**** BEGIN LOGGING AT Tue Sep 01 02:59:57 2020 Sep 01 04:01:33 mariadb-10.5.4-r0 do_install - 30m50s and still running wow! I have built full distros in 30mins Sep 01 04:33:10 yikes Sep 01 06:30:18 good morning Sep 01 06:34:26 good morning Sep 01 06:44:06 LetoThe2nd: good news are on the way (spoiler) Sep 01 06:45:10 mckoan: ? Sep 01 06:45:48 ah. saw the mail now. :) Sep 01 07:14:15 Hi there! Does anybody have experience with building nano for Poky-tiny? I run into configuration errors that are not present when I use the normal Poky, or my vendor-specific distro. Sep 01 07:14:42 miklosj: pastebin a log of the errors, please. Sep 01 07:17:13 This is the log of the do_configure step:https://pastebin.com/BFyWSqwp Sep 01 07:18:14 miklosj: missing cursesw5 on your host machine ? Sep 01 07:19:31 well nano has the ncurses DEPENDS properly set Sep 01 07:19:42 It is installed. "libncursesw5-dev is already the newest version (6.1-1ubuntu1.18.04)." Sep 01 07:20:19 nah, thats unrelated. i guess that this breaks it: http://git.yoctoproject.org/cgit/cgit.cgi/meta-yocto/tree/meta-poky/conf/distro/poky-tiny.conf?h=master#n54 Sep 01 07:20:59 probably nano checks for wchar support under the ncurses hood, which poky-tiny disables. Sep 01 07:23:07 Thanks! I was suspecting something in poky-tiny.conf, just wasn't aware of the exact setting. Sep 01 07:33:07 hello people of this world Sep 01 07:34:18 How can I update the applications on my yocto image? On ubuntu I can do apt update... Sep 01 07:35:17 hehe, funny nick :) Sep 01 07:35:33 letothethird: in a nutshell: not at all, it doesn't work like that. Sep 01 07:35:41 LetoThe2nd Thanks :) Sep 01 07:36:05 it has no package manager and I don't want to rerun my build Sep 01 07:36:32 well then, no update. Sep 01 07:37:17 and a package manager alone wouldn't save you too, as it would have no repositories to pull from. Sep 01 07:37:48 conclusion: in order to update everything I have to rerun my build and then somehow deploy it to the target Sep 01 07:37:57 yes. Sep 01 07:38:38 ok then I have another question, suppose I rerun my build, is there a way of wirelessly deploying this to the target (without sd card, supposing the target is behind a wall I can't access)? Sep 01 07:39:01 *technically* speaking you could set up a world build plus repository server, create an image including a package manager to pull from it, and then use that, but... well, thats more like a theoretical option. Sep 01 07:39:27 if your target has something in place that can accept updates wirelessly, then you can. if not, then not. Sep 01 07:39:34 this is not something yocto provides. Sep 01 07:40:01 totally clear, good that you're here(y) Sep 01 07:40:04 thanx Sep 01 07:40:10 (y) Sep 01 07:43:19 hello, how do I properly configure connman to use iwd instead of wpa-supplicant under Yocto? Sep 01 07:45:29 eduardas: probably needs quite some hammering at the recipe, the usage of wpa in case of wifi support is hardcoded at the moment: http://cgit.openembedded.org/openembedded-core/tree/meta/recipes-connectivity/connman/connman.inc?h=master#n40 Sep 01 07:45:33 as far as I understand connman has an iwd option in its configure.ac , but it is not exposed in the bitbake recipe via PACKAGECONFIG Sep 01 07:46:33 thats what i said in other words :) Sep 01 07:47:43 I just wanted to be sure that is the case and its not just me misreading the recipe. thanks for telling me Sep 01 07:48:20 so I assume no one is using iwd in production for Yocto? Sep 01 07:48:44 if someone could share their experience with it, would be nice Sep 01 07:49:48 I am currently having an issue with an uBlox LILY-W132 module where kernel spams ieee80211 phy1: mwifiex_cfg80211_sched_scan_start : Invalid Sched_scan parameters Sep 01 07:50:37 wpa-supplicant outputs mlan0: Failed to initiate sched scan Sep 01 07:51:04 wonder if this would be different with iwd Sep 01 07:51:31 Wi-fi actually works eventually, but this still feels wrong Sep 01 08:02:35 Hello together :-D, Sep 01 08:02:35 I want to build my java and go applications with yocto. Sep 01 08:02:35 The steps I did so far: Sep 01 08:02:35 * Follow the awesome youtube videos from theyoctojester Sep 01 08:02:35 * Create a dummy layer and build a c application Sep 01 08:02:35 