**** BEGIN LOGGING AT Mon Apr 10 03:00:01 2017 Apr 10 04:35:27 How to update busybox in poly 2.1 . Getting error ps -e Invalid option Apr 10 06:29:33 d Apr 10 07:01:38 Hello, I'm new in bitbake and have an understanding problem. I read a lot about patches, but I didn't found any documentation how I can write a patch. My problem is that I have a CMakelists.txt which search for a package I will not need on embedded. My first approach was with a cmake flag. But I think I can use a patch with removes this package out of the Cmakefile or? Apr 10 07:11:11 You can check the documentation of "git format-patch" Apr 10 07:21:14 Hunk: your idea sounds correct. if you've never worked with patches there's a bit of a learning curve, but you might want to try devtool: see http://www.yoctoproject.org/docs/latest/dev-manual/dev-manual.html#using-devtool-in-your-workflow, especially 4.3.1.2 Apr 10 07:21:53 (that last bit is assuming you have a recipe already) Apr 10 07:55:02 @jku ty I will read it Apr 10 08:38:08 Hi. I think I've fixed a small bug in the latest meta/lib/oe/package_manager.py script. I guess I should open a bugzilla for it? Apr 10 08:42:14 please send a patch to the openembedded-core mailing list, let me try and find a link for you Apr 10 08:45:16 arkver: http://www.openembedded.org/wiki/How_to_submit_a_patch_to_OpenEmbedded Apr 10 08:48:21 joshuagl: Thanks. The code in question is in poky though, committed with this commit: http://git.yoctoproject.org/cgit.cgi/poky/commit/meta/lib/oe/package_manager.py?id=2b77735c72cb5100dbd850e79d4f639e82629842 Apr 10 08:48:46 So should this still go to OE, or via yoctoproject's bugzilla? Apr 10 08:49:29 arkver: right, the poky repository is a combined repository which "bundles" multiple upstream repos into one. That code is from OE-Core Apr 10 08:49:38 you can tell from the "(From OE-Core rev: 3a9e2fdef9316e24b52ce99ac355fc2b09786c72)" tag in the log Apr 10 08:50:02 ah, right. Thx. Apr 10 08:50:47 to the mailing list is preferable, than we can get it reviewed in a more timely fashion and your contribution will be correctly tracked Apr 10 08:51:00 ok, will do. Apr 10 08:53:34 much obliged! Apr 10 08:59:23 joshuagl: Is the OE ML going to ask for the patch against an OE tree or will the one against poky's version be OK? Apr 10 09:00:38 arkver: poky patch is ok Apr 10 09:00:52 Hi everyone, what is the best way to add an image in a image ? I read somewhere that doing IMAGE_INSTALL_append in the conf/local.conf of the build isn't the best, but nice for testing. Thanks by advance ! Apr 10 09:00:53 ed2: super, ta. Apr 10 09:01:03 to add a package in an image* Apr 10 09:07:35 Chrys: depends. If you have your own image then the obvious place is in the image recipe Apr 10 09:10:08 jku : And if I use an image already made adding my meta, the best is IMAGE_INSTALL_append in the build directory then ? Apr 10 09:10:41 jku : The best practice is to do my own image recipe if I understand. Apr 10 09:13:36 jku : Thanks jku. Apr 10 09:22:28 Maybe I should have included an [OE-Core] tag? Apr 10 09:25:21 the mailing list adds that prefix automatically Apr 10 09:31:53 ok, good. Apr 10 10:28:21 good morning Apr 10 11:04:17 hello, I'm currently facing a problem on poky's master branch. During do_fetch of a recipe with an svn SRC_URI following error occurs: "failed with exit code 127, output: /usr/bin/env: svn: No such file or directory" Apr 10 11:06:34 I'm currently migrating from jethro to master (=> pyro) and the recipes worked fine since daisy... Any help would be greatly appreciated! Apr 10 11:10:33 g0hl1n: sounds like svn-native doesn't get built. Apr 10 11:10:48 or that svn isn't in HOSTOOLS? Apr 10 11:11:15 rereading the message I think LetoThe2nd is right Apr 10 11:11:28 joshuagl: RSS hitting? Apr 10 11:16:57 LetoThe2nd: since when does a "svn-native" recipe exists/gets used? IMHO until now svn was taken from the host? Furthermore I'm unable to find any svn-native related stuff in poky.git... Apr 10 11:17:41 meta/classes/base.bbclass: d.appendVarFlag('do_fetch', 'depends', ' subversion-native:do_populate_sysroot') Apr 