**** BEGIN LOGGING AT Mon Jul 03 03:00:02 2017 Jul 03 08:00:47 Hello guys !! Does anibody know how to add an user to a bitbaked image along with root ? Jul 03 08:00:57 i'm struggling Jul 03 08:02:12 fberg: here's your example: https://git.yoctoproject.org/cgit.cgi/poky/tree/meta-skeleton/recipes-skeleton/useradd/useradd-example.bb Jul 03 08:03:47 ty Leto ! Jul 03 08:34:11 would OE-core upstream be interested in merging an image rootfs test for file ownership? Checking that the uid/uname of each file is present in /etc/passwd, the gid/gname in /etc/group, that id/names are conistent between tar metadata and shadow db etc Jul 03 08:35:10 we have this check in our local OE branch and we're wondering if it would provide any value upstream because there were bugs leading to inconsistent permissions in the past Jul 03 08:36:32 like https://bugzilla.yoctoproject.org/show_bug.cgi?id=7724 Jul 03 08:36:34 Bug 7724: major, Medium, 2.4 M1, leonardo.sandoval.gonzalez, RESOLVED FIXED, useradd.bbclass users cannot be safely referred to by dependent recipes due to sstate-induced race with base-passwd Jul 03 08:38:14 sounds sensible Jul 03 08:40:40 right now we have implemented that logic as a stand-alone script in our local repo, I'm wondering what's the best way to integrate this check into oe-core, are there QA checks for images? Jul 03 08:43:18 yes Jul 03 08:43:52 look up IMAGE_QA_COMMANDS Jul 03 08:44:24 thank you Jul 03 09:01:03 pyro nightly-oecore seems to be using bitbake branch 1.32? Jul 03 09:04:07 yeah, my fault. I triggered it with the wrong branch Jul 03 09:10:06 hello ! I've generated a systemd based image with yocto. Does anibody knows why if i "cat /var/log/messages" the systemd daemon repeatidly says: serial-getty@ttyS0.service: Service has no hold-off time, scheduling restart. Stopped Serial getty on ttySo. Started serial getty on ttyS0. /dev/ttyS0: not a tty. Jul 03 09:10:41 the daemon generetes these log filese continuosly. by the wya /dev/ttyS0 is listed under the /dev folder Jul 03 09:10:49 fberg: are you logged in through the serial port? or rather ssh? Jul 03 09:11:04 now trhough ssh Jul 03 09:12:06 I've disabled the serial port. The login is now on tty1 instead of serial0 Jul 03 09:12:30 fberg: then i guess your kernel is missing some configuration thing. don't remember it completely, but i think it was fhandle Jul 03 09:12:35 cause i need the serial port to use the gps Jul 03 09:13:16 fberg: here you go: https://cgit.freedesktop.org/systemd/systemd/tree/README -> REQUIREMENTS Jul 03 09:13:21 mmm sounds complicated.. Jul 03 09:13:37 I'll buy you a beer Leto Jul 03 09:14:20 fberg: no beer needed. :-) Jul 03 09:15:18 so basically I need to set CONFIG_FHANDLE ? Jul 03 09:15:38 fberg: you basically have to make sure your kernel config has all of those enabled. Jul 03 09:16:24 ops... ok ! there is a way to have it done using the local.conf file or must be done with bitbake only ? Jul 03 09:17:18 fberg: no, this means you have to modify your kernel config. technically, this is totally unrelated to bitbake and your local.conf Jul 03 09:18:24 as i know to change the kernel configuration I have to run something like "$ bitbake -c menuconfig virtual/kernel" Jul 03 09:19:04 fberg: um.... well its a means to access the kernels configuration TUI, but actually more like a wrapper. Jul 03 09:19:21 fberg: where does your kernelconfig come from at the moment? Jul 03 09:19:31 linux-raspberry Jul 03 09:19:59 shipped from meta-raspberry Jul 03 09:21:02 those seem to pull in the upstream defconfig: https://git.yoctoproject.org/cgit.cgi/meta-raspberrypi/tree/recipes-kernel/linux/linux-raspberrypi.inc Jul 03 09:21:44 i personally would replace teh dummy defconfig file https://git.yoctoproject.org/cgit.cgi/meta-raspberrypi/tree/recipes-kernel/linux/linux-raspberrypi/defconfig Jul 03 