**** BEGIN LOGGING AT Wed Jul 11 03:00:02 2018 Jul 11 06:42:26 New news from stackoverflow: How to build yocto sdk (meta-toolchain) on Dockerfile for Eclipse Che and get permissions favourable for meta-toolchain? Jul 11 06:49:34 hi guys Jul 11 06:50:08 I am trying to bitbake an core-image-sato image but I get this error: Jul 11 06:50:18 Problem: package openssh-7.6p1-r0.i586 conflicts with dropbear provided by dropbear-2017.75-r0.i586 - package packagegroup-core-ssh-dropbear-1.0-r1.noarch requires dropbear, but none of the providers can be installed - conflicting requests (try to add '--allowerasing' to command line to replace conflicting packages or '--skip-broken' to skip uninstallable packages) Jul 11 06:51:08 if I try to run the image with the --allowerasing or --skip-broken then I get a message that the option does not exist Jul 11 06:51:22 any points on how I can fix this? Jul 11 07:13:42 Alexp: do you want dropbear or openssh? You probably need to choose which, then find out what is pulling in the one you don't want and stop it Jul 11 07:42:50 hello, I have built core-image-minimal, and ran it with runqemu qemux86, all good. then I found out it doesn't have usb-host support so I have added libusb to PACKAGECONFIG... rebuild... runqemu kernel fails with "FS: Cannot open root device "vda" or unknown-block(0,0): error -6", I have deleted whole build directory and rebuild. but the error persists, what could cause qemu config file to be wrong? (if that is the case) Jul 11 07:45:02 since then I have removed libusb from PACKAGECONFIG of qemu-native... but machine still doesn't find rootfs Jul 11 07:58:51 relevant qemu arguments: -drive file=/...file-exists.rootfs.ext4,if=virtio,format=raw Jul 11 08:03:28 good morning/* Jul 11 08:04:00 what is export ding in a .bb recipe line like this: export VAR="value" ? Jul 11 08:05:05 is the exporting to any subshell started by that recipe? Jul 11 08:17:33 morning Jul 11 08:17:50 cornel: yes Jul 11 08:18:12 cornel: in particular any shell based tasks like do_configure/compile/install Jul 11 08:18:22 good morning Jul 11 08:21:12 RP: morning Jul 11 10:20:01 I am no longer able to build the latest Yocto release (sumo branch) when downloading the source freshly - anyone else having this problem? Jul 11 10:20:21 Building not the branch but the tag works - because there are other/older versions of some packages used. Jul 11 10:20:55 I can show the problem easily when attempting to just build pseudo-native: https://pastebin.com/AY9rnVrY Jul 11 10:26:21 rettichschnidi: looks like your local clone of pseudo is bust. that sha is in pseudo master Jul 11 10:26:46 if that was a clean tmpdir then delete all the pseudo bits in DL_DIR, specifically the clone in the git2/ folder Jul 11 10:27:51 rburton: this was totally fresh (logs are in the pastebin) Jul 11 10:27:59 and you don't have any funky proxies? Jul 11 10:28:56 I found a conversion on the ML (https://lists.yoctoproject.org/pipermail/poky/2018-May/011338.html) regarding this - but I am using my home connection, no proxy Jul 11 10:29:59 But I will try again using another ISP Jul 11 10:31:23 Using tags/yocto-2.4 works, while using branch sumo or tags/yocto-2.5 does not Jul 11 10:33:31 the sha is refers to exists: https://git.yoctoproject.org/cgit/cgit.cgi/pseudo/commit/?id=fddbe854c9db058d5a05830d3bcdd4233d95ee2e Jul 11 10:33:50 outside of yocto can you clone that repo? Jul 11 10:34:04 just git clone git://git.yoctoproject.org/pseudo Jul 11 10:34:06 yes, I can Jul 11 10:34:16 Gonna post some more, likely relevant information soon Jul 11 10:34:53 i guess go to the clone that it made in dldir and poke at it. is it totally corrupted, or just old, or what Jul 11 10:35:10 It seem that (for unkonwn reasons) bitbake can not fetch git, then gets the tarball, unpacks it, deletes the remote (in the git repo), adds the new one twice and fail the second time Jul 11 10:35:32 hm yeah maybe the git tarball on the mirror is old Jul 11 10:35:38 yes, it looks like Jul 11 10:35:45 pastebin the *full* fetch log for pseudo-native please Jul 11 10:36:09 but it also seem that bitbake tries to using this tarballs state and fetch to update it Jul 11 10:36:16 (from a clean dldir) Jul 11 10:36:27 will do Jul 11 10:38:47 i suspect that you've got local git config which is breaking in the sort-of-jail we run builds in Jul 11 10:45:06 hey, me again with same problem, runqemu doesn't find rootfs, there are no partitions listed after "here are available partitions", just after VFS: Cannot open root device "vda" or unknown-block(0,0): error -6 Jul 11 10:45:46 I have build core-image-minimal, with runqemu qemux86 Jul 11 10:46:10 rburton: You are right! Deleting my ~/.gitconfig and the builddir (including sstate and dl dir) allows me to build pseudo-native successfully! Jul 11 10:55:20 ok, virtio blk device was missing, added it with bitbake linux-yocto -c menuconfig, how is that possible, I have removed build directory and rebuilt everything Jul 11 10:57:19 rettichschnidi: curious what in there was breaking it Jul 11 10:57:49 rburton: I am bisecting Jul 11 11:02:16 This changes to my ~/.gitconfig fixes the problem: Jul 11 11:02:16 -[branch "master"] Jul 11 11:02:17 - remote = origin Jul 11 11:02:17 - merge = refs/heads/master Jul 11 11:02:50 I am not even aware of ever having put this lines into ~/.gitconfig Jul 11 11:11:05 looks like a git config went wrong Jul 11 11:11:10 that should never be in the global config Jul 11 11:18:31 rburton: indeed Jul 11 11:18:37 rburton: thanks a lot for your support! Jul 11 11:23:36 rburton: there is an open bug with halstead about the tarballs Jul 11 12:07:06 RP: pushed a new mut Jul 11 12:15:07 RP: thank you very much Jul 11 12:15:45 and without that 'export' the variable will be ignored by those scripts? Jul 11 12:16:02 i mean, why not export all by default? Jul 11 12:39:35 cornel: our number of variables is large and rebuilds would be more frequent as we couldn't tell which tasks depend upon which variables Jul 11 12:40:52 rburton: thanks. Need to distill the known broken or not to be merged buts out of that to run a new test build Jul 11 12:41:47 RP: thank you very much Jul 11 12:44:36 RP: i take offence at the idea that mut may contain non-perfect changes Jul 11 12:44:50 oh there's new systemd patches aren't there Jul 11 12:45:05 rburton: and one you've rejected ;-) Jul 11 12:45:21 from tip to fstests is good Jul 11 12:45:40 rburton: so my point stands? ;-) Jul 11 12:45:48 that's a tweaked libdrm as i can't reproduce JaMa's util-macros issue Jul 11 12:46:21 rburton: ok Jul 11 12:48:10 rburton: I was just checking it now :) Jul 11 12:48:11 configure.ac:# Require xorg-macros minimum of 1.12 for XORG_WITH_XSLTPROC Jul 11 12:48:11 configure.ac: [m4_fatal([must install xorg-macros 1.12 or later before running autoconf/autogen])]) Jul 11 12:48:22 but this part isn't included in meson configuration afaik Jul 11 12:48:47 yeah, meson basically has native support for all the util-macros stuff Jul 11 12:49:06 it was a bug that we didn't have the dependency but its moot with meson Jul 11 12:49:21 should I resend it? Jul 11 12:49:46 or the tweaked one is without the dependency? Jul 11 12:52:30 \/ Jul 11 12:52:34 | Jul 11 12:53:46 rburton: -next is running Jul 11 14:42:39 able to bitbake swupdate. Jul 11 14:42:50 see happy man ^^^^ Jul 11 14:48:47 in the last section, Template for recipe using the class, should this be a new .bb file or appended into an existing .bb file, or otherwise? https://sbabic.github.io/swupdate/building-with-yocto.html Jul 11 14:57:03 yates: it is a new .bb Jul 11 15:07:17 mckoan: thanks. apparently it is not enough to simply create the file (i called it swu.bb and placed it into the "recipes-support" folder of our company's meta- layer. when i build my normal image, i see no extra files in my deploy Jul 11 15:09:35 if i "bitbake swupdate-image" i do see some new files prefixed with swupdate-image. but i see no .swu file Jul 11 15:11:41 you'd have to put the recipe somewhere bitbake will actually parse it (read conf/bblayers.conf in that layer, specifically BBFILES) and then actually bitbake that recipe Jul 11 15:12:30 kergoth: minibase seems interesting project, have you fiddled with it yet ? Jul 11 15:13:05 not yet, no, just seemed like it had interesting priorities and could be worth playing with as another alternative for minimal rootfs for embedded or containers, but never tried it Jul 11 15:13:26 kergoth: yes indeed Jul 11 15:21:22 khem, hi Jul 11 15:21:57 I am slowly progressing with the musl/gcc regression (LDRD unaligned addr) Jul 11 15:22:24 what I never tested is bfd vs. gold Jul 11 15:56:41 after making a small change to a bbappend bitbake rebuilt many packages, Just for curiiousity I called bitbake-diffsigs -t myimage do_image and now it seems that command went into an endless loop. It has been running for well over an hour and from stracing it it appears it reads the same stamp files again and again Jul 11 15:56:55 ant_home: I think its a gcc bug Jul 11 15:57:21 is that something that I could have caused by bad recipes??? Jul 11 16:06:15 khem, after Ithis https://bugs.chromium.org/p/chromium/issues/detail?id=468473 read Jul 11 16:07:15 khem, I am tryng to lure some hacker on #gcc and #asm Jul 11 16:07:51 the disassembly is clear Jul 11 16:07:52 https://pastebin.com/ZY94mS1R Jul 11 16:08:20 I suspect that musl patch reducing the volatile int array from 2 to 1 did misalign Jul 11 16:08:44 somehow the dirstream structure / buffer Jul 11 16:09:36 now, I am almost lost at the problematic instruction Jul 11 16:09:38 e280c020 add ip, r0, #32 Jul 11 16:10:00 here is good, #28 (array -1) is bad Jul 11 16:10:29 maybe add some padding? Jul 11 16:14:25 ant_home: its bad because 28 is not 8byte aligned and LDRD needs to be 8byte aligned Jul 11 16:14:54 it is 28 because we had lock[2] and now is lock[1] I guess Jul 11 16:14:55 so gcc is ignoring the LDRD alignment fact Jul 11 16:15:01 right Jul 11 16:15:09 so earlier the issue was hidden Jul 11 16:15:25 what is type of lock Jul 11 16:15:34 is it unsigned long ? Jul 11 16:16:05 - volatile int lock[2]; Jul 11 16:16:05 + volatile int lock[1]; Jul 11 16:17:25 this is the 'bad' dump Jul 11 16:17:26 <2><165e1>: Abbrev Number: 7 (DW_TAG_member) Jul 11 16:17:26 <165e2> DW_AT_name : (indirect string, offset: 0x73b): lock Jul 11 16:17:26 <165e6> DW_AT_decl_file : 4 Jul 11 16:17:26 <165e7> DW_AT_decl_line : 7 Jul 11 16:17:26 <165e8> DW_AT_type : <0x1668c> Jul 11 16:17:28 <165ec> DW_AT_data_member_location: 24 Jul 11 16:17:50 <2><165ed>: Abbrev Number: 6 (DW_TAG_member) Jul 11 16:17:50 <165ee> DW_AT_name : buf Jul 11 16:17:50 <165f2> DW_AT_decl_file : 4 Jul 11 16:17:50 <165f3> DW_AT_decl_line : 8 Jul 11 16:17:50 <165f4> DW_AT_type : <0x16691> Jul 11 16:17:50 <165f8> DW_AT_data_member_location: 28 Jul 11 16:20:31 khem, in the 'good' dump the last line is Jul 11 16:20:34 <1662a> DW_AT_data_member_location: 32 Jul 11 16:20:48 I am sure the problem lies here Jul 11 16:21:00 why? Jul 11 16:21:02 heh Jul 11 16:58:41 so .bbclass code is python? Jul 11 16:59:03 obviously. Jul 11 17:00:58 It can contain python code, but it isn't exclusively python Jul 11 17:01:42 I beleive it allows anything that a .bb file allows Jul 11 17:01:57 correct Jul 11 17:02:03 .bbclass is bbclass Jul 11 17:02:12 it can contain python functions, or shell functions Jul 11 17:10:46 is it python 2.x or 3.x? Jul 11 17:19:14 3 Jul 11 18:02:19 bitbake produces shortened symlink names to several of the files under tmp/deploy/... - where/how is this specified in the recipes? i am misisng a symlink Jul 11 18:03:03 i need a symlink to core-image-full-cmdline-imx6ul-var-dart-20180711153252.rootfs.ubifs: https://paste.fedoraproject.org/paste/P-aZopNCfKRK7EQMh02XLw Jul 11 18:07:47 nm Jul 11 18:14:45 Hmm, occasionally i'd like to see recipetool create allow sourceless recipe creation, just to create a recipe template in a given layer as a time saver Jul 11 18:16:33 What is the advantage over an editor template like: https://github.com/kergoth/vim-bitbake/blob/master/plugin/newbb.vim Jul 11 18:32:08 my biggest issue with those is i don't always need or want to use a template, and ideally there'd be multiple templates available depending on the needs, so i'd rather it was opt-in / explicit Jul 11 18:39:54 kergoth: Makes sense... I haven't used recipetool much Jul 11 19:25:35 is there a place to see the script that is run with bitbake -c