**** BEGIN LOGGING AT Mon Sep 30 03:01:03 2019 Sep 30 05:31:59 I am creating a recipe for a binary package that is 99% 64-bit stuff, but it has a couple of 32-bit libraries in it. This makes Yocto QA unhappy: "QA Issue: Architecture did not match (ARM, expected AArch64)". I'm wondering what the best solution to this would be. Is there a way to disable this particular warning or do I have to create a separate recipe for the 32-bit libs? Sep 30 05:34:27 INSANE_SKIP_${PN} += "arch" did the trick Sep 30 05:34:31 its better for you to package it into separate 32/64 bit packages Sep 30 05:35:03 you can disable the checks but they can bite in maintainance Sep 30 05:37:20 khem: thanks, I will consider what will be the best solution in this case. Have two possible ways of solving it now. Sep 30 06:43:03 good morning Sep 30 06:50:48 New news from stackoverflow: Differences between openembedded-core and poky Sep 30 07:08:58 good morning! Sep 30 07:14:50 morning! Sep 30 07:20:53 New news from stackoverflow: "Meson failed" while generating core-image-selinux in yocto warrior branch Sep 30 07:25:45 Good morning! Sep 30 10:33:26 I am trying to make my root filesystem read-only by enabling EXTRA_IMAGE_FEATURES += "read-only-rootfs". That part is working fine, but during boot I get a few error messages: "rm: can't remove '/tmp': Read-only file system Sep 30 10:34:35 "rm: can't remove '/etc/resolv.conf': Read-only file system" and "ln: /etc/resolv.conf: File exists" Sep 30 10:34:41 iceaway: then find out which script tries to do things there. it is almost certainly some custom thing, as we're using the read-only-rootfs in production since years witout problems. Sep 30 10:34:51 I am trying to figure out where these are coming from but so far no luck. Sep 30 10:35:20 LetoThe2nd: I am on that path and checking all init scripts, but have not found anything so far. Figured it might be a common issue. Sep 30 10:36:16 iceaway: it might be cause through something that comes with poky/OE of course, but its definitively no read only inherent problem. Sep 30 10:39:09 iceaway, its been fixed in master Sep 30 10:39:46 kroon: Oh? Any link to the actual fix? I am on sumo at the moment without possibility to upgrade right now. Sep 30 10:40:49 iceaway, https://patchwork.openembedded.org/patch/152602/ Sep 30 10:41:39 https://patchwork.openembedded.org/patch/152602/ Sep 30 10:41:44 https://patchwork.openembedded.org/patch/152602/ Sep 30 10:41:52 oops sorryy about that Sep 30 10:42:44 iceaway, commit fb028a0a25c02ab56ad262afd8e6e9495a44673d in oe-core Sep 30 10:47:41 Seems like it has not been backported to sumo (yet). Can I add a .bbappend which patches the populate-volatile.sh file, or perhaps I can copy the original, make the fix, and add it to my bbappend recipe? Sep 30 10:49:42 iceaway, id say go for the option 2, should be the simplest thing to do Sep 30 10:54:52 i wonder why selftest didn't notice the read only rootfs proble Sep 30 11:06:14 <__angelo> hi, gm. Is there a table that suggest i.e. ubuntu version to use for poky ? Sep 30 11:07:46 <__angelo> *for sumo Sep 30 11:08:27 __angelo: https://www.yoctoproject.org/docs/2.1/ref-manual/ref-manual.html#detailed-supported-distros Sep 30 11:08:32 damnit safari Sep 30 11:08:41 <__angelo> rburton, thanks Sep 30 11:08:44 https://www.yoctoproject.org/docs/2.5/ref-manual/ref-manual.html#detailed-supported-distros Sep 30 11:08:50 sumo is 2.5 not 2.1 Sep 30 11:09:15 newer distros may work but are unsupported, what with sumo being out of support now Sep 30 11:12:01 rburton, is that the test in meta/lib/oeqa/selftest/cases/buildoptions.py ? Sep 30 11:12:12 yes Sep 30 11:12:18 if its a warning maybe the test needs to check the logs Sep 30 11:18:50 rburton, "rm: can't remove '/tmp': Read-only file system" Sep 30 11:19:07 rburton, "rm: can't remove '/etc/resolv.conf': Read-only file system" Sep 30 11:21:35 it is a runtime error btw, if that wasn't clear Sep 30 11:22:06 hmm maybe buildtime aswell Sep 30 11:24:20 don't see it in the core-image-minimal temp/ logs though Sep 30 11:31:41 ah the test doesnt do a boot test Sep 30 11:31:42 maybe it should Sep 30 12:02:24 Hi all, Is it possible to include conf files ? Sep 30 12:04:01 prabhakarlad: would you please elaborate what you mean by "include" and "conf files"= Sep 30 12:16:47 I added a new recipe with a .bbappend file called initscripts_%.bbappend, in this I have SRC_URI += "file://populate-volatile.sh". I also have a "files" subdir where the actual file is located. The recipe is listed under bitbake-layers show-appends. Still I do not get the modified version of populate-volatile.sh in my image. Am I missing some step here? Sep 30 12:18:32 iceaway: have you checked the output of bitbake -e initscripts if everything is actually applied the way you think? Sep 30 12:18:50 LetoThe2nd: good idea, I haven't. Will do! Sep 30 12:21:40 New news from stackoverflow: Network Manager Build Error configure: error: rl_echo_signal_char() is required (install readline6?) Sep 30 12:22:10 LetoThe2nd I have local.conf file for a board - A, now we have one more board -B which similar to board- A with few changes, so is it possible to include the local.conf from A to B and overwrite a few variables ? Sep 30 12:23:59 prabhakarlad: totally wrong approach. your local.conf is ONLY meant to setup the build, in terms of DISTRO, MACHINE, TMPDIR and SSTATE (amongst some other minor things, but thats the gist). if you're having board specific things in your local.conf, thats a total red flag that you're doing it wrong. Sep 30 12:24:49 prabhakarlad: what you can (and probably should) do is, pour those things into a distro, a machine, and an image. and then those can easily share similarities Sep 30 12:25:47 prabhakarlad: the machine conf files for mach-foo and mach-bar for example can both include the common stuff from mach-foobar, and only overwrite their respective bits. Sep 30 12:26:08 LetoThe2nd thank you for the pointers. Sep 30 12:50:40 JPEW: Unfortunately that cleanup race is still there :( Sep 30 12:51:45 New news from stackoverflow: Yocto: when enable hwcodecs in IMAGE_FEATURES || SYSTEMD_SERVICE_swupdate value swupdate.service does not exist Sep 30 12:53:25 iceaway, did you also prepend to FILESEXTRAPATHS ? Sep 30 12:55:05 iceaway, and I dont think you should append to SRC_URI at all, the file is already listed there I guess Sep 30 12:58:44 RP: Arg, really? Have a link? Sep 30 13:02:39 JPEW: see 1d in https://autobuilder.yoctoproject.org/typhoon/#/builders/56/builds/732 Sep 30 13:04:03 RP: OK, I'll look into it... I'm out the rest of the week, so I may not have anything until next week, will that be OK? Sep 30 13:04:31 JPEW: yes, understood Sep 30 13:07:55 JPEW: have a look at meta/lib/oeqa/core/utils/concurrencytest.py:removebuilddir() Sep 30 13:08:04 JPEW: this was a simlar problem in oeqa Sep 30 13:08:09 OK Sep 30 13:21:50 New news from stackoverflow: Yocto: removing an existing layer leads to "no layers matching" error Sep 30 13:23:04 does yocto have any sort of luarocks support? Sep 30 13:51:54 New news from stackoverflow: Enable pcre support in clamav yocto Sep 30 14:21:58 New news from stackoverflow: No internet access without running 'ifup eth0' or restarting /etc/init.d/networking on jetson nano with yocto/poky Sep 30 15:39:02 Ran into an interesting issue...we're still building some images on the morty branch and we started to get some intermittent problems building curl. It looks like curl automatically detects whether or not brotli is in the global sysroot and it wasn't in its DEPENDS list. This meant that it was a race condition on whether or not curl would use Sep 30 15:39:02 