**** BEGIN LOGGING AT Thu Mar 16 03:00:02 2017 Mar 16 04:03:36 I have this error Mar 16 04:03:40 when trying the create a minimal immage Mar 16 04:03:42 https://dpaste.de/J6Xx Mar 16 04:04:23 I need to increase the inotify number, but I don't have idea if that may have bad consequences Mar 16 06:05:30 cfoch-al1: clean your /tmp and reboot the machine Mar 16 06:05:35 does that help Mar 16 08:16:32 tlwoerner: RP: yeah, I automatically fixed the sentence, I got that a "not" was missing Mar 16 08:21:14 zeddii: ping Mar 16 08:21:47 (is he in EU or US now?) Mar 16 09:18:16 I should keep on this trend: 14 files changed, 1 insertion(+), 737 deletions(-) Mar 16 09:19:22 jku: certainly, just saw the patch list :) Mar 16 09:27:11 I wonder if there's QA check we could do... carrying unused patches in repo is not critical but carrying needed patches that aren't applied would be Mar 16 09:28:21 (that's in relation to the patchset that removes load of unused patches) Mar 16 10:20:40 I built an agl image but it won't boot Mar 16 10:20:43 https://drive.google.com/file/d/0B_RdLVA7cJo_LTE4LUtRaWJCa1U/view Mar 16 10:20:58 which is weird - the minimal yocto image I tried earlier had no issues Mar 16 10:21:18 how can I make it use the same boot configuration Mar 16 10:36:32 Hi, I currently working in a multi-domain project. We have encountered that some other domains changes recipes via bbappend, but they should not do that. It is possible to prevent bbappends from other layers? Mar 16 10:37:11 s/encountered/detected/ Mar 16 10:42:09 How do I place LICENSE_FLAGS_WHITELIST in .bbapend file of a recipe file ? Mar 16 10:49:43 Ramose: basically you can't as whatever you do in a recipe (and that includes the append) is only visble in tasks that belong to that recipe Mar 16 10:54:50 Letothe2nd: I can't do it in recipe file, so there is no way to add LICENSE_FLAGS_WHITELIST in .bbapend file ? Mar 16 10:55:30 Ramose: yes. things that have to be visible everywhere can only be set in .conf files. Mar 16 11:04:23 LetoThe2nd : I tried this actullay, PACKAGECONFIG += "${@bb.utils.contains("LICENSE_FLAGS_WHITELIST", "commercial", "gstreamer1.0-libav", "", d)}" in .bbapend file Mar 16 11:04:43 but it did not seems to be worked. Mar 16 11:40:30 Ramose: "gstreamer1.0-libav" does not sounds like a reasonable PACKAGECONFIG name Mar 16 11:40:44 Ramose: what are you trying to do? Mar 16 11:50:35 jku: I just wanted to set a LICENSE_FLAGS_WHITELIST flag in gstreamer1.0-libav.bbappend file Mar 16 11:51:21 Ramose: put LICENSE_FLAGS_WHITELIST += "commercial" Mar 16 11:51:22 in local.conf perhaps? Mar 16 11:52:08 it's not a recipe thing, it's a distro thing Mar 16 11:52:10 kanavin: That works but these changes needs to be done in gstreamer1.0-libav.bbappend file only Mar 16 11:52:45 kanavin: or something we can prepare a patch of Mar 16 11:54:37 Ramose: what is pulling in gstreamer1.0-libav? Mar 16 11:55:08 sorry, didn't get you ? Mar 16 11:55:38 Ramose: gstreamer1.0-libav recipe is not installed by default, it is pulled into images as a dependency of something. What is listing it as its dependency? Mar 16 11:56:55 Ramose: generally people set that flag distro wide in distro or local.conf. if you want to set it in just that recipe, just set it in the bbappend as LICENSE_FLAGS_WHITELIST=commercial (or use a recipe-specific override in your distro conf) Mar 16 11:57:52 rburton : I tried to set up in bbappend but getting issue with install Mar 16 12:01:11 such as? Mar 16 12:03:09 how to make neovim run in devshell? Mar 16 12:03:58 rburton: https://paste.ubuntu.com/24188235/ Mar 16 12:04:23 | NOTE: make -j16 DESTDIR=build/tmp/work/cortexa9hf-vfp-neon-linux-gnueabi/gstreamer1.0-libav/1.4.5-r0/image install Mar 16 12:04:26 thats very wrong Mar 16 12:04:31 what was your bbappend? Mar 16 