 <JaMa> normally .so should be just symlink to the actual versioned library, that's why it's packaged in PN-dev by default, and it should show you QA warning or error when it's not symlink
Apr 22 14:29:26 <nacknick> JaMa: How to fix it?
Apr 22 14:30:17 <JaMa> nacknick: https://stackoverflow.com/questions/7553184/how-to-do-versioning-of-a-shared-library-in-linux
Apr 22 14:32:15 <darwi> smurray, perfect. Thanks a lot
Apr 22 14:35:06 <nacknick> JaMa: Thanks for the link, but in my case, we don't build that share library "normally". The recipe is downloading that *.so file from a remote server and copy it as is to `/lib` dir. How to fix it in that case?
Apr 22 14:35:07 <rburton> nacknick: you should have got a qa warning
Apr 22 14:35:28 <nemgti-og> I have a local file that I want to use in my recipe and that file cannot be found in github or anywhere else. Can I use devtool to create a new recipe that makes use of this file? Can devtool create an initial recipe that places suh file on the FILES subfolder of the recipe folder and make it calculate the checksum?
Apr 22 14:35:32 <rburton> nacknick: https://wiki.yoctoproject.org/wiki/TipsAndTricks/Packaging_Prebuilt_Libraries#Non-versioned_Libraries
Apr 22 14:36:18 <JaMa> nacknick: or grep FILES_SOLIBSDEV to find the examples of packaging of bad libraries
Apr 22 15:10:57 <yocti> New news from stackoverflow: Bitbake/Yocto simple script tasks <https://stackoverflow.com/questions/61368015/bitbake-yocto-simple-script-tasks>
Apr 22 16:01:56 <nameclash> what is the preferred way to exclude a recipe from a populate_sdk task?
Apr 22 16:03:45 <nameclash> I have an application that's in IMAGE_INSTALL and when I run bitbake -c populate_sdk, I can see that the application gets compiled, i.e. do_compile is executed
Apr 22 16:04:23 <nameclash> how do I avoid that?
Apr 22 16:05:10 <nameclash> sry, bitbake -c populate_sdk core-image-minimal
Apr 22 16:09:40 <nameclash> in the case of an application recipe that inherits cmake, the app even lands in the SDK
Apr 22 16:11:23 <nameclash> I could separate the sdk recipes from the image recipes and run different bitbake targets, e.g. bitbake core-image-minimal vs. bitbake -c populate_sdk core-sdk
Apr 22 16:12:20 <nameclash> but what about recipes that provide both image and sdk artifacts? there is a more elegant way to solve this, right?
Apr 22 16:13:21 <neverpanic> Isn't this the entire point of the SDK, to have the headers and debug symbols for the libraries and applications you deploy in the image?
Apr 22 16:13:55 <ingonus> Hi guys, i'm trying to write some words with special characters like "ã" in yocto bash via ssh, but bash write a "#" and break line instead. If I open vim or tmux, this characters can be written with no problem.
Apr 22 16:14:27 <nameclash> I dont think it's the point of an SDK to contain applications that play no role in SDK usage
Apr 22 16:14:31 <ingonus> I also set my locale to pt_BR.UTF8
Apr 22 16:14:41 <nameclash> which is the case for me
Apr 22 16:14:43 <ingonus> via export LANG and via localectl
Apr 22 16:16:37 <ingonus> nameclash, you can append the do_populate_sdk task changing what you want, is that what you mean?
Apr 22 16:17:36 <neverpanic> nameclash: That depends on what you want to do with the SDK. If you want to debug a coredump generated by that application, it does play a role in the SDK usage.
Apr 22 16:17:56 <ingonus> The SDK has an entire "chroot" of the image to use for cross-compiling stuffs, so it needs an image as base
Apr 22 16:17:58 <nameclash> ingonus: you mean by deleting the application from the sysroot again after it was installed in do_install?
Apr 22 16:18:29 <rburton> nameclash: you must have bad dependencies if the *application* ends up in the sdk
Apr 22 16:18:42 <ingonus> i think you can do that, but I don't know if the SDK will work without sysroot
Apr 22 16:18:52 <rburton> iirc
Apr 22 16:19:05 <nameclash> rburton: yes, this is what's happening
Apr 22 16:20:39 <nameclash> I have a cmake/qt application thats gets added to IMAGE_INSTALL from core-image-minimal.bbappend. When I run bitbake -c populate_sdk, the application ends up in the same location in the SDK as it does in the image rootfs
Apr 22 16:21:26 <nameclash> I just noticed that and it kinda smells to me..
Apr 22 16:24:02 <nameclash> neverpanic: what you say about coredump analysis is definitely a point, but I still wouldnt go so far to say an SDK should contain the entire rootfs of an image. This is not how I'd define an SDK
Apr 22 16:28:56 <ingonus> nameclash the SDK needs the entire rootfs as a toolchain to build binaries to target S.O. using libs and everything necessary to build all you have on your S.O.
Apr 22 16:30:26 <nameclash> ingonus: does it need my application to build binaries?
Apr 22 16:30:28 <neverpanic> ingonus: No it doesn't. Binaries of the rootfs aren't required to build other binaries, only libraries and headers are needed.
Apr 22 16:30:42 <nameclash> neverpanic: exactly
Apr 22 16:31:12 <nameclash> I suspect there's something misbehaving, as rburton indicated..