10 11:18:55 g0hl1n: ^^^^^^ Apr 10 11:19:05 if you have an long-lived build directory it's possible you have subversion-native in ASSUME_PROVIDED ? Apr 10 11:19:47 joshuagl: resulting in requesting the one from the host and not building it? but then, why would it have worked before? Apr 10 11:19:58 HOSTTOOLS Apr 10 11:20:03 I'd expect Apr 10 11:20:28 so if you have the ASSUME_PROVIDED you'll need to either a) remove that or b) add svn, and probably env, to HOSTTOOLS Apr 10 11:21:38 joshuagl: you wouldn't need env since it's being invoked by absolute path Apr 10 11:21:46 joshuagl: Ah, thanks... haven't looked for "subversion" (only svn)...; Just grepped through the whole builddir for ASSUME_PROVIDED and found nothing related to svn or subversion; The build dir was set up 2 days ago with the current HEAD from master... Apr 10 11:22:08 bluelightning: ah, good to know. thanks Apr 10 11:24:00 g0hl1n: hrmm, has subversion-native been built? Apr 10 11:25:02 I've another question re: dnf. I've got my repo of rpms now and dnf makecache is happier. Apr 10 11:25:34 But now dnf upgrade -y is complaining about a lack of GPG public keys installed. Apr 10 11:25:58 Do I need to inherit "sign_package_feed" or something? Apr 10 11:26:06 nrossi: Am I right in assuming that the morty branch for meta-xilinx is quite OK? Apr 10 11:26:43 JoiF: should be, hitting an issue? Apr 10 11:26:45 joshuagl: fyi, also no HOSTTOOLS set in the builddir. Yes, svn-native was built in tmp/work/x86_64-linux/subversion-native/1.9.5-r0... Apr 10 11:27:14 what's the subversion using recipe? how can I replicate it? Apr 10 11:27:44 nrossi: Not at all. Just starting a new thing.. and I wanted to have recent stuff. :-) Apr 10 11:28:37 kanavin: ^^ Apr 10 11:30:03 nrossi: i.e. we finally have our custom boards ready and I'm officially starting our Yocto stuff (i.e. for building for our custom boards instead of the dev-bards) and I wanted to make sure I'm basing it on the latest and greatest. :-) Apr 10 11:30:57 JoiF: latest and greatest? want me take a selfie? ;-) Apr 10 11:31:06 Apr 10 11:31:51 LetoThe2nd: Hmm.. I could run it through an edge-detection filter and put your face in copper on.. say... layer 9 of the board? ;) Apr 10 11:32:16 JoiF: let me guess, on a 8 layer pcb ;-) Apr 10 11:32:23 :p Apr 10 11:32:32 [aside - I just picked an SVN recipe at random (bitbake wmiconfig) and it downloaded and built subversion-native-1.9.5] Apr 10 11:33:46 joshuagl: It's basically the following recipe: https://gist.github.com/rleitner/458196f6d2b11f21555a328ab17068c1. I've obfuscated the names and paths because it references our internal dev servers... Apr 10 11:36:21 nrossi: Btw, can you remember (from the top of your head) what the difference is between "zybo-zynq7" and "zybo-linux-bd-zynq7"? Apr 10 11:37:42 JoiF: zybo-linux-bd-zynq7 is a machine setup for the linux-bd reference design from Digilent, 'zybo-zynq7' is ps7 only Apr 10 11:39:43 arkver: on current master? I'll try that in a fresh builddir on my host. Just give me a minute ;-) Apr 10 11:40:49 g0hl1n: yes, master as of yesterday afternoon. Apr 10 11:40:56 this rootfs to ${D}/home/root guy around here? ;-) Apr 10 11:41:07 633ad6c9f4 Apr 10 11:42:27 arkver: ok. that would be the same rev as I'm on... fresh build is already running... Apr 10 11:42:45 nrossi: Ahh, cool. Thanks! Apr 10 11:44:22 but I do have svn installed on my host, so I dunno which the wmiconfig recipe is actually using. So I'm probably not helping here. Apr 10 11:46:01 bluelightning, does poky add any pythong modules to native-sdk not in an oe-core sdk? Apr 10 11:48:55 Crofton|work: the pythong module? that does exist :) Apr 10 11:53:03 g0hl1n: I just looked at the do_fetch log and it failed the same way as yours (/usr/bin/env svn: NSFD) Apr 10 11:53:22 The package built ok 'cos it found a mirror tarball. Apr 10 11:53:40 arkver: yes, I also looked it up and it built successfully using the mirror ... Apr 10 11:54:14 RP, diversity! Apr 10 11:54:24 do not highlight my typos Apr 10 11:57:11 seems something is wrong with the svn fetcher