**** BEGIN LOGGING AT Wed Jun 25 02:59:58 2014 Jun 25 06:31:14 since I have bothered #oe with this the past days, I just want to say that for 720p@60 the qt5 performance is decent on my wandboard :) Jun 25 06:31:43 1080p@60 was just totally outside its comfort zone.. Jun 25 06:31:48 good morning! Jun 25 07:57:38 good job men! Jun 25 08:05:29 morning ! Jun 25 08:18:07 good morning Jun 25 08:18:52 IMAGE_INSTALL_append = does the same as IMAGE_INSTALL += ? Jun 25 08:23:13 no Jun 25 08:23:18 *g* Jun 25 08:23:32 append appends to the variable Jun 25 08:23:47 += appends the string and replaces the variable Jun 25 08:24:24 so thats why you ahve to take care about the whitspaces when using append Jun 25 08:24:33 bye till later Jun 25 08:26:33 woglinde: so different behaviour but identical result Jun 25 08:28:51 https://lists.yoctoproject.org/pipermail/yocto/2012-June/007363.html Jun 25 08:33:46 when using the built distro, the term is unable to display UTF8 i feel like i forget to declare something in the distro... i have LOCALE_UTF8_ONLY = "1" Jun 25 08:33:46 and ENABLE_BINARY_LOCALE_GENERATION ?= "1", DISTRO_FEATURES has libc-locales and libc-locale-code, any idea ? Jun 25 08:34:21 mckoan, i think one also needs to consider that the _append's are applied _after_ the recipe and its dependencies has been parsed, at least *i think* this is the case Jun 25 08:35:07 it all goes hideously wrong when someone does BLAH_append += "" Jun 25 08:35:13 mckoan, so setting foobar_append in a .bbappend would overwrite and wipeout a foobar_append set in the main recipe Jun 25 08:35:23 which for some reason used to be really common Jun 25 08:36:11 XorA, is BLAH_append += "" discouraged ? Jun 25 08:36:43 kroon: it should be used with care, in most cases it was just wrong as what they meant was _append without the += Jun 25 08:36:56 kroon: but it evaluates twice at different times Jun 25 08:41:25 XorA, I guess the thing I'm struggling with in understanding in general, is how and when bitbake evaluates variables ... Jun 25 08:41:40 kroon: aim for the simplest solution Jun 25 08:42:19 * mckoan likes the enlightening comments above Jun 25 08:42:34 thx Jun 25 08:47:25 :) Jun 25 08:51:22 bluelightning, hi. do you happen to know if it is correct to state that "_append"'s are applied _after_ a recipe has been parsed ? Jun 25 08:52:31 kroon: well, the recipe parsing step is really just about populating the cache with values of variables that bitbake needs to be able to look up quickly Jun 25 08:52:44 kroon: variable continues to occur after that Jun 25 08:52:53 er, variable expansion I meant to type Jun 25 08:53:30 references in variable values are expanded upon demand, basically Jun 25 08:54:05 although you can force it immediately by using := Jun 25 08:54:19 but generally on-demand is the right thing for almost all cases Jun 25 08:55:30 when we talk about operations on variable values, those do happen at different times however Jun 25 08:56:08 = += =+ .= and =. will change the value immediately Jun 25 08:56:52 _append, _prepend and _remove will store the value to be appended / prepended / removed and then later each time the value is expanded the operations will be applied Jun 25 08:57:23 ??= is deferred in the same way as well Jun 25 08:57:51 ( ?= is immediate, it either takes effect or not depending on whether the variable has a value at the time the statement is parsed) Jun 25 10:12:11 how do you enable v4l2 in gstreamer? Jun 25 10:13:09 IMAGE_INSTALL_append = " gstreamer gst-meta-base gst-plugins-good gst-plugins-base v4l-utils " Jun 25 10:13:27 I have been googling for hours Jun 25 10:13:44 ERROR: pipeline could not be constructed: no element "v4l2src" Jun 25 10:14:15 mckoan: I think the gst-plugins-good / gst-plugins-base packages are actually empty Jun 25 10:14:30 mckoan: you want gst-plugins-good-meta and gst-plugins-base-meta Jun 25 10:14:40 (there is a bug open for that, it's pretty annoying) Jun 25 10:14:57 is there some base recipy that contains variables for all other recipies? Jun 25 10:15:31 pompomJuice: not a base recipe, but there is a meta/classes/base.bbclass that is inherited implicitly by every recipe Jun 25 10:15:40 perfect Jun 25 10:15:42 thanks Jun 25 10:15:50 bluelightning: thx, bitbaking Jun 25 10:16:07 pompomJuice: but you wouldn't modify that if that's what you're thinking of Jun 25 10:16:34 I presume I can add to it somehow... I want to save a variable somewhere that can be switched on inside the recipies for prod and dev mode Jun 25 10:16:45 ok Jun 25 10:16:47 pompomJuice: all recipes start with the global configuration, that's typically what you change (or, if it's something you want to change only for a group of recipes, create a new class the recipes can inherit) Jun 25 10:17:14 the problem I have is merging between dev and prod refs Jun 25 10:17:19 the dev tree contains dev settings Jun 25 10:17:26 there's DISTRO_TYPE = "debug" for things like that Jun 25 10:17:31 thats what I want Jun 25 10:17:35 ok Jun 25 10:17:37 perfect Jun 25 10:17:40 I called mine IMAGE_MODE Jun 25 10:17:54 I can switch on DISTRO_TYPE in all recipies then Jun 25 10:18:08 thanks Jun 25 10:23:08 is DISTRO_TYPE something we still use? Jun 25 10:23:47 I mean, there's nothing to stop you using it - but there's nothing intrinsic to that variable in OE-Core at least Jun 25 10:24:01 aah Jun 25 10:24:10 it looks used Jun 25 10:24:25 not sure I understand how to use it properly Jun 25 10:24:30 where to overwrite it Jun 25 10:25:16 currently I pulled a trick where I have a /conf/distro/angstrom-v2014.06.conf that is overriding the one in angstrom Jun 25 10:25:30 instead of this include mechanism that I dont have time for Jun 25 10:28:38 unless it as easy as putting .inc files in /conf/distro/include? Jun 25 10:28:51 in my layer Jun 25 10:30:48 bluelightning: AFAYK is it possible to send gstreamer on framebuffer? Jun 25 10:47:17 mckoan: I don't know, sorry Jun 25 10:52:28 yes, at last I suceeded Jun 25 10:53:10 bluelightning: actually I needed only to tacke a shoot Jun 25 11:15:46 IMAGE_MODE = '${@base_conditional("DISTRO_TYPE", "release", "prod", "debug", "dev")}' Jun 25 11:15:46 AC_BUILD_STORER_STREAM = "${IMAGE_MODE}_storer" Jun 25 11:15:46 AC_BUILD_DOTNET_STREAM="${IMAGE_MODE}_dotnetfinal" Jun 25 11:15:46 AC_BUILD_MISC_STREAM="${IMAGE_MODE}_misc" Jun 25 11:16:48 , d Jun 25 11:17:39 nm Jun 25 12:54:53 bluelightning ping Jun 25 12:55:00 woglinde: pong Jun 25 12:55:16 whois the dora maintainer? Jun 25 12:55:25 of OE-Core or meta-oe? Jun 25 12:55:56 oe-core Jun 25 12:56:15 Robert Yang Jun 25 12:56:29 FYI: https://wiki.yoctoproject.org/wiki/Stable_branch_maintenance Jun 25 12:57:27 ah okay Jun 25 12:57:44 thanks Jun 25 12:58:08 will give him some time for the perf stuff Jun 25 13:05:46 i built lighttpd from oe, and when i started it with debug.log-request-handling = "enable", i noticed that the mesages contain path to sources because they use __FILE__,__LINE__. But as it has been built with oe, the path it wide ! So i am wondering if it would be a good idea to use -iremap src:dst in conjunction with --sysroot=path in order to hide the path, to make the output of program using __FILE__ more human readable, and to save some more bytes in binari Jun 25 13:18:27 afournier hm what? Jun 25 13:18:36 :) Jun 25 13:20:36 lighttpd uses __FILE__ in its source code, as sources are built somewhere on the host, __FILE__ contains a very long path + path to lighttpd source code Jun 25 13:20:59 it's not a problem for lighttpd only, but i use it as an example Jun 25 13:22:55 woglinde: do you get my point ? :) Jun 25 13:23:24 hm yes Jun 25 13:23:52 but better write to oe-core Jun 25 13:24:25 ok Jun 25 13:55:41 when i try out the aotutool example on http://www.yoctoproject.org/docs/1.6/dev-manual/dev-manual.htm i get "No rule to make target install" is the example incomplete? Jun 25 14:03:42 rick___: it depends - if your Makefile does not provide an "install" target, you'll need to write your own do_install Jun 25 14:04:08 the example probably assumes that the Makefile does provide that target Jun 25 14:30:52 in what case do i use -b and not? Jun 25 14:31:02 bitbake -b Jun 25 15:08:25 rick___: generally, you should not use -b at all Jun 25 15:08:40 I rarely use it Jun 25 15:09:49 bluelightning: a simple helloworld example (helloworld.c locallystored with no autotools) builds only with this option.. Jun 25 15:15:59 rick___: that's probably