**** BEGIN LOGGING AT Wed Mar 01 03:00:03 2017 Mar 01 10:22:39 Hi, since recently when I rerun an image recipe without any change there are some tasks (including do_rootfs) being re-executed, any idea from where it come ? (I tested on core-image-minimal and core-image-sato on a new clone repo and same result) Mar 01 10:47:19 Hi All, Mar 01 10:47:20 I'm writing a receipe for a CMake project. This project allows to set some options. How can I set this cmake options in the receipe ? Mar 01 10:51:07 jobro: http://www.yoctoproject.org/docs/current/ref-manual/ref-manual.html#ref-classes-cmake Mar 01 10:52:50 rburton: Thanks ! Mar 01 11:11:03 kanavin: +PR = "r1" <— no need Mar 01 11:11:10 but yay! Mar 01 11:11:29 i do like a one line patch that lets you fix up several recipes Mar 01 11:11:44 rburton: there is a need, without it the change will not be picked up by the compile step Mar 01 11:11:54 damn well should Mar 01 11:12:09 can't be cleaning properly Mar 01 11:12:54 rburton: feel free to test, but I wasted quite some time pulling hair when this change that should've fixed the issue, did not Mar 01 11:13:46 hmm the prepend changed, so it would re-run compile. it should then do the cp correctly. Mar 01 11:13:48 mysterious Mar 01 11:14:13 rburton: it will do the cp, but the subsequent make will not figure out that the file changed, and will do nothing Mar 01 11:14:40 idiot make Mar 01 11:14:48 so, missing dependencies in the makefile? Mar 01 11:15:14 I guess Mar 01 11:15:40 I wouldn't have looked into this, but this issue prevented dnf from running on mips64 image :) Mar 01 11:16:40 haha Mar 01 11:18:27 it's one thing to have introspection broken for webkit, and another to have package management broken Mar 01 11:18:42 yes Mar 01 11:19:58 rburton: so, other than (unfortunately, multiple) multilib failures, the rest of the AB run is in pretty good shape Mar 01 11:20:18 I will try not to spend any more time on this until the end of explorathon Mar 01 11:20:30 which I guess pushes the whole thing to M1 Mar 01 11:21:01 which might be a good thing, because no one except marquiz actually tried the patchset, and the only way to make people do it is to force it on them - sad but true Mar 01 11:21:12 pretty much Mar 01 11:21:17 RP: ^^^ Mar 01 11:21:24 m3 is late for such an invasive change. if it were a gcc major bump we'd say no. Mar 01 11:22:26 rburton: I'm still kind of tempted to try but that is me... Mar 01 11:22:42 If the patches aren't ready, that gives me a problem though Mar 01 11:22:54 (and I know its not for want of trying) Mar 01 11:23:05 RP: multilib is broken, or at least it is incompatible with the assumptions the AB tests make Mar 01 11:23:27 multilib is a *nightmare* Mar 01 11:23:33 and pretty important Mar 01 11:23:46 hi! I'm newby in yocto. I'm using the yocto on nitrogen6x board. I have to reduce my boot time. So I need to compile my kernel. I have changed my kernel files a little bit. How can I compile the new kernel? Do you have any document about this? Mar 01 11:24:04 rburton: the basic use cases work, but AB tries advanced stuff too, unfortunately :D Mar 01 11:26:34 rburton: who is a multilib specialist, should I need one? Mar 01 11:26:51 RP: ^^^ Mar 01 11:27:38 kanavin: probably me or fray Mar 01 11:28:05 which timezone fray is actually in? Mar 01 11:28:10 where is he? Mar 01 11:31:03 kanavin: mid US Mar 01 11:38:39 hello Mar 01 11:38:54 we have issues trying to download iqvlinux from sf.net Mar 01 11:39:10 is there a way to add an alternative URI for download? Mar 01 11:39:16 (in the recipe) Mar 01 11:43:44 found a bitbake page about mirrors Mar 01 12:16:24 RP: I was wondering about base_conditional being deprecated (it isn't yet) then I started to compare oe.utils and bb.utils. Iirc long ago the idea was that bitbake is not necessaary used to build OE so it has its set of utils. Now I see there are even duplicate func between the two utils.py (i.e. prune_suffix) Mar 01 12:17:16 What is