**** BEGIN LOGGING AT Mon Sep 09 02:59:58 2019 Sep 09 06:25:42 RP: just a quick test for log collection... gcc logs are 120M, g++ 180M... so storing them in the json as strings = big testresults. So log collection and compression! -> https://github.com/nathanrossi/openembedded-core/commit/cb63deddcaccfae6306a3d7e3571d9aa3eb22867 Sep 09 06:26:44 RP: they compress well, totalling an extra ~10M on top of ~44M of just result data Sep 09 07:38:39 nrossi: sounds good! :) Sep 09 08:02:14 fyi: I'm @ linux plumbers in Lisbon, if anyone needs someone tracked down .. let me know! :D Sep 09 08:02:44 zeddii: my mind. Sep 09 08:03:06 i'll start searching the local pubs ASAP Sep 09 08:03:23 thx Sep 09 08:04:43 "of all the things that i have lost, i miss my mind the most" - ozzy osbourne Sep 09 08:06:22 :D Sep 09 08:18:58 zeddii: proper attribution: https://www.youtube.com/watch?v=4kQ0zynbzh4 Sep 09 09:29:13 New news from stackoverflow: how to build OS in RCAR M3 Starter Kit pro? Sep 09 11:29:30 I'm trying to incorporate swupdate in my yocto build, but I'm having a bit of a hard time understanding how to create/define my image. I have a working image (created with wic), and now I want to include a swupdate "initrd", for performing upgrades. There is nothing in the image recipes provided by swupdate that seems to actually create an initrd, just a normal rootfs. Sep 09 11:37:59 jonmason: are the slides for https://twitter.com/jonmason00/status/1170992260617572352 available do you know? Sep 09 11:42:05 rburton: Not yet, but it will be at https://linuxplumbersconf.org/event/4/contributions/469/ when posted Sep 09 11:42:13 awesome Sep 09 11:42:28 wasn't aware they were talking at LPC Sep 09 11:42:41 zeddii already got his out of the way but I see no slides either Sep 09 11:42:53 https://linuxplumbersconf.org/event/4/contributions/467/ Sep 09 11:43:01 2 Yocto talks, back-to-back Sep 09 11:43:14 micdrop Sep 09 11:43:39 I think he said people were sick of Yocto at that point ;-) Sep 09 12:08:33 jesus Sep 09 12:08:52 it is interesting they are by names I do not recognize Sep 09 12:09:56 ? Sep 09 12:10:38 jon mason's links Sep 09 12:11:09 Who is this Bruce Ashfile guy? Sep 09 12:11:39 Crofton|work: hrhrhrhrhr Sep 09 12:20:04 Crofton|work: ms have been beavering away quietly Sep 09 12:20:31 I understand this is a thing Sep 09 12:21:55 Crofton|work: did I missspell it? Sep 09 12:22:58 no Sep 09 12:23:01 I did Sep 09 12:23:05 Ashfield Sep 09 12:23:45 ok, my spelling is always terrible Sep 09 12:24:34 this distro track is really good. lots of talk about how to make all of testing, confs, etc common for distros (and meta-distros like yocto) Sep 09 12:25:47 RP: so if i fork yocto-ab-helper somewhere and push some changes, i can just test them on the AB by changing the path/rev right Sep 09 12:27:14 rburton: yes Sep 09 12:27:39 rburton: curious what you're changing Sep 09 12:31:23 just going to see if i fixed the last thing breaking a world api-documentation run Sep 09 12:31:35 because i think we should change from sato to world for that Sep 09 12:33:12 RP: uhhhh maybe it should not print every duplicate ptestresult when reporting :) Sep 09 12:33:41 nrossi: we shouldn't have that many :/ Sep 09 12:34:06 RP: qemux86-64+selftest Sep 09 12:34:38 nrossi: I do see the report you mean, its unreadable :/ Sep 09 12:34:53 nrossi: these shouldn't be running multiple times though Sep 09 12:36:45 nrossi: oe-selftest --skip-tests distrodata.Distrodata.test_checkpkg buildoptions.SourceMirroring.test_yocto_source_mirror -T machine -T toolchain-user -T toolchain-system -j 15 Sep 09 12:36:57 nrossi: that shouldn't be running toolchain-* should it? Sep 09 12:37:12 https://autobuilder.yoctoproject.org/typhoon/#/builders/79/builds/368/steps/8/logs/step2d from https://autobuilder.yoctoproject.org/typhoon/#/builders/79/builds/368/steps/8/logs/stdio Sep 09 12:37:21 RP: shouldn't be Sep 09 12:37:30 