12:07:24 rburton: gstreamer1.0-libav_1.4.5.bbappend Mar 16 12:09:05 Ramose: he means can you share the contents of the file Mar 16 12:10:05 LICENSE_FLAGS_WHITELIST = "commercial" that in only thing I had in bbappend Mar 16 12:28:08 Stupid me, it was some spelling mistake, Sorry everyone . Mar 16 14:39:02 Hi, how do i update the yocto os? Mar 16 14:43:37 BarBQ: big question. can you be more precise please. Mar 16 14:44:11 I have a Yocoto that seems to lose its connection after a few hours. Thinking that perhaps upgrading the drivers might make a difference. Mar 16 14:46:05 its running Linux e3815 4.4.16-yocto-standard #1 Thu Aug 18 16:15:21 CEST 2016 x86_64 x86_64 x86_64 GNU/Linux Mar 16 14:47:15 BarBQ: speak to the vendor who provided your image Mar 16 14:55:10 how can I find the package that provides the "xxd" linux tool? Mar 16 14:55:26 rburton: I was reviewing your ross/mutt and you missed musl update patch Mar 16 14:55:50 rburton: xserver-xf86-config patch depends on musl patch Mar 16 14:56:01 which you have staged there Mar 16 14:58:22 yeah thanks Mar 16 14:58:27 thought i marked both Mar 16 14:58:29 well spotted Mar 16 14:58:37 np Mar 16 14:59:15 a bit amazed that musl implemented lazy loading :) Mar 16 15:01:10 "bit amazed" you were swooning half the afternoon Mar 16 15:02:47 rburton: persistence always prevails :) Mar 16 15:03:14 although its not implemented same like glibc Mar 16 15:03:35 its a lot better and I wont like it to take newer cruft Mar 16 15:05:00 joshuagl: not quite swooning, just fainting in shock Mar 16 15:10:07 67615e01751bd (Alexander Kanavin 2017-02-13 16:44:48 +0200 558) exclude_pkgs = (bad_recommendations.split() if bad_recommendations else []) + (package_exlcude.split() if package_exclude else []) Mar 16 15:10:21 "exclude" vs. "exlcude" Mar 16 15:10:49 results in a nice big screenful of red for me. :) Mar 16 15:10:56 nice Mar 16 15:11:06 paulg: got a patch? :) Mar 16 15:11:23 credit to fray ; he spotted the fat finger typo before me Mar 16 15:11:44 I'd prolly still be staring at it scratching my head. Mar 16 15:12:12 rburton, I was hoping someone would tell me it is already fixed. :) Mar 16 15:12:50 $ git grep package_exlcude.split Mar 16 15:12:50 meta/lib/oe/package_manager.py: exclude_pkgs = (bad_recommendations.split() if bad_recommendations else []) + (package_exlcude.split() if package_exclude else []) Mar 16 15:12:51 nope Mar 16 15:13:15 all right ; guess I'll send one along Mar 16 15:21:23 sent. Mar 16 15:23:09 Only needed to send nine patches to finally get out the bottom of my do_rootfs. :) Mar 16 15:26:18 woot! Mar 16 15:26:52 i'm sort of impressed that this typo survived Mar 16 15:27:26 almost certainly means we don't have a test case for that code Mar 16 15:27:28 hint paulg ;-) Mar 16 15:28:30 I wouldn't have a clue how to add a super specific test case for that. Mar 16 15:28:55 but building meta-overc root file system tests a lot of stuff apparently Mar 16 16:05:56 rburton, khem: didn't we already take the musl upgrade? Mar 16 16:08:23 apparently not Mar 16 16:09:10 RP: more stuff in mut, testing here now Mar 16 16:09:15 including the musl :) Mar 16 16:09:27 rburton: I just merged some bits to -next but I'm likely missing musl :/ Mar 16 16:09:34 yes Mar 16 16:10:09 i see mut just halved in size again :) Mar 16 16:10:10 repushed Mar 16 16:11:20 khem: You need to use more unique shortlogs as my filtering batched this in the same place as some previous updates :/ Mar 16 16:11:26 rburton: I just found and added musl Mar 16 16:12:00 rburton: about half of -next is nearly through testing after which I could merge and trigger a round2 test on what is in -next, or we could do mut? Mar 16 16:43:39 RP: yeah I paste exact outputs Mar 16 16:44:57 RP: musl got a fresh patch which drops the xorg server preloading that we did for musl, so I was inclined to get this since it was ugly and error prone Mar 16 16:45:29 when one hooked other BSPs you have to override this file since each one of them had different xserver mods to load Mar 16 16:45:50 khem: I don't blame you, just wanted to mention why it got lost Mar 16 16:56:17 looking at scripts/qemu-ifup: $IFCONFIG addr add 192.168.7.