the logic behind the split? Mar 01 12:18:46 ant_work: there is no idea that bitbake is not necessary to build OE that I'm aware of Mar 01 12:19:10 ant_work: In general we tend to prefer bitbake but it really depends on the context of the functions Mar 01 12:19:23 I rephrase: bitbake could build other stuff Mar 01 12:22:15 better phrasing :) Mar 01 12:23:24 iirc there was some example of bitbake standalone...my ram is defective by now Mar 01 12:25:27 there was a talk at elc sort of related Mar 01 12:29:22 Right, there was a talk where they used bitbake but their own metadata Mar 01 12:29:35 and this does support the idea of some of the common functions being in bitbake Mar 01 12:29:43 kanavin: I like the mips64 patch :) Mar 01 12:29:58 the pr bump is bothering me :) Mar 01 12:30:09 rburton: well, could be worse Mar 01 12:31:33 RP: I'd say in the oe utils.py should remain only functions somehow bound to the metadata Mar 01 12:32:53 ant_work: ultimately I'd agree Mar 01 12:45:43 * abogani waves all Mar 01 12:45:52 If I build the sdk (aka -c populate_sdk) for two machines with different kernel but the same DISTRO and TUNE_PKGARCH (i.e. cortexa8hf-neon) I incur this build error: Mar 01 12:46:14 ERROR: XXXX-sdk-1.0-r0 do_populate_sdk: The recipe XXXX-sdk is trying to install files into a shared area when those files already exist Mar 01 12:46:38 The simplest solution is set TOOLCHAIN_OUTPUTNAME to something derived from MACHINE but it creates as many SDK as MACHINE which I would prefer avoid.... What I'm doing wrong? Mar 01 12:59:15 abogani: can you be more specific about the kinds of files its trying to overwrite? Mar 01 12:59:58 abogani: can't you just build and use one of these if they're all the same? Mar 01 13:00:18 I'm guessing since the kernel is different and it could include the kernel, its building each time Mar 01 13:04:48 RP: Are there a way to build an SDK *without* kernel so it can targeting any hw boards of TUNE_PKGARCH type? Mar 01 13:05:26 RP: It tries to overwrite the 3 files in build/tmp/deploy/sdk/. Mar 01 13:42:57 is there a way to add more than one mirror to SOURCEFORGE_MIRROR ? Mar 01 13:48:01 I'm using the yocto on nitrogen6x board. I have to reduce my boot time. So I need to compile my kernel. I have changed my kernel files a little bit. How can I compile the new kernel? Do you have any document about this? Mar 01 13:48:51 cornel: you should be able to add new ones to MIRRORS, see meta/classes/mirrors.bbclass Mar 01 13:49:02 jku, thank you Mar 01 13:52:30 jku, ok, it seems i've needed the \n :) Mar 01 13:52:41 jku, thank you very much Mar 01 13:54:14 jku, is the order important? Mar 01 13:54:59 jku, if i've placed first the sf.net mirror var then the hardcoded mirror, it failed at the former without trying the latter Mar 01 14:00:20 cornel: if you add a line to MIRRORS, the original should be tried first, then the one you added Mar 01 14:00:37 jku, that was also my expectation Mar 01 14:51:56 abogani: you are correct that if you don't have kernel dependencies in the sdk, it should then have the same stamps in those different configurations and not be overwriting those files Mar 01 15:06:43 Hi, is there a way to get a prompt to type my git credentials within bitbake. I want to use private git repositories, unfortunately key authentication is not possible on this git-server. To write my credentials to the recipe is not wanted. Mar 01 15:07:05 no -- bitbake process time does not permit stopping and asking the user for questions.. Mar 01 15:07:13 there are other ways to deal with this.. Mar 01 15:07:42 you can use a program that intercepts GIT_ASKPASS and returns credentials, or you can put your credentials into your local .nc file.. Mar 01 15:08:50 (at Wind River we use a program called 'anspass', https://github.com/WindRiver-OpenSourceLabs/anspass) Mar 01 15:13:44 fray: Thanks, I will check it Mar 01 15:39:28 does vnc work on yocto? Mar 01 15:40:44 fqtw: clien? Mar 01 15:40:49 