Now I want to modify that layer so I can build my java application with gradle. Is this possible? If that can not be done, can I create a layer that takes a binary from my host machine (some local path) and bakes it into the final yocto image? Sep 01 08:02:35 Thank you for your help :D. Sep 01 08:17:05 jUnG3: many things are possible, its only software after all. yet, java has been known to be particularly painful Sep 01 08:18:15 jUnG3: basically your first starting point is this: http://git.yoctoproject.org/cgit/cgit.cgi/meta-java Sep 01 08:18:53 and yes, you can "bake in" things. the magical keywords to find the needed information is "externally produced binaries" in the dev-manual. Sep 01 08:20:27 Thx for the advice, I am gonna read the stuff I try my best :D Sep 01 08:20:34 Btw great videos :D Sep 01 08:24:04 Thanks! Sep 01 09:21:59 hmm.. PACKAGES =+ "${PN}-x" doesn't seem to add a package in bbappend... is that normal? Sep 01 09:22:56 ak77: check the evaluation order with bitbake -e, guessing that you need _append instead of =+ Sep 01 09:26:19 LetoThe2nd, thank you for reminding me of "-e", i need prepend (to capture FILES before main package does) - I was missing a space between packages! Sep 01 09:26:33 :) Sep 01 10:43:29 kanavin_home: are you prepping a patch bomb with the AUH results? Sep 01 10:43:37 I'll start collecting this week if not Sep 01 10:53:06 rburton: we're past feature freeze so we'll need to sort sooner than later... Sep 01 10:55:27 anyone building (oci) containers with oe ? Sep 01 10:56:04 testing if having "containers" as a machine (e.g. qemux86-64-container) is viable Sep 01 11:01:33 cengiz_io, Sep 01 11:01:38 nvm Sep 01 11:08:32 ak77: https://youtu.be/jPbcQEffzJo Sep 01 11:19:05 rburton: I am, but holding it for a couple of days to let others send those patches Sep 01 11:19:19 rburton: just want to avoid giving the impression that auh updates are all alex's problem Sep 01 11:19:46 anyway, the set is here as always http://git.yoctoproject.org/cgit/cgit.cgi/poky-contrib/log/?h=akanavin/package-version-updates Sep 01 11:19:57 ak77: thought that might be what your plan was Sep 01 11:31:12 rburton: you mean kanavin_home? Sep 01 11:32:33 erm, yes Sep 01 11:38:01 kanavin_home: thanks for filing the selftest issue. Was https://autobuilder.yoctoproject.org/typhoon/#/builders/110/builds/1277/steps/8/logs/step1c identified? Sep 01 11:44:26 RP: yes, I added a link to existing bug https://bugzilla.yoctoproject.org/show_bug.cgi?id=13646#c19 Sep 01 11:52:35 LetoThe2nd: I followed the instructions in the dev-manual for "Externally Produced Binaries". Now when I try to bitbake my layer I got this error message: "do_install: bin_package has nothing to install. Be sure the SRC_URI unpacks into S.". The ".tar.gz" file that is supposed to be extracted is placed in "my-layer/recipes-core/my-application/my-application/${BP}.tar.gz" Sep 01 11:52:58 LetoThe2nd: Have you some tips? Sep 01 11:54:08 jUnG3: if the tar.gz file contains a dir called foo/, then S should be set in the recipe to ${WORKDIR}/foo Sep 01 11:55:35 jUnG3: My guess is that S is pointing to an empty of non-existent directory, resulting in the error you see Sep 01 11:56:11 erbo: the "tar.gz" contains just "usr/bin/my-application" Sep 01 11:57:21 jUnG3: Then I think there's some parameter you can pass via SRC_URI to make it unpack into a subdir anyway. Let me google :) Sep 01 11:59:57 erbo: ty for the advice. There is a subdir parameter. It has to be set and then point S to it. Now the build runs fine Sep 01 12:04:50 jUnG3: right, subdir is the one, so appropriately named Sep 01 12:05:34 jUnG3: either S="${WORKDIR}" or SRC_URI="file://${BP}.tar.gz;subdir=${WORKDIR}/${BP}" (not so sure abvout the last one) Sep 01 12:05:47 ah, erbo beat me to it Sep 01 12:07:12 erbo: I have done it so SRC_URI = "file://${BP}.tar.gz;subdir=my-app" S = "${WORKDIR}/my-app" Sep 01 12:07:25 jUnG3: otherwise.. just follow implicit conventions and add a subdir in your tarball named the value of ${BP} Sep 01 12:08:00 jUnG3: I think that's the best approach, if you don't want to change the tarball as qschulz suggest Sep 01 12:08:19 That is also an cool option Sep 01 12:14:29 and now if I run "runqemu qemuarm64" I should be able to see the