**** BEGIN LOGGING AT Tue Nov 20 02:59:57 2012 Nov 20 08:36:49 otavio: would be nice to keep all qt patches we have also in git repo for easier upgrade to next version https://github.com/meta-qt5/qtbase/tree/oe-core-5 Nov 20 08:37:14 otavio: you did some changes to zecke patch, so with rebase I meant to update it in qtbase repo Nov 20 09:13:11 morning all Nov 20 09:13:18 mornin Nov 20 09:14:49 morning all, hi bluelightning hi afournier Nov 20 09:14:57 hi Nov 20 09:26:11 Someone know about trouble at startup scripts that doesnt be executed with angstrom? Nov 20 10:26:20 good morning Nov 20 10:30:44 morning Nov 20 11:07:59 I had a short chat with GNUtoo about it yesterday, just curious, is anyone doing any work or intending to, to build stuff for android with OE core? Nov 20 11:08:18 he was saying there was a fork that someone did for classic, but he did not know anything about core Nov 20 11:08:37 and what caveats would you expect, except messing with bioniclib? Nov 20 11:13:38 HI all, I'm trying to add a device driver to the kernel and recompile it. I can do this to add a .bb file to recipes kernel with the right patch right? Nov 20 11:17:27 koennie_: patch to kernel? Nov 20 11:18:11 koennie_: I think you can just add your own .bbappend and point to your own defconfig Nov 20 11:18:27 ...unless you it is some driver that is not in the kernel, then you'd probably need a patch Nov 20 11:20:20 koennie_: if you want to add the driver in kernel using patch then again you can use .bbappend to add that patch to kernel instead of creating separate recipe file Nov 20 11:20:35 morning all Nov 20 11:20:52 udev 174 changelog says:"The udev daemon moved to /lib/udev/udevd. Non-systemd init systems Nov 20 11:20:52 and non-dracut initramfs image generators need to change the init Nov 20 11:20:52 scripts. Alternatively the udev build needs to move udevd back to Nov 20 11:20:52 /sbin or create a symlink in /sbin, which is not done by default." Nov 20 11:21:12 With above update default udev init script is looking into /lib/udev/ to get udevd but for mips64 arch it is present in /lib64/udev. I think we need to fix it. Correct? Nov 20 11:32:09 Jin^eLD_: the defconfig is automaticaly generated so i should't edit it? Nov 20 11:32:33 not automatically - it contains *your* settings Nov 20 11:32:42 i.e. drivers you want to have, etc Nov 20 11:39:00 coowl, thanks Jin^eLD Nov 20 11:40:03 koennie_: I'd pick the defconfig that is used to build the kernel that you are trying to modify, and then use this defconfig as your basis to enable whatever that is you want to enable there Nov 20 11:50:43 Hi all! I wrote a recipe that fetches a git project from my local server. It works, but after i make a new commit, the sources in bitbake are not updated bitbake -c clean works but -c fetch takes always old sources. Is this the right place to ask this question? Thank you in advance. Nov 20 11:52:01 I think you want some autorev magic in your recipe Nov 20 11:58:35 oenewcomer: what do you set SRCREV to in your recipe? Nov 20 12:14:14 * jackmitchell really wishes he understood automake properly :( Nov 20 12:14:37 bluelightning: "master" Nov 20 12:14:52 oenewcomer: ok, set it to "${AUTOREV}" Nov 20 12:15:00 that will always use the latest revision Nov 20 12:15:49 if you need to you can set ;branch=xxxx within SRC_URI to change the branch and still get the latest revision from that branch Nov 20 12:16:33 bluelightning: thank you, it worked! Nov 20 12:16:39 :) Nov 20 12:16:47 great :) Nov 20 13:35:56 is there a way to set ${datadir} to be different for a single recipe Nov 20 13:36:14 it is an autotools recipe and I want the resulting files to go into /opt Nov 20 13:36:28 so I have EXTRA_OECONF ="--datadir=/opt" Nov 20 13:36:58 which is fine, but then my FILES_${PN} = "${datadir}" is wrong, as ${datadir} still points