client? Mar 01 15:44:30 lsandov, server Mar 01 15:45:42 fqtw: https://layers.openembedded.org/layerindex/branch/master/recipes/?q=vnc just shows client stuff Mar 01 15:46:41 lsandov, isn't x11vnc the server? Mar 01 15:47:56 fqtw: you are right. give it a try Mar 01 15:48:11 ok thanks Mar 01 15:49:38 lsandov, btw, where should I start if I want to compile binaries for yocto with a layer, (but I don't want to build a yocto image)? Mar 01 15:51:51 i'd say the only meaningful way to use a layer without doing an acutal build is by using an esdk that fits the image in use. Mar 01 15:52:54 fqtw: why dont you want a build? you can try the vnc stuff with a qemu machine Mar 01 15:53:16 fqtw: but you need to build a image including your package Mar 01 15:54:35 of course you can crosscompile for the image if you have only a toolchain. but thats unrelated to any layer, then. Mar 01 16:08:17 joshuagl: do both ABs have the same sstate mirrors etc setup now? Mar 01 16:09:16 rburton: both should be setting the same sstate mirror in eSDK installer configuration, if that's what you're asking Mar 01 16:10:24 jku: do you still intend to do anything for 2.3 on https://bugzilla.yoctoproject.org/show_bug.cgi?id=9527 ? Mar 01 16:10:25 Bug 9527: enhancement, Medium, 2.3, jussi.kukkonen, NEW , Design support for a stateless OS Mar 01 16:17:07 lsandov, I don't have the recipe for the image I want to compile the binaries for. only the OEM has it Mar 01 16:17:39 fqtw: sounds like the OEM wasnt thinking very much about how they expect people to use the software to be honest Mar 01 16:18:31 rburton, they just expect people to use the existing functionality but not add to it Mar 01 16:18:54 they are selling mainly the hardware Mar 01 16:36:27 god, archiver.bbclass is so freaking broken Mar 01 16:37:32 kergoth: tag you're it! Mar 01 16:37:40 do_ar_configured is completely non-functional for non-linux-yocto kernels, as far as i can tell. it overrides WORKDIR nad re-runs do_unpack/do_patch, but kernel.bbclass arranges to wipe STAGING_KERNEL_DIR and move S to STAGING_KERNEL_DIR regardless of what WORKDIR is set to Mar 01 16:38:01 which means do_unpack_and_patch breaks do_configure, depending on task execution order Mar 01 16:38:08 i think i was the last person to rewrite it all, but that predates the kernel magic Mar 01 16:38:15 maybe we need better selftest for archiver ... Mar 01 16:38:56 also, just running do_unpack and do_patch alone isn't guaranteed to get you the same source tree as S, when other tasks have been introduced to prepare the sources further Mar 01 16:39:31 i feel like it should just make devtool do it, if we can't hook the existing tasks, and at least get that source preparation logic consolidated in one place that way Mar 01 16:40:11 maybe archiver shoud run all tasks up to do_compile instead Mar 01 16:40:36 hmm Mar 01 16:41:09 one issue is archiver tries to satisfy all use cases, so ti has a ton of configuration options, and i'm guessing we aren't testing every combination Mar 01 16:41:25 but i think fundamentally ohw it's implemented today is flawed Mar 01 16:42:46 * kergoth ponders Mar 01 16:45:11 kergoth: it does sound broken. We do have the "tasks between" logic now so it could be potentially fixed Mar 01 16:45:22 not that I like the code Mar 01 16:46:24 i think we need to add a marker along the line sof do_image_complete, but for the sources being fully prepared Mar 01 16:46:37 then devtool woudln't need its hardcoded handling of linux-yocto so much Mar 01 16:58:19 kergoth: doesn't do_configure effectively act as that? Mar 01 16:58:43 kergoth: I guess you could just have a variable specifying the task with configure as the default? Mar 01 17:11:22 true, i always forget about the - suffix support you added, guess just running do_configure- would get the job done Mar 01 17:12:35 i forgot too Mar 01 17:12:42 the syntax is too horrible to remember ;) Mar 01 17:57:50 Has anyone seen Jefro? Mar 01 17:58:44 outreachy? Mar 