$n/32 broadcast 192.168.7.255 dev $TAP Mar 16 16:56:55 hi, I'm trying create a recipe for a Makefile project... in its source it include , and during compilation it fails with errors such as "unknown type name 'uint32_t'"... I'm not sure how I can specify it to include the correct sys/types.h Mar 16 16:57:39 32 bit mask means single ip, right? We have at least 2 for host->guest network. How does it work? Mar 16 16:59:57 so far my understanding was that we need at least /30 for two hosts in the network Mar 16 17:03:05 RP: right, your automation is broken ;) Mar 16 17:03:58 khem: if only I had the time to fix it... Mar 16 17:05:48 ssalenik: Add #include Mar 16 17:06:42 ed2: /31 would be ideal for 2 addresses Mar 16 17:08:52 khem: true. however, runqemu uses /32 on the host interface. How does this work? Mar 16 17:10:16 khem: it also adds "ip=:::255.255.255.0" to the kernel command line. Looks like some dangerous mess to me. Mar 16 17:10:26 or I'm missing something fundamental here. Mar 16 17:13:55 ed2: It does sound slightly wrong but happens to work? :/ Mar 16 17:18:31 kscherer scrum Mar 16 17:19:55 ed2: I wonder if it even respects /32 Mar 16 17:21:19 khem: i configured tap0 manually the same way and now can't ping guest :) Mar 16 17:32:05 RP: webkitgtk needs nproc installed on host Mar 16 17:32:53 khem: why haven't we seen that fail? :/ Mar 16 17:33:15 its failing on musl/aarch64 Mar 16 17:34:35 RP: http://paste.debian.net/plain/920519 Mar 16 17:34:40 this is the snippet Mar 16 17:35:45 khem: I guess we add it to the list then :/ Mar 16 17:36:38 also seems to need taskset Mar 16 17:37:08 khem: I can see that annoying people more :( Mar 16 17:37:29 (people trying to make minimal containers) Mar 16 17:37:55 make them nonfatal with a comment that they're only needed for specific setups? Mar 16 17:38:14 or patch out that craz Mar 16 17:38:55 isn't that running on arm hosts? Mar 16 17:39:52 actually this test is supposed to be run when building webkitgtk natively Mar 16 17:40:06 yeah, i was going to say - that looks like tweaking thats only relevant for native Mar 16 17:40:30 cmake has a cross compiling variable that test needs to check Mar 16 17:41:55 look at this https://bugs.webkit.org/show_bug.cgi?id=138499 Mar 16 17:41:56 Bug 138499: was not found. Mar 16 17:42:15 silly yocti Mar 16 17:42:44 yeah sad that nobody noticed that wasn't cross friendly Mar 16 17:42:53 yeah surely Mar 16 17:43:44 its made specifically for rpi I believe Mar 16 17:44:29 because building webkit on the rpi is such fun Mar 16 17:45:02 and this define should be enabled otherwise webkitgtk wont work on pi3/64bit Mar 16 17:45:04 https://bugs.webkit.org/show_bug.cgi?id=138315 Mar 16 17:45:05 Bug 138315: was not found. Mar 16 17:49:35 khem: good to know about taskset, I got reports about util-linux-native failing to build on Ubuntu 14.04 and indeed it does because Ubuntu 14.04 shiped with incompatible combination of linux-libc-dev and libc6-dev (missing syscall number for sched_getattr) Mar 16 17:49:58 khem: so one of the "sollutions" was to add --disable-schedutils which doesn't build taskset Mar 16 17:57:16 don't suppose anyone has a reasonable way to shoehorn a firmware toolchain/blob build into the overall build process? Mar 16 18:05:39 strange; getting fetch errors on git://github.com/file/file.git Mar 16 18:07:33 bad SRCREV it seems ; fetched it manually and 79814950aafb81ecd6a910c2a8a