in current master. Apr 10 11:58:25 Crofton|work: that one just seemed particularly worrying :) Apr 10 11:58:43 RP: Crofton|work: typo, what typo? https://pypi.python.org/pypi/pythong Apr 10 12:02:03 so the real question is why an sdk from poky can build libfdt in u-boot and the sdk from a non-poky build can't Apr 10 12:03:49 nrossi: meta-xilinx comes with a bitstream for the zybo board, is the vivado project that created that bitstream available? Apr 10 12:06:28 yourfate: meta-xilinx doesn't come with the bitstream, it just sources it from a github repo. The project is available here: https://github.com/Digilent/ZYBO Apr 10 12:06:52 thanks! Apr 10 12:08:09 kanavin: I'm trying to figure out how to deal with 'working-scripts' packageconfig in gtk-doc. My fixed-native-DISTRO_FEATURES build breaks when api-documentation is enabled Apr 10 12:08:13 https://autobuilder.yocto.io/builders/nightly-qa-extras/builds/218/steps/BuildImages_2/logs/stdio Apr 10 12:09:39 kanavin: think it would be ok to set 'working-scripts' in PACKAGECONFIG_class-native? Apr 10 12:12:12 kanavin: if it wasn't clear: the breakage happens because my patch sets DISTRO_FEATURES="" for native. So working-scripts does not get set Apr 10 12:24:01 jku: I come to your desk in a moment Apr 10 12:25:32 comparing files named "svn" in the builddir it seems pretty ok to me: https://gist.github.com/rleitner/3f44c46d278f33df4a8557598edcdb34; does anybody of you have an idea on how to solve that problem? Apr 10 12:25:40 jku: the issue is that we can run builds with and without docs. In the "without docs" case, we don't want all the extra dependencies Apr 10 12:26:27 jku: this might be a case where the native flag needs to reflect the main flag? Apr 10 12:27:06 arkver: do you still need help with dnf? it has a command line switch to disable gpg verification on packages, or you can set it in /etc/dnf configuration file Apr 10 12:27:27 arkver: it's enabled by default, whether or not you actually sign your packages Apr 10 12:27:44 arkver: and you should sign them :) Apr 10 12:43:23 For me it seems env is unable to find svn because the sysroots/ dir is empty. Instead everything lies in sysroots-components/. Could that be possible? Apr 10 12:49:22 RP: we were just talking about that... this might be a case like that Apr 10 13:17:26 yeah, "working-scripts" packageconfig for gtk-doc brings in 7 new native deps (including boost-native) for core-image-minimal. That would be nice to avoid if possible Apr 10 13:22:25 kanavin: Thanks, yes still working on it. Trying to get the packages signed. Apr 10 13:23:32 kanavin: Getting: Exception: NotImplementedError: Package feed signing not yet implementd for rpm Apr 10 13:24:10 Am I on the "bleeding edge" here or is all this expected to work? Apr 10 13:25:36 arkver: package feed signing does not work, but package signing should work Apr 10 13:25:41 they are two different things Apr 10 13:26:01 configured through a different set of variables Apr 10 13:28:37 I'm just trying to mend a build that was working OK with smart. A "how-to-migrate" doc would be super, but I haven't found one. Apr 10 13:30:19 I see sign_package_feed depends on signing-keys which seems to be installing keys to the rootfs. Apr 10 13:36:46 So how does one install the necessary keys so that dnf can install packages from the package feed? Apr 10 13:39:09 nvm, I just rtfm'd Apr 10 13:39:27 (RPM_GPG_* and INHERIT += " sign_rpm" Apr 10 13:39:28 ) Apr 10 13:58:09 the device tree compiler should be in the kernel/scripts/dtc directory, right? Apr 10 13:58:19 there is a dtc.c file in there, but no binary... Apr 10 14:04:31 ah, right, I need to have it build it Apr 10 14:04:50 however, that would then be an arm executable. do I need some sort of cross-dtc? Apr 10 14:23:48 RP: around? g0hl1n has discovered a failure with the svn fetcher. AFAICT base.bbclass adds subversion-native:do_populate_sysroot should include svn in the recipe-sysroot-native for the recipe? Apr 10 14:26:14 joshuagl, Do you have the path the the slow kernel clone handy? I could break up the mount points too