extracted binary inside the image? Sep 01 12:15:29 only if you added it to the image Sep 01 12:19:06 LetoThe2nd: I added the layer to build/conf/bblayers.conf. Is this enough? Sep 01 12:19:54 thats enough to build the recipe. :) Sep 01 12:20:54 jUnG3: but it won't go into any image unless the image recipe actually includes it. you might want to watch live coding #2 and/or #7 and/or #15, they all should explain that :) Sep 01 12:22:32 LetoThe2nd: I am right now watching #15 of the series. And now I get it... sry I am a bloody noob and need time to understand things and ask silly questions :D Sep 01 12:22:47 jUnG3: all is well :) Sep 01 12:24:26 jUnG3: everyone started from somewhere at some point, no worries. LetoThe2nd don't forget to explicit that it's a *package* that should be added in an image, not a recipe. (the recipe vs package was one of the fundamental notions that made me understand Yocto better) Sep 01 12:24:54 qschulz: very good point. Sep 01 12:25:15 isn't that #3 or #4? ;-) Sep 01 12:27:09 gonna watch them again, now that a light poped on in my head Sep 01 12:28:07 * LetoThe2nd totally read "a light pooped in my head". Sep 01 12:28:51 LetoThe2nd: that's what you get when you have kids in daycare age :) Sep 01 12:29:32 erbo: here its the other way round. finally i have kids as excuses for my own stupid way of humour. Sep 01 12:29:51 Hahahahahahahha :D Sep 01 12:30:19 Yeah, becoming a dad finally made all dad-jokes totally legit Sep 01 12:54:06 Its me again :D. I searched the web how to enable SocketCAN and found this E-Mail https://www.yoctoproject.org/pipermail/yocto/2017-September/038159.html. Where can the "CONFIG_CAN_VCAN" config variable be found and changed? Sep 01 12:54:37 jUnG3: it's a kernel config Sep 01 12:55:24 CONFIG_CAN_VCAN would enable you to use a virtual CAN interface, which can be suitable for testing without an actual CAN interface available Sep 01 12:56:03 But it's not needed "to enable SocketCAN" Sep 01 12:56:25 Maybe it's better if you describe a bit more on what you want to achieve Sep 01 13:03:54 LetoThe2nd: dad jokes are the best bit about parenting Sep 01 13:04:20 erbo: The application I wrote connects to an CAN (virtual or real). For testing I use a virtual, on my host machine I set the vcan like in this documentation https://elinux.org/Bringing_CAN_interface_up. My goal is to have a virtual can device inside the image because I want to connect my application with the vcan and test it. :D Sep 01 13:05:33 rburton: ++ Sep 01 13:33:19 I'm getting an error make: *** No rule to make target 'fdts/stm32mp153a-raichu-cubemx-mx.dts' Sep 01 13:34:01 Is anyone familiar with the stm32mp addons layer and how to provide custom DTS files for the kernel/u-boot/tf-a? **** BEGIN LOGGING AT Tue Sep 01 13:36:19 2020 Sep 01 13:36:50 how did you provide yours? **** BEGIN LOGGING AT Tue Sep 01 13:39:17 2020 **** BEGIN LOGGING AT Tue Sep 01 13:43:30 2020 Sep 01 13:43:48 Basically, I created a new layer called "meta-raichu". I created a new machine file by copying meta-st-stm32mp-addons/conf/machine/stm32mp1-mx.conf -> conf/machine/stm32mp1-raichu-cubemx.conf Sep 01 13:43:54 shan1: python has been on the list of possible topics for quite a while, but i never got round to it so far. maybe i will, but its not planned. Sep 01 13:44:08 also there won't be a session in september, just so you all know. Sep 01 13:44:12 thecomet: creating a new BSP for Yocto is not so straightforward and we need more details to help you Sep 01 13:44:35 mckoan: I can dump a directory listing of my layer if that helps Sep 01 13:44:41 thecomet: and the puszzling ST documentation won't help Sep 01 13:45:18 s/puszzling/puzzling Sep 01 13:45:34 thecomet: that could be a starting point Sep 01 13:46:32 LetoThe2nd I know you tend to be busy so maybe I can write something in the mean time and after you come back maybe you can live stream it. I have never made live streams but am very well with blog posts and write ups. Sep 01 13:46:48 mckoan: Here you go: https://pastebin.com/mY9kxZcH Sep 01 13:48:04 you also seem to have forgotten completely about [yocto-cheatsheet](https://github.com/shantanoo-desai/yocto-cheatsheet) which you had forked centuries ago :> Sep 01 13:48:17 Here is the machine conf file: https://pastebin.com/ywq0NVJV Sep 01 13:49:10 shan1: there's a couple of competent folks in here, so if you have a write up on something and are not afraid of feedback we can probably find somebody to check it first and then promite it. and no, i've not forgotten but never really made it high enough on my to-do-prio list Sep 01 13:50:05 @LetoThe2nd - Do you have a couple of minutes - A comment to your last Jester Session ;) Sep 01 13:50:16 thecomet: the tree is not valid. Who is supposed to use the content of mx directory? Sep 01 13:50:23 RobertBerger: sure, of course. Sep 01 13:51:12 mckoan: i guess there is trickery going on in the recipes to refer to mx as some form of included source. already seen such. Sep 01 13:51:19 So - Remember I asked about auto enabling the systemd service, which I was missing in the recipe. Sep 01 13:51:28 *[yoctoproject-cheatsheet](https://github.com/shantanoo-desai/yoctoproject-cheatsheet) Sep 01 13:51:39 And indeed it is missing - the doc says by default it's enabled ;) Sep 01 13:51:50 > no, i've not forgotten but never really made it high enough on my to-do-prio list Sep 01 13:51:59 thecomet: however the first advice about BSP creation during my courses is: prepare it out of Yocto Sep 01 13:52:03 understandable (add sad cat meme) Sep 01 13:52:22 mckoan: In the machine file you can provide a path to the dts files located in the mx directory using the CUBEMX_PROJECT variable. If you look at the meta-st-stm32mp1-addons layer in the file that's being included it makes more sense Sep 01 13:52:25 RobertBerger: yeah i remember. didn't realize that it was meant as important remark, i checked the auto-enable in systemd.bbclass before the session. Sep 01 13:52:33 You might want to add SYSTEMD_AUTO_ENABLE="enabled" or "disabled" in your recipe, of I can send you a pull request if you prefer once I cook up something Sep 01 13:52:36 I created the layer using bitbake-layers and co Sep 01 13:52:44 or I can Sep 01 13:53:03 jUnG3: Ok, then it's just about making sure your kernel is built with support for VCAN. How that is accomplished depends a bit on your BSP, basically which recipe that provides the kernel for you. **** BEGIN LOGGING AT Tue Sep 01 13:56:12 2020 Sep 01 13:56:42 Assuming you have a working service file it's fine. I just don't have them on the first attempt :) Sep 01 13:57:00 RobertBerger: if you have a couple of spare minutes, send a PR which adds a comment explaining this to the recipe :) Sep 01 13:57:06 thecomet: I don't use cubemx-stm32mp.bbclass, sorry Sep 01 13:57:17 @LetoThe2nd - will do Sep 01 13:57:32 Is there anyone here who has used cubemx-stm32mp.bbclass? Sep 01 13:57:38 erbo: Thx for the answer :D. Sep 01 13:57:52 thecomet: you can try on https://community.st.com Sep 01 13:57:54 paulbarker: i have great crystal balls on sale... want me to prepare an offer? Sep 01 13:58:03 The st community is the worst Sep 01 13:58:14 thecomet: LOL Sep 01 13:58:15 Hmm great balls out of crystal ???? Sep 01 13:58:25 and on sale? Sep 01 13:58:28 LetoThe2nd: I did take the bait a bit there but felt it was better than no reply Sep 01 13:59:05 thecomet: the easy way would be to have SRC_URI = "file://mydts;subdir=arch/arm/boot/dts/" and that would make it available in your source dir Sep 01 13:59:15 otherwise, no, never used st stuff Sep 01 13:59:47 shan1: i think the absolutely easiest way for me is if you send a PR for the yoctojester website, and file it under articles. then it goes instantly live once i pull and i can promote it. don't forget to add your name/url/whatever in the article so the credit is yours! Sep 01 14:00:04 jUnG3: If you use a kernel recipe based on linux-yocto, it's quite easy to do. Have a look at https://www.yoctoproject.org/docs/latest/kernel-dev/kernel-dev.html and the "Changing the Configuration" section Sep 01 14:03:18 paulbarker: i know. but the answres to the first mails already suggested absolutely no own effort, hence i won't spend my time there. he even replied to me in private first. Sep 01 14:03:46 LetoThe2nd: shouldn't you answer with a link to your blogpost from yesterday :D Sep 01 14:04:09 qschulz: i actually thought about it. Sep 01 14:13:11 qschulz: Which file would I put that in? Would that be a new .bbappend