brotli. Luckily this isn't a problem on thud because there's no longer a global sysroot. :) Sep 30 15:40:19 marler899721: one of the many reasons we switched :) Sep 30 15:40:41 yeah, local sysroots solves alot of issues I've noticed, very good move Sep 30 17:17:45 marler899721:morty was last release with combined staging sysroot, these issues are gone there after we still might have components looking into build system itself for stuff but thats rare since we also whilelist only needed stuff from buildhost Sep 30 17:18:13 updating might be a better deal Sep 30 17:20:17 does anyone have an example of a dracut/systemd initramfs recipe for yocto? Sep 30 17:45:19 is it normal for bitbake to hang when i have a syntax error? Sep 30 17:46:23 mischief: that happens to me.. and the server doesn't die. Sep 30 17:46:30 correct Sep 30 17:46:34 i have to kill -9 it :\ Sep 30 17:46:36 I have to kill it manually Sep 30 17:46:37 yes Sep 30 17:46:55 sigterm doesn't do it Sep 30 17:46:57 well, i'm glad it's not just me :-) Sep 30 17:47:04 same here. :-) Sep 30 18:31:16 slightly off topic but is there some trick to viewing graphviz output from bitbake -g? whenever i try to run dot on them, it takes forever.. much longer than im willing to wait Sep 30 18:40:45 mischief: open in text editor of your least distrust. dotfiles are pretty easy to manually interpret Sep 30 18:44:05 <__angelo> hi, enabled busybox systemd trough some config options in a .cfg . Btw, in the final rootfs there is not trace of systemd. Why ? Sep 30 18:46:16 __angelo: huh? Sep 30 18:46:36 __angelo: busybox and systemd do not have much to do with each other Sep 30 18:46:38 <__angelo> oh :) busybox syslogd :) Sep 30 18:46:51 <__angelo> (systemd takes mind control) Sep 30 18:47:06 :-) Sep 30 18:47:11 in that case, no idea. Sep 30 18:52:12 mischief: theres a tool in master or maybe even warrior in scripts/ to prune the deps Sep 30 18:52:44 __angelo: check if the busybox recipe creates multiple packages (WORKDIR/recipe-packages?) and try to find the one which is adding syslod. If it's not there, check that your .cfg is taken into account. If it is, check that it's not conflicting other .cfg (assuming .cfg are fragments (I have no experience in this though) Sep 30 18:53:56 <__angelo> qschulz, well, checking busybox version, it says syslogd is defined. But strangely, there is no syslogd link to busybox Sep 30 18:54:00 __angelo: perhaps try DISTRO_FEATURES_append += " systemd" and VIRTUAL-RUNTIME_init_manager = "systemd" and VIRTUAL-RUNTIME_initscripts = "systemd-compat-units" Sep 30 18:54:16 in local.conf Sep 30 18:54:52 oh, wait, you don't want full systemd. Sep 30 18:55:06 fullstop: systemd != syslogd :) Sep 30 18:55:10 <__angelo> fullstop, no, was a lapsus, i need syslogd :) Sep 30 18:55:14 haha Sep 30 18:55:23 let me get my glasses. :-) Sep 30 18:55:54 no, my glasses are fine. "hi, enabled busybox systemd ... " Sep 30 18:56:18 <__angelo> there was then an : oh :) busybox syslogd :) Sep 30 18:56:44 too many words. :-) Sep 30 18:56:47 <__angelo> btw, syslogd seems built in, but for some reason the symlink to busybox is missing, investiganting Sep 30 19:09:40 <__angelo> in what order are processed a recipe_1.1.bbappend and recipe_%.bbappend ? Sep 30 19:54:34 __angelo depends on layer priority Sep 30 19:56:36 khem: we are working on upgrading to thud. We have millions of lines of code that depend on the older packages on thud though, so it's a big effort Sep 30 19:56:49 I meant, that depend on the older packages that are on morty Sep 30 20:04:10 RP, bitbake selftest is failing. want me to bug it? Sep 30 20:14:46 marler899721:certainly, you are not alone.. Sep 30 20:15:45 marler899721: this in general is paying off technical debt along so the effort is actually quite beneficial in long term maintainance of your sw Sep 30 20:16:31 