maybe. Apr 10 14:26:59 joshuagl: I am here, yes Apr 10 14:27:19 joshuagl: yes, subversion-native should get included in the sysroot Apr 10 14:27:45 RP, I'm starting the NFS changes now. Cluster down for a bit. Apr 10 14:28:08 halstead: do you agree we should use 4.2? Apr 10 14:28:39 halstead: I was really meaning to start a discussion rather than saw we should switch. The experience with f24 seemed good though Apr 10 14:28:57 RP: it appears not to be :-/ I'm sadly unfamiliar with rss code, can you point me at some files/functions? Apr 10 14:29:05 RP, Yes I think that is the right thing to try. 4.1 has failed in the past and 4.0 isn't working. Debian 8's default kernel doesn't support 4.2 is the only reason we haven't switched yet. Apr 10 14:29:05 * joshuagl could grep the git logs too Apr 10 14:29:18 But I can install the backport kernel. Apr 10 14:29:18 joshuagl: the log from the recipe where this fetch hasn't happened would be usefu Apr 10 14:29:57 halstead: or just leave debian8 doing what its doing and move the rest to 4.2? Apr 10 14:30:57 RP that's where we started months ago. We could try it again and see if the other changes are enough but I'd rather be uniform if possible. Apr 10 14:31:32 btw: RPM_GPG_PASSPHRASE isn't happy with phrases with whitespace and shell special characters in them. Not sure why this was changed from RPM_GPG_PASSPHRASE_FILE. Apr 10 14:32:40 RP: I'm wrong, svn is linked there in the recipe-sysroot-native Apr 10 14:32:46 g0hl1n: can you share some logs with us? Apr 10 14:33:27 RP: admittedly not a full log, but from the backlog "failed with exit code 127, output: /usr/bin/env: svn: No such file or directory" Apr 10 14:34:25 * joshuagl tries to replicate with a recipe from meta-oe Apr 10 14:35:28 halstead: I think this bz item has the details https://bugzilla.yoctoproject.org/show_bug.cgi?id=11211 Apr 10 14:35:29 Bug 11211: major, Medium+, 2.3 M4, michael, NEW , Git clones to NAS are very slow and have significant impact on build times Apr 10 14:39:47 joshuagl, Hrm... Looks like the current_sources dir which is the name spot we need the attribute cache disabled for .lock files. Apr 10 14:41:11 halstead: I'll leave it with you Apr 10 14:41:36 Thanks for the update RP. I'm glad to be back. Apr 10 14:43:25 halstead: the good thing about the slow clones is that its easy to test/reproduce outside of builds Apr 10 14:44:13 Silver linings. :) Apr 10 14:45:49 joshuagl: I put some svn log here, if it's any good. This is the same error as g0hl1n was seeing. Apr 10 14:45:53 https://pastebin.com/AjfH3JLb Apr 10 14:47:34 arkver: thanks Apr 10 14:47:43 I think the issue is probably /home/ian/tx6/XXX/build/tmp/work/cortexa9hf-neon-poky-linux-gnueabi/wmiconfig/0.0.0+svnr5394-r2/recipe-sysroot-native/usr/bin/arm-poky-linux-gnueabi Apr 10 14:47:52 when the svn binary is in /home/ian/tx6/XXX/build/tmp/work/cortexa9hf-neon-poky-linux-gnueabi/wmiconfig/0.0.0+svnr5394-r2/recipe-sysroot-native/usr/bin Apr 10 14:48:30 for some reason that seems to have an target suffix on the end of that path Apr 10 14:49:09 joshuagl: that other path is later in the log Apr 10 14:49:16 joshuagl: the first path is the cross compiler dir Apr 10 14:49:38 RP: right you are Apr 10 14:49:45 I'm not doing very well at debugging this :-/ Apr 10 14:49:49 ok, svn is indeed there in the plain usr/bin dir. Apr 10 14:50:06 btw, I don't want to steal g0hl1n's thread if his problem is different. :) Apr 10 14:51:11 svn fetcher just WFM on oscam from meta-oe :-/ Apr 10 14:51:36 arkver: it appears to be similar/the same Apr 10 14:58:40 arkver: can you run the svn binary in /home/ian/tx6/XXX/build/tmp/work/cortexa9hf-neon-poky-linux-gnueabi/wmiconfig/0.0.0+svnr5394-r2/recipe-sysroot-native/usr/bin ? Apr 10 14:58:47 (right permissions?) Apr 10 14:59:21 yes Apr 10 14:59:26 also via env... Apr 10 14:59:42 arkver: you can see how we're puzzled :/ Apr 10 14:59:45 