that we want isn't a valid object/commit. Mar 16 18:08:24 heh, somebody rebase a public branch? Mar 16 18:09:14 kind of smells that way, doesn't it? Mar 16 18:09:35 ugh Mar 16 18:11:07 If I go to another machine that hasn't been updated, I might be able to find out what that commit was and map it sanely onto whatever it is now. Mar 16 18:12:56 ha. Mar 16 18:13:02 old commit we reference.... Mar 16 18:13:07 commit 79814950aafb81ecd6a910c2a8a3b8ec12f3e4a6 Mar 16 18:13:07 Author: Christos Zoulas Mar 16 18:13:07 Date: Fri Feb 10 17:25:14 2017 +0000 Mar 16 18:13:07 welcome to 5.30 Mar 16 18:14:00 ...and what is in the repo now.... Mar 16 18:14:04 co-op:~/poky/build/file$ git show FILE5_30 Mar 16 18:14:04 commit 3050419355566d2a96c5be97fef0ffae097bbb96 Mar 16 18:14:04 Author: Christos Zoulas Mar 16 18:14:04 Date: Fri Feb 10 17:25:14 2017 +0000 Mar 16 18:14:04 welcome to 5.30 Mar 16 18:14:17 Stellar. Mar 16 18:14:39 Guess I get to send another patch today. Mar 16 18:17:45 ugh Mar 16 18:19:54 strange thing is that "--pretty=fuller" shows them having the same commit/applied date. Mar 16 18:20:11 I really expected one to be way more recent than the other. Mar 16 18:20:29 anyway, will cc the repo author/maintainer with the oe patch. Mar 16 18:23:27 rburton: RP posted a patch for webkitgtk Mar 16 18:32:14 I working currently in a multi-domain project. We have detected that some other domains changes recipes via bbappend, but they should not do that. It is possible to prevent bbappends from other layers for some recipes in a special layer? Mar 16 18:35:03 unrelated; I noticed a deployment today had /bin/bash as a symlink to /bin/bash.bash Mar 16 18:35:12 as if update-alternatives lost its marbles. Mar 16 18:35:36 bad pixel - I mean big blocks, the video quality is nearly unuseable. Mar 16 18:35:51 bad: wrong windo Mar 16 18:36:00 that broke my ability (and presumably build-appliance too) to build since we require /bin/bash to be a real file as a sanity check. Mar 16 18:36:42 I fixed it manually for now, and will have to investigate the install logs later.... Mar 16 18:36:49 silviof: maybe with BBMASK, with some generic expression; untested, something like: .*/yourrecipe_.*\.bbappend ? Mar 16 18:37:57 (assuming that you now the consequences of that, not something i would recommend. you should fix the other layers instead ideally :)) Mar 16 18:39:30 zibri: I know, with 500 developers and 20 or more layers its nearly unmanageable. ^^ Mar 16 18:39:58 :) Mar 16 18:42:46 file srcrev fix sent. Mar 16 18:45:44 paulg: thats because someone turned on busybox's ability to pretend to be bash Mar 16 18:46:04 paulg: master has a fixed sanity that doesn't make that test Mar 16 18:47:33 interesting. I was pretty sure I hit busybox with fire and death and a silver stake through the heart in our layer. Mar 16 18:47:46 sure, but you'll still get the alternative names Mar 16 18:47:48 damn thing is like a bad zombie movie. Mar 16 18:47:59 ah, right. Mar 16 18:48:34 * paulg is on master, so wonders why he didn't get the fixed sanity test. Mar 16 18:51:41 hm, might be in -next Mar 16 18:54:02 rburton, that SRCREV change explode for you? Mar 16 18:54:25 I tested it on a machine without the old srcrev and one with it... Mar 16 19:52:27 RP: https://github.com/openembedded/bitbake/compare/master...kergoth:shallow — the change to support multiple mirror tarballs is now less risky (keeps the logic the same), and unit tests are included. won't be submitting it until early in the next cycle after this release, of course, but in case you're curious or have free time (hah!) :) Mar 16 19:52:59 the first three commits are general fixes/cleanup, will submit that as a separate series, but these depend on those to avoid conflicts Mar 16 20:00:38 nice Mar 16 20:01:53 so i take it's