**** BEGIN LOGGING AT Mon Aug 07 03:00:03 2017 Aug 07 05:49:41 Is it possible to disable /sys (sysfs) in image ? Aug 07 05:54:40 ranran: possible? certainly. will the system still work? probably no :-) Aug 07 05:55:45 LetoThe2nd: how should it be done in yocto ? Aug 07 05:56:01 ranran: disable it in your kernel config, see what happens Aug 07 05:56:24 but there is also a mount command in the startup script, right ? Aug 07 05:56:41 it should be part of the fstab. Aug 07 05:57:31 how should I update fstab ? Aug 07 05:57:47 um, just remove the line that mounts the sysfs? Aug 07 05:57:55 but why would you want to do that? Aug 07 05:58:24 do you mean, find & update the recipe for fstab ? Aug 07 05:58:53 i would just try it on the target first. my bets are on "system will not work anymore" Aug 07 05:59:53 it is embedded small linux, we've done it in other disctributions (not in yocto). Aug 07 06:00:46 then just do whatever you did in the other distributions here too. openembedded is no different Aug 07 06:06:59 LetoThe2nd , thanks. Aug 07 06:07:24 What is the difference between IMAGE_INSTALL & IMAGE_FEATURES ? Aug 07 06:07:42 Aren't both responsible for adding a package into image ? Aug 07 06:09:17 ranran: image_install lists packages to be installed. image_features are, well.. think more general image funktionality. like, all debug images, or all docs. see the dev manual, it has an extensive list of all available image_features. Aug 07 06:12:27 How can I know the exact package name to be specified in IMAGE_INSTALL, for example tftp server/client , How to find its package name ? Aug 07 06:14:29 ranran: start searching at https://layers.openembedded.org Aug 07 06:14:55 ranran: usually the package name is the recipe name (but there might be special cases) Aug 07 06:17:29 LetoThe2nd, thanks! Aug 07 06:19:19 LetoThe2nd, I will need to make a minimal linux, initramfs, with custom init, is the best approach will be to make image changes by append core-image-minimal recipe ? Aug 07 06:20:41 ranran: close. we also have a "core-image-minimal-initramfs" :-) http://cgit.openembedded.org/openembedded-core/tree/meta/recipes-core/images/core-image-minimal-initramfs.bb?h=master Aug 07 06:32:46 Is it also possible in do_install to remove binaries from rootfs ? Aug 07 06:32:59 sort of cleanup Aug 07 06:36:16 that is more like a rootfs post process. but why would i need to remove anything? just do not install it earlier :-) Aug 07 06:38:25 morning all Aug 07 07:21:19 Does "image" in yocto terminology means building rootfs, or building the final image (for example hdd image) ? Aug 07 07:22:05 for example ROOTFS_PREPROCESS_COMMAND image is before building the rootfs or before building the final image ? Aug 07 07:23:35 ranran: "image" mwans the contents of the image, e.g. in the vast majority of cases the rootfs. tar.gz, extX, any kind of binary blob is an IMAGE_FSTYPE then Aug 07 07:27:06 LetoThe2nd, thanks ! Aug 07 07:32:33 I am a bit confused between ${D} , and IMAGE_ROOTFS. Is it the same thing ? (ROOFS /) Aug 07 07:36:13 ranran: D is the _recipe specific_ installation directory Aug 07 07:36:16 /win 12 Aug 07 07:39:03 LetoThe2nd, Thanks, that really help me to clear some uncertainties ! Aug 07 08:30:50 hmm, there was a "rpm: add explicit dependency on bzip2-replacement-native for native builds" patch in june but I still just saw that failure -- rpms debugedit fails to load libbz2.so.1 because it's not in sysroot Aug 07 08:33:18 rerunning the task succeeds :/ Aug 07 09:09:35 What can cause glibc-dbg to get installed into the image, when its manifest does not mention it? Aug 07 09:15:12 sveinse: imagefeature debug, in the simplest case Aug 07 09:16:33 I have to admit I had expected that it would reflect on the manifest Aug 07 09:17:01 hm. good point. Aug 07 09:17:51 is glibc the only dbg package being installed? Aug 07 09:18:21 LetoThe2nd: I think so. Difficult to tell when the manifest isn't reliable :( Aug 07 09:18:45 sveinse: you could inspect the dependency thing of the image Aug 07 09:21:53 right, enable buildhistory and it'll give you the dep graph for the image Aug 07 09:25:11 I