09:21:53 with my custom, nailed-down config Jul 03 09:24:32 as now I'am using the default configuration Jul 03 09:25:10 I'll try to add the systemd required configuration to the defconfig file Jul 03 09:27:39 please be patient with me. I'm a noob Jul 03 11:53:12 I have to run an installer file from do_install of a recipe but it requires sudo permission, can one one please point me out, how it should be done ? Jul 03 11:53:39 Ramose: by not using it. Jul 03 11:53:44 Ramose: by pathing it. Jul 03 11:53:49 *patching Jul 03 11:53:57 Ramose: or by replicting the logic inside. Jul 03 11:54:29 its .bin file(installer) that needed to be run Jul 03 11:54:34 inside do_install the installer will think it is root anyway Jul 03 11:54:46 mwhahahaha Jul 03 11:54:47 etc Jul 03 11:54:48 if it starts out like this, then the installer script is most likely to be broken for crossompilation / non-native installation anyways. Jul 03 11:55:53 *if* the installer is essential to run and it doesn't check if it is root already and forcibly runs sudo, then I can only suggest putting a fake sudo binary on $PATH somewhere before running the installer :) Jul 03 11:57:54 i remember too many people claiming "i really need to run that installer script" to believe it anymore - when in the worst case said installer actually only was a bunch of sudoed commands in a shell script that would only work on a specific ubuntu revision. Jul 03 11:57:59 http://ortc.io/workspace/meta-boot2qt/sources/meta-ti/recipes-ti/devtools/ti-ccsv6-native_6.1.3.00034.bb Jul 03 11:58:12 .bin is installer here Jul 03 11:58:53 its starts executing with sudo permission but few of package inside it work with sudo only Jul 03 11:59:45 sorry , starts executing *without* sudo permission Jul 03 12:00:38 to me it looks like it is related to https://git.yoctoproject.org/cgit.cgi/meta-ti/tree/recipes-ti/devtools Jul 03 12:00:43 moan at ti? :) Jul 03 12:00:59 threaten denix with headnuts! Jul 03 12:01:09 define "with sudo only", do you mean it forcibly runs sudo Jul 03 12:01:16 if it does then fakesudo it Jul 03 12:01:48 how ? Jul 03 12:02:05 write a script that has the same calling interface as sudo but doesn't actually call sudo Jul 03 12:03:09 as i said, do_install is a fakeroot task so the environment is effectively (not but really) root Jul 03 12:03:24 but first moan at TI :) Jul 03 12:04:51 rburton: I actually run it from do_install but it didn't make any diference though Jul 03 12:15:01 rburton: can installer be run using fakeroot from recipe file ? Jul 03 12:15:21 don't have any clue what should I do ? Jul 03 12:15:31 do_install is fakeroot already Jul 03 12:15:40 oh thats a native recipe Jul 03 12:15:44 which isn't fakeroot Jul 03 12:16:08 try setting do_install[fakeroot]="1" in the recipe to force that Jul 03 12:16:20 assuming you're having problems with the recipe you posted btw Jul 03 12:16:21 ok . let me try Jul 03 12:18:51 yeah, I am trying to run same recipe. Jul 03 12:25:33 rburton: still seeing the same issue Jul 03 12:34:18 Is there a way to build SE kernel modules into Yocto? Jul 03 12:35:38 tcpdump: i think there's nothing that keeps you from changing the kernel config. the question is more, are the corresponding userspace tools already packaged. Jul 03 12:38:28 I see Jul 03 12:42:37 Alright - one more issue (for now). Im trying to create a service script for an openvpn client. I put this unit file in /etc/systemd/system/opevpn@.service : https://2048-bit.com/bin/?6626aa9de7dbab3e#iCKejun50D/fyW0/I/SFme5kJjUJjYU9+gyHGHcNTyQ= Jul 03 12:44:34 1 Jul 03 12:46:41 When I issue systemctl start openvpn@client it just hands and produces the following: https://2048-bit.com/bin/?b7912b498095e9b0#MktCArcZUE5jv5n7YcmvYS36to3wduBydqnEIQfslwQ= Jul 03 12:46:45 but never drops back ot the CLI Jul 03 12:46:53 so it looks like its working but