PATH="/home/ian/tx6/XXX/build/tmp/sysroots-uninative/x86_64-linux/usr/bin:/home/ian/tx6/XXX/sources/poky/scripts:/home/ian/tx6/XXX/build/tmp/work/cortexa9hf-neon-poky-linux-gnueabi/wmiconfig/0.0.0+svnr5394-r2/recipe-sysroot-native/usr/bin/arm-poky-linux-gnueabi:/home/ian/tx6/XXX/build/tmp/work/cortexa9hf-neon-poky-linux-gnueabi/wmiconfig/0.0.0+svnr5394-r2/recipe-sysroot/usr/bin/crossscripts:/home/ian/tx6/XXX/build/tmp/work/ Apr 10 14:59:45 cortexa9hf-neon-poky-linux-gnueabi/wmiconfig/0.0.0+svnr5394-r2/recipe-sysroot-native/usr/sbin:/home/ian/tx6/XXX/build/tmp/work/cortexa9hf-neon-poky-linux-gnueabi/wmiconfig/0.0.0+svnr5394-r2/recipe-sysroot-native/usr/bin:/home/ian/tx6/XXX/build/tmp/work/cortexa9hf-neon-poky-linux-gnueabi/wmiconfig/0.0.0+svnr5394-r2/recipe-sysroot-native/sbin:/home/ian/tx6/XXX/build/tmp/work/cortexa9hf-neon-poky-linux-gnueabi/wmiconfig/0.0.0+ Apr 10 14:59:46 svnr5394-r2/recipe-sysroot-native/bin:/home/ian/tx6/XXX/sources/poky/bitbake/bin:/home/ian/tx6/XXX/build/tmp/hosttools" /usr/bin/env svn --version Apr 10 14:59:49 svn, version 1.9.5 (r1770682) Apr 10 14:59:51 compiled Apr 10 2017, 12:30:42 on x86_64-pc-linux-gnu Apr 10 15:00:07 yes, puzzling indeed. Apr 10 15:00:46 alimon: you added -k to "bitbake -S none" when testing the layer in test_signatures. What was the motivation and where did that work? I'm looking at a layer that triggers ERRORs, and then bitbake returns a non-zero error code and the test aborts, despite the -k. Apr 10 15:02:16 * zeddii enjoys the random jenkins spam Apr 10 15:02:40 pohly: it was to figured out what signatures changes Apr 10 15:02:40 RP: Indeed if I run the exact command from the "Running ..." line in that log in a (subshell) it works fine. Apr 10 15:02:43 -k doesn't tell bitbake to ignore errors, or that they're okay, only that it should keep running tasks as much as possible rather than exiting immediately. the exit code would be non-zero regardless. Apr 10 15:02:54 pohly: may be with your diffsigs isn't useful atm Apr 10 15:03:08 alimon: I understand the test, and it is useful. But what's the -k for? Apr 10 15:03:41 pohly: to avoid abort on errors, but may be needs to be remove Apr 10 15:04:46 alimon: then it doesn't work as intended. We need to keep it and if we want to continue (which probably is better than aborting), check whether we got the locked-sigs.inc despite the non-zero return code. Apr 10 15:05:17 Whether a "world build" works needs to be a different test. Apr 10 15:05:49 pohly: ok Apr 10 15:05:57 pohly: yes Apr 10 15:06:14 I can change that. Just wanted to get a better understanding of the current code first. Apr 10 15:06:17 pohly: i can't code the world build test because will be a long time consume Apr 10 15:06:27 pohly: may be add a flag to run in a long test mode... Apr 10 15:06:29 alimon: could be a --dry-run. Apr 10 15:06:58 pohly: but i doesn't test that actually builds Apr 10 15:07:09 only will test that the tasks are created and parsing Apr 10 15:07:23 No, but it already catches exactly the errors why layers break a world build. Apr 10 15:08:01 RP: What creates recipe-sysroot-native? Could there be a race here? Apr 10 15:15:51 RP: arkver: addtask do_prepare_recipe_sysroot before do_configure after do_fetch Apr 10 15:16:19 joshuagl: after do_fetch? Apr 10 15:17:40 arkver: right :-/ Apr 10 15:17:51 * joshuagl is waiting for RP to tell him what he missed Apr 10 15:19:15 btw: the wmiconfig recipe I picked at random carries on and builds only because it finds a mirror tarball. In this case I guess the post do_fetch tasks are called and the sysroot is made. Apr 10 15:19:31 Maybe oscam does the same thing? Apr 10 15:20:32 correct — it does, I didn't notice the NOTE: first time around Apr 10 15:21:31 well, at least this is easy to reproduce. :) Apr 10 15:22:08 Hmm, anyone done ptests using the kyua test runner/engine? Apr 10 15:30:54 RP, Is there anything new ready to start on https://autobuilder.yocto.io/builders/nightly? Apr 10 15:46:51 halstead: add some musl world builds :) Apr 10 15:47:41 ...another day... another person with a misconfigured jenkin's server that spewed everyone that contributed to meta-openembedded Apr 10 15:48:55 ...I really don't care you are having build issues with your 'Merge branch 'develop' into feature/ipvideo' Apr 