**** BEGIN LOGGING AT Mon Feb 22 02:59:58 2010 Feb 22 04:06:14 hi all Feb 22 04:06:31 am facing some prblm with Clutter and OE Feb 22 04:06:36 getting the error Feb 22 04:06:58 /clutter_0.8+git.bb' failed Feb 22 05:18:10 heyo Feb 22 05:30:44 hi Feb 22 05:36:01 sup yall Feb 22 07:31:22 good morning Feb 22 07:43:00 gm Feb 22 07:44:19 www.search2.net Feb 22 08:33:08 good morning Feb 22 08:37:31 recalcati: hi Feb 22 08:38:13 hi Feb 22 08:58:29 03Koen Kooi  07org.openembedded.dev * rff1e09b5e7 10openembedded.git/recipes/u-boot/ (6 files in 2 dirs): Feb 22 08:58:29 u-boot-git: update beagleboard patches for new XM revision Feb 22 08:58:29 * needs more work, but it sets the correct mux now for DVI output Feb 22 08:58:30 03Koen Kooi  07org.openembedded.dev * r276ffe064a 10openembedded.git/conf/distro/include/sane-srcrevs.inc: opkg-*: move to r522 Feb 22 08:58:31 03Koen Kooi  07org.openembedded.dev * rd9e2b8c01e 10openembedded.git/recipes/udev/ (udev-151/local.rules udev-151/udev.rules udev_151.bb): udev 151 OE rules: update to drop deprecated options Feb 22 08:58:32 03Koen Kooi  07org.openembedded.dev * re691a31f3c 10openembedded.git/recipