**** BEGIN LOGGING AT Fri Feb 07 02:59:57 2020 Feb 07 03:35:43 RP: ah! got it. I was talking about an installer image. So the losetup stuff etc is on the target hw Feb 07 05:39:10 Hi all, My solution vendor has provided me some HW with Yocto-Poky. I need there port forwarding via ssh, but can't find the sshd_config file. Any help would be appreciated Feb 07 05:45:26 BMWE: Do you know if you're using dropbear or openssh? Feb 07 05:49:34 no idea Feb 07 05:49:54 version of the Poky will assist? Feb 07 05:50:49 I don't think dropbear ship any sshd_config, so you might want to look into moving to openssh. There's an "image feature" you can use called "ssh-server-openssh" . Try setting EXTRA_IMAGE_FEATURES += "ssh-server-openssh" in your image recipe. Feb 07 05:50:57 Poky have both Feb 07 06:00:44 As I'm not the SW owner, I not sure that I can include it. Feb 07 06:01:12 any idea how can I enable port forwarding with dropbear? Feb 07 06:40:06 New news from stackoverflow: Unable to start bitbake server Feb 07 07:09:17 khem: are you on archlinux, too? Feb 07 07:15:02 khem: Actually docker would not be too bad here (it's not like I need an IDE), but I'll pin the file for now. Just because I don't like the idea of being on arch linux but having to run everything on an ubuntu container xD Feb 07 07:18:06 palate: I run all yocto builds in an ubuntu docker, even though I'm also using Ubuntu or Debian on the host. It's so little overhead once you're used to it, and ensures that build will work the same regardless of which host system I'm on. I can recommend it. Feb 07 07:18:40 I do all work outside of docker, I just use docker for the shell where I run bitbake Feb 07 07:19:12 erbo: so you mount your working directory as a volume with the same user id? Feb 07 07:19:53 yes Feb 07 07:20:23 erbo: may I ask how you run the container? I never know how to deal with the user id :/ Feb 07 07:20:35 I wrote a blog about it once, in an attempt to convince some colleagues at the time: http://blog.unixshell.se/2018/03/simplify-your-yocto-builds-using-docker/ Feb 07 07:21:09 Basically the CROPS team have solved it for you, if you use the right docker image as a base Feb 07 07:24:20 palate: yes I use archlinux and debian container if you are interested https://github.com/yoedistro/yoe-distro Feb 07 07:24:22 Since that blog I've changed the Dockerfile a bit: https://pastebin.com/SghP8Hi5 Feb 07 07:24:56 And I also use -v to mount in a folder where I share sstate and dl_dir for all builds Feb 07 07:25:59 erbo: somehow we all kind of use variations of that technique nowadays. Feb 07 07:26:52 * LetoThe2nd tries to push erbo in front of a webcam to show the world! Feb 07 07:28:42 erbo: checking you blog :) Feb 07 07:29:08 sstate and dl_dir are some kind of caches then? Feb 07 07:29:19 LetoThe2nd: hehe, I'm not very good with public speaking. But it's a thing I would sort of like to change, so maybe I'll try recording a video as a step :) Feb 07 07:30:34 erbo: neither am i. hint: people expect way higher quality of prepared, recorded stuff than they do of "live" situations. (thats why i do the latter!) Feb 07 07:30:51 khem: yoe seems pretty cool. May be a nice way to get started with yocto actually Feb 07 07:30:52 palate: yes, DL_DIR is where downloads are stored and SSTATE_DIR is where the "build cache" are stored. Sharing those between builds will save lots of time. Feb 07 07:31:55 got it Feb 07 07:34:53 palate: yes docker on yoe should work out of box pretty much Feb 07 07:55:56 tgamblin_: I wouldn't try, just create a fix Feb 07 09:40:40 New news from stackoverflow: Bitbake OpenCV 3.4.5 for ArmV7 fails Feb 07 10:49:00 