and at same time it might result in patches for OE too, so all in all its a progressive excercise Sep 30 20:30:45 __angelo: to be sure, run bitbake -e on your recipe, you'll see which bbclass/inc/bb/bbapend is doing what and what's the outcome Sep 30 20:43:54 <__angelo> qschulz, many thanks ! Sep 30 20:52:52 Yocto project got dropped on my lap, don't have any experiance with it. Haveing lots of fun. Sep 30 20:53:07 having* Sep 30 21:04:01 <__angelo> real a mistery. I see CONFIG_SYSLOGD in busybox enabled, i see the script that copies busybox stuff into rootfs executed, but there is no syslogd, nor busybox-syslogd.service in the final rootfs Sep 30 21:17:30 <__angelo> oooh, mistery solved. need to add busybox-syslog package to the image ... Sep 30 21:25:00 Can you add extended file attributes to package files in yocto? Sep 30 21:25:12 I'm having a hard time finding any examples and setcap doesn't seem to be available Sep 30 21:25:36 __angelo: that was my first suggestion, happy you found out :) Sep 30 21:27:24 <__angelo> qschulz, ooh, ok thanks :) Sep 30 21:32:43 marler899721: use libcap see http://layers.openembedded.org/layerindex/recipe/547/ Sep 30 21:33:35 machinehum: its quite naughty, if it troubles feel free to report here :) Sep 30 21:33:53 ok, so I just need to add DEPENDS += "libcap-native" Sep 30 21:34:09 khem: Thanks I really need to learn more before I ask some really silly questions lol. Sep 30 21:35:04 marler899721: yes and whatever you want to do during install/compile Sep 30 21:35:41 machinehum: no worries, its ok to ask stupid questions, everyone starts somewhere Sep 30 21:41:41 armpit: its the "FileNotFoundError: [Errno 2] No such file or directory: 'hashserve.sock'" so there is already a bug open for it Sep 30 21:41:56 armpit: JPEW has it but its looking like I may need to look into it Sep 30 21:42:16 ah Sep 30 21:45:29 marler899721: beware though; extended attributes might not get preserved through packaging and installation from packages. Sep 30 21:56:15 gotcha we'll see what happens Sep 30 22:18:54 So when bitbake give a big: ERROR: error msg here. Why does it keep building? Sep 30 22:19:47 Is it because it's running multiple threads, so only the thread with the error is killed Sep 30 22:25:33 How would I start to investigate this? https://pastebin.com/2Q6mAE0G Sep 30 22:26:57 error breaks the current task but doesn't abort the build straight away Sep 30 22:27:01 that would be fatal Sep 30 22:27:19 rburton: Got it, thanks Sep 30 22:27:20 marler899721: no packaging format supports extended attributes afaik Sep 30 22:31:15 I suppose that would mean I would need to set the file attribute somewhere in the do_rootfs step Sep 30 22:31:23 or postinst Sep 30 22:31:37 typically a chattr in postinst works Sep 30 22:31:47 postinst runs on the device? Sep 30 22:31:56 This file is packaged into a squashfs so I can't modify it at runtime Sep 30 22:32:40 during rootfs Sep 30 22:33:12 oh that would work them Sep 30 22:38:22 marler899721: there is no technical limitation to getting xattrs preserved, though -- it's all tarballs anyway. You just need to decide on one implementation (either bsdtar/libarchive or gnutar) and use that everywhere Sep 30 22:38:57 Which will take some fiddling with the toolchain. It's on my radar, too, because I'd really like to see that happen (and I don't think it's too complicated to actually make it happen either), but for now I'm busy with other things. Sep 30 23:23:26 New news from stackoverflow: What is the device tree overlay ? its use and its benefits? Sep 30 23:45:02 does anyone know how to make systemd-initramfs invoke initrd.target? for some reason the default target ends up being multi-user.target in my initramfs.. Sep 30 23:55:28 d'oh, of course, there's a variable for that. **** ENDING LOGGING AT Tue Oct 01 02:59:57 2019