might have been a little quick. The manifest /does/ mention libc6-dbg. I was confused by the naming of glibc vs libc it seems Aug 07 09:26:03 ...which is good. Having an inconsistent manifest kind of scared me Aug 07 09:29:48 A small poll: What pkg class do you guys use? ipk, deb or rpm? Aug 07 09:30:04 ipk, no runtime package managemenger :) Aug 07 09:47:02 Isn't it common to disable installation of rrecommends at all in Yocto? In full fledge debian systems, you can configure it to never install recommends on a global level. The manual rather hints at specifying individual packages with BAD_RECOMMENDATIONS, NO_RECOMMENDATIONS and PACKAGE_EXCLUDE Aug 07 09:48:29 If you're building a minimal system, you'd want to minimize the inclusion of rrecommends, right? Aug 07 09:53:14 sveinse: NO_RECOMMENDATIONS does exactly that, disables all RRECOMMENDS Aug 07 09:54:02 if the manual isn't clear on that we should probably improve it Aug 07 10:45:49 if I modify a libinput so the main package no longer includes binaries, the package gets automatically renamed (libinput -> libpinput10). I thought in this case I'd have to handle RPROVIDES/RREPLACES/RCONFLICTS but my patch in June 1 wasn't merged... Aug 07 10:46:02 rburton: remember if there was some other problem or have I misunderstood? Aug 07 10:47:16 (also: hi, I'm back from vacation) Aug 07 10:48:52 jku: huh. ping the patch please :) Aug 07 10:49:19 rburton: oh that one is no longer valid. I'll send a new one Aug 07 12:59:54 replying to mailing list threads defaults the ¨TO:¨ field to the sender´s address, should I change it to the mailing list address? Aug 07 13:00:41 probably yes Aug 07 13:00:46 fadeway: correct behaviour is to sender, cc list Aug 07 13:01:01 thanks Aug 07 13:15:46 hm wish runqemu wouldn't blast all over the selftest log Aug 07 13:26:16 hi is there possibility to have one lib in few versions in one layer and have some images that depends on different versions of that lib? example i need mylib in versions 1.0 and 1.1 and i have imageA that depends on mylib 1.0 and imageB that depends on mylibB Aug 07 13:26:40 exetyrhing should be in one layer and if possible on the same branch Aug 07 13:30:30 write two recipe, one for each version. *assuming* you can install them in parallel, i.e. no file names that clash. Aug 07 13:33:04 rburton: mylib in both versions provides the same files but the content differs Aug 07 13:33:34 in theory with the latest release that *might* work Aug 07 13:33:40 well, no Aug 07 13:33:50 because you can't do versions in build dependencies Aug 07 13:34:15 fix the library so you don't need two versions? Aug 07 13:34:30 if you need both versions as there was a massive API change then the version of the library should have changed, so the files won't conflict Aug 07 13:35:27 rburton: the problem is that because of api change I cant use the same lib in both images Aug 07 13:35:44 atleast for some time Aug 07 13:35:47 api change should mean soname version change should mean filename change Aug 07 13:36:00 if they didn't do that then whoever wrote the library needs to be told to do it properly Aug 07 13:38:29 rburton: ok thanks, I hoped that some hack in layer will be sufficient for this.. Aug 07 14:03:18 Does anyone know what controls the size in wordsize.h when generating a toolchain with bitbake meta-toolchain? Aug 07 14:56:03 RP: any ideas on this? https://gist.github.com/kergoth/1c7a38c5a0ab1a2a84cb4b17971781d1 Aug 07 14:59:12 The manual states that NO_RECOMMENDATIONS only is supported for ipk and rpm, but not deb. What this imply for deb? That is installs every RRECOMMENDS for deb, or that is doesn't? Aug 07 15:01:49 kergoth: Is libgcc-external in PACKAGES? Might it not be seeing the RDEPENDS leading to a build race? Aug 07 15:02:18 sveinse: that it does install every RRECOMMENDS Aug 07 15:02:26 RP: thanks Aug 07 15:02:35 RP: it is, yeah, though iirc it's renamed with PKG_ Aug 07 15:02:54 good thought, though, maybe i'll drop the package renaming bits to see if that's a factor Aug 07 15:10:10 * kergoth also checks for task ordering issues, could be the pkgdata wasn't yet populated, and