RP: have you looked at the latest autobuilder run yet? Feb 07 11:04:38 rburton: my debug patch wreaked the overnight one, waiting on a fast rebuild Feb 07 11:04:49 ah, its failed now Feb 07 11:05:07 rburton: so no Feb 07 11:10:30 rburton: the list is looking better, libevdev, ncurses-doc, perl-module-config-heavy/perl/perl-ptest, sudo and sysvinit Feb 07 11:11:00 evdev is new, sudo was fixed so may be a new issue :/ Feb 07 11:27:48 rburton: I did find a raft of other issues, a new style of them, last night :/ Feb 07 11:28:14 bah Feb 07 11:28:20 lets stop it using sstate again :) Feb 07 11:34:43 rburton: libevdev looks like generated header sorting issue Feb 07 11:37:10 sudo is "/bin/mv" vs "/usr/bin/mv" Feb 07 11:37:31 host paths leaking in? Feb 07 11:38:03 rburton: yes, will be a configure setting Feb 07 11:38:16 perl is differing in /usr/lib/perl/ptest/cpan/Encode/Byte/byte_t.fnm with a sorting issue Feb 07 11:41:45 i'll take sudo once i've done some branch cleanup for something else :) Feb 07 11:42:23 rburton: the easy one? :) Feb 07 11:42:29 damn straight Feb 07 11:42:50 maybe we should just give alex the perl one after yesterday ;) Feb 07 11:52:02 rburton: maybe. My guess with sudo is ac_cv_path_MV=${bindir}/mv ;-) Feb 07 11:54:35 yeah Feb 07 11:54:46 i'll check it doesn't look for more paths Feb 07 11:56:09 Do we have the videos of YP summit from ELCE 2019 on Youtube somewhere? Feb 07 12:00:23 qschulz: not yet because $REASONS Feb 07 12:00:33 LetoThe2nd: echo $REASONS Feb 07 12:00:39 Feb 07 12:00:43 qschulz: "scooters" Feb 07 12:35:00 Hi i like to use some features from modemmanager-qt from the qt-extra layer. but it is locatetd in the $includedir$/kf5/modemmanagerqt and the compilation fails because the kf5 dir inbetween Feb 07 12:35:06 what am i doing wrong Feb 07 12:38:26 use, in which form? like, is it an include for a c file? or what? Feb 07 12:41:18 New news from stackoverflow: TOOLCHAIN_HOST_TASK Vs TOOLCHAIN_TARGET_TASK Feb 07 12:50:04 i try to use it in a c++ file inside qt creator Feb 07 12:53:45 and the kf5 is non-standard, or what? Feb 07 12:54:10 because this sounds mostly like static includes instead of pkg-config'ed ones (or whatever one uses there) Feb 07 12:54:42 qt uses qt += for it Feb 07 12:56:00 this actually suggests that this is the canonical path: https://packages.ubuntu.com/de/xenial/amd64/modemmanager-qt-dev/filelist Feb 07 12:56:09 * i need to do a QT += ModemManagerQt for qmake Feb 07 12:57:08 i amno qt user, but it sounds like the qmake as you are using it is not properly looking at the target sysroot. Feb 07 13:02:59 any id how to fix that ? other than moving the files manualy ( in the installed Feb 07 13:03:23 sdk) Feb 07 13:04:33 hmw1: fix the building system? Feb 07 13:05:01 i mean, if it behaves exactly the same if you are building for ubuntu, then chances are that ubuntu+sdk are right, and your way is wrong. Feb 07 13:07:13 LetoThe2nd: my idea was already that my way would be wrong. Feb 07 13:08:38 LetoThe2nd:I think I'm missing a package Feb 07 13:11:24 New news from stackoverflow: try to build bsp recipe but an error occured required build target 'sysdig-driver' Feb 07 13:12:38 rburton: ok, libevdev and two perl issues sorted. That leaves ncurses, sudo and sysvinit Feb 07 13:19:18 RP: my build failed because i failed to drop my perl and it conflicted with alex's. rebuilding now to sort sudo. Feb 07 13:21:03 yeah sudo looks more several binaries Feb 07 13:24:03 RP: cab you stage the bison 2.5.1 upgrade to master-next as well at your convenience ? my builders are green again Feb 07 13:32:09 khem: ok, been trying to avoid rebuilds to sort these repro