**** BEGIN LOGGING AT Thu Jan 17 03:00:00 2013 Jan 17 09:11:13 good morning Jan 17 09:23:07 morning all Jan 17 09:29:48 hmm, the systemtap recipe in danny points to a non-existent repo. Jan 17 09:31:08 (git://sources.redhat.com vs git://sourceware.org in master) Jan 17 09:31:42 hmm, channel confusion disorder Jan 17 13:10:08 hello all Jan 17 13:11:24 i'v defined a python function on a bbclass and i want to be able to call this function on other recipes using the syntax ${@my_function()} what should i do to make my_function visible from other recipes ? Jan 17 13:14:30 I belive you can use "inherit class_name" Jan 17 13:15:11 i'v tested but this does not work, is there any export name of something like that ? Jan 17 13:17:13 b737800: iirc your class needs to have a line "EXPORT_FUNCTIONS myfunction" Jan 17 13:20:07 othanks i will try that Jan 17 13:30:42 this does not seems to work Jan 17 13:32:14 i found that function such as getVar setVar are defined in explode_dep_versions2 Jan 17 13:32:28 sorryn in bitbake/lib/bb/utils.py Jan 17 13:34:23 seebs: around? Jan 17 13:38:53 b737800: it depends how you define it. I think you need to use the def xxx(): syntax to be able to call it like that Jan 17 13:39:04 Zagor: EXPORT_FUNCTIONS is not the answer here Jan 17 13:39:20 yes it s exactly what i'm doing right now ;) Jan 17 13:42:08 hi, not sure if I should ask this here or in #poky but here goes: downloaded danny stable and checked out meta-fsl-arm 'danny' branch, but when I try to build something it fails on missing udev 173.bbappend (danny has udev 164), any suggestions? Jan 17 13:48:27 markos_: harass the freescale people as they broke it. moving _173 to _164 will help. Jan 17 13:48:56 rburton, thanks Jan 17 14:01:01 Anyone back-ported bitbake -c populate_sdk from 1.3 to 1.2 and would like to share ? Jan 17 14:32:35 lumpi_: fray might have I think Jan 17 14:42:17 @bluelighnting: thanks, now I need a way to make him wake up ;) Jan 17 14:47:14 fray: I read in some news archive you backported bitbake -c populate_sdk from 1.3 to 1.2 .. any hints or source code on how to do it ? Jan 17 14:50:47 otavio: are the non-X11 amd-gpu-bin-mx51 drivers not available in meta-fsl-arm? any specific reason? Jan 17 14:51:20 morning Jan 17 15:06:19 hi. i created a .bbappend for psplash, containing 'FILESEXTRAPATHS_prepend := "${THISDIR}/files:"' and 'SRC_URI += "file://psplash-colors.h"'. problem is that the file is copied to WORKDIR only, but not to S, which is "${WORKDIR}/git". how can i work around this, e.g. copy it by hand? problem is that THISDIR in do_compile_prepend expands to the directory of the original .bb, but not to the directory of the .bbappend Jan 17 15:07:32 files are always unpacked to workdir. Jan 17 15:07:38 if you want to copy in a task, copy it from workdir Jan 17 15:07:45 cp ${WORKDIR}/foo ${S}/ Jan 17 15:07:46 or whatever Jan 17 15:08:01 but you can influence how the unpack process works, iirc the subdir= parameter on the url will tell it where to unpack the file to Jan 17 15:08:16 if your S is ${WORKDIR}/git, you should be able to set file://psplash-colors.h;subdir=git Jan 17 15:08:30 oh that's a neat option i didn't know about Jan 17 15:09:23 yeah, it's underused Jan 17 15:12:20 fatal^_: well, the libraries are there Jan 17 15:15:20 kergoth: copying from WORKDIR worked, thanks. i'm wondering why i have to do this though, as other files seem to be replaced without manual copying Jan 17 15:15:30 I don't understand what you mean Jan 17 15:15:43 as i already explained, urls in SRC_URI are always unpacked to WORKDIR Jan 17 15:15:45 not S Jan 17 15:18:57 i have a .bbappend for lighttpd e.g.. it simply does FILESEXTRAPATHS_prepend and 'SRC_URI += ... and the original file gets replaced with my version. but for psplash i have to add an extra copy step Jan 17 15:21:06 otavio: ok... trying to wrap my head around all of this.... would like to build without X. Guess I'll need to dig a bit deeper to understand it all... Jan 17 15:21:36 schramae: no, you're misunderstanding how stuff works. First of all, if the file is already in SRC_URI in the main recipe, you don't need to add it again yourself, that'll just result in the file being unpacked twice. all you need is the FILESEXTRAPATHS Jan 17 15:22:14 bitbake just assembles SRC_URI into a single variable. it has no knowledge of which parts came from which appends, so adding af ile a second time there doens't actually tell it it came from that file Jan 17 15:23:38 kergoth: i see, thanks. and so far i only changed cfg files, which always seem to be placed in WORKDIR, thus overwriting the original one without manual interference Jan 17 15:23:54 again, all files in SRC_URI are always unpacked into WORKDIR Jan 17 15:23:59 you seem confused about what unpacking does Jan 17 15:25:14 * kergoth cheers at the "Remove obsolete update-modules" patch series Jan 17 15:26:50 kergoth: seems so, will RTFM again. thanks Jan 17 15:28:09 all unpack does is takes the url from where it lives (DL_DIR if fetched, the existing local path if not) and deploys it (extracting if necessary) into workdir. that's it. Jan 17 15:28:22 anything else beyond that is handled elsewhere in the recipe Jan 17 15:28:25 read the recipe. Jan 17 15:31:29 yes, the other header file is as well copied within the recipe. as said: i confused this because i only changed files already being present in the original recipe so far, but they are as well handled manually Jan 17 15:32:03 if the other header is copied from workdir in a step after unpack, then yours will be as well, because the file in workdir will be yours. Jan 17 15:38:34 hello all Jan 17 15:39:02 does somebody know how to create a user in core-image-base built ? Jan 17 15:39:58 peterdev: on the running image, or at image build time? Jan 17 15:40:45 at the image built time Jan 17 15:41:42 peterdev: if there's a package that the user is related to, you can create it there using the useradd class. see xserver-nodm-init which creates the "xuser" user Jan 17 15:43:04 rburton: thanks very much ! Jan 17 15:46:19 kergoth: i got 2 files, psplash-poky-img.h (already present and handled in the original recipe) and psplash-colors.h (not present in original recipe). so now i only added the new file to the SRC_URI. my next mistake was doing the copying in do_compile_prepend - do_unpack_append seems to be the better choice, as it finally works now. thanks again :) Jan 17 15:48:47 peterdev: there is also an advanced example in meta-skeleton/recipes-skeleton/useradd Jan 17 15:50:43 schramae: That looks really interesting, thanks for the hint! Jan 17 15:50:54 np Jan 17 15:54:50 Another thing... somebody may have a clue about what to try about the following issue. I built a core-image-base image and running it on pandaboard. On startup , none of the drivers getting initialised so I modprobe the wifi related drivers but its complaining that wlcore_sdio.ko do not find the firmware in the file system. Everything is clearly in place, it must be some very obvious reason but I;m out of ideas Jan 17 16:04:07 peterdev: are the firmware files really present/installed in the final image? i had to do 'IMAGE_INSTALL_append = "kernel-firmware*"' to make this work. search for '*firmware*' in build/XXX/tmp/deploy/, there should be a package AFAIK Jan 17 16:07:43 schramae: Thanks for the answer first of all. The thing you saying is totally new to me. I was trying the following so far. Generally my goal is to get the wifi working on pandaboard so I enabled the kernel drivers through bitbake -kernel version- -c menuconfig.. than baked the whole image and Ive got 2 output. the main image .bz2 and the modules which is contains the drivers as modules. The next step was to manually copy the ti-con Jan 17 16:08:16 schramae: to the /lib/firmware/ti-connectivity folder in the fs Jan 17 16:09:11 after all I try to modprobe the driver and its saying , it can not find the firmware/ti-connectivity folder Jan 17 16:09:46 is this a wrong method to go with ? Jan 17 16:10:58 peterdev: the right method is to use a package for the firmware, which probably already exists Jan 17 16:13:07 rburton: my idea came from a tutorial in this topic. I do not know how else I can include a firmware pack which is not part of the meta-ti recipes Jan 17 16:14:22 peterdev: i guess manually copying the files should be ok as well, but i'm no yocto guru either. as said: i created a .bbappend file for the desired image recipe and added the 'IMAGE_INSTALL_append = "kernel-firmware-YOURNAME"' to it. simply enabling the firmware isn't enough, as this means it will get built, but not deployed to the final image/filesystem. search your build dir for kernel-firmware - there might already be a package as rburton said Jan 17 16:14:24 rbuton: I think my main problem is related to the messy firmware installation. At some point the privilages must getting twisted.. that could explain the situation like I'm in at the moment Jan 17 16:15:12 thanks for the help , I'll got for this solution ! Jan 17 16:15:50 np Jan 17 16:17:07 does sb. know if there is a package