10 15:49:01 lol Apr 10 15:49:36 halstead: give me a few minutes and we can have something Apr 10 15:49:53 khem: I have enough problems with the existing tests :( Apr 10 15:50:24 joshuagl: the code also adds a postfunc for any task with a populate_sysroot dependency Apr 10 15:51:21 joshuagl: the function is extend_recipe_sysroot and its a prefunc, sorry Apr 10 15:51:32 joshuagl: see staging_taskhandler Apr 10 15:52:01 RP: aha, yes. I shouldn't have stopped reading at that line :-/ Apr 10 15:52:07 RP: apologies for the red herrings Apr 10 16:00:54 Hey all. I am building an image based on the Freescale Community BSP on the master branch. The meta-browser layer has recipes for Chromium-wayland 53x and 54x, but the BSP wants Chromium-wayland 48.x and 52.x. Apr 10 16:01:09 Any suggestions on resolving this difference in versions? Apr 10 16:03:55 halstead: we can run a master next on the new cluster Apr 10 16:10:44 fwiw, it looks like the answer is to use the morty branch of meta-browser Apr 10 16:18:30 RP: joshuagl: I put in a DEPENDS += " subversion-native" in the wmiconfig recipe. No difference. Apr 10 16:19:16 RP: joshuagl: Also, if I do cleanall and then bitbake -c prepare_recipe_sysroot wmiconfig, it does the do_fetch task first, then pop sysroot. Apr 10 16:21:10 On the plus side, dnf now works for me. :) Apr 10 16:21:37 khem: Do you have a linux 4.9 upgrade patch in progress for meta-raspberrypi? Just checking before I start working on one Apr 10 16:22:41 paulbarker: yes its in kraj/master Apr 10 16:23:46 RP: Just wanted to add to your worries just in case there was room Apr 10 16:24:02 arkver: you don't have subversion-native in ASSUME_PROVIDED do you? Apr 10 16:24:02 I thought there was a builder which was idle :) Apr 10 16:24:31 RP: btw. world build with musl for all qemu machines supported in oecore is clean Apr 10 16:24:39 and with security flags turned on Apr 10 16:24:58 khem: that is very cool. We do test at least one musl world build now to try and stop things regressing Apr 10 16:24:59 nope, only libsdl-native. Also, it does fetch & build svn ok. Apr 10 16:25:00 khem: I'll pick that patch up then and include it in my testing Apr 10 16:25:15 paulbarker: gr8 Apr 10 16:25:21 it just doesn't make the sysroot early enough. Apr 10 16:25:32 there's been a few CVEs fixed since 4.9.13 so it'd be good to get that patch onto the mailing list Apr 10 16:25:37 RP: yeah, may be once a week build for all arches may not be a bad deal Apr 10 16:25:44 arkver: does adding do_fetch[prefuncs] += "extend_recipe_sysroot" to the recipe help? Apr 10 16:25:58 arkver: that is supposed to happen automatically Apr 10 16:26:00 paulbarker: I never got to do a satsfactory testing done yet Apr 10 16:26:12 exit Apr 10 16:26:54 stephano: sorry, you can't leave ;-) Apr 10 16:26:57 https://en.wikipedia.org/wiki/No_Exit Apr 10 16:27:01 :) Apr 10 16:27:19 khem: I'll bump it to 4.9.21 first as that's the latest and then do some basic testing Apr 10 16:27:31 paulbarker: ok Apr 10 16:27:46 paulbarker: there are bunch of other patches too Apr 10 16:27:50 thats on the branch Apr 10 16:28:15 cool, I'll have a look, should be able to pick them all up Apr 10 16:29:39 RP: no. I assume bitbake -c fetch wmiconfig should do all the necessary steps up to there? Apr 10 16:30:14 I did a cleanall wmiconfig, and after the fetch the recipe-sysroot and recipe-sysroot-native directories exist, but are empty. Apr 10 16:30:33 arkver: you assume correctly Apr 10 16:33:28 Diffs to the recipe, fyi: https://pastebin.com/SxBLFy4q Apr 10 16:44:59 arkver: does bitbake wmiconfig -g show a dependency on subversion-native in task-depends.dot ? Apr 10 16:45:16 arkver: can you share the log.do_fetch file for that having run it? Apr 10 16:46:12 "wmiconfig" -> "subversion-native" Apr 10 16:48:03 arkver: do_fetch specifically though? Apr 10 16:49:09 yes, uploading Apr 10 16:50:18 https://pastebin.com/6bX2NHwU Apr 10 16:50:48 There's a lot of junk in there, partly due to my premirrors, but the fail is at line 335 Apr 10 16:51:07 (or before) Apr 10 16:52:05 RP: Was that what you were