issues Feb 07 13:32:21 but I've changed perl so its probably a moot point now Feb 07 13:35:50 khem: queued Feb 07 13:40:36 patch: /lib/x86_64-linux-gnu/libattr.so.1: version `ATTR_1.3' not found (required by patch) Feb 07 13:40:40 This is new :/ Feb 07 13:43:15 RP: checking for sh... /data/poky-tmp/master/hosttools/sh Feb 07 13:43:16 \o/ Feb 07 13:47:57 RP: patch sent Feb 07 13:49:14 rburton: fun Feb 07 13:49:17 tgamblin: thanks! Feb 07 13:51:19 * RP fast tracks that into -next Feb 07 13:52:24 the testimage was stuck doing some valgrind test (today I've let it run for over 40000s) and after manually killing drb-amd64-linux couple times the testimage finished with "All required tests passed (successes=47, skipped=14, failures=0, errors=0) Feb 07 13:53:09 JaMa: sounds like bad error handling in valgrind's ptest :( Feb 07 13:57:08 RP: I'm still trying to reproduce the issues you reported with the artifacts naming pull request, now using poky directly instead of nodistro build Feb 07 13:58:22 JaMa: I have to admit I didn't dive into them in detail, if you continue to be unable to I guess I can take a look Feb 07 13:58:27 JaMa: just a bit overloaded :( Feb 07 14:00:05 RP: don't worry, I'll continue from my end, the longest delay are e.g. the gold issue in new kernel which was breaking my nodistro builds for a while Feb 07 14:02:08 hmm, I can break sysvinit but not quite the same as the autobuilder difference. "bitbake sysvinit; bitbake sysvinit -c clean; bitbake sysvinit -c package_write_ipk -f" Feb 07 14:08:07 it is the same issue, diffoscope is lying to me Feb 07 14:11:35 New news from stackoverflow: Can I make MSYS2 suitable for yocto compilation? Feb 07 14:29:41 RP: sudo fix posted Feb 07 14:29:47 hopefully better than my perl one Feb 07 14:32:47 rburton: thanks, it was late :) Feb 07 14:34:40 * RP has a fix for sysvinit but it has bad implications Feb 07 14:59:41 You are all going to get the reproducible fixes in before I get diffscope in to make them easier :) Feb 07 15:02:51 JPEW: I think I know what the ncurses issue issue is btw, its not actually a big issue Feb 07 15:04:23 (0-without-manpages-renames to configure) Feb 07 15:10:44 RP: OK. I haven't been as productive as I'd like on the reproducible stuff this week, had a few fires come up and I had to setup a new desktop Feb 07 15:11:48 New news from stackoverflow: Cannot change keyboard layout on Yocto/Poky linux Feb 07 15:11:53 JPEW: np, I'm just kind of ploughing through them as I figure them out Feb 07 15:14:39 JPEW: old one was on fire? :P Feb 07 15:25:43 haha Feb 07 15:26:04 LetoThe2nd: No, my work finally sprang for decent machines. Went from 4 cores to 18, so I'm pretty happy Feb 07 15:26:12 only 18 Feb 07 15:26:15 you should be insulted Feb 07 15:26:29 rburton: Ya, I know :) Feb 07 15:27:14 i'm quite happy to not have such stuff on my desk anymore Feb 07 15:27:16 my perspective is skewed since i did a build on the beasty in the office. 88 cores, core-image-sato in 25 minutes. Feb 07 15:27:30 not sure i want to be within 100 meters of it when its building obviously Feb 07 15:27:37 JPEW: were you distributing the builds over multiple 4-core machines at least? Feb 07 15:27:52 machines like that ought to go into some server cabinet. Feb 07 15:27:59 tgamblin: Ya, there is a reason I have such an interest in icecream :) Feb 07 15:28:25 JPEW: does everyone else know they were contriubting to the cluster or did HR have mysterious slowdowns to their desktops Feb 07 15:28:31 rburton: I feel like the number of cores you have is always like those fishing stories; each time I ask you have more ;) Feb 07 15:29:00 JPEW: I figured :) I'm