nrossi: appears it is :/ Sep 09 12:37:31 RP: can i push to that repo? Sep 09 12:37:59 rburton: ssh git@git.yoctoproject.org should tell you Sep 09 12:38:10 i always forget that! Sep 09 12:38:11 rburton: I've no issue if that is what you're asking Sep 09 12:38:30 i'll follow the scheme and map contrib/ross to the pushed branches Sep 09 12:40:25 RP: can i just kick qa-extras using a custom y-ab-h? is that what the nameless branch/repo/rev box is for? Sep 09 12:41:04 if i press start new a-quick i get a y-ab-h branch/repo/rev selector Sep 09 12:41:11 but the same box isn't there for qa-extras Sep 09 12:41:32 rburton: yes Sep 09 12:42:21 rburton: its the nameless box Sep 09 12:42:38 leave project and rev empty, does the right thing i guess Sep 09 12:44:31 rburton: yes, has defaults Sep 09 12:48:39 any pointers on the correct approach to have samba enable its systemd.services? Sep 09 12:53:16 RP: ok so if we can't rename steps from step3b, can we add a description field for the AB to show? Sep 09 13:07:34 if a recipe sets SYSTEMD_SERVICE_${PN} and SYSTEMD_AUTO_ENABLE = "enable", then all service files in the former variable should be enabled in the resulting image - do i read this correctly? Sep 09 13:07:49 rburton: you'd have to dive into the code to figure that out Sep 09 13:19:34 (any comment if i'm readng this right? https://git.yoctoproject.org/cgit/cgit.cgi/poky/tree/meta/classes/systemd.bbclass#n191 ) Sep 09 13:26:46 rburton: I like the idea but not the time right now Sep 09 13:28:40 kanavin_: i'd forgotten that gstreamer had moved away from gtkdoc to https://hotdoc.github.io Sep 09 13:28:46 yay another doc framework to beat up Sep 09 13:29:00 eg 'doc disabled in cross' comments in the gst modules Sep 09 13:29:22 oh c'mon it needs llvm too Sep 09 13:34:30 RP: figured out why the filtering was not working.... my bad should have read argparse better :) - https://github.com/nathanrossi/openembedded-core/commit/a4cdd5bc1a548aa46b9aaf1d1ef30049405c935a Sep 09 13:35:31 nrossi: ah, makes sense. I will admit I was worrying about other things than spotting that! Sep 09 13:35:41 nrossi: I'll rerun with that applied and see where we end up, thanks Sep 09 13:36:47 RP: im introspecting the failing gcc runs atm, slow internet :( Sep 09 13:37:13 nrossi: I was about to ask if the logs gave any insight there Sep 09 13:44:36 RP: "error while loading shared libraries: libmpfr.so.6: cannot open shared object file: No such file or directory" thats the error for all gcc-cross execs Sep 09 13:45:59 nrossi: aha. sysroot dependencies missing Sep 09 13:46:10 nrossi: that could explain a lot! Sep 09 13:46:22 nrossi: or maybe a relocation issue Sep 09 13:46:37 RP: yer uninative related? Sep 09 13:47:06 nrossi: probably more the standard sysroot relocation code Sep 09 13:47:32 nrossi: I suspect its more sysroot poulation dependencies thinking about it Sep 09 13:50:26 RP: just looking at my ubuntu build, i see the issue... on xgcc in gcc-runtime -> Library runpath: [/home/nathan/build/tmp/work/x86_64-linux/gcc-cross-x86_64/ Sep 09 13:51:02 RP: which is why i did not see the issue locally, cause the gcc-cross directory is still there Sep 09 13:51:18 nrossi: ah. This does explain a few things at least Sep 09 13:52:08 RP: is there a function for fixing up rpaths with staging? Sep 09 13:52:41 nrossi: I think you should just need to extent PREPROCESS_RELOCATE_DIRS Sep 09 13:52:57 (and that should let you find chrpath.bbclass and relocatable.bbclass Sep 09 13:54:21 RP: as in put the ${B}/* paths into PREPROCESSS_RELOCATE_DIRS? Sep 09 13:58:21 nrossi: have a look at how its used in chrpath.bbclass Sep 09 13:58:37 nrossi: it will be the path to the files inside the sysroot Sep 09 13:59:05 RP: sure but xgcc is not in the sysroot Sep 09 13:59:29 nrossi: it comes from the stashed gcc build directory doesn't it? Sep 09 14:01:03 RP: ah ok, i see gcc-runtime needs to have