**** BEGIN LOGGING AT Mon May 18 02:59:59 2015 May 18 08:01:23 morning all May 18 08:14:02 morning May 18 08:27:17 good morning bluelightning, AlexVaduva, all May 18 08:33:49 morning May 18 08:47:31 the best thing about Hob being put out to pasture is that I will no longer have to read people type it as HOB May 18 08:53:13 joshuagl: word May 18 08:54:00 :)) May 18 10:24:18 joshuagl: I completely missed the Build Appliance news May 18 10:24:57 does exist any useful link showing how does it work without necessarily install it? May 18 10:28:24 mckoan: https://www.yoctoproject.org/tools-resources/projects/build-appliance ? May 18 10:28:33 not sure how useful it is, but that's *a* link May 18 10:28:39 there are docs and everything May 18 11:20:53 Hey May 18 11:21:12 What should I do when I cant build an image cause of this error? May 18 11:21:27 ERROR: Nothing PROVIDES 'syslinux' May 18 11:21:29 can two recipes have the same name? I need to include two layers, both have a recipe with the same name, will this work? May 18 11:35:29 abdurr: yes May 18 11:37:59 rburton: Do you know what I should do with my ERROR maybe? May 18 11:39:25 cart_man: find out why it can't find the syslinux recipe i guess. what machine are you using? May 18 11:43:24 Im using Ubuntu x64 and im trying to compile for beaglebone May 18 11:43:32 Ubuntu 14 May 18 11:43:38 rburton: ^ May 18 11:44:53 cart_man: the syslinux recipe will only build for x86… May 18 11:45:33 so either stop trying to use syslinux if you're explicitly trying, or don't build an image that needs syslinux. are you trying to build a live image for beaglebone? May 18 11:47:12 the images that need syslinux shouldn't be attempting to use it on non-x86 May 18 11:47:20 Well a live image would be nice and helpfull although I could not even get the Beaglebone to boot a normal SD card version. So no im not trying to build a live version May 18 11:47:30 I used HOB May 18 11:47:37 after using HOB it seems to have this problem\ May 18 11:48:13 Is there live bootable version of Yocto for beaglebone? May 18 11:48:29 personally, i recommend not using hob - it does odd things. it probably enabled live images for you. May 18 11:49:52 rburton: Im starting to think the same thing...Problem though is I am not really familier with getting my custom package totally right. theres always something wrong and it crashes... No ive kind of given up on the idea. I first thought that adding packaages is as easy as adding them to the list on the local.conf file May 18 11:49:56 well, an image with the "live" type won't work with non-x86 May 18 11:50:05 but it kept on crashing when I build it May 18 11:50:20 it is that easy - CORE_IMAGE_EXTRA_INSTALL_append = " my-recipe" May 18 11:50:33 for the "quickly add a recipe" situation May 18 11:51:29 So one literrally just add your packages to that list? May 18 11:51:33 whow ok May 18 11:52:13 http://www.yoctoproject.org/docs/1.8/dev-manual/dev-manual.html#usingpoky-extend-customimage May 18 11:54:03 rburton: and what about the " EXTRA_IMAGE_FEATURES " in the local.conf ? because the mostly the things I want on the Image May 18 11:54:21 Ohh nevermind I guess you just add them there May 18 11:54:54 features != recipes May 18 11:55:00 features map to package groups May 18 11:56:02 out of interest, is hob gaining additional traction (again)? i always thought it was going to be superseded by toaster. May 18 12:06:09 LetoThe2nd: So should one rather use Toaster? May 18 12:07:53 cart_man: at least at this very moment, one should probably not listen to me specifically in this respect - i'm all in for the cli anyways May 18 12:11:10 I don't know which is the more usable option right now, but certainly Toaster is receiving the bulk of engineering effort May 18 12:12:33 I hear the LF training course is still teaching hob May 18 12:12:35 weird May 18 12:13:21 well, i personally am using some script magic to setup layers and config, and build on jenkins - so i'm probably not the blueprint May 18 12:14:35 * Crofton|work uses a script to build images and run wic May 18 12:18:36 for reference, this is what i'm using: https://github.com/LetoThe2nd/blubber May 18 12:39:26 rburton: In which file do I switch off the "Live" function? May 18 12:59:48 cart_man: it would have been a modification to your local.conf May 18 13:00:03 rburton: Ok thanks allot May 18 13:05:03 hi.. i try to build qt5-base. i got wayland and opengles in