because without -b, bitbake expects to find recipes via BBFILES Jun 25 15:16:53 (and each layer you add to conf/bblayers.conf will extend BBFILES in its own conf/layer.conf to search directories within the layer for recipes) Jun 25 16:38:58 How do I set the location of the bitbake logfile? I'm getting "ERROR: Function 'staging_packager' failed (see /dev/null for further information)" Jun 25 16:39:21 afaik thats just a bug in old versions of bitbake Jun 25 16:39:25 really old versions of bitbake Jun 25 16:39:29 * kergoth shrugs Jun 25 16:39:44 what versions are you building? Jun 25 16:49:29 kergoth: BitBake Build Tool Core version 1.12.0, bitbake version 1.12.0 Jun 25 16:50:07 what yocto/poky/oe-core version? Jun 25 16:50:20 kergoth: oe-classic Jun 25 16:50:42 there's no way to fix that message short of cherry picking the fixes from a future branch into your bitbake Jun 25 16:50:51 and crossing your fingers :) Jun 25 16:51:53 kergoth: slightly confused by your response. I want to see the "logfile" that it thinks its writing to "/dev/null" There's no way to set the location of the logfile? Jun 25 16:52:07 its not a configurable path Jun 25 16:52:09 its just a bug Jun 25 16:52:16 it was always intended to write to a real logfile Jun 25 16:52:56 so the logfile really exists somewhere, I just need to go find it. Jun 25 16:53:25 no Jun 25 16:53:27 well, maybe Jun 25 16:54:11 https://github.com/openembedded/bitbake/commit/90fc09ee050a24095a488d8ce98c7945bf282f6d Jun 25 16:54:36 afaik thats what fixed it Jun 25 16:54:46 from some quick searching of the git history with git log —grep=/dev/null Jun 25 16:54:53 kergoth: thanks, that looks super useful. Jun 25 16:55:41 np Jun 25 16:55:59 also, you may wish to move to OE-Core when convenient ;) Jun 25 16:56:13 indeed, thats pretty ancient Jun 25 16:57:37 wish I could. This project is in maintenance mode, and there's lots of stuff I'm not allowed to do. Jun 25 17:02:33 it does look like that commit will cherry pick without conflicts onto 1.12.0, so hopefully.. Jun 25 17:09:48 awozniak: understood Jun 25 17:31:45 kergoth bluelightning: If I switched from bitbake 1.12 to bitbake 1.22, do you foresee any problems with an oe-classic build? Jun 25 17:54:26 awozniak: afaik there are at least some things that were deprecated and then removed along the line Jun 25 17:54:31 so there'd almost certainly be stuff to fix Jun 25 17:55:19 kergoth: Thank you Jun 25 17:56:29 np Jun 25 18:55:15 JaMa, have you seen any more ice issues? Jun 25 19:02:42 denix, have you used wic? Jun 25 19:17:17 Crofton: not today IIRC Jun 25 19:17:29 Crofton: wouldn't zeromq better fit in meta-networking? Jun 25 19:21:55 I guess I'm thinking it is pretty niche oriented Jun 25 19:22:42 and not focused on small networking devices Jun 25 19:23:24 does Joe Macdonald hang out on irc? Jun 25 19:26:31 It has seemed to me that meta-networking is not a catch all for everything that uses a network :) Jun 25 19:27:02 Crofton: right, it should not be that Jun 25 19:27:23 pretty much everything works over a network in some way these days :) Jun 25 19:28:02 lets move everything to meta-networking! Jun 25 19:28:25 do you know if wic has been used to make SD cards for things like the beagle stuff? Jun 25 19:29:19 there's nothing particularly special about the beagle SD card layout, so I can't see why it wouldn't work Jun 25 19:29:35 I'm dumb, I need examples to work from :) Jun 25 19:29:40 and pizza arrived Jun 25 19:29:45 mmm pizza Jun 25 19:30:09 btw, are congratulations in order? Jun 25 19:31:30 you could use the galileo SD card wks file from the IoT devkit as an example Jun 25 19:31:41 I guess so :) Jun 25 19:32:55 example: http://git.yoctoproject.org/cgit/cgit.cgi/meta-intel-iot-devkit/tree/scripts/lib/image/canned-wks/iot-devkit.wks Jun 25 19:33:34 (the comment in the second part is out of sync with the partition definition line) Jun 25 19:42:01 Internet of Toilets? Jun 25 19:42:56 and yes, this is a thing: https://twitter.com/IoToilets Jun 25 19:45:11 if you can imagine it, someone on the internet has probably done it :) Jun 25 19:46:19 that has variable flush amounts Jun 25 19:46:35 I'm thinking on the fixed amount USian ones, you could use this