file? Sep 01 14:18:58 thecomet: don't know, couldn't find a git repo with the bbclass you were talking about so /me shrugs Sep 01 14:19:19 but yeah, normally you have a bbappend for the kernel and you give it your dts Sep 01 14:19:40 (well, the best way is to own the kernel git repo and push to it directly, no tricks required then) Sep 01 14:28:08 @LethoThe2nd PR is out ;) Sep 01 14:31:45 RobertBerger: I would remove https://github.com/TheYoctoJester/meta-quicksilver/blob/master/recipes-quicksilver/simpleechoserver/simpleechoserver_git.bb#L5-L10 in your PR as well :) Sep 01 14:36:06 @qschulz - I think I fixed it before LetoThe2nd took it Sep 01 14:36:59 yup Sep 01 14:36:59 RobertBerger: you're actually right.. I'm awfully nitpicky by moments :/ Sep 01 14:37:09 thanks both of you! Sep 01 14:37:27 ;) Sep 01 14:37:36 And now to something completely different. Sep 01 14:37:50 i just realized the mail setup of gh is not what i expected. plus, i am totally not used to the workflow. Sep 01 14:38:03 I am hunting for benchmarks to show differences in compiler tune flags ;) Sep 01 14:38:10 Do you have any suggestions? Sep 01 14:38:38 benchmarks in terms of performance or assembly output? Sep 01 14:38:44 I played around with phoronix-test-suite which has some interesting ones, just it tries to compile them. Sep 01 14:38:54 I want to run a program and get some numbers. Sep 01 14:38:55 Wait Sep 01 14:39:44 Like e.g. this one, which I was running on 2 boards: https://openbenchmarking.org/result/2008307-HU-2008306HU29 Sep 01 14:40:03 i would see if https://quick-bench.com has something that helps Sep 01 14:41:41 Can you run this somehow on a target? Sep 01 14:41:47 arm32/arm64 Sep 01 14:42:31 Maybe I should elaborate a bit more. Sep 01 14:43:04 We usually tune the compiler tune flags for every specific target board and rebuild everything. Sep 01 14:43:24 The container people (e.g. docker) don't really do anything like that. Sep 01 14:44:29 There are e.g. arm32/v7 arm64/v8 kind of and not all those compiler tunes we care about. Sep 01 14:44:56 ah, so the question is if the tunes are actually worth it? Sep 01 14:45:21 So I am wondering what are performance differences between running stuff native and in a container - I guess maybe I/O and not so much CPU. Sep 01 14:45:36 And as you say is it worth it. Sep 01 14:46:04 And if it's worth it with oci images I could use them. Sep 01 14:46:57 from my point of view its not about if its worth it. its about if it works at all. example: vfp switches on arm cores Sep 01 14:47:58 So I would really like to collect some data about it. Sep 01 14:48:07 no idea, sorry. Sep 01 14:48:38 At the moment all my arm32 have a single (sub optimal) compiler tune and so maintenance/testing is much easier. Sep 01 14:48:46 One kernel/rootfs for all the boards. Sep 01 14:57:44 YPTM: Jan-Simon is on Sep 01 15:00:46 YPTM: armpit is on Sep 01 15:01:15 YPTM: Scott Murray is on Sep 01 15:01:37 RP: ERROR: elfutils-0.180-r0 do_package_qa: QA Issue: No GNU_HASH in the ELF binary /home/ross/Yocto/build/tmp/work/core2-64-poky-linux/elfutils/0.180-r0/packages-split/elfutils-ptest/usr/lib/elfutils/ptest/tests/test-nlist, didn't pass LDFLAGS? [ldflags] <-- seen that? Sep 01 15:04:08 rburton: no :/ Sep 01 15:04:35 it was with next, will reproduce/bisect Sep 01 15:09:06 Hi Sep 01 15:11:25 I would like to checkout some git submodules during the fetch step of a recipe. What is the best way to achieve this? A first try with do_configure_prepend seems to work, but I think the git submodules are not cached. Sep 01 15:13:03 ykrons: there is a fetcher for that, you can specify gitsm:// instead of git:// Sep 01 15:17:34 insert meme Sep 01 15:19:52 rburton: \m/ O \m/ Sep 01 15:23:13 YPTM; over Sep 01 15:23:27 RP bind /dhcp patches sent Sep 01 15:24:52 armpit: thanks Sep 01 15:25:18 smartin: gitsm seems interesting but it seems it has some issue with mirroring. Did you know what are these problems? And is there a way to list the submodules we want to not check all submodules? Sep 01 15:26:28 ykrons: Mirroring and the archive bbclass should work with gitsm on the master branch, it's probably worth testing it out Sep 01 15:27:16 gitsm requires you to either mirror all submodules -- OR no submodules.. If