my sysroot. but qtbase is configured for -opengl desktop May 18 13:05:13 instead of -opengl es2 May 18 13:06:11 if i modify to use es2 it works, but why it does not by itself? May 18 13:06:34 i use meta-qt5 May 18 13:16:44 Crofton|work: we should probably reach out to LF and ask them why … May 18 13:17:44 yeah May 18 13:17:59 I found out from a guy who had the class last week May 18 13:18:03 need to double check May 18 13:20:28 Crofton|work: I've told jeffro about it. Let's see what he says May 18 13:30:49 H, May 18 13:32:16 Hm, I don't seem to be able to over-ride /etc/default/ssh via a bbappend as image.bbclass redirects into it in the read_only_rootfs_hook() .... :-/ I guess that means I have to modify image.bbclass ? May 18 13:35:53 pev: are you using read-only-rootfs in IMAGE_FEATURES? May 18 13:36:16 Yep May 18 13:37:44 bluelightning: It looks like it's half been thought about..! I need to move the key storage into the r/w mount which is easy to do via updating sshd_config / sshd_config_readonly but I can't also update the defaults to match... May 18 13:38:11 I suppose I could always mount-bind over the top instead but it's kludgey May 18 13:38:40 I would have thought generating the post-process hook in the openssh recipe would have been more appropriate? May 18 13:39:03 pev: it is probably an oversight... you should probably email the list about it and CC the person who worked on that (Chen Qi ) May 18 13:42:42 bluelightning: Yep, for sure - although I try to not get too involved in the list as I'm flat out and my builds are based on daisy which makes it trickier to submit changes... May 18 13:43:47 or can I submit against daisy and let them find their own way upstream? May 18 13:45:17 I've got a few things stacked up I should pass on I guess and quite a few read-only rootfs tweaks May 18 13:46:16 pev: we do prefer things to be against master unfortunately May 18 13:47:11 pev: the other alternative is you send them to the list clearly marked as being WIP and needing adaptation for master, and CC Qi asking him to take care of them May 18 13:47:32 I was just about to suggest pushing a branch and asking whether anyone wants to own upstreaming them to master May 18 13:47:34 at minimum it would be good to let Qi know of the issues May 18 13:47:49 joshuagl: right, that would be an idea May 18 14:17:20 hi there. is there a way to populate headers from "non-standard" paths (/opt/foobar/include) into sysroot? May 18 14:18:43 isn't it up to the buildsystem of the software in question? May 18 14:19:13 or at worst, your own installation rules into the sysroot? May 18 14:19:40 (i.e. do_install() { .. }) May 18 14:20:08 lpapp: would you install into $some_sysroot_var there? May 18 14:20:47 yes, there are special variables. May 18 14:21:10 i tried installing into ${D}${includedir}/... which works but that duplicates the files in my manually defined -dev package which then holds opt/ and usr/include with the same files May 18 14:21:55 ${sysconfdir}, ${bindir}, ${libdir}, etc. May 18 14:23:08 seezer: list the files in your package May 18 14:23:19 seezer: or use mv instead of copy :-) May 18 14:24:02 is it some proprietary software coming with preinstalled headers like that or why can you not specify it for the buildsystem, by the way? May 18 14:25:20 actually i could. for the sake of "simplicity" for others depending on the framework, some argued putting all into /opt would be "easier". May 18 14:26:01 ok, that is not too linux-y, but it is possible. May 18 14:26:26 yeah i know - it's currently mostly a POC, porting the framework to arm May 18 14:26:30 it's a cmake based receipt and therefore i use a CMAKE_INSTALL_PREFIX of /opt/foo May 18 14:27:04 which fits for everything but the -dev package in sysroot. but i guess i'll just move it and be done with it. simple = sexy. May 18 14:27:24 hello, anyone has accomplished building gutenprint for yocto? May 18 14:27:38 seezer: sorry, I do not follow May 18 14:27:42 i'm trying to add ppd files to cups May 18 14:27:57 seezer: why cannot you use the right path to cmake if it already allows customisation? May 18 14:28:18 and your colleagues will use this /opt/foo for their own experiment. May 18 14:28:27 do you need to provide the headers at both places? If so, why not symlink? May 18 14:29:49 lpapp: yeah moving/linking are what i'm gonna do. i was just thinking i'm missing some knobs to "populate /opt". but that will work