never forking? Jul 03 12:48:15 hmm Jul 03 12:48:27 when i did systemctl enable openvpn@client Jul 03 12:48:32 tcpdump: not really, Type=notify means that the daemon should call sd_notify() internally Jul 03 12:48:33 and rebooted it seems towork. Jul 03 12:48:35 so whatever... Jul 03 12:49:07 zero_note: any sd_notify? is that to notify the system of a new interface? Jul 03 12:54:10 Hi guys ! Jul 03 12:54:35 I'm facing some trouble installing pulse-audio, I can't find alsa-plugins-pulseaudio-conf ... Has anyone face this problem ? Jul 03 12:54:42 tcpdump: no, it is used by systemd to handle units dependencies. When a .service have Type=notify we expect that the related daemon (or whatever) in ExecStart= call the sd_notify() API to inform systemd about its "ready" state, so systemd can consider the unit as started, and the related dependencies could be launched as well Jul 03 12:55:10 hmm Jul 03 12:55:12 "ERROR: Required build target 'rpi-basic-image2' has no buildable providers. Missing or unbuildable dependency chain was: ['rpi-basic-image2', 'pulseaudio-module-bluez5-discover', 'alsa-plugins-pulseaudio-conf']" Jul 03 12:55:20 So am I using that wrong? Jul 03 12:55:23 before that sd_notify, the unit will be "Starting...", until timeout Jul 03 12:55:45 It looks like the tunnel is coming up but dieing shorly after. Jul 03 12:55:58 tcpdump: seems not, because ever the original .service has that Type=notify Jul 03 12:56:53 tcpdump: could be related to your vpn config Jul 03 12:57:43 https://2048-bit.com/bin/?a3c3ba0ffce2ab7f#n3mKcLc08ipZg7CitO3tKsWdS9UP7gQKwJRwzwhNCBE= Jul 03 12:57:58 thats journalctl -xe Jul 03 12:58:04 it seems to run, then die. Jul 03 12:58:05 :/ Jul 03 12:58:08 Hmm... Jul 03 13:01:54 YCN-: you need alsa-plugins, which provides alsa-plugins-pulseaudio-conf Jul 03 13:02:40 Yes I added it in IMAGE_INSTALL ! Jul 03 13:03:01 should I add it elsewhere or add something specific in DISTRO_FEATURE? Jul 03 13:05:31 YCN-: are you trying to install pulseaudio, right? Jul 03 13:09:17 yes indeed Jul 03 13:12:49 YCN-: you should add it to DISTRO_FEATURES Jul 03 13:13:21 YCN-: DISTRO_FEATURES_append += " pulseaudio" Jul 03 13:13:52 I did that too ^^ Jul 03 13:19:26 YCN-: have you already added packages like pulseaudio-server and pulseaudio-modules-* ?(the latter depends on you needed) Jul 03 13:20:22 YCN-: I mean to your IMAGE_INSTALL Jul 03 13:25:26 I did add those in IMAGE_INSTALL : pulseaudio-server \ pulseaudio-module-bluetooth-discover \ pulseaudio-module-bluetooth-policy \ pulseaudio-module-bluez5-device \ pulseaudio-module-bluez5-discover \ Jul 03 13:25:43 and also those one : alsa-dev \ alsa-lib \ alsa-utils \ alsa-utils-scripts \ alsa-plugins \ Jul 03 13:27:12 YCN-: ok, and despite so you have no pulseaudio instance running on your rpi ? Jul 03 13:49:37 no I don't have anything... Jul 03 13:49:41 that's pretty weird Jul 03 13:51:08 But there's another thing weird in my distro it's that I should use systemd because I did : VIRTUAL-RUNTIME_init_manger = "systemd" Jul 03 13:51:27 but I don't have sustemd, but sysctl Jul 03 13:52:39 YCN-: systemctl != sysctl Jul 03 13:52:53 YCN-: too many werid things Jul 03 13:52:59 weird* Jul 03 13:54:03 oh well, but then why don't I have anything in : /etc/systemd/system/ Jul 03 13:54:13 I don't even have /systemd btw Jul 03 13:55:52 YCN-: hum, do you have that set... where? Jul 03 13:56:21 hopefully in your local.conf Jul 03 13:56:54 YCN-: did you read the docs on how to switch to systemd? Jul 03 13:59:31 Hm, I put that in my receipe. May be that's the first problem I should work out even before thinking about pulseaudio Jul 03 14:00:37 Indeed I do have in my local.conf sysvinit ! Jul 03 14:00:50 I'll change that first and check if that's working better Jul 03 14:03:00 YCN-: you can take a look here