to /usr/share Nov 20 13:37:43 would ${PN}_datadir = "/opt" work? Nov 20 13:39:37 i'd try datadir = "/opt" before the inherit, but i am not sure Nov 20 13:44:02 Jin^eLD Noor, I'm seeing files to edit in the /build/tmp/work/devicexx/linuxxx/package of git... Nov 20 13:44:57 afournier: cheers, that did the trick Nov 20 13:45:16 in both these directories I can edit/add drivers and makefiles. If I bitbake again at which of these directorys will it take source from? Nov 20 13:45:19 if anyone knows an any issues with this method shout up Nov 20 13:46:45 Jin^eLD: kindly run bitbake -c unpack this will unpack sources in required folder Nov 20 13:47:14 kernel does not compile from package folder Nov 20 13:47:18 guess that that was meant for koennie_ :) Nov 20 13:47:35 thanks ^^ Nov 20 13:47:41 jackmitchell: great Nov 20 13:48:08 oopsorry Jin^eLD .... I want to point it to koennie_ Nov 20 13:48:09 :) Nov 20 13:49:16 koennie_: infact if you run bitbake -c patch that will be good Nov 20 13:49:39 coz in that way it has already applied the patches and then you modify sources Nov 20 13:50:27 realy? just add the patch file in the bitbake command? no need to add anything else? Nov 20 13:52:02 koennie_: no patch is the bitbake task that will run for your recipe Nov 20 13:52:20 it will apply all the patches that has been listed in SRC_URI Nov 20 13:52:45 if you have a patch you can add your patch in SRC_URI and it will be applied via bitbake task Nov 20 13:54:58 nice, thanks Noor Nov 20 13:55:39 koennie_: you are welcome Nov 20 14:07:15 So i'm looking at the helloMod kernel module example Nov 20 14:07:20 and it is a very simple makefile Nov 20 14:07:30 I currently have a git repo with my module source code Nov 20 14:08:22 which also has a makefile which I use for cross compiling on my host system Nov 20 14:08:50 how do I get a Makefile which will either do both, or can I over write the Makefile in the git repo, with a simple OE one? Nov 20 14:33:37 So I am trying to build python-cheetah-native Nov 20 14:33:54 and the cheetah recipe needs python-pickle and python-pprint Nov 20 14:34:03 which seem ok for the non-native case Nov 20 14:34:11 but are not provided for the native case Nov 20 15:12:58 question: did patchwork had accounts cleanup or just I forgot my login/pass? Nov 20 15:14:43 I can't remember... I think khem preserved accounts during upgrades the last two times but it was quite a while ago now Nov 20 15:16:41 i'd like to .bbappend the base-files recipe with my layer, what will happen when a new recipe of base-files with a new version will be pushed to the oe-core ? my .bbappend won't be use ? Nov 20 15:18:28 what's the last commit to "oe-commits" list? Nov 20 15:18:36 afournier: it goes by filename. if the filename doesn't change, it'll work fine, if it does, your build will error. Nov 20 15:18:42 I'm suspicious about my filters Nov 20 15:18:43 afournier: you'll get an error saying there is no recipe for the bbappend, which is your cue to update it for the new version Nov 20 15:19:25 I see one in Nov 19: Saul Wold : sysstat: Update to 10.1.2 Nov 20 15:19:31 I would be really happy if someone checks Nov 20 15:19:37 bluelightning: ok so it will work if the recipe is "renamed" but what if another one is added ? Nov 20 15:19:56 afournier: it's rare that we do that in OE-Core Nov 20 15:20:01 nice Nov 20 15:23:05 oh ok, I've checked the list archives, seems normal Nov 20 15:23:41 afournier: also, you should be able to run bitbake-layers show-appends and it should warn if an append is being applied to one version of the recipe, but not to the current preferred version, unless that's changed recently (haven't played with bitbake-layers in depth in some time) Nov 20 15:24:03 kergoth: was just about to mention that... last I checked that still worked Nov 20 15:24:15 ah, nice Nov 20 