to detect leaks Jun 25 20:18:17 this wic thing is clear as mud Jun 25 20:19:47 Is it possible to whitelist env-variables on a per-recipe basis ? Jun 25 20:28:27 hello Jun 25 20:28:30 PACKAGECONFIG_remove = "${@bb.utils.contains('DISTRO_FEATURES', 'x11', '', 'eglgles', d)}" Jun 25 20:28:34 is this correct? Jun 25 20:28:37 it isnt working Jun 25 20:28:56 I want to remove "eglgles" from the packageconfig if x11 isnt being used Jun 25 20:42:34 is there a way to query bitbake for which package is pulling in a specific dependency? Jun 25 21:05:40 staylor, bitbake -g -u depexp $targetimage Jun 25 21:05:54 staylor, find the package you are curious about, and see the box labeled RDEPENDS Jun 25 21:58:44 it looks like koen has the only solution for sd card generation that works with loop file systems? Jun 25 22:00:14 why would you want to use loop file systems if you don't have to? They generally cause a dependency upon running as root, or at least ensuring your user has the necessary access to loopback devices, which imposes a host setup dependnecy Jun 25 22:00:42 because I do not know what device the sd card reader is Jun 25 22:00:50 and it may not be attached to the build machine Jun 25 22:01:22 there are better ways to create a disk image than to use loop devices Jun 25 22:01:36 e.g. iirc parted can perform operations against a disk image, as can sfdisk Jun 25 22:01:40 see some of the alternative sd card bbclasses Jun 25 22:01:54 yeah, the problem is there are so many of them Jun 25 22:02:07 agreed Jun 25 22:02:17 wic seems like it's best positioned to replace them, but there aren't all that many examples of usage yet Jun 25 22:02:24 yep Jun 25 22:02:38 though the example bluelightning just linked looks extremely promising Jun 25 22:02:58 but I could n't work out wtf was really going on Jun 25 22:03:05 it's not that complex, just a language for describing the partitioning scheme Jun 25 22:03:14 it was just a piece and can't work out how it was tied into the rest of the problem Jun 25 22:03:26 partitions are only one piece of the problem Jun 25 22:03:29 bitbake produces a filesystem image, wic turns that into a disk image Jun 25 22:03:35 i don't really see a misisng piece there Jun 25 22:03:41 * kergoth shrugs Jun 25 22:04:12 maybe I'm missing something Jun 25 22:04:24 like he said the docs are seriosly lacking Jun 25 22:04:57 it's a standalone tool that takes a disk image and description file to do the partitioning as well as splitting up the filesystem into the multiple partitions (e.g. splitting out what's currently in /boot onto th emsdos /boot partition) Jun 25 22:08:33 hmmm Jun 25 22:08:53 the —ondisk bit is slightly concerning, i can see what you're worried about there, it seems to want to operate directly, which wouldn't be ideal Jun 25 22:11:45 yeah Jun 25 22:12:47 it seems like there are ways of setting up stuff in fstab to work around the loop root issue Jun 25 22:12:52 https://github.com/canatella/bitbake-el/blob/master/bitbake.el#L573-L624 may be of interest Jun 25 22:13:15 if somethign writes into random places, I'd like them to be safe Jun 25 22:13:17 emacs function which wraps bitbake image + wic create + dd, might serve as an example of usage, if not wks contents Jun 25 22:14:12 emacs mode for bitbake Jun 25 22:14:15 madness Jun 25 22:15:20 https://www.yoctoproject.org/docs/1.5.2/dev-manual/dev-manual.html#creating-partitioned-images Jun 25 22:16:58 5.6.7. OpenEmbedded Kickstart (.wks) Reference seems of use, though it could be more fleshed out Jun 25 22:17:10 yeah Jun 25 22:17:21 I need to lower my head and run into the pile Jun 25 22:17:32 reseraching stuff is just frustrating me Jun 25 22:17:55 doesn't seem like it'd be too bad to get going with this, i'll have to play with it for my minnow, omap5-evm, or beagle Jun 25 22:18:02 wic is a crappy name the google spacce is overloaded Jun 25 22:18:15 http://www.fns.usda.gov/wic/women-infants-and-children-wic Jun 25 22:18:29 needless to say, lots of web pages about wic Jun 25 22:18:32 the explanation for the name is even worse Jun 25 22:19:31 beat me, OpenEmbedded Image Creator (oeic) is sort of unheard Jun 25 22:19:33 agreed Jun 25 22:19:38 i never heard of oeic before Jun 25 22:19:42 