01 18:11:20 Yes. Mar 01 18:44:54 bluelightning: devtool add doesn't work for me when trying to create a npm based recipe :( Mar 01 18:45:32 bluelightning: latest poky master, meta-oe is included in layers, I get this: Mar 01 18:46:48 'Host distribution "Ubuntu-14.04" has not been validated with this version of the build system' http://www.yoctoproject.org/docs/2.2.1/mega-manual/mega-manual.html#detailed-supported-distros Mar 01 18:47:35 I am curious if that message is accurate somehow Mar 01 18:48:01 you're looking at the 2.2 manual, are you on the morty branch? Mar 01 18:48:10 master dropped 14.04 for pyro, afaik Mar 01 18:50:21 bluelightning: http://pastebin.com/HcQmFGdy Mar 01 18:54:14 kanavin_home: yep, there's a bug covering that Mar 01 18:54:19 kanavin_home: it's a result of RSS Mar 01 18:55:00 bluelightning: at the moment my interest is mostly in getting shrinkwrap and lockdown files out of it Mar 01 18:55:09 bluelightning: I can write the rest of the recipe myself :) Mar 01 18:55:12 https://bugzilla.yoctoproject.org/show_bug.cgi?id=10992 Mar 01 18:55:13 Bug 10992: normal, Medium+, 2.3 M3, paul.eggleton, IN PROGRESS IMPLEMENTATION , devtool: add: check for nodejs-native broken since recipe specific sysroots changes Mar 01 18:55:23 you may wish to try with morty instead Mar 01 18:56:20 bluelightning: and then the plan it to see how Go does these things (modules installation), and whether we can abstract it into a generic thing with plugins Mar 01 18:58:37 kergoth: I *think* this build I am attempting is for Dylan (1.4) Mar 01 18:58:50 bluelightning: the problem should be similar regardless of language: fetching, checksumming and license check needs to be generalized to language modules installed with a custom tool Mar 01 18:58:55 JesperHan: every release has its own set of tested / supported distros. Mar 01 18:59:05 you'd have to examine the manual corresponding to the release you're using Mar 01 18:59:20 note the '2.2.2' in the manual url you linked Mar 01 18:59:24 oh, http://www.yoctoproject.org/docs/1.4/mega-manual/mega-manual.html#detailed-supported-distros Mar 01 18:59:25 er 2.2.1 Mar 01 18:59:39 starting to understand Mar 01 19:00:15 thank god for vagrant Mar 01 20:32:52 is there a way to change PV after do_configure(), for example in a do_configure_append()? Mar 01 20:33:10 i'd like to embed inside PV the version of something that is downloaded in do_configure() Mar 01 20:35:09 nope, PV is used by bitbake, it has to be set at parse time Mar 01 20:36:00 hm. can i fetch something and run a parse-time recipe to set it? Mar 01 20:36:23 there's a build number embedded within a dependency that i'd like to extract and put into PV Mar 01 20:47:42 miceopede: so you could do it in anonymous python, although that'll be ugly to have that running at that time, and it may run more often than you'd like Mar 01 20:48:11 miceopede: one alternative solution would be to set PKGV instead of PV - that can be set at packaging time, so the output packages have the desired version on them Mar 01 20:52:08 @bluelightning if i do it at packaging time, just override do_package(), or add do_package_append()? Mar 01 20:52:15 and set PKGV? Mar 01 20:55:09 miceopede: it'd need to be a do_package_prepend() (and it'll be python since that's what do_package is) Mar 01 20:55:38 populate_packages_prepend would also work Mar 01 20:56:35 got it, thanks. how do i set PKGV from python? PKGV = "foo"? not really sure how the scoping works in bitbake. Mar 01 20:58:53 there are effectively two scopes.. 'global' (config space) and local (recipe space).. local always has priority Mar 01 20:59:07 so PKGV = "foo" will set PKGV inside the recipe.. Mar 01 20:59:22 as long as something 'later' doesn't come in an do 'PKGV = "bar"' it will do what you want Mar 01 20:59:33 @fray thank you Mar 01 21:00:07 (there are a lot more operators then python, such as ??=, ?=, +=, =+, _append, _prepend, .= and =. Mar 01 21:00:17 ??