rpath_replace run on B Sep 09 14:02:00 nrossi: I'd have thought it would happen as the binaries are stashed in gcc-cross Sep 09 14:02:31 the links would be made relative so they work whereever Sep 09 14:03:23 RP: I guess it would be better to do it then, i will test on the extract to make sure the replacement is correct first Sep 09 14:29:39 I recall recently that someone mentioning about how much faster OE was at building up a system compared to some competitors... does anyone have a reference for that? Sep 09 14:30:28 JPEW: cisco talking about yocto vs their old system using openwrt Sep 09 14:30:35 at elc, presumably Sep 09 14:30:57 rbutron: That was it! Thanks! Sep 09 14:30:59 Hi guys, quick question. What is the difference between DEPENDS and RDEPENDS for *native* packages? I've this native recipe which is installing a script but not running it. This script requires a few native tools. Right now, we depends on the latter in the native recipe and we DEPENDS on the native recipe in places where we use this script. Could RDEPENDS be a better candidate? Sep 09 14:31:04 rburton, JPEW: yes, was a cisco presentation at ELC compared to their old openwrt Sep 09 14:31:32 I wish I had gone to that one now :( Sep 09 14:50:15 Hi, someone knows the better solution to share DL_DIR between 2 Docker containers ? Sep 09 14:51:05 PinkSnake: We bind mount them into our build containers Sep 09 14:52:15 JPEW no issue if the folder is mounted inside several containers in the same time ? Sep 09 14:53:55 PinkSnake: Not sure... I don't have the horsepower to do multiple builds at once :) I *think* the download code uses lock files that allow it to be shared? Sep 09 14:57:03 JPEW Thx for replying(y) ;D I will make a try to be sure :] Sep 09 14:58:08 yeah you can share DL_DIR no problem Sep 09 15:04:44 (as long as filesystem locking is working properly.. which it doesn't always on NFS) just an FYI Sep 09 15:16:35 zeddii, 4.14 kernel cache. any reason why it is not updated to .120 ? Sep 09 15:16:53 lol my guess, nobody has sent him .120 :) Sep 09 15:17:23 the kernel is a .120 but the cache no match Sep 09 15:18:14 I'll send it, just no sure if was left out on purpose Sep 09 15:30:16 New news from stackoverflow: Yocto - List providers of a package Sep 09 15:39:23 RP: looks like this does the trick, https://github.com/nathanrossi/openembedded-core/commit/788a07024e72ade10dfd50b949599e8cac285050 Sep 09 15:40:19 RP: will clean up and sort out the WIP patches tomorrow, feel free to test that rpath fix if theres time free on the autobuilder Sep 09 15:41:12 nrossi: great, thanks. Will submit for testing Sep 09 15:42:15 JPEW: after much head scratching and a few experiments, we can't use this idea of a nodeps hash, it won't work Sep 09 15:51:04 vmeson: Would we be able to find someone to help with https://bugzilla.yoctoproject.org/show_bug.cgi?id=13504 Its merged into master (systemd on mips issue) Sep 09 15:51:05 Bug 13504: normal, Undecided, ---, ross.burton, NEW , [master-next] qemumips-lsb sytemd failure Sep 09 15:51:35 * RP is also worried about https://bugzilla.yoctoproject.org/show_bug.cgi?id=13506 Sep 09 15:51:36 Bug 13506: normal, Undecided, ---, ross.burton, NEW , [master-next] strace ptest timed out Sep 09 15:51:49 (new timeouts appeared in ptest results) Sep 09 15:52:03 that one is x86 and arm Sep 09 16:03:17 nrossi: that didn't work so well on the autobuilder :/ Sep 09 16:22:03 RP: ah right yer.... completely forgot that it trashes the rpaths so that installed binaries have broken rpaths now :( back to the drawing board Sep 09 16:23:30 RP: i did test it locally but i only built without uninative Sep 09 16:29:39 RP: Oh? Why is that? Sep 09 16:44:52 JPEW: key input information that governs the hash equiv may be in an earlier dependent task so we have to have the previous task checksums included Sep 09 16:45:47 RP: OK, makes sense. I hadn't yet had a change to really