15:24:17 i never used bitbake-layers :/ Nov 20 15:25:36 it's a bit painful to make a read only rootfs :/ Nov 20 15:36:24 why do we have /media and /mnt ? Nov 20 15:36:51 anyone know why pythong native recipes have DEPENDS issues with python modules? Nov 20 15:37:15 this keeps annoying me and i forget the proper fix Nov 20 15:39:58 Crofton|work: python-native doesn't end up in the PATH unless you inherit pythonnative or distutils (which inherits pythonnative indirectly). it's not often that you're building something that needs our python-native but doesn't inherit distutils, though.. Nov 20 15:40:33 afournier: there are bugs open on read-only root FYI, it's being worked on Nov 20 15:40:38 the issue is the recipe DEPEDNS on pythin-pprint and something else, which the python native does not provide Nov 20 15:40:43 but seems to build anyway Nov 20 15:40:59 bluelightning: it can work, just need some hacks Nov 20 15:41:11 I had this happen before and forget how I fixed it Nov 20 15:41:17 afournier: indeed, for simple images it's not far off working already Nov 20 15:41:33 bluelightning: sure Nov 20 15:41:38 afournier, one of the project goals is get ro roots working properly Nov 20 15:41:51 bluelightning: it boots and i have a shell, the rest is configuration mostly Nov 20 15:42:21 basically, in meta-oe, pythong cheetah works, but python-cheetah-native does not Nov 20 15:43:30 btw, i don't understand why the base-files recipe has ${localstatedir}/volatile/lock, ${localstatedir}/volatile/tmp, those are not supposed to be created my populate-volatile ? Nov 20 15:43:34 s/my/by/ Nov 20 15:43:45 RDEPENDS_${PN} = "python-pickle python-pprint" Nov 20 15:43:53 afournier: not sure about that I'm afraid Nov 20 15:44:01 Crofton|work: that's runtime, won't affect native at all Nov 20 15:44:03 commenting out this line lets python-native-cheeath build (and appear to work fine) Nov 20 15:44:22 I know that :) Nov 20 15:44:47 Crofton|work: so I'm a bit puzzled as to what's happening then... Nov 20 15:44:49 but it feels like the the native recipe wants to fins python-pickle-native etc Nov 20 15:44:58 bluelightning: so maybe i should not add this in my fstab : "tmpfs on /var/volatile type tmpfs (rw,relatime)" Nov 20 15:46:42 bluelightning, NOTE: Runtime target 'python-pickle-native' is unbuildable, removing... Nov 20 15:47:07 afournier: if you have a read-only root though populate-volatile won't work - that's one of the issues Nov 20 15:47:10 so it feels like RDEPENDS is translated for -native Nov 20 15:47:46 is there a way I can kill the RDEPENDS only for native? Nov 20 15:48:32 bluelightning: it almost work, for me, it does not because it starts writing users stuff in /tmp that is RO, but i am thinking about mounting it in tmpfs Nov 20 15:48:36 that should fix the problem Nov 20 15:48:52 Crofton|work: RDEPENDS_class-native = "" Nov 20 15:48:52 Crofton|work: this behaviour is news to me... Nov 20 15:49:05 Crofton|work: both depends and rdepends are translated for -native, yes Nov 20 15:49:30 it feels like python-native builds the stuff, but does not provide it Nov 20 15:50:04 ah. i was going to say, if it doesn't need those things, then why depend on them.. and if it does, then why does it work without rdepending on it.. Nov 20 15:50:15 we really should make all recipes emit packages, even native.. Nov 20 15:50:21 I think it does need them Nov 20 15:50:28 (not just pretend to) Nov 20 15:50:38 but -native installs them where native ptyon can find them? Nov 20 15:52:12 kergoth, agreed Nov 20 16:06:38 anyway talking about read-only rootfs, is there a way to overload _all_ pkg_{pre,post}{inst,rm} so they remount the rootfs as needed ? Nov 20 16:07:24 or does [oi]pkg has to handle this ? Nov 20 16:08:23 don't the initscripts just obey fstab? Nov 20 16:08:33 should be trivial to do that bit, i'd think