looking for? Apr 10 16:54:33 RP: btw, joshuagl saw the same error with the oscam recipe from meta-oe. Apr 10 16:55:08 arkver: NOTE: subversion-native exists in sysroot, skipping Apr 10 16:55:21 arkver: so it thinks its already there? Apr 10 16:57:08 RP: I guess, but those dirs are still empty. Apr 10 16:58:08 RP: Checking in a different sysroot? Apr 10 16:58:10 arkver: something isn't adding up here and I really don't have the time to go through this step by step with you :/ Apr 10 16:59:17 RP: I don't think I use any SVN recipes, so I'm only trying to help debug this. Other SVN users may be unhappier. Apr 10 17:35:35 RP: joshuagl: fwiw I added some debug to staging.bbclass. At the time it checks for the existence of those packages, they do exist in recipe-sysroot-native, but by the time the fetch is complete, they're gone. I don't know what's removing them. Apr 10 17:39:20 Nice of the people at jenkins@savant.com to spam anyone who made a commit in the last two years with 9MB of useless yocto changeset summaries. Apr 10 17:39:35 idiot savant, I guess. Apr 10 17:40:09 paulg, not the first people to do it :) Apr 10 17:40:25 probably won't be the last either.... :-/ Apr 10 17:40:44 usually I get that from people mis-handling auto Cc: on kernel commits. Apr 10 17:42:34 but the last 4.8.x stable I did for yocto got some "netcfg" tester spamming me and zeddii with megs of useless shit too. Apr 10 17:43:15 ouch Apr 10 18:03:42 I lied. It was "netconf" https://lists.yoctoproject.org/pipermail/linux-yocto/2017-March/006310.html Apr 10 18:17:37 hello, anyone using fuego around here? Apr 10 18:23:21 arkver: that is very very strange :/ Apr 10 18:23:42 arkver: does this recipe happen to do ${S} = "${WORKDIR}" ? Apr 10 18:24:45 RP: S = "${WORKDIR}/AR6kSDK.build_sw.18/host/tools/wmiconfig" Apr 10 18:25:15 arkver: ok, scratch that theory then :) Apr 10 18:26:08 I put a call to ls at the end of extend_recipe_sysroot and I think the sysroot is intact at that point. Apr 10 18:29:19 RP: Specifically, I can see a /usr/bin/svn in there. Apr 10 18:32:11 arkver: so something is wiping out bits of WORKDIR? :/ Apr 10 18:33:13 RP: So it would seem. I can watch it happen from another shell window. It's quick, so can't spot what's doing it in the BB output. Apr 10 18:33:45 RP: Removed my PREMIRRORS just in case - no difference. Apr 10 18:34:16 arkver: I've just suddenly wondered if this is bitbake itself :/ Apr 10 18:35:16 arkver: try disabling this line staging.bbclass:do_fetch[cleandirs] += "${RECIPE_SYSROOT} ${RECIPE_SYSROOT_NATIVE}" Apr 10 18:36:28 RP: That surely looks a suspicious line. Apr 10 18:36:48 arkver: well, its very necessary but I can start to see how this might break Apr 10 18:37:20 RP: And suddenly it works. Apr 10 18:37:35 DEBUG: Fetcher accessed the network with the command /usr/bin/env svn --non-interactive --trust-server-cert co --no-auth-cache -r 5394 http://svn.openmoko.org/trunk/src/target/AR6kSDK.build_sw.18@5394 AR6kSDK.build_sw.18 Apr 10 18:37:56 arkver: hmm. how to fix this properly I wonder. We'll probably need anon python... Apr 10 18:38:47 RP: I need to call it a day - dogs are hungry. :-) I can look at this a little more tomorrow, but really need to get back to customer oriented work. Apr 10 18:39:17 RP: Is there anything else you need that I can provide in a couple of mins? Apr 10 18:39:20 arkver: thanks for looking at it, I think you've helped solve it Apr 10 18:39:31 arkver: I should have a proper fix shortly fwiw but it'll need testing Apr 10 18:39:42 RP: ok, no probs. Apr 10 18:41:29 arkver: the likely fix: http://git.yoctoproject.org/cgit.cgi/poky-contrib/commit/?h=rpurdie/wip-rss2&id=d524b74531d8441c518bfd648153e1da7f7fba63 Apr 10 18:42:38 RP: I'll give it a go tomorrow - got to dash. Apr 10 18:44:58 arkver: thanks Apr 10 19:11:21 g0hl1n: patch on the mailing list for the svn issue fwiw Apr 10 19:11:27 joshuagl: ^^^ Apr 10 19:22:47 Chromium 52.x keeps failing at compile complaining it can Apr 10 19:22:54 't find gperf. It's definitely installed...? Apr 10 19:27:09 is it properly populated in chromium specific