for me. thanks! May 18 14:41:38 When I enable -fsanitize=address for some programs in my yocto build I see this error https://gist.github.com/843a0e86725d6a6f53cc . The same program appears to work fine with libsanitizer when built for my local system May 18 14:41:55 Anyone else run into that failure? Or have debugging suggestions? May 18 14:44:15 limotec-dev: perhaps you can make your own .bbappend? May 18 14:45:17 lpapp: i'm not that experienced in writing recipes my own May 18 14:45:46 lpapp: I found a recipe for gutenprint, but it's 5years old May 18 14:46:34 limotec-dev: hmm, that sounds like a good starting point then to customize, no? May 18 14:47:42 lpapp: yes, thats true May 18 14:48:00 lpapp: but I don't have any clue what cups needs May 18 14:48:32 where did you find the recipe? May 18 14:48:52 https://github.com/openembedded/openembedded/tree/master/recipes/gutenprint May 18 14:50:40 limotec-dev: so it no longer works? May 18 14:50:59 lpapp: not for me, i'm getting quite a lot errors May 18 14:51:35 lpapp: like do_stage is deprecated May 18 14:52:14 it is possible that it could be migration to meta-openembedded with some fiddling. May 18 14:52:19 migrated* May 18 14:53:06 limotec-dev: if you do want to migrate that, you may find this useful: http://www.openembedded.org/wiki/Migrating_metadata_to_OE-Core May 18 14:54:00 i'll look into that May 18 14:54:12 thanks bluelightning May 18 14:54:16 np May 18 15:06:59 morning May 18 15:10:14 kergoth: you once gave my some idea to find out which packages are missing from: "Multiple .bb files are due to be built which each provide" .. "This usually means one provides something the other doesn't and should" May 18 15:10:37 what was that? May 18 15:10:39 sry May 18 15:10:58 generally you should inspect both PACKAGES and PROVIDES for those cases, as it can be either a build time or runtime dependency pulling them in. bitbake -g may be of use here May 18 15:11:15 ok May 18 15:13:21 kergoth: PACKAGES what i am missed.. thanks! May 18 15:16:07 np May 18 15:43:30 any idea of how can i define a PREFERRED_PROVIDER in the recipe of a package (.bb) (for any machine) ? May 18 15:44:37 clopez: you can't; PREFERRED_PROVIDER must be set at the configuration level May 18 15:50:20 :\ May 18 15:53:14 clopez: you can depend explicitly on the preferred provider that you would have selected, if that helps May 18 15:58:01 good idea, thanks :) May 18 16:10:25 kergoth: hmm.. that was not all.. how to remove mesa? all i need is provided by pre-built binaries that i copy. but i get the multiple provide errors. last time i managed to get all the packages that i should provide listed on the console.. but i can not remember.. May 18 16:14:39 bluelightning: maybe you got some idea also? Pls see my recipe here: http://paste.ubuntu.com/11208750/ May 18 16:45:37 Hi, I'm having a problem trying to get initramfs/initrd to work. I'm never able to get yocto to generate an initramfs image (it always hangs on what appears to be an internal script), and if I try to load the rootfs separately through initrd, the kernel just hangs. It never boots. I have little idea of these things (I've been mostly following guides and what I can find in the Internet), so I can't even provide more information about this May 18 16:45:49 I'm trying to load everything through u-boot, if that helps. May 18 16:51:50 This is the error that appears when trying to bundle everything for the initramfs: http://pastebin.com/J5fLms4x May 18 18:39:51 Hello Friends. I have a bitbake recipe with RDEPENDS+${PN} and name of few packages. When I install the ipk generated with this recipe, I get this error --> "satisfy_dependencies_for: Cannot satisfy the following dependencies for" May 18 18:40:27 and list of packages listed in RDEPENDS_${PN} May 18 18:40:49 does this mean I have to install all the packages listed in RDEPENDS_${PN} May 18 18:40:54 before installing this package: May 18 18:40:55 ? May 18 18:41:36 Is there a way to install all dependencies (dependent packages) as part of single installation? May 18 18:42:27 how exactly did you 'install the ipk generated with this recipe'? May 18 18:42:40 if you want to use opkg to install something and its dependencies, you need a package feed May 18 18:42:45 see the opkg documentation May 18 18:43:20 opkg install package.ipk May 18 18:48:02 thanks for the direction May 18 18:48:20 Is there a way to pack all these dependencies