my examining the files was after they all run Aug 07 15:12:56 btw, rumour is that YPDD on oct 26, right? Aug 07 15:18:18 denix: is there a pyro-branch in meta-ti ? Aug 07 16:20:23 If I want some functions and variables set in the global scope of bitbake, but does not want to clutter local.conf, where is a natural location and/or naming for it? Aug 07 16:21:12 in your distro config, or a class that you inherit at the distro level Aug 07 16:21:35 Specifically, it contains scripts for getting the central CM controlled info such as fetching the release version number Aug 07 16:21:42 rburton: thanks Aug 07 16:25:41 what is "uninative" poky releases? Aug 07 16:29:56 so, someone patched diffutils recently to no longer use its internal copy of getopt if it can't find a good one, yet we don't set gl_cv_func_getopt_gnu=yes in our siteinfo, which means it always assumes we don't have a valid gnu getopt, which causes the build to fail Aug 07 16:29:58 lovely Aug 07 16:30:02 The better question is what is uinative Aug 07 16:32:54 sveinse: uninative is just a rebuild glibc that can be used to abstract the host os if you want to share sstate between many distros. Aug 07 16:33:05 sveinse: poky uses it so the autobuilder cluster can share native sstate Aug 07 16:34:56 for ALTERNATIVE_PRIORITY, is a higher or lower value prefered when update-alternatives runs? Aug 07 16:38:35 The best I can tell is that higher trumps lower based on https://linux.die.net/man/8/update-alternatives Aug 07 16:39:45 think that's correct, yes Aug 07 16:39:46 I wonder if bitbake -e files can be sourced by sh safely Aug 07 16:39:59 At first glance it seems so Aug 07 16:40:06 sveinse: yeah the format was designed to be shish Aug 07 16:40:12 great Aug 07 16:43:28 sveinse: note he said sh *ish*. it also emits the python functions for reference, which won't be valid. so some of it, yes, the entirety, possibly not Aug 07 16:43:32 shish apparently. bash doesnt like the empty syntax like DISABLE_STATIC_pn-nativesdk-openssl="" Aug 07 16:43:52 fails with command not found :( Aug 07 16:43:53 empty is irrelevent Aug 07 16:43:55 it's the dashes Aug 07 16:43:59 dash isn't valid in shell variable names Aug 07 16:44:03 ah, right, yes Aug 07 16:45:15 not sure i want to know why you're running -e into a shell ;) Aug 07 16:45:47 if you want to grab some values then the format is sh like so its trivial to parse Aug 07 16:47:21 I'l tell you anyways :) Its our CM's pre-build setup scripts. Rather than duplicating settings in the CM config, its better to fetch them from yocto's local.conf & co. Aug 07 16:47:29 Parsing the -e output is trivial, no worries Aug 07 17:13:44 can also just grep out the particular vars you need and eval just those parts Aug 07 17:14:57 kergoth: not too fond of eval, so I grep and sed it into a variable. But essentially the same thing, yeah Aug 07 18:03:28 Does anybody know when/how should I run depmod when adding recipes for out-of-tree kernel modules to an image? Aug 07 18:14:54 just inherit the appropriate classes (generally 'module') and its hould be done for you by setting up a postinst to do it when the package is installed Aug 07 18:14:58 afaik anyway Aug 07 18:19:05 hmmm, strange. I inherit the module class, but no modules.dep is generated for the final image so maybe I need a different class or I borke things somewhere Aug 07 20:14:58 I have an overgrown local.conf, and are planning to move most of it into a custom distro. Are there any guidelines on what parameters belong in a distro and what should be left in local.conf? Aug 07 20:16:26 local.conf is meant to be local to that particular build, likely not source controlled, and is intended to be easy for the user to modify. i'm assuming you mean what should be left in a local.conf *template*, and the answer there depends on what you want to make easy for the user doing a build to be able to change, or what you want to be very visible to the user, rather htan hidden away in the distro Aug 07 20:20:53 right. I understand. In our context, the purpose of the builds is not to prepare for generic use, but to build and replicate exact product builds. So