considering grabbing either a lot of USFF ThinkCentres/Optiplexes off of Ebay to play around with distributed builds, or maybe a Picocluster Feb 07 15:29:06 JPEW: my own machine has 20. the 88 core beast is half way across the planet which means its not great for interactive building Feb 07 15:29:21 JPEW: I have a similar build machine. I think this is why I don't notice these parsing issues people keep talking of... Feb 07 15:29:46 RP: do builds on your thinkpad now and again Feb 07 15:29:52 rburton: I do! Feb 07 15:29:52 RP: Heh, I've been meaning to ask about that Feb 07 15:30:05 rburton: mainly when travelling Feb 07 15:30:10 in a vm on my laptop bitbake takes minutes to parse :/ Feb 07 15:30:12 RP: same here, but i avoid it Feb 07 15:30:25 rburton: I just can't have email and builds :/ Feb 07 15:30:29 i'd rather spin up an ec2 instance Feb 07 15:30:48 my build machine can be remotely woken Feb 07 15:30:50 I'm going to have to benchmark my 2700X to see how fast it can make core-image-sato Feb 07 15:31:18 tgamblin: my reference is literally just poky sample conf with dl_dir set Feb 07 15:31:38 that is that what our buildperf benchmark is on the test machines too Feb 07 15:31:42 rburton: It was opt in... we have very few Linux users :) Feb 07 15:31:53 JPEW: docker ftw Feb 07 15:32:15 rburton: i still read ftw as f*** the world Feb 07 15:32:19 rburton: RP: good to know. I'll try and remember to spin it up this weekend Feb 07 15:32:28 LetoThe2nd: and now i will too thanks for that Feb 07 15:32:41 rburton: YW. Feb 07 15:33:03 tgamblin: I think you could even run the standard benchmark tests from oe-core Feb 07 15:33:36 tgamblin: you can see what https://autobuilder.yoctoproject.org/typhoon/#/builders/92/builds/2030 does Feb 07 15:35:20 how to get history of an variable that has an override? Feb 07 15:35:37 for example, i'm interested in the history of ALTERNATIVE_systemd but all i can see is Feb 07 15:35:40 # override[systemd]:rename from ALTERNATIVE_${PN} data.py:169 [expandKeys] Feb 07 15:38:55 RP: interesting. Maybe I'll be indirectly responsible for the cost of AMD CPUs skyrocketing when it turns out they're better cost-to-performance than some other popular options :) Feb 07 15:40:17 tgamblin: we used to joke we were driving cpu sales Feb 07 15:41:53 New news from stackoverflow: simlink in eSDK sysroot points to a non existent path Feb 07 15:48:57 RP: https://twitter.com/TheYoctoJester/status/1225808616902602762 Feb 07 15:50:21 LetoThe2nd: back in the old days, pre CoW it was all about driving memory sales ;-) Feb 07 15:50:46 RP: pre-cow.... mhh Feb 07 15:50:54 Feb 07 15:51:36 LetoThe2nd: ;-) Feb 07 15:52:34 RP: tweeting random stuff suits me! Feb 07 15:52:55 LetoThe2nd: its why I'm feeding you ammo :) Feb 07 15:53:44 i shuold note "RPs PR wingman" in my CV Feb 07 16:59:49 RP: thx, I think one problem I see is that when I use -j N -l XX.YY to make, GNU make ignores BUILT_SOURCES from deps, i cant reproduce it without -l Feb 07 17:00:04 RP: perhaps we should upgrade to make 4.3 ;) Feb 07 17:02:55 khem: heh, I think we assume_provided make ? Feb 07 17:03:45 hello there. i am having trouble using yocto on a raspberry pi 4. i used it before on a raspberry pi 3 and it worked quite well, but i am not getting any video output ( official display, hdmi0 and 1) from the pi with yocto. is this a known issue with the rpi4 being quite new? Feb 07 17:11:24 RP: https://git.openembedded.org/openembedded-core/commit/?h=master-next&id=e83d5109efdef0034e9a9455892a1ec6815f2678 drop that one, I am sending a v2 Feb 07 17:11:59 RP: that will not cause a full rebuild thankfully :) Feb 07 17:13:28 maybe a more