you chose the later, then it is up to you to manually add the submodules needed for the package to function to your SRC_URI in the recipe. Sep 01 15:27:34 Both mechanisms are actively used, the later just gives the recipe developer more control over exactly what is downloaded Sep 01 15:29:29 paulbarker, I'm far from master :( Sep 01 15:29:36 rburton I just noticed a patch to the list to fix this.. Sep 01 15:29:54 oh good Sep 01 15:30:04 "insane: fix gnu-hash-style check" Sep 01 15:30:21 hm, that was integrated Sep 01 15:30:23 (or if that is in master-next, it fixed the checked and this is a 'real' problem) Sep 01 15:30:43 yeah... Sep 01 15:30:51 likely this was a long standing bug, and it's finally been found/fixed.. :( Sep 01 15:30:53 fray: is there a documentation on to proper way to use gitsm and to select all or none submodules Sep 01 15:42:31 fray: the test is doing a better job now, that file doesn't get the linker flags explicitly Sep 01 15:46:19 Anyone have a good files.watcherExclude for vscode for oe workspaces? remote - ssh is acting up presumably due to a too large workspace. but it's a pain to debug, since it's not clear which files are actually matching your patterns and which are not Sep 01 15:48:11 seems like a question alejandrohs might be able to answer Sep 01 15:52:15 kergoth: my workaround is to put build/ up a level Sep 01 15:55:49 hrmph, https://github.com/microsoft/vscode/issues/3998 Sep 01 15:55:51 so vscode just sees the layers Sep 01 15:56:17 hmm i'll check into that, would have to adjust my workspace setup scripts to use a different layout. i like to set up a fresh set of repo clones and everything for each development task to isolate the changes Sep 01 15:56:53 https://github.com/microsoft/vscode/issues/29711 Sep 01 15:57:00 and https://github.com/microsoft/vscode-remote-release/issues/1257 Sep 01 15:57:01 lovely Sep 01 16:01:51 Hey, I built meta-rapsberrypi and meta-browser for dunfell. I can't seem to get hardware acceleration working for chromium. This is on a Pi3b. Not sure what I'm missing Sep 01 17:01:46 havok101: are you using mesa/vc4 or userland graphics Sep 01 17:03:28 good.. so it was a real bug Sep 01 17:03:42 'er.. previous was rbuton.. :) Sep 01 17:04:37 khem: vc4 Sep 01 17:05:58 My image_install has these: packagegroup-rpi-test, sudo, systemd, chromium-x11, matchbox-keyboard, mesa-demos Sep 01 17:06:18 My distro_features has these: x11 opengl systemd Sep 01 17:06:45 image_features has these: x11-base hwcodecs Sep 01 17:12:40 RP: have you used the buildbot rest api before? Sep 01 17:17:15 rburton: yes, technically Sep 01 17:17:29 rburton: I've even patched it Sep 01 17:17:43 see my question in #buildbot :) Sep 01 17:17:45 rburton: not sure I remember anything right now :/ Sep 01 17:17:50 fiar **** BEGIN LOGGING AT Tue Sep 01 17:38:42 2020 **** BEGIN LOGGING AT Tue Sep 01 17:41:35 2020 Sep 01 17:42:50 Has anyone seen the SDK tests crashing the host before? **** BEGIN LOGGING AT Tue Sep 01 17:46:57 2020 **** BEGIN LOGGING AT Tue Sep 01 17:51:16 2020 Sep 01 18:16:04 RP: Ya, Annoyingly I'm not quite sure whats wrong because I loose all output on Jenkins when it dies.... I suspect it's something on our end but it's really hard to debug the cause Sep 01 18:18:20 I think part of that is because the output gets confused when stderr is "lost" on the SDK tests, which is why I posted that patch Sep 01 18:19:12 JPEW: yes, I'd noticed that problem elsewhere but hadn't gotten to root causing it Sep 01 18:19:57 I feel like i'm going crazy here. Sep 01 18:20:10 Error: Sep 01 18:20:10 smartpdu-1.0.0-r0.aarch64 Sep 01 18:20:12 aghhhh Sep 01 18:22:31 khem, armpit: I've merged a fix for dhcp-client Sep 01 18:24:54 ok cool Sep 01 18:25:36 RP: the u-a fix I think I will withdraw it, because it can only work if we add virtual/u-a to images unconditionally, are you ok with such a change ? Sep 01 18:26:12 since glibc is going to use it and musl too which means its pretty much given in any image Sep 01 18:32:23 thanks Sep 01 18:33:12 any suggestions or things I can check to get chromium to run with hw acceleration. All the glx demos seems to work well. vc4 module is also loaded Sep 01 18:44:19 havok101: Does chrome://gpu work? Sep 01 18:44:59 JPEW: it just says