in one package? May 18 18:48:43 so as part of one ipk, install the actual package + all its dependencies May 18 18:50:35 noe May 18 18:50:36 nope May 18 18:50:49 either use a feed, or manually copy them and install them May 18 19:10:47 hmm, devtool assumes SHELL is absolute when set, that's not guaranteed May 18 19:13:55 I would tend to consider a non-absolute $SHELL to be a Very Bad Idea. May 18 19:14:05 I will take a stand against shellativism. May 18 19:15:04 it seems some versions of zsh set SHELL=zsh May 18 19:15:37 fish doesn't set it at all May 18 19:15:53 hmm May 18 19:16:15 Huh. That seems like a strange choice. I would not expect a shell to *set* $SHELL, which I always thought was supposed to be set to the shell field from passwd. May 18 19:17:28 ah, no, it wasn't zsh setting it, was a lurking profile remnant. eep, that's no good May 18 19:18:31 case $0 in May 18 19:18:33 [...] May 18 19:18:37 interesting, "This variable shall represent a pathname of the user's preferred command language interpreter." .. doesn't indicate that it has to be absolute, however May 18 19:18:41 *csh) SHELL=/bin/csh May 18 19:18:41 # This should never happen; little do we know. May 18 19:18:43 echo "What the \$SHELL?" >&2 May 18 19:18:45 as opposed to PWD, which is "This variable shall represent an absolute pathname of the current working directory." May 18 19:18:50 I don't think that code ever executed. May 18 19:19:02 Huh. I ... hmm. May 18 19:19:15 Yeah, that's odd. Because "a pathname" is almost useless. May 18 19:19:35 there's no guarantee PATH is what it was when the shell was spawned.. May 18 19:19:39 Yeah. May 18 19:19:40 so having to search to find it could find the wrong thing May 18 19:19:43 weird May 18 19:19:55 Or $PWD, for a path with at least one slash but no leading slash. May 18 19:20:00 Or consider insanity like May 18 19:20:00 i'm guessing most things make it absolute an yway, but it's strange that the standard doesn't see mto require it May 18 19:20:31 ln -s /bin/sh /tmp/dummy; export SHELL=/tmp/dummy; rm /tmp/dummy May 18 19:20:51 There you go. SHELL was sent to a value representing a pathname of the preferred command-line interpreter. May 18 19:21:04 Heck, if I want to be dogmatic, I can point out that it doesn't say "a pathname on this host". May 18 19:21:21 "/usr/xpg4/bin/sh", which is "a" pathname for the desired command interpreter. On Solaris. I think. May 18 19:21:22 huh ,yeah, it doesn't say it has to exist or anything May 18 19:21:28 basically its value is useless May 18 19:21:40 Be careful when you stare into the POSIX spec, for the POSIX spec also stares into you. May 18 19:21:45 hahaha May 18 19:21:46 true May 18 19:22:13 Standards interpretation is very vulnerable to overthinking. May 18 19:22:18 i was hoping/expecting it to be sane in this context... that was silly of me May 18 19:22:39 it's kind of scary what *isn't* defined/explicit in there May 18 19:22:41 I play D&D-like-games, and I note that they have the exact same problem. May 18 19:22:45 and we just assume May 18 19:23:04 And indeed, one of the major recentish releases was built around *dropping* a ton of specification and saying "look, the reason we have a person run the game is so they can make reasonable judgment calls". May 18 19:23:18 Because the attempts at formal specifications just made even weirder debates about what was specified. May 18 19:23:49 hah, yeah, that makes sense. there'd be creative ways of working around the ambiguities anyway, might as well give them full lattitude May 18 19:24:55 There were these insanely complicated arguments that would arise out of combinations of bits and pieces of wording. May 18 19:25:16 There was one I found where it is *completely obvious* what the rule means. May 18 19:25:31 And by that I mean, every gamer I know thinks that it is totally obvious. May 18 19:25:36 And it's about a 50-50 split on which of two answers they think is "totally obvious". May 18 19:25:44 nice. May 18 19:25:45 And that ambiguity has been in the rule system for probably 30 years. May 18 19:26:16 And in all that time, it's never really been made completely clear, but everyone reads it a given way and finds it consistent. May 18 19:29:43 hmm, I wonder if devtool should get some functionality added to make it easier to override bits coming from outside the source tree, e.g. defconfig or other file:// bits **** ENDING LOGGING AT Tue May 19 02:59:58 2015