direct question. i think the raspberrypi4 machine image uses warrior, how can i use zeus? Feb 07 17:16:48 lastaid: checkout zeus for all your layers Feb 07 17:20:54 rburton i am really new ... how would i do that? i followed that tutorial https://doc.qt.io/QtForDeviceCreation/qtee-custom-embedded-linux-image.html Feb 07 17:21:33 never used repo, dunno. i'd just 'git checkout origin/zeus' in all the layers Feb 07 17:21:43 no idea if boot2qt even supports zeus Feb 07 17:21:51 ask qt Feb 07 17:22:54 ok, i'll try Feb 07 17:23:04 lastaid: very likely you won't like to use Boot to Qt Feb 07 17:23:21 mckoan elaborate? Feb 07 17:23:43 i did it without b2qt before and it was fine, this just looked quite handy Feb 07 17:23:48 lastaid: Boot to Qt requires a commercial support from Qt Feb 07 17:23:58 well, thats bad Feb 07 17:24:24 isn't it all that gpl stuff, as long as i don't change it or statically link? Feb 07 17:25:05 lastaid: learn more from my presentation https://koansoftware.com/koan-at-qtday-2019/ Feb 07 17:25:43 smooooth :) Feb 07 17:28:07 also read https://devclass.com/2020/01/28/qt-makes-lts-releases-commercial-only-and-accounts-mandatory-for-binaries-again/ Feb 07 17:28:50 thanks a lot!!! will be good reading Feb 07 17:29:01 so just build qith meta-qt5? Feb 07 17:29:05 lastaid: glad to be helpful Feb 07 17:32:48 khem: exactly Feb 07 17:35:39 mckoan this is a really good presentation! it is really hard to get any information about the licensing from qt directly Feb 07 17:35:50 if in doubt, pay up Feb 07 17:36:00 is the message they try to convey it seems Feb 07 17:44:44 i just had my build server with 48 gigs of ram start swapping ... :D Feb 07 17:46:04 lastaid: thanks Feb 07 17:55:29 lastaid: sounds like the legal version of the planning approval in hitchhikers guide Feb 07 18:12:26 I think the large deployerrs like Automotive Tier 1 vendors should buy commercial so QT company can feed their engineers, it seems desperate attemp from them to get some revenues Feb 07 18:12:57 I think they will back track on it Feb 07 18:19:44 well, i do understand that thez want to get paid for their work. but the licensing cost is quite prohibitive for small or medium companies Feb 07 18:19:59 who do other stuff apart from qt, or want to try qt Feb 07 18:20:26 and i just found out the raspberrypi4-64 recipe does not build an image ... i am going home now Feb 07 18:43:59 lastaid: yes thats my point, in the larger fight smaller ones are also impacted.. not good Feb 07 19:26:19 is "devtool upgrade linux-libc-headers" a bad idea? Feb 07 19:32:52 hmmm Feb 07 19:33:03 ideally we dont like it Feb 07 19:33:39 but then people have done it, as long as the kernel is unmodified Perhaps is ok Feb 07 19:34:17 but most of times people point it to their heavily modified SOC kerrnels, with 3379 patches changing UAPIs etc. which is not cool Feb 07 19:39:36 I tried. It did not work cleanly at all, so I backed off. Feb 07 19:39:56 My kernel is 5.5 and I'm trying to build libgpiod 1.5 (which requires 5.5 kernel headers to build) Feb 07 19:48:15 I tried adding do_configure[depends] += "virtual/kernel:do_shared_workdir" but that did not work Feb 07 19:59:07 is virtual/kernel:do_shared_workdir still a thing? Feb 07 20:32:42 adding that to the recipe adds _nothing_ to run.do_configure Feb 07 21:12:57 New news from stackoverflow: Can I make MSYS2 suitable for yocto compilation? [closed] Feb 07 22:43:12 New news from stackoverflow: Here is another type problem of '“do_populate_sdk:Could not invoke dnf. Command” #2 Feb 08 02:13:16 RP: sent a v3 for the cpio QA patch which hopefully works **** ENDING LOGGING AT Sat Feb 08 02:59:58 2020