software only Sep 01 18:45:14 I tried launching with various modes still not luck Sep 01 18:45:28 chromium --use-gl=egl --gpu-no-context-lost --ignore-gpu-blacklist Sep 01 18:45:39 havok101: Not sure. That's about the extent of my chromium acceleration knowledge. We use webkit :) Sep 01 18:47:34 I'm curious to try it. is this the meta-webkit layer you're using? https://github.com/Igalia/meta-webkit Sep 01 18:48:22 Yep. We use it because we're a little too resource constrained to run chromium and we also use wayland. Sep 01 18:48:47 at the time, I don't think chromium supported wayland very well Sep 01 18:49:05 ok i'm assuming it should work with x11 Sep 01 18:49:32 I have not experience with wayland and not sure I want to switch Sep 01 18:50:23 havok101: I suspect webkitgtk would work, but thats not the one we use Sep 01 18:51:27 We use cog, which is a very simple frontend for displaying page; theres none of the normal browser trappings (even lighter than a "kiosk" mode) Sep 01 18:51:59 Interesting, I'll poke around. Sep 01 18:52:27 I'd still like to figure out why chromium won't work with hardware acceleration though :( Sep 01 18:52:29 Homing in on the cause of my build errors. I have a recipe that's supposed to simply supposed to extract a tarball (containing a node/react web application) to it's needed location on rootfs `(/www/appfolder)`. Bitbake (i guess?) is analyzing the files and thinking it's missing dependencies ( like /usr/env) when it's totally fine. Sep 01 18:53:09 havok101: FWIW, I think there is a webkitgtk recipe in oe-core these days. The one in meta-webkit is usually more up-to-date Sep 01 18:55:49 JPEW: That might be better to try first before I pull in other sources. Sep 01 19:00:10 havok101: do you have HA enabled in chrome://settings/ Sep 01 19:00:34 and force enable GPU Rasterization Sep 01 19:03:56 I havent tried x11 version in a while but it should be a simple config error Sep 01 19:05:09 khem: In setting ha is enable. In chrome://flags override software rendering list, GPU rasterization, Out of process rasterization is enabled Sep 01 19:05:33 havok101:are you using master ? Sep 01 19:06:39 yes Sep 01 19:07:09 havok101: and using clang to compile right / Sep 01 19:07:16 This is on a raspberry pi 3 so I'm not sure if I need additional config there. I have set the gpu_mem to 128 Sep 01 19:07:27 yes meta-clang is using the dunfell branch though Sep 01 19:07:34 thats ok Sep 01 19:08:01 that should be ok, but for tests change it to 512 Sep 01 19:14:29 havok101: you can try to add PACKAGECONFIG_append_pn-chromium-x11 = " impl-side-painting" in local.conf Sep 01 19:15:55 ok let me try that, any way I can make these builds shorter. component-build seems to fail on an assert for me Sep 01 19:36:28 If a .bbappend file inherits a class, and there are python blocks inside the class file, when do those get executed? Sep 01 19:49:05 TheComet_: Anonymous python blocks (like `python () {`) I beleive get executed at the end of parsing Sep 01 19:49:26 Ok. Then there's something weird going on here Sep 01 19:50:09 I have a class file that gets inherited by recipes-kernel/linux/linux-stm32mp_*.bbappend and it never gets executed Sep 01 19:50:44 TheComet_: Is that the literal file name: "recipes-kernel/linux/linux-stm32mp_*.bbappend" ? Sep 01 19:51:04 Sorry, it's recipes-kernel/linux/linux-stm32mp_%.bbappend Sep 01 19:51:11 I'm doing "bitbake linux-stm32mp" Sep 01 19:51:14 TheComet_: OK, just checking :) Sep 01 19:56:48 Is there a standard way to create your build folder so everyone who wants to build my distribution has the same bblayers.conf file? Sep 01 19:58:24 Is it good practice to add bblayers.conf and the build directory to version control? Sep 01 19:58:43 TheComet_: no real standard, but a couple of prominent approaches. look up the live coding session on kas for an example. and no, nothin in build goes into VCS Sep 01 19:59:34 LetoThe2nd: What's kas short for? Sep 01 20:00:05 maybe... Krass Automatisches Setup? Sep 01 20:00:14 read: i have no clue. Sep 01 20:01:16 I'm trying to google "live coding session kas" and I'm getting some turkish website lol Sep 01 20:01:29 yup, so it's in fact some files within this archive i've made of my node application. I just can't