Nov 20 16:08:40 if they don't, it's arguably just a bug in the initscripts Nov 20 16:09:24 i was thinking about installing a .ipk at runtime, as the rootfs would be readonly Nov 20 16:09:52 seems to me that's something your installation/update frontend ought to handle really Nov 20 16:10:06 bluelightning: ok Nov 20 16:10:39 it's better to do it just once, when many packages must be updated at once :) Nov 20 19:03:28 how do you fix such QA issues? QA Issue: non -staticdev package contains static .a library: python-numpy path '/work/armv7a-vfp-neon-oe-linux-gnueabi/python-numpy-1.4.1-r4/packages-split/python-numpy/usr/lib/python2.7/site-packages/numpy/core/lib/libnpymath.a' Nov 20 19:03:59 the same way you change anything else in packaging Nov 20 19:04:07 adjust the FILES variables Nov 20 19:04:17 that would be to obvious Nov 20 19:04:32 qt is building, so I ahve time for idle chatter Nov 20 19:04:55 likely all you n eed to do is add an appropriate glob to FILES_${PN}-staticdev Nov 20 19:05:02 iirc staticdev is before the rest in PACKAGES Nov 20 19:05:33 ah Nov 20 19:06:01 we have FILES_${PN}-staticdev += "${PYTHON_SITEPACKAGES_DIR}/numpy/core/lib/*.a" Nov 20 19:06:52 why doesn't that pick it up Nov 20 19:07:18 that should be all you need, assuming PYTHON_SITEPACKAGES_DIR is set to somethin sane (use bitbake -e to check), and assuming staticdev is before ${PN} in PACKAGES (use bitbake -t o check) Nov 20 19:07:26 crfoton some regexp earlier? Nov 20 19:07:31 whenever something seems inexpliable, it's time to b ust out bitbake -e and check some assumptions Nov 20 19:07:33 I'll do that when the build finishes Nov 20 19:08:28 crofton is the python fast enough for the sdr stuff? Nov 20 19:08:51 in gnuradio most of the python is out of the signal processing path Nov 20 19:08:58 just used to glue thing together Nov 20 20:07:13 kergoth, can you explainwhat bitbake -t o check is supposed to do? Nov 20 20:09:42 ? Nov 20 20:10:08 the only bitbake -t argument is used to select whether you wnat the xmlrpc or process based bitbake server Nov 20 20:12:47 bitbake -e and grepping for PACKAGED suggests there is no staticdev package in packages Nov 20 20:13:45 crfoton so add it Nov 20 20:14:32 yeah, that is what I am thinking Nov 20 20:15:36 interesting Nov 20 20:17:14 crofton pythonclass overwrites PACKAGES? Nov 20 20:32:19 ah Nov 20 20:32:40 no -staticdev in PACKAGES in distutils-common-base Nov 20 20:33:21 ;) Nov 20 20:33:36 I am wondering that jama didnt catch that yer Nov 20 20:33:38 yet Nov 20 20:34:23 otavio: ping Nov 20 20:34:28 probably not many python packages shipping static libraries Nov 20 20:35:13 hm right Nov 20 20:35:17 mostly its .so Nov 20 20:35:39 yeah Nov 20 20:36:01 there are likely three people that care about numpy Nov 20 20:36:19 and i am trying to be a better citizen by clobbering OE QA issues when I see them Nov 20 20:36:53 crofton hm there are more Nov 20 20:37:10 you forgot the python-opencv users Nov 20 20:37:19 they need numpty too Nov 20 20:37:41 but good catch anyway Nov 20 20:50:38 woglinde: why me? :) Nov 20 20:51:39 jama hm didnt you fixed some python recipes? Nov 20 20:51:51 I remeber something Nov 20 20:54:48 probably yes :) Nov 20 20:55:01 PACKAGES = "${PN}-dev ${PN}-dbg ${PN}-doc ${PN}" Nov 20 20:55:19 needs a ${PN}-staticdev on the front Nov 20 20:56:02 to clear a QA warnign in numpy Nov 20 20:56:06 maybe it should filter the existing PACKAGES rather than overriding it? Nov 20 20:56:09 will that break things for others? Nov 20 20:56:25 I am over my head here :) Nov 20 20:57:45 are empty -staticedev packages OK? Nov 20 20:59:24 they arent generated nor? Nov 20 20:59:32 just try it Nov 20 21:09:53 works for numpy Nov 20 21:10:05 hi, what are the new requirements for the new systemd Nov 20 21:10:09 I imported that patch: Nov 20 21:10:29 cgroupfs: create /sys/fs/cgroup to mount cgroupfs on Nov 20 21:10:39 but it still fails for 2.6.35 on bug20 like that: Nov 20 21:11:25 [ 7.479614] <29>systemd[1]: systemd-udevd.service: main process exited, code=exited, status=206/OOM_ADJUST Nov 20 21:11:25 [FAILED] Failed to start udev Kernel Device Manager. Nov 20 21:12:13 <30>systemd[1]: systemd 195 running in system mode. (-PAM -LIBWRAP -AUDIT -SELINUX +IMA +SYSVINIT -LIBCRYPTSETUP +GCRYPT +ACL +XZ; angstrom) Nov 20 21:12:22 hi gnutoo Nov 20 21:12:32 hi woglinde Nov 20 21:12:37 Does oe core have a recipe yet for building wayland? Nov 20 21:12:52 I was looking for it, but could not find it Nov 20 21:13:27 woglinde, I guess you don't use systemd? Nov 20 21:13:33 Just wanted to confirm, before I submit a recipe for it... Nov 20 21:15:22 gnutoo not on such old kernels Nov 20 21:15:42 ok Nov 20 21:15:51 the problem is that whole bmi stuff Nov 20 21:15:54 it's non-standard Nov 20 21:16:17 so If I would port quickely a new kernel to it Nov 20 21:16:23 I expect to have a lot of issues Nov 20 21:16:26 because of that Nov 20 21:16:41 if you dont switch to dts you are doomed anyway Nov 20 21:16:47 why? Nov 20 21:16:57 * kergoth rolls eyes at the comments about upgrading a device on http://wiki.baserock.org/faq/vs-openembedded-yocto/ Nov 20 21:17:11 I'm switching the work kernels to dts Nov 20 21:17:23 but I did nothing for the other non-work machines Nov 20 21:17:51 levonmaa: there's a recipe in meta-guacamayo you could presumably use as a starting point. never looked at it, just did a find against all the layers in the LayerIndex Nov 20 21:18:03 levonmaa: ./meta-guacamayo/meta-guacamayo/recipes-graphics/wayland/wayland_0.95.0.bb Nov 20 21:18:39 ah wayland support Nov 20 21:20:30 kergoth: right, looks abt the same what I rolled up Nov 20 21:20:32 ;) Nov 20 21:20:49 woglinde, do you know why I'll be doomed without dts? Nov 20 21:21:05 will all mach-* get dts? Nov 20 21:21:17 Should that be brought into openembedded-core/meta/... Nov 20 21:22:19 levonmaa: meta-oe might be a better choice, for now. oe-core is intended to be a core baseline, after all, and I can't imagine wayland is near ubiquitous yet Nov 20 21:22:46 arms one Nov 20 21:23:04 ok Nov 20 21:23:18 so they will remove platform stuff that is not dts eventually? Nov 20 21:23:37 like in drivers? Nov 20 21:24:41 kergoth: right Nov 20 21:26:30 woglinde, do you have more details? Nov 20 21:31:21 gnutoo yes Nov 20 21:31:26 thats what dts is for Nov 20 21:31:49 so I guess I'll drop the bug20 and port the gta02 on dts Nov 20 21:32:20 the problem is that I've serious booting issues with the GTA02 with > 3.2 kernels Nov 20 21:32:28 init is slow as hell Nov 20 21:32:31 same for systemd Nov 20 21:32:41 hm Nov 20 21:32:49 you need the kernel profiling stuff Nov 20 21:33:03 but I never used it sofar Nov 20 21:33:05 kergoth, baserock is a tool for sellig their arm build hardware Nov 20 21:33:53 saw that. they do have some interesting ideas though. converting all the upstream sources to git and using that is rather nice, I've been wanting to try doing that with oe metadata for years now Nov 20 21:35:02 there's a wrapper around git-svn/hg-fast-import/etc for populating the git mirrors from actual upstream that seems interesting. http://git.baserock.org/cgi-bin/cgit.cgi/baserock/baserock/lorries.git/tree/active Nov 20 21:36:27 woglinde, about profiling it's just the first userspace program that is problematic Nov 20 21:36:34 if I run it on bootchart it's fine Nov 20 21:36:53 ??? Nov 20 21:37:07 when bootchart is fine whats wrong than? Nov 20 21:37:29 the first init process can takes really too long to boot Nov 20 21:37:34 before it took hours Nov 20 21:37:38 I'll recheck Nov 20 21:37:57 but if I run it under bootchart it works fine Nov 20 21:59:41 hi **** ENDING LOGGING AT Wed Nov 21 02:59:58 2012