= is set if nobody else -ever- sets it.. Mar 01 21:00:22 ?= is set if not already set Mar 01 21:00:30 += append to the string with a leading 'space' Mar 01 21:00:36 =+ preend to the string add a space Mar 01 21:00:53 append and prepend just do it, but -last- after all other operations have been performed.. Mar 01 21:01:06 .= and =. are same as += and =+ but without the additional space.. Mar 01 21:01:32 _remove = assume this is a space separated list, and remove the item fromt he list (does it very late) Mar 01 21:02:06 @fray but setting PKGV inside do_package_prepend() only sets it for the do_package_prepend() scope, no? Mar 01 21:02:23 correct.. you need to do it at ther ecipe level.. Mar 01 21:02:34 so I guess there are three levels.. config (global), recipe and task Mar 01 21:03:33 i don't know what PKGV should be at parse-time, only at do_package_prepend(). Mar 01 21:04:16 oh, i guess i could d.setVar()? Mar 01 21:05:01 doesn't really matter what PKGV is at parse time (it can stay the default of "${PV}" Mar 01 21:05:37 d.setVar() is what you'd use within do_package_prepend or populate_packages_prepend yes Mar 01 21:05:57 there are actually three* scopes Mar 01 21:06:07 (* because there are extras in some parts of the code) Mar 01 21:06:27 the third not covered is each task... you can't just setVar() in a task and expect it to apply to other tasks Mar 01 21:06:44 here we're just doing it within do_package so it'll be fine (populate_packages is called from within do_package) Mar 01 21:06:59 the 'anonymous' python is the only pyhton section that covers teh recipe scope right? Mar 01 21:07:07 yep Mar 01 21:07:28 d.setVar() works perfect. thanks all Mar 01 21:07:31 well, aside from declaratively setting a variable from the recipe (A = "value") Mar 01 21:07:57 and of course, you can set a variable declaratively to a python expression, so you can get a dynamic value that way as well Mar 01 21:08:12 what the hell is anonymous python? as i understand it, bitbake recipes are parsed by some custom parser, and parts are compiled to python and the rest are run as shell? Mar 01 21:08:25 reminds me I forgot the ':=' which is 'evalute this expression -now- Mar 01 21:08:25 python() { ... } Mar 01 21:08:47 that's an anonymous python section which will execute as part of parsing the recipe Mar 01 21:08:48 the anonymous python function runs (effectively) at parse time Mar 01 21:09:16 you can name it "__anonymous" and you'll still see that in places in the metadata, but that's entirely optional Mar 01 21:09:21 you can define multiple anonymous python chunks to run (but run order is not promised) Mar 01 21:09:48 http://www.yoctoproject.org/docs/current/bitbake-user-manual/bitbake-user-manual.html#anonymous-python-functions Mar 01 21:10:08 time to rtfm :D Mar 01 22:44:19 second question: can i set a global variable in do_package_prepend() that will be picked up in ROOTFS_POSTPROCESS_COMMAND() later? i know i could manipulate files on the rootfs and then do it that way, but would be easier to just set a number and get it out at rootfs creation time Mar 01 22:58:24 can i expect to d.setVar in a recipe's do_package_prepend() and be able to get that out in a ROOTFS_POSTPROCESS_COMMAND()? wit ${@d.getVar...}? Mar 01 23:44:03 halstead: https://autobuilder.yocto.io/builders/nightly-intel-gpl/builds/169/steps/Git%20checkout%20of%20poky/logs/stdio - not good :/ Mar 01 23:44:50 halstead: failures on three different workers Mar 01 23:45:05 halstead: in case I'm not around, we need a master build... Mar 01 23:52:20 halstead: I'm around tonight after I get home from JF. Ping me if you need anything. Mar 01 23:59:00 * RP has fired the build on the old cluster, it will just take a lot longer there as the sstate cache is cold Mar 01 23:59:25 Incidentally, I think we sped up rebuilds hugely with that recent runqueue.py change Mar 02 00:00:10 and the recent selftest improvements thanks to mariano **** ENDING LOGGING AT Thu Mar 02 03:00:01 2017