probably not worth it to look at making custom toolchain tasks? Mar 16 20:03:06 that should probably stay separate and go into a bin package/deploy-blob thing Mar 16 20:06:43 custom toolchain tasks? Mar 16 20:36:02 nerdboy: you can do custom toolchain tasks but they're harder than you think. Your question earlier was very vague which made it hard to write a meaningful answer Mar 16 20:37:15 hi folks, I've started seeing some strange opkg failures when generating images with poky master, with opkg generally syaing something like "package X requires Y, but none of the providers can be installed". https://gist.github.com/anonymous/0e3a3cee10357de49dd2647fa7012f9c is an example. Mar 16 20:38:08 kergoth: might as well send up the first three :) Mar 16 20:38:09 I'm currently moving back through the dep tree by replacing X with Y in my append, but I was wondering if there was a way to get opkg to just tell me what the root cause is? Mar 16 20:38:12 kergoth: that os.path.join(".") is lovely Mar 16 20:38:18 hehe Mar 16 20:39:10 kergoth: can we kill off mirrortarball in favour of mirrortarballs ? Mar 16 20:40:43 jmesmon: you should report this to Alejandro. I suspect the recent switch to using libsolv Mar 16 20:42:59 adelcast: ^^^ Mar 16 20:50:46 jmesmon: one of the changes that come with the libsolv switch is that opkg now follow dpkg policy more closely. For Provides, this means that if a package has a dependency which includes a version, then that dependency will only be satisfied by a package that exactly matches the dependency name. Mar 16 20:50:58 and when I get to the last package, it installs without issue :( [final item is a kernel module which PROVIDES & RPROVIDES_${PN} a virtual. Mar 16 20:51:06 from https://www.debian.org/doc/debian-policy/ch-relationships.html: "If a relationship field has a version number attached, only real packages will be considered to see whether the relationship is satisfied (or the prohibition violated, for a conflict or breakage). In other words, if a version number is specified, this is a request to ignore all Provides for that package name and consider only real packages." Mar 16 20:51:23 hm... Mar 16 20:51:37 I am guessing this might be the issue you are seeing? Mar 16 20:52:34 basically, opkg will not satisfy the libsettingsbroker1 >= 1.0.0+git0+558b0f6c4b dependency with a virtual package Mar 16 20:56:41 So I guess I'm trying to figure out how I translate that into a fix in yocto: my metadata isn't explictily referencing a version, and there is only an RDEPENDS_${PN} refering to the virtual/kernel-module-mcc. Can RDEPENDS_${PN} just no longer contain virtuals? Mar 16 20:57:01 s/metadata/bitbake metadata/ Mar 16 20:58:13 RP: it's a fair point, we certainly could. the only reason i avoided it was compatibility, as there is some metadata using mirrortarball. but given it's targeted for 2.4 anyway, can always adjust the metadata to match. i'll add it to my list Mar 16 20:58:34 It can. The restriction is that if a RDEPENDS has a version constrain, then it means that only real packages will be able to satisfy the dependency Mar 16 20:58:57 problematic package is https://github.com/Freescale/meta-freescale-3rdparty/blob/master/recipes-bsp/libmcc/libmcc_1.05.1.bb , FWIW. Mar 16 20:59:12 so, in this case sounds like audio-1.0+git0+e1d257015e_8a255e1f4a-r0.cortexa5hf-neon has a dependency with version restrictions on libsettingsbroker1 Mar 16 20:59:40 I would check who provides libsettingsbroker1 to confirm that is indeed the problem. If it is a virtual package, then that would explain the failure. Mar 16 21:00:32 kergoth: I think in this case I'd prefer to try and simplify. We can't have much metadata relying on that do we? Mar 16 21:00:45 adelcast: no, actually it is: audio -> libsettingsbroker :: settingsbroker -> libmcc-wrapper :: mcc-wrapper -> libmcc -> virtual/kernel-module-mcc :- kernel-module-mcc-X.Y.Z Mar 16 21:00:50 issue is way down the chain Mar 16 21:01:27 unless the packages like `libsettingsbroker1` are considered virtuals? (I'm not too familiar with that) Mar 16 21:01:43 adelcast: part of the problem here is that it doesn't sound obvious where the problem is :/ Mar 16 21:02:43 RP: there's also an altera fpga blob that's part of this build Mar 16 21:02:56 RP: i doubt it, probably just a couple of the source archival / mirror population classes floating around, iirc. i'll see about cleaning it up Mar 16 21:02:57 well, technically not part of it Mar 16 21:03:06 thanks for taking a look Mar 16 21:03:18 but it needs quartus toolchain to build the blob Mar 16 21:03:52 kergoth: its not a proper look, it was what just jumped out at me. I think I'm allergic to if statements now as it gives me testing headaches :) Mar 16 21:04:04 jmesmon: what are the dependencies of audio-1.0+git0+e1d257015e_8a255e1f4a-r0.cortexa5hf-neon? the error seems to point to that package, not further down the chain.... Mar 16 21:04:07 plus they switched from socfpga to freescale for the arm side so now it's *really* separate... Mar 16 21:05:36 hah. yeah, it's hard to get full coverage. i tried to keep the changes limited, but i had to modify needs_update and the like to handle certain cases. there is one possibility, which is to add a new url scheme and new fetcher which subclasses git. downside is existing MIRRORS/PREMIRRORS rules woudln't apply, and you'd have to add a hook to the metadata to modify the scheme in SRC_URI to enable it across the board Mar 16 21:06:15 by the way, a virtual package is a package that PROVIDES something else. Kinda make sense to have this restriction since if package A PROVIDES, let's say ssh, then it makes sense that the PROVIDES doesn't rely on any version, since other packages could also provide ssh (and there is no common versioning across packages that provide the same thing) Mar 16 21:07:58 adelcast: the error pops up for every item on that chain as I replace a CORE_IMAGE_EXTRA_INSTALL_append in my local.conf. Here's one for mcc-wrapper: https://gist.github.com/d480676e622628bce124530036027a1f Mar 16 21:08:10 RP: Yeah, usually if there is a dependency problem at the package manager level in OE, it's hard to de-tangle how that translates to a bug on a recipe. Not sure how we can improve that.... Mar 16 21:08:18 I hate submitting large changes of any kind for review now, there's a tendency to just +1 the thing and not thoroughly review it, of necessity. that's part of why i broke out git-make-shallow into a separate script. don't think i could have broken up the rest much more, other than to split up tarball creation vs usage Mar 16 21:09:15 adelcast: having opkg print the reasoning down the chain could be useful. In this case I've had to manually walk down the dependency chain. Mar 16 21:10:32 notably, libmcc complains a bit differently: https://gist.github.com/b312f41a7e3f8684aa418240f0a19287 , doesn't note versions at all Mar 16 21:10:47 (well for the requires, anyhow) Mar 16 21:11:59 jmesmon: can you confirm that the problem is what I described? that seems like the first thing to do. I believe opkg is giving the correct information: libmcc1 >= 1.05.1 is not going to be able to be satisfied if libmcc1 is provided by a package that is not called libmcc1 Mar 16 21:13:57 what are you asking me to look at? If libmcc1 exists? https://gist.github.com/ec158e1b045eef53476a8b008224a285 ? ./tmp/pkgdata/fitcore2/runtime-reverse/libmcc1 https://gist.github.com/60f55132066528253811d23625f9b92c ? Mar 16 21:16:50 there is a recipe called `libmcc`, and it doesn't explictly create any packages. `libmcc1` appears to be something generated