http://www.yoctoproject.org/docs/2.2/dev-manual/dev-manual.html#using-systemd-exclusively Jul 03 14:10:04 okay thanks I'll check that. And I hope it will help solve my problems :) Jul 03 14:18:05 hello guys. anybody knows whoch is the bestway to tell yocto not to load a bunch of modules ? I usually set a blacklist under /etc/modprobe.d/blacklisted-module.conf. There is a better way different from generating a recipe ? Jul 03 14:21:28 fberg: exclude those modules from your kernel defconfig if dont need them Jul 03 14:40:11 khem: https://autobuilder.yocto.io/builders/nightly-qa-systemd/builds/315/steps/BuildImages/logs/stdio Jul 03 14:50:29 okay it's seems that's there's something wrong with that : PACKAGES += "${@bb.utils.contains('PACKAGECONFIG', 'pulseaudio', 'alsa-plugins-pulseaudio-conf', '', d)}" Jul 03 14:50:44 because in the end nothing did the installation Jul 03 14:50:55 of the package alsa-plugins-pulseaudio-conf Jul 03 14:51:15 so I added in the hard way by modifying the line as forllowing : Jul 03 14:51:29 PACKAGES += "alsa-plugins-pulseaudio-conf" Jul 03 14:51:51 (that's taken from alsa-plugins_1.1.1.bb) Jul 03 14:52:27 YCN-: check your rootfs manifest Jul 03 14:52:35 It's seems like others than me had the issue with RPI. https://www.raspberrypi.org/forums/viewtopic.php?f=71&t=171080&p=1182283#p1182283 Jul 03 14:53:47 in the end that was only reversing this patch : https://patchwork.openembedded.org/patch/94605/ Jul 03 14:53:49 YCN-: that only happens if PACKAGECONFIG contains pulseaudio, which happens by default if DISTRO_FEATURES contains pulseaudio Jul 03 14:53:52 so what's wrong finaly ? Jul 03 14:54:08 yes but DISTRO_FEATURE did have pulseaudio Jul 03 14:55:47 rburton: seems easy to fix. I will add the patch to systemd on top of hardening branch Jul 03 14:59:58 I think that was my fault in the end. I did put pulseaudio in DISTRO_FEATURE_append Jul 03 15:00:10 in my receipe and not in my local.conf Jul 03 15:00:16 where I should have. Jul 03 15:00:44 Why doesn't it works when it's done in the receipe with a DISTRO_FEATURE_append? Jul 03 15:01:00 what recipe Jul 03 15:01:20 the recipe of my image, where I add all the package Jul 03 15:01:45 because the assignment is only valid inside the image scope, at which point the packages have been built already Jul 03 15:02:01 an image is just another recipe Jul 03 15:02:10 and assignments inside recipes only impact the recipe, not other recipes Jul 03 15:03:49 oh okay that makes sense thanks a lot ! Jul 03 15:07:03 the what should one put in local.conf and not in the recipe? And vice versa? Jul 03 15:07:07 *then Jul 03 15:07:39 DISTRO_FEATURES is global so unless you've very specific needs it should go in your distro configuration, or local.conf Jul 03 15:09:03 Well in my case I wanted to do a bluetooth A2DP station using pulse audio and alsa Jul 03 15:09:26 untill I had to install pulseaudio everything was good and then it kind of get very messy Jul 03 15:09:46 since I had those dependency problems which are not that easy to understand. Jul 03 15:09:56 becase you've not enabled pulseaudio ;) Jul 03 15:10:10 turn it on in local.conf and it should all work Jul 03 15:10:41 yep that's what I'm doing now, and I even add alsa stuff because my pulse audio modules needs alsa too Jul 03 15:11:23 so I don't know exactly what I should or shouldn't add inside my recipe Jul 03 15:12:59 But anyway thanks for your help that was really usefull :) Jul 03 15:18:30 YCN-_: in short: whenever something affects more than one recipe, it needs to go into a conf file Jul 03 15:28:46 rburton: i think we should replace sysklogd with rsyslog in lsb images. no distro uses sysklogd now a days since last 5 years Jul 03 15:29:45 i will open a bug for this, since it will need few changes on testing side too Jul 03 15:34:17 rburton: What tool did you use to find that openssh vs openssh-sshd package issue? Jul 03 16:00:17 rburton: http://git.openembedded.org/openembedded-core-contrib/commit/?h=kraj/hardening-fixes&id=1ca93f97b2face0e89b33e645bd7359e3f080add should fix the xlocale issue Jul 03 16:00:30 I pushed it to branch as well as to mailing list Jul 03 16:09:04 thanks khem Jul 03 16:09:17 jpew: buildhistory-diff Jul 03 16:11:50 rbuton: Thanks! I'll get that fixed tonight. Jul 03 16:16:51 rburton: I am looking at https://autobuilder.yocto.io/tgrid, everything else seems to be chugging along except this systemd niggle Jul 03 16:16:55 yeah Jul 03 16:17:18 rburton: btw. is yocto.io new AB or replacement ? Jul 03 16:17:58 i believe the plan is to make it the main one at some point Jul 03 17:08:45 hey everyone - I have taken OpenVPN and tried to add a unit file to automatically start the client as a service. When I do systemctl start openvpn@client it never drops back to the CLI, but rather sits for about 35 seconds total. During that duration the tunnel comes up, can pass traffic, etc. After about 30 seconds the service seems to shutdown, the tunnel drops, etc. Jul 03 17:09:06 https://2048-bit.com/bin/?b564dbbe02368d8f#FmORY5gzTtzk+aXhzeLLROxfloEKj22wleKNPXU737E= Jul 03 17:09:06 Heres the unit file, and the journalctl logs... Jul 03 17:09:28 Its like something is preventing the service from staying running - the vpn itself seems to be working OK. Jul 03 17:09:29 between here: Jul 03 13:55:05 yocto openvpn[2691]: Initialization Sequence Completed Jul 03 17:09:29 and Jul 03 13:56:37 yocto systemd[1]: [[0;1;39mopenvpn@client.service: Start operation timed out. Terminating.[[0m Jul 03 17:09:36 Is there something wrong with my unit file? Jul 03 18:28:16 got it - so no one wastes any time reading my question. Jul 03 19:59:08 rburton: I think mixing glibc 2.26 with hardening has caused the ppc segfault issue and x86 ovmf failures, Jul 03 19:59:30 yes, that definitely my guess too Jul 03 19:59:34 ovmf build system is very hacky and hard to understand Jul 03 19:59:53 rburton: let me pull out the glibc 2.26 from that branch Jul 03 20:00:05 I was too optimistic Jul 03 20:02:23 rburton: I have dropped the glibc 2.26 and dependent patches Jul 03 20:02:31 and refreshed the branch Jul 03 20:03:04 that will get the hardening solved Jul 03 20:03:21 I can see whats going on with glibc 2.26 regressions Jul 03 20:08:37 rburton: I fixed a mesa issue for meta-freescale as well as rebased 17.1.4 upgrade Jul 03 20:09:10 rburton: v2 is sent Jul 03 20:09:24 thanks, khem and otavio Jul 03 20:24:02 khem: fired again Jul 03 20:27:55 gr8 Jul 03 21:27:00 RP: I am testing multiconfig support and it seems it has some problems in sharing tmp among different machines Jul 03 21:27:05 RP: is it known? Jul 03 21:47:40 otavio: are you changing libc too ? Jul 03 21:49:07 otavio: There is a fix for some issues I'm working on Jul 03 21:49:34 otavio: http://git.yoctoproject.org/cgit.cgi/poky-contrib/commit/?h=rpurdie/wip-rss2&id=92d0d24c27da147be8ddb7a7ea8e056fd0294235 Jul 03 21:57:24 khem: I am not; same distro Jul 03 22:01:03 ok Jul 03 22:01:16 rburton: it seems ppc failure is happening with 2.25 too ... Jul 03 22:01:26 damn ppc Jul 03 22:01:49 RP: I applied it and I will leave it building over night Jul 03 22:03:57 otavio: patch isn't quite right yet but shouldn't hurt anything. We may need to create a new separate event... Jul 04 01:38:43 rburton: figured and fixed the ppc test failures seen on AB, pushed a patch to pull branch Jul 04 01:38:45 http://git.openembedded.org/openembedded-core-contrib/commit/?h=kraj/hardening-fixes&id=084aec82f283f06e584b175b0388fc273a25a702 Jul 04 01:40:23 now kraj/hardening-fixes should be good to go, I think you can try to just rerun nightly-ppc-lsb Jul 04 01:40:34 since this patch only affect ppc **** ENDING LOGGING AT Tue Jul 04 03:00:01 2017