Apr 22 16:32:48 <ingonus> nerverpanic: yes, but the packages "-dev" depends on the main package witch makes the whole package and dev to be installed in the sysroot
Apr 22 16:34:12 <ingonus> I thought yocto just put the rootfs as sysroot in the SDK
Apr 22 16:40:00 <rburton> im doing a build myself to double check assumptions
Apr 22 16:40:03 <rburton> i thought /usr/bin was ripped out
Apr 22 16:41:30 <nameclash> rburton: the application is installed at /<AppName>/app/<AppName>
Apr 22 16:41:42 <nameclash> maybe that's the misconfiguration?
Apr 22 16:41:51 <rburton> quite likely
Apr 22 16:41:53 <ant__> JaMa, rburton it is indeed like this with nodistro, handy to test musl and glibc
Apr 22 16:42:03 <rburton> nameclash: is there anything in /usr/bin/ in the sdk sysroot?
Apr 22 16:42:43 <rburton> ant__: but with recipe-specific sysroots and so on, does that actually solve anything? also, if you wanted split tmpdir, easy to set tmpdir as such
Apr 22 16:43:09 <rburton> but we periodically do get questions from people wondering why tmp is 'suddenly' tmp-glibc because they changed distro
Apr 22 16:43:56 <ant__> actually I do set TMPDIR = "/tmp/build/tmp"
Apr 22 16:43:56 <nameclash> rburton: yeah /usr/bin is quite busy actually
Apr 22 16:49:55 <ant__> so I keep it in tmpfs
Apr 22 16:52:31 <neverpanic> nameclash: Ah, that explains things. Yocto expects you to stick to the filesystem hierarchy standard for your applications. Weird things will happen in multiple places if you don't.
Apr 22 16:54:41 <nameclash> I'll take that one with me as lesson learned. However still wondering why applications like openssl are also in /usr/bin ...
Apr 22 16:55:35 <rburton> my impression was the target sdk was pruned like we do for sysroots already
Apr 22 16:55:37 <rburton> but maybe not
Apr 22 16:56:43 <nameclash> rburton: it's rocko over here, to be fair
Apr 22 16:57:13 <nameclash> so maybe that's something that was improved in later releases?
Apr 22 18:32:45 <rburton> nameclash: yeah usr/bin is fairly full, i thought it was pruned
Apr 22 19:11:45 <yocti> New news from stackoverflow: Do not understand error with packagegroup <https://stackoverflow.com/questions/61372769/do-not-understand-error-with-packagegroup>
Apr 22 19:23:34 <shurelous> Hi guys, anyone knows how to set the default locale of a image? letting it with locale and keymap set to a specific language and layout?
Apr 22 19:24:12 <shurelous> I've tried to set IMAGE_LINGUAS, but it just let the locales available, it do not set the current locale
Apr 22 20:13:15 <nameclash> rburton: thanks
Apr 22 20:37:35 <nameclash> shurelous: maybe this will be of help: https://stackoverflow.com/a/48925706/13280907
Apr 22 20:40:09 <nameclash> though I'd rather bbappend base-files instead of creating a dedicated recipe for it
Apr 22 20:48:13 <njemb> Is there any way to remove the terminal boot messages from showing after the OS has booted?
Apr 22 21:14:55 <paulg> njemb, there isn't really anything yocto specific about that -  I'm guessing a google of "tty clear login"  will get you what you want.
Apr 22 21:15:47 <paulg> this is one of those "100 ways to skin a cat" things - though.
Apr 22 21:16:40 <njemb> paulg Yeah, I saw setting CONFIG_FRAMEBUFFER_CONSOLE=n as well as many other things
Apr 22 21:16:54 <njemb> I guess I was just wondering what the cleanest way would be
Apr 22 21:17:24 <paulg> there is also a boot arg for linux  "loglevel=<nnn>"
Apr 22 21:17:37 <paulg> see Documentation/admin-guide/kernel-parameters.txt
Apr 22 21:17:37 <njemb> I am also unsure what recipe to append for CONFIG_FRAMEBUFFER_CONSOLE
Apr 22 21:17:51 <njemb> Will do
Apr 22 21:17:56 <paulg> if you set it to the minimum, you'll get no boot messages.
Apr 22 21:18:31 <paulg> see also the man page for "dmesg" on your typical linux build host to set the context for what loglevel does.
Apr 22 21:19:12 <paulg> basically instead of controlling what the console device is, you are controlling what messages go there.
Apr 22 21:19:40 <njemb> Yup. Thanks
Apr 22 21:19:42 <paulg> This has no impact on pre-linux stuff, like u-boot or any other boot loader, of course.
Apr 22 21:20:33 <paulg> for that, you really do want to send a ^L  (control-L) to the console device if you want to clear off all the boot messages that are pre-Linux.
Apr 22 21:24:54 <njemb> Right
Apr 22 21:43:22 <yocto_newb> could any body share a hint about compiling with shared headers and g++ cant see to get yocto to find -lpthread -lbluetooth
Apr 22 22:23:14 <khem> describe your problem a bit more perhaps lot of recipes link with pthreads and other libraries
Apr 22 23:12:35 <yocti> New news from stackoverflow: Remove ROS to Reduce Image Size <https://stackoverflow.com/questions/61376427/remove-ros-to-reduce-image-size>
**** ENDING LOGGING AT Thu Apr 23 02:59:58 2020