sysroot? Apr 10 19:27:54 Er, this is week 2 of working with Yocto for me, can you explain what you mean by that? Apr 10 19:28:05 recently things switched to package specific sysroots, and if the chromium bb doesn't explicitly call it out in depends, it won't be in the sysroot. Apr 10 19:28:21 so basically "bitbake -c devshel chromium" Apr 10 19:28:43 then you'll see recipes-sysroot dir. Go snooping in there for gperf goodies. Apr 10 19:29:12 Cool, I'll dig into that a bit. Thanks. Apr 10 19:29:23 if you don't find any, then the chromium bb is missing the proper dependency foo on gperf. Apr 10 19:29:38 "devshell" Apr 10 19:30:02 It dropped me into a shell, [do_devshell was the option], and gperf auto-completes...? Apr 10 19:30:07 So it seems like it's there Apr 10 19:31:33 I don' Apr 10 19:31:38 in temp dir you'll find run_do_compile ; you can re-run that in your devshell and snoop around some more. Apr 10 19:31:41 t see a recipes-sysroot dir Apr 10 19:32:16 maybe a hard coded path in a makefile like /usr/bin/ so the sysroot doesn't get searched/used... Apr 10 19:32:44 anyway - gives you a start on how to snoop around on stuff in yocto. Apr 10 19:33:20 Ok, I found the sysroot directory and there is no gperf in it's ./usr/bin dir Apr 10 19:34:15 So I need to add it to the bb file for Chromium Apr 10 19:34:21 ...somehow. :D Apr 10 19:35:18 ya, the devshell can be typically one dir deeper than where the sysroot is. Apr 10 19:35:28 Yea, that was the case. Apr 10 19:44:33 paulg: Got it copied into that directory and it appears to be compiling again. Thanks for the pointers. Apr 10 19:46:00 majuk: np. proper long term fix is to find the chromium-.bb file and append the parent package name of gperf to the DEPENDS line. Apr 10 20:10:19 paulg_: Noted. Are you aware of somewhere I could report this issue? Apr 10 20:11:04 I can't imagine I'm the first person to find this issue, but I guess I could be Apr 10 20:12:39 majuk - it mostly depends on where the chromium.bb file is ; it might be a yocto issue, it might be an open-embedded issue. Apr 10 20:12:50 guessing the latter though. Apr 10 20:15:05 majuk, I cheat and use wildcards to find recipes in layers.. like this: Apr 10 20:15:08 paul@yow-cube1:~/poky$ ls -ld meta*/rec*/chro* Apr 10 20:15:09 drwxrwxr-x 5 paul paul 4096 Apr 10 00:27 meta-browser/recipes-browser/chromium Apr 10 20:15:09 paul@yow-cube1:~/poky$ Apr 10 20:15:44 and then if you go into meta-browser, there is a README with maintainers who you can report issues to ; send patches to, etc. Apr 10 20:16:08 Cool. I'll run that down. Thanks again. Apr 10 20:16:16 no problem. Apr 10 20:17:37 hi all, i'm looking at packaging the vmware guest extensions (which allow vmware images to be much, much faster than they would be without the extensions) into some standard layer. currently, they are in the repo for the open-switch project, and they are a bit stale and don't build properly against master. i have fixed them up to work but am wondering where the most appropriate layer would be. any suggestions? Apr 10 20:17:38 here's where they are right now: Apr 10 20:17:38 https://github.com/open-switch/ops-build/blob/master/yocto/openswitch/meta-foss-openswitch/recipes-extended/open-vm-tools/open-vm-tools_10.0.5.bb Apr 10 22:13:22 So I am mid-build and I filled up the drive I'm mounted to. Are there any files I can remove from the ./build/tmp directory in order to free up some space that won't reset the build operation? Apr 10 22:17:01 Building on an EC2 instance, so I can make a new, bigger volume, mount, copy and build again. That's option B Apr 10 22:25:46 majuk: delete tmp/ dir in your build Apr 10 22:25:52 and restart the build Apr 10 22:26:19 this will reuse sstate and not rebuild already built packages and save some space that way Apr 10 22:26:40 khem: Nice, thanks, I'll do that. Apr 10 22:27:18 Already made a new volume and am copying the whole thing over, so I'll have a backup if I bork something. Apr 10 22:27:19 :D Apr 10 22:43:18 khem: Awesome, worked perfectly. Thanks again. **** ENDING LOGGING AT Tue Apr 11 03:00:02 2017