the concept of user is somewhat moot. And because of this our local.conf is version controlled. Aug 07 20:21:14 because of precisely this, the local.conf has grown to host settings that does not belong there Aug 07 20:22:57 in an ideal world local.conf changes are limited to setting DISTRO (and actual host settings such as DL_DIR), and distro can be set via a local.conf template in your layer Aug 07 20:27:11 In that context, the sample local.conf is quite elaborate. Or can be made quite elaborate. Aug 07 20:27:39 the word you're looking for is "sample" :) Aug 07 20:27:56 its part of the documentation Aug 07 20:28:40 But I suppose what a distro image is varies from user to user. E.g. some tweak on a distro like Poky (most users?) and other, like us, want to build a fixed image. Aug 07 20:28:55 i hope most users are not tweaking poky via local.conf Aug 07 20:28:58 that would be doing it very wrong Aug 07 20:29:02 (they're not) Aug 07 20:29:39 uhm, there is a reason I'm making a distro right now. *cough* Aug 07 20:30:40 you should have a distro that sets the global variables, and your own images which are possibly constructed from your own package groups Aug 07 20:31:13 the only things in local.conf should be host settings (site.conf exists for a reason...) and per-build tweaks. Aug 07 20:31:37 (for the normal use, obviously my job is testing the patches so my local.conf is about 400 lines of commented out tweaks) Aug 07 20:35:57 yocto or bitbake (or toaser) never writes or changes auto.conf, right? Aug 07 20:38:02 no, nothing writes to auto local or site.conf Aug 07 21:23:42 <_jrsharp> hey all Aug 07 21:24:26 <_jrsharp> what would this error message ("ERROR: Nothing PROVIDES ''") indicate when attempting to set up and build a yocto image? Aug 07 21:24:45 <_jrsharp> (generally) Aug 07 21:28:32 _jrsharp: something is depending on '' but nothing is providing it. The system doesn't do that by default in our testing so it has to be something you've added or configured Aug 07 21:29:57 _jrsharp: I can make the system do that if I type "bitbake ''" Aug 07 21:31:01 <_jrsharp> ok... well, what would the empty string indicate, then? That one of my .bb/recipe files is referencing an empty string? if I had a name, I could track it down, but with an empty string, what do I look for? Aug 07 21:37:58 _jrsharp: You could try the -k option, see if that gives any more hints Aug 07 21:41:23 If bitbake throw a DISTRO 'sp-poky' not found at me, how can I debug what it's complaining about? Aug 07 21:41:41 I have a layer with conf/distro/sp-poky.conf in it Aug 07 21:43:37 please bear with a basic and/or stupid question... Aug 07 21:43:39 sveinse: is the layer in bblayers? does the layer.conf say there are files in conf/? Aug 07 21:44:23 i'm using Varascite's yocto imx6ul project as a starting poitn for our project Aug 07 21:44:29 variscite. Aug 07 21:44:33 sveinse: you're probably missing a BBPATH assignment in the layer.conf Aug 07 21:45:15 i want to modify the GPIO1_IO03 muxing, so I searched on that string under /sources Aug 07 21:45:57 and i only got hits on it under /sources/poky/build_x5/tmp/work-shared/... Aug 07 21:46:12 this confuses me. Aug 07 21:46:34 where do these .dts files originate from? Aug 07 21:46:36 <_jrsharp> thanks RP Aug 07 21:47:15 _jrsharp: when you find it I'd be interested to know how to trigger that and see whether we can make the cause more obvious, that might be worth a bug if/when you find it Aug 07 21:47:33 yates: kernel source I'd guess (which is placed in a shared work directory) Aug 07 21:47:38 and i only got hits on it under /sources/poky/build_x5/tmp/work-shared/.../imx6ull-var-dart-common.dtsi, for example. Aug 07 21:48:44 RP: ok. so if i want to modify something in that source, i need to generate a .patch ? Aug 07 21:49:00 yates: yes Aug 07 21:49:40 rburton_: it was, thanks Aug 07 21:50:49 there are several files with this string in the kernel-source/arch/arm/boot/dts/ directory which could be used, e.g., imx6ul-14x14-ddr3-arm2.dts, or imx6ull-14x14-ddr-arm2.dts, for example. how do i know which one the project is utilziing? Aug 07 21:54:11 RP: is there a document that describes how to generate such a patch file? would it go in my custom layer somewhere? is it applied prioer to the files getting "generated" info the /tmp/work-shared folder? Aug 07 21:56:15 <_jrsharp> yates: btw, there is a webinar scheduled tomorrow: Demystifying Device Tree for NXP i.MX Processors Aug 07 21:56:54 yates: The dts should generate a dtb file, so in my limited knowledge, I *think* you could just find the dtb file that gets generated Aug 07 21:57:27 yates: And if the dts file is in tree for the kernel, just patch that corresponding file Aug 07 21:57:35 _jrsharp: Is it open to anyone? Do you have an URL to it? Aug 07 21:59:25 rewitt: right, but a) how do i know which files in the kernel source tree are getting pulled in (as I asked previously), and b) even knowing how to generate a .patch file, where exactly does it go in my sources ? Aug 07 22:00:07 * _jrsharp looks for URL... Aug 07 22:00:20 i know these are basic,, i don't know how elswe to be begin. i've done some reading of the yocto reference manual but haven't seen the answer (yet) Aug 07 22:01:12 <_jrsharp> Here it is: https://www.toradex.com/webinars/demystifying-device-tree-for-nxp-imx-processors Aug 07 22:01:29 thanks _jrsharp Aug 07 22:01:56 <_jrsharp> np Aug 07 22:02:38 i've also tried to parse the dev-manual - nye onto impossible! Aug 07 22:03:02 s/nye/nigh/ Aug 07 22:03:14 s/nigh onto/damn near/ Aug 07 22:06:46 <_jrsharp> RP: my empty string issue was related to RECOVERY_INITRAMFS_IMAGE not being set... looks like I wasn't specifying the DISTRO appropriately Aug 07 22:13:02 _jrsharp: do you know specifics on the imx6ul/imx6ull processors? Aug 07 22:13:45 namely, where does the REF_CLK_32K come from in GPIO1_IO03 mux? i don't see it in the reference manual Aug 07 22:13:53 <_jrsharp> I'm fairly new to the imx6, but am working with the imx6q/imx6dl cpus on a project presently Aug 07 22:15:19 <_jrsharp> have you looked at the relevant .dtsi files in your kernel source? (arch/arm/boot/dts) ? Aug 07 22:16:26 yes, i am looking ther enow Aug 07 22:17:31 <_jrsharp> it may also trace back to an included header file, in the case of the pin mux options... Aug 07 22:18:17 regardless of what is in those .dts[i] files, how can you select a non-existent mux?!? Aug 07 22:18:23 <_jrsharp> I don't know about that signal specifically, but I recall having to peel back several layers before I found all of the alt-function pin mux options Aug 07 22:18:43 <_jrsharp> yeah... that's strange if not in the manual Aug 07 22:19:10 <_jrsharp> t Aug 07 22:19:21 <_jrsharp> I know the ul parts differ a fair bit from the q/dl parts Aug 07 22:19:58 what is a "keeper"? Aug 07 22:21:30 _jrsharp: see 33.6.10 of the rm. do you see a mux mode ALT3? Aug 07 22:21:47 that's what they're selecting (according to the scehmatic) for REF_CLK_32K Aug 07 22:23:18 yates: generically speaking, and given I understand the context right, many SoC has a bus-keep function. It ensures that the pin won't float, but stay either high or low to prevent excessive current consumption Aug 07 22:23:35 aha. Aug 07 22:23:46 Its an input, not low Z driven output Aug 07 22:24:18 basically they connect a resistor network for a logic low or high? Aug 07 22:24:29 as an input .. Aug 07 22:25:25 damn cmos. gimme the good-ole days of LS-TTL... Aug 07 22:25:34 and wire-wrap Aug 07 22:26:11 No, its a little more refined than that. If left floating it will, say low. If you drive it high, it will stay high after you've removed the driver. A kind of schmitt trigger into positive feedback Aug 07 22:26:37 hm Aug 07 22:28:38 <_jrsharp> yates: good luck, signing off... Aug 07 22:28:42 <_jrsharp> thanks all! Aug 08 01:57:53 where can i find a current guide on building yocto for a galileo gen2 board? most builds / guides i'm finding aren't for poky. Aug 08 02:20:42 hello? Aug 08 02:27:13 the intel forums is the only place I've seen info like that. Aug 08 02:27:28 and I don't think it's that specific as you want. Aug 08 02:27:59 and intel's forum was a mess and support kinda poor from what I remember. **** ENDING LOGGING AT Tue Aug 08 03:00:00 2017