seem to figure a way for bitbake to ignore them. I've tested that it works as is by manually extracting the archive onto rootfs after building the image (not including the recipe for this node application) and everything runs out-of-the-box. guess Sep 01 20:01:29 it's off to SO to ask Sep 01 20:01:51 TheComet_: https://youtu.be/KJHJlOtTdaE Sep 01 20:02:10 Thanks Sep 01 20:09:05 So far he said you need to create local.conf and bblayers.conf manually Sep 01 20:09:09 every time Sep 01 20:10:47 just keep on watching. Sep 01 20:20:30 @LetoThe2nd: kas as in "A so a kas" Sep 01 20:21:11 RobertBerger: :) Sep 01 20:22:14 Kicking And Screaming Sep 01 20:23:43 Konfabuliert Autokratisches Schnapsmarketing Sep 01 20:25:23 I had to look up "Konfabuliert" ;), which should be obvious with 6 year of Latin ;) Sep 01 20:25:33 anyways, time for e´me to call it a day. Sep 01 20:26:05 CU **** BEGIN LOGGING AT Tue Sep 01 20:32:21 2020 **** BEGIN LOGGING AT Tue Sep 01 20:35:52 2020 Sep 01 20:39:19 Sweet, thanks LetoThe2nd kas looks exactly like what I needed **** BEGIN LOGGING AT Tue Sep 01 20:57:50 2020 Sep 01 21:16:14 khem: so I just noticed when I open chrome://gpu on the bottom under log messages I have three messages of GPUProcessHost: The GPU process crashed! Sep 01 21:22:39 havok101: hmm yeah I guess thats the problem then perhaps mesa issues, can you try dunfell version of oe-core/poky Sep 01 21:24:29 khem: I'm on dunfell for everything except meta-browser which is on master Sep 01 21:27:37 oh ok Sep 01 21:27:45 then I think it should be debugged Sep 01 21:28:22 havok101: are you using chromium from X11 session ? is there something strange in chromium logs like "libgl error: MESA-LOADER" Sep 01 21:28:23 yea was thinking of doing that once I got a new build so most likely tomorrow. Sep 01 21:41:03 dev1990: checked the debug log. No sign of it Sep 01 21:43:20 It goes invalid-client-secret since it's using invalid keys followed by the GPU Process crashing 3 times Sep 01 21:45:35 ok, my recommendation is to check with latest mesa (20.2 is about to be released) and pretty recent kernel (linux-raspberrypi-5.4.x) Sep 01 21:47:55 https://github.com/koenkooi/meta-kodi/issues/24 - this problem is not releated to X11 and chromium but it was a research why kodi not working with i945GSE Sep 01 21:49:33 maybe helpful somewhat Sep 01 21:50:29 dev1990: Thanks I'll check it. Sep 01 22:02:17 khem: not keen on that, no Sep 01 22:03:06 armpit: I used the typo from your commit :( Sep 01 22:12:40 LOL... really? package podman-2.0.7+git0+ce7b78a43a-r0.corei7_64 conflicts with docker in meta-virtualization Sep 01 22:12:48 i find that quite interesting Sep 01 22:31:02 reality isnt it :) Sep 01 22:42:32 RP: yeah I was not happy either Sep 01 22:43:38 khem: Sep 01 22:43:42 oops haha Sep 01 22:44:21 khem: do you know why it was decided to use $CC on the sdk instead of calling gcc Sep 01 22:44:36 khem: btw I do know why I'm trying to get a proper argument Sep 01 22:45:38 I guess my question is when did we decide to mandate the use of $CC for the sdk Sep 01 22:49:14 alejandrohs: many reasons, we configure kernel with a build time sysroot, but install time sysroot is only decided when SDK is installed so we need to fix that during SDK install secondly we want use ABI options e.g. using hardfp via cmdline but default compiler is not hf to keep it common across hardfp and softfp ABIs Sep 01 22:49:40 and we also add right tune options for the given machine Sep 01 23:13:40 bind looks like its failing in do_package in master-next https://autobuilder.yoctoproject.org/typhoon/#/builders/52/builds/2370 :( Sep 01 23:14:05 armin isn't here, I should sleep too Sep 01 23:17:33 RP: let me look Sep 01 23:18:52 khem: thanks, I've not looked at the problem, I need to sleep Sep 01 23:20:58 no worries either of us will take a look Sep 01 23:21:13 I think we never do multilib builds with meta-oe Sep 01 23:21:19 thats why we never saw it Sep 01 23:22:34 mv: cannot stat '/home/pokybuild/yocto-worker/qemux86-world/build/build/tmp/work/core2-64-poky-linux/bind/9.16.5-r0/package/usr/bin/bind9-config': No such file or directory Sep 02 01:49:30 khem: thanks! **** BEGIN LOGGING AT Wed Sep 02 02:26:55 2020 **** BEGIN LOGGING AT Wed Sep 02 02:30:46 2020