by poky to fill out RDEPENDS with linked libraries. Mar 16 21:17:05 kergoth, may I bother you about one of your commits of years ago? Seems the issue you fixed is back :/ Mar 16 21:18:30 mm, why is RPROVIDES_libmcc: libmcc (=1.05.1) needed? Mar 16 21:18:56 ant_home: i doubt i'll remember, but sure Mar 16 21:18:57 :) Mar 16 21:19:00 it's about cml1.bbclass Mar 16 21:19:16 f you read the log history of cml1.bbclass you'll see two distinct fixes specific for for ncurses Mar 16 21:19:37 b036adf cml1: set and export TERMINFO, to handle ncurses-native relocation Mar 16 21:19:50 followed by Mar 16 21:19:52 889e026 ncurses, busybox, cml1.bbclass: Fix menuconfig display corruption Mar 16 21:20:23 if it's failing, i'm guessing it's fallout from recipe specific sysroots Mar 16 21:20:31 now what happens is, do_menuconfig works for busybox but fails for kernel Mar 16 21:20:55 I bet it's STAGING_KERNEL_DIR = "${TMPDIR}/work-shared/${MACHINE}/kernel-source" Mar 16 21:21:17 vs. STAGING_DATADIR_NATIVE = "${STAGING_DIR_NATIVE}${datadir_native}" Mar 16 21:21:29 RP: should think about using combo-layer or subtrees or peru or *something* to import the 3rd party python modules in the bitbake repo rather than doing so manually, so it's easier to maintain Mar 16 21:27:02 jmesmon: need to run, but do you mind filing a bug report on https://bugzilla.yoctoproject.org? If you include a way to reproduce the issue, I can take a look at it tomorrow Mar 16 21:27:46 kergoth, first failure seems about /usr/share/tabset/stdcrt not found, log http://pastebin.com/8RKy1zQN Mar 16 21:28:52 best part of this bug is, if I hack (counterintuitively) cml1.bbclass it works... Mar 16 21:29:01 -do_menuconfig[depends] += "ncurses-native:do_populate_sysroot" Mar 16 21:29:10 +do_menuconfig[depends] += "ncurses:do_populate_sysroot" Mar 16 21:29:20 so it peeks in recipe sysroot Mar 16 21:30:24 too much for me atm... it's very complicated interaction between bitbake.conf, kernel.bbclass and cml1.bbclass plus ncurses Mar 16 21:47:45 ant_home: FWIW I think Brian and sgw might have figured this out Mar 16 21:48:05 ant_home: Its a problem with pkg-config and when to use the target pkg-config and when to use the native one Mar 16 21:53:06 RP: I'm testing now nodistro/qemux86 to see if it oes reproduce Mar 16 21:54:35 RP: it has been broken long before RSS, back then the band-aid was 'bitbake ncurses' Mar 16 21:54:59 now I'm fed up and hope to debug it to th eend... Mar 16 21:59:05 rburton: I posted 3 patches on ml, that now fixed world builds for musl on mips/aarch64 Mar 16 21:59:08 all clean now Mar 16 22:02:38 RP: heh..doesn't happen with nodistro/qemux86/linux-yocto_4.10 Mar 16 22:03:30 our kernel does set S = "${WORKDIR}/linux-${PV}" Mar 16 22:04:32 hm...this is fine to build but not for menuconfig? Mar 16 22:07:06 khem, pls let me know wheter your raspberry kernel recipe suffers the issue Mar 16 22:10:41 hello Mar 16 22:10:53 trying my way around yocto, but this has tripped me up Mar 16 22:11:42 http://pastebin.com/hg7Z3BjU Mar 16 22:12:32 trying to work out how the stamps work Mar 16 22:16:40 ant_home: it probably depends if you've built ncurses before or after you build menuconfig Mar 16 22:17:33 I don't think so, I did -c cleansstate before Mar 16 22:17:50 for both, ncurses and -native Mar 16 22:22:58 it's just the different source dir and the processing done in kernel.bbclass Mar 16 22:23:04 # Old style kernels may set ${S} = ${WORKDIR}/git for example Mar 16 23:01:40 gn Mar 17 00:00:03 * paulg is intrigued by yocti chasing random links. Mar 17 00:00:47 https://www.google.com/show_bug.cgi?id=31337 Mar 17 00:00:48 Bug 31337: was not found. Mar 17 00:01:01 * paulg laughs Mar 17 00:02:21 poor thing; probably not nice to tease the robots. **** ENDING LOGGING AT Fri Mar 17 03:00:01 2017