containing/creating "/etc/acpi/events/"? the acpid recipe/package does not and acpid does not start without it Jan 17 16:17:36 otavio: you around? Have a couple of bugs that I need to get Priority for Jan 17 16:21:54 zeddii, if I start with the defconfig from the evil vendor tree, can I use the config fragments to modify the entries that need changing? Jan 17 16:22:34 I am struggling with how to show the path from a the allegedly known good config, to one that runs with an oe based file system Jan 17 16:22:57 yep. that should work as an approach. Jan 17 16:25:36 anyone using meta-systemd with hddimg? mine is sitting at "waiting for removable media" Jan 17 16:25:40 udevd is actually running which is a good start Jan 17 16:52:24 I assume we have seen this? http://free-electrons.com/blog/experiment-with-yocto/ Jan 17 17:01:16 RP: Around now, saw your improvements, will have a look at the data structures and why they are crazy like that. I think it was mostly "didn't even consider the question, really". Jan 17 17:01:46 Huh. You know, I could probably buy some significant time by making it copy-on-write. Jan 17 17:02:09 In a copy, set a flag that the history isn't locally owned yet, and then IF something causes me to try to log an event, make copies then. Jan 17 17:02:41 Although basically any finalize will requires copies, I think there are a few CreateCopy that don't end up getting followed by changes. Jan 17 17:04:21 seebs: My main question is whether I can separate out the include and variable history structures Jan 17 17:04:42 seebs: I've tried it and have some patches - seems to work ok Jan 17 17:04:57 well, more than ok, it performs betetr Jan 17 17:05:00 I don't remember why those were bundled up. Jan 17 17:05:06 There was doubtless a reason at the time. Jan 17 17:24:29 My guess is that it was something really well-considered, like "this way I am only filling up one slot of namespace". Jan 17 17:29:53 msm: I hope that you can work on the oprofile issue today, as we would like a patch for M3 Jan 17 17:30:37 sgw1: ok, im going to start looking it Jan 17 17:30:43 i had to stop and fix a danny issue Jan 17 17:30:50 which i think is solved Jan 17 17:31:05 msm: thanks Jan 17 17:47:40 So, followup on some stuff I was looking at yesterday involving SKIP_FILEDEPS. Jan 17 17:48:01 Right now, SKIP_FILEDEPS is a normal variable, with overrides per recipe with SKIP_FILEDEPS_pn-package-name. Jan 17 17:48:25 But there's no way to do an override by package name; you can't do something like SKIP_FILEDEPS_-foo-dev. Jan 17 17:49:02 It seems like it'd be pretty easy to add a check for this in the per-package loop, doing something like getVar("SKIP_FILEDEPS_pkg-" + pkg, True). Jan 17 17:49:43 But since I don't see any other idiom for overrides per-package, I don't know whether "_pkg" is the right thing, or whether it'd be something else. FILES_, RDEPENDS_, and so on all just use the package name raw, with no prefix, so that might be preferable just for parallelism. Jan 17 17:50:12 override is _pkgname or pn- Jan 17 17:50:54 Yes. Thing is, the SKIP_FILEDEPS_ doesn't do either per-package. So I think I propose to add a check for that to the per-package filedeps generation. Jan 17 17:56:58 seebs: http://git.yoctoproject.org/cgit.cgi/poky-contrib/log/?h=rpurdie/incvartrack - I've not changed the commit headers but I did hack the patches around a bit Jan 17 17:57:27 seebs: These perform a lot better Jan 17 17:58:41 Cool. I think I am officially Too Busy to work on it this week, but I might sneak some time in on the weekend to look at it and see if I can find any other questionable assumptions I made. Jan 17 17:59:57 seebs: Are you ok with me merging those if I decide they're ready or do you want to review them first? Jan 17 18:00:31 Go ahead, I'd rather have something decent that works acceptably in and refine later. Jan 17 18:01:42 seebs: ok, I've not 100% finished/decided yet but I'll continue to play... Jan 17 18:14:07 RP: any thoughts/comments on the inspection tool I've started on as a background task? See https://github.com/kergoth/bb#readme and https://gist.github.com/4455568 for details. It's a work in progress, but I'm looking for an opinion on the direction and potential inclusion in the bitbake repo in the future Jan 17 18:14:36 hello everybody Jan 17 18:14:57 could somebody give me a hint about how to apply this patch ? I never done something like that, would be a great help http://permalink.gmane.org/gmane.linux.embedded.yocto.meta-ti/937 Jan 17 18:37:13 I've noticed that there are several cases of multiple packages that work to get rid of python dependencies. Is this primarily to remove the dependency from the end device to also to remove the dependency from building? Jan 17 18:40:31 to also/or to also Jan 17 18:43:21 blloyd_: both Jan 17 19:21:17 halstead: ping Jan 17 19:21:26 Hi wmat. Jan 17 19:51:58 sgw1: i have a fix for if perf is disabled… i think i need to do some more testing though after enabing perf which seems to require libpfm Jan 17 19:52:06 sgw1: the short is im still looking at it Jan 17 19:52:19 sgw1: i wanted to ask does someone test oprofile is working? Jan 17 19:53:34 i understand theoretically *we* should do that - lol Jan 17 20:04:59 rburton: are there flags that can be added to conf.local to enable/disable python dependencies? Jan 17 20:09:12 anyone seeing quota build errors on master Jan 17 20:09:13 https://gist.github.com/4559264 Jan 17 20:10:20 sgw1: Hi Jan 17 20:10:46 sgw1: sorry, I was out lookg at a customer board Jan 17 20:11:30 kergoth: Personally, I've been wanting to see a return a bitbake shell so that looks like it has a lot in common Jan 17 20:19:17 blloyd_: only if the recipe is setup to do that, with say PACKAGECONFIG Jan 17 20:21:09 RP: yep Jan 17 20:23:16 kergoth: I've wanted to get the shell back for a while for exactly those kinds of things Jan 17 20:25:29 Hey, silly newbie question: Does yocto/bitbake have strong opinions on umask? Jan 17 20:25:42 Like, does it take into account the possibility of users setting umask to strange values? Jan 17 20:26:13 seebs: I think we force it to something "sane", or at lest sanity check it Jan 17 20:26:34 Context: want to copy things. Want to preserve modes. Don't want to preserve ACLs. Jan 17 20:42:19 zeddii, you around? Jan 17 21:07:25 otavio: back, just wanted to get your priority updates on the 2 new bugs filed Jan 17 21:11:25 sgw1: sure :) Jan 17 21:24:06 is PROVIDES related to package? Jan 17 21:49:41 can i put PREFERRED_PROVIDER inside a recipe? Jan 17 21:51:29 I don't think that'd work, because recipe stuff is processed only when we've already decided that we're buillding that recipe. I think. Jan 17 22:00:18 bitbake checks the global config metadata for that Jan 17 22:13:13 so usually i will add that on the machine conf or distro conf files, right? Jan 17 22:14:17 yes ftonello Jan 17 22:14:47 thanks Jan 17 22:15:11 it affects all packages that might depend on the package, so it's more global than just a single package. Jan 17 22:37:23 how come im using rpm as package manager and still i cant list any packages in the target Jan 17 22:37:34 i try to do sometihng like this: rpm -l x11-common Jan 17 22:37:55 i try to use smartrpm and it shows that no package is installed Jan 17 22:38:03 with 'smart stats' Jan 17 22:38:26 with deb was working 'dpkg -l' Jan 17 22:38:39 is there anything that i need to do to enable rpm packages? Jan 17 23:55:12 hi, how can I config the e1000e driver as kernel module and use "modprobe e1000e" after boot? Jan 17 23:56:16 have add a kernel recipe to set "CONFIG_E1000E=m" but the "e1000e.ko" not found in the target "lib/modules/xxx" Jan 18 00:02:59 eddy-win: check to see if final config has CONFIG_E1000E=m in it or not. I'm thinking that may be set to =y by another fragment. Jan 18 00:03:30 in which case if it is y, you don't need a modprobe. Jan 18 00:04:21 is the "final config" in the tmp/work/xx-poky-linux/linux-yocto-3.4xxx/linux-xx-standard-build/ directory? Jan 18 00:04:59 yes it's a .config file Jan 18 00:05:14 may be in the linux directory nextto that one. Jan 18 00:26:03 I have use "INHERIT += "rm_work"" in my local.conf, this will delete the .config file after build, right? Jan 18 01:13:35 hi, I use "depmod" on my yocto target, but there's errors: Jan 18 01:13:36 root@jasperforest:/# depmod Jan 18 01:13:37 WARNING: could not open /lib/modules/3.4.11-yocto-standard/modules.order: No such file or directory Jan 18 01:13:37 WARNING: could not open /lib/modules/3.4.11-yocto-standard/modules.builtin: No such file or directory Jan 18 01:15:19 and found "modules.builtin.bin modules.symbols.bin" in the /lib/modules/3.4.11-yocto-standard/ Jan 18 01:15:32 should I rename these 2 files? Jan 18 01:17:38 eddy-win, huh, I honestly don't know Jan 18 01:17:51 eddy-win, would you mind opening a bug for this? Jan 18 01:17:58 ok Jan 18 01:17:59 if we install depmod, this should work Jan 18 01:18:37 thanks dvhart Jan 18 01:18:55 eddy-win, yeah, sorry I don't have a good answer for you right off **** ENDING LOGGING AT Fri Jan 18 02:59:59 2013