seems pointless to retain that history Jun 25 22:19:43 me niether Jun 25 22:20:11 i'm no good at names, i tend to just describe what hte thing does verbatim :) Jun 25 22:25:14 this is x86 only stuff ? Jun 25 22:25:35 boot type seems to be 'efi' or 'pcbios' (syslinux) Jun 25 22:26:09 the bootloader command seems to be, but the partitioning stuff seems generic, afaict Jun 25 22:26:24 * kergoth shrugs Jun 25 22:26:56 ant__ those are just the implemented templates Jun 25 22:27:08 ant__ others can be added for other layouts, bootloaders, etc. Jun 25 22:27:36 I see Jun 25 22:29:02 still does need better docs, but whats in the dev manual is a good starting point, it seems Jun 25 22:29:29 definitely an area we would all like to see improved Jun 25 22:29:42 directdisk seems to be the thing Jun 25 22:29:44 deployment is quite challenging with the "can't be root" restriction Jun 25 22:29:58 directdisk.bbclass? heh... that's a disaster Jun 25 22:30:23 well, we can't use mkefidisk Jun 25 22:30:42 mkefidisk fulfills a very specific purpose Jun 25 22:30:46 what are you looking to do? Jun 25 22:30:50 i'd create a custom wks for your needs, but the existing ones work as examples Jun 25 22:31:15 That would be my suggestion as well Jun 25 22:31:15 make images I can dd to an sd card Jun 25 22:31:34 for which ARCH/bootloader? Jun 25 22:31:42 arm u-boot Jun 25 22:31:45 right Jun 25 22:31:54 So you have two choices really Jun 25 22:32:05 fish around the existing sd card classses Jun 25 22:32:08 the wks or write your own version of mkefidisk Jun 25 22:32:11 I don't recommend the latter Jun 25 22:32:32 the problem with anything that is a bbclass is it has to work around the no root restriction Jun 25 22:32:39 yep Jun 25 22:32:45 which means a lot of error prone code Jun 25 22:33:12 by working outside of bitbake, things like wic can use the well tested established tooling to do that Jun 25 22:33:28 anyway, that's a major win in my book Jun 25 22:33:32 i've been hoping for image creation to move out of bitabke for many years, so i'm very glad to see this Jun 25 22:33:33 but I'm lazy ;-) Jun 25 22:33:45 now i just want to see rootfs creation from package feeds out too, but not everyone might agree :) Jun 25 22:33:57 kergoth, what would you do instead? Jun 25 22:34:06 kergoth, your wish list is a mile long :) Jun 25 22:34:12 oh, same thing, just not with bitbake? Jun 25 22:34:44 yeah, just conceptually different, not functionally, now that we've gained the ability to build images from existing feeds Jun 25 22:34:54 things like narcissus made images from feeds for ages :) Jun 25 22:35:41 I'm tired of copying bits and pieces onto an sd card Jun 25 22:36:57 * kergoth nods Jun 25 22:38:51 ok, I need a break from this Jun 25 22:39:09 now to look into openbts state of the disunion Jun 25 22:51:49 how do you avoid the --ondisk bit Jun 25 22:51:56 or will that take a file name? Jun 25 22:52:32 i'm not clear on what exactly that does, given it still ends up emitting a disk image file anyway, not writing to that device, unless it's used by the bootloader setup Jun 25 22:52:36 * kergoth shrugs Jun 25 22:52:41 hopefully someone else knows better Jun 25 23:00:23 I need to figure out where the wks files go Jun 25 23:00:46 it seems like people are copying the wic files into their layer Jun 25 23:02:21 this seems like a bad idea Jun 25 23:02:47 that does seem unfortunate, it should either be able to accept a full path to the wks, or should be able to search alternate paths via BBPATH Jun 25 23:02:54 yep Jun 25 23:03:12 also need to grok --ondisk Jun 25 23:05:30 on the surface it doesnt' make much sense other than potentially to configure the bootloader, could examine the wic code and find out for sure, i suppose Jun 25 23:11:27 maybe ondisk refers to the partitions on the target device? Jun 25 23:12:01 this could be a breakthrough! Jun 25 23:12:24 that does seem likely, as it doesnt seem like it should touch host devices anyway Jun 25 23:12:31 (one would hope, anyway) Jun 25 23:12:47 yeah Jun 25 23:13:01 and I think I can give it a path to a wkx file Jun 25 23:14:44 ah, nice Jun 25 23:37:35 I suppose I report wic bugs via yp bugzilla? **** ENDING LOGGING AT Thu Jun 26 02:59:58 2014