dig into it. Sep 09 16:46:07 RP: Looks like we need a streaming equivalence server then? Sep 09 16:50:19 is there a good example of using pypi.bbclass? Sep 09 16:51:36 fullstop: meta-python is 99% pypi users Sep 09 16:51:51 let me poke around Sep 09 16:55:01 hmm. mine looks similar, but it's not getting python3 into sysroot Sep 09 16:56:38 JPEW: yes :/ Sep 09 16:56:55 JPEW: I got as far as coding the nodeps thing only to realise it was a deadend Sep 09 16:57:36 RP: That's unfortunate. On the bright side, the streaming changes should be minimally invasive. I'll work up some patches to try on the AB Sep 09 16:58:14 RP: I also have the patch to fix the reproducible build race condition, have you seen it since Sep 09 16:58:19 ? Sep 09 16:58:26 oh, a clean fixed it up Sep 09 17:12:03 hey guys! I want to override /etc/systemd/system.conf but it seems like I can't just provide another file as it gets taken from systemd's sources by systemd-conf recipe ;-( Sep 09 17:12:30 what's the best course of action? as soon as I just install it in my systemd_%.bbappend I got conflicts with systemd-conf Sep 09 17:13:36 I guess, stupid question. I can override it in do_install_append Sep 09 17:19:25 yeah, override it in the systemd-conf recipe Sep 09 17:19:30 or just use recipetool appendfile Sep 09 17:19:36 armpit. I have stable updates for all the kernels pending. once I'm back from plumbers I'll send them out and fixup the kernel-cache kver, etc. Sep 09 17:20:08 zeddii, k. thanks Sep 09 17:21:47 oh, recipetool. need to learn that. thanks! Sep 09 17:25:18 appendfile does exactly this, overrides a target file, but it figures out hte recipe automatically using oe-pkgdata-util and appends it for you Sep 09 18:12:22 is the repo at https://github.com/openembedded/meta-openembedded official? Sep 09 18:12:24 kergoth, thank you! will learn more about this. meanwhile overrode systemd-conf manually to my success :-) Sep 09 18:12:29 * kergoth nods Sep 09 18:12:30 np Sep 09 18:16:22 if that repo is official, the warrior branch is messed up. Sep 09 18:19:11 that depends on what you mean by official Sep 09 18:19:27 it's not part of the yocto project directly, it is part of OE Sep 09 18:19:41 Well, I take all of that back. :-) Sep 09 18:20:08 cgit branch switching in my browser wasn't actually switching the branch. Sep 09 18:43:33 well warrior in OE supplies YP poky so if its broking in OE, its broken in Poky (YP) Sep 09 18:47:55 I'm getting better with this stuff.. but, wow, some packages (not included) are just hell. Like scipy. Sep 09 18:50:29 my toolchain does not have fortran support, which matplotlib requires. If I add that everything will rebuild and I'm out for ~12 hours. :-( Sep 09 18:52:22 fullstop: cough up a couple of bucks, spin up high powered ec2 instance, be done in 2hrs :) Sep 09 18:52:46 haha, it's ~12 hours if it finishes. I still need to keep an eye on it. Sep 09 18:53:06 I can't believe that fortran is still in use for new things Sep 09 18:53:12 but I guess people like it Sep 09 18:55:14 fullstop: it has its uses, especially for highly optimized corner cases Sep 09 20:48:27 fullstop:github.com repos are mirrors for easy read access, I can see warrior is fine on it, how it is messed up ? Sep 09 21:04:16 hi guys, i'm trying to customize the initrdscript init-install, so i created a new initramfs-module and I'm adding this module to core-image-minimal-initramfs on INITRAMFS_SCRIPTS Sep 09 21:04:38 then i'm setting the LABELS_LIVE to "custom" Sep 09 21:04:54 my initramfs-module creates /init.d/custom.sh Sep 09 21:05:15 but when I build the image, the label still install Sep 09 22:47:33 khem: https://autobuilder.yocto.io/pub/non-release/20190909-13/testresults/testresult-report.txt (at least the top part looks good but its a long log with all the failures underneath) Sep 09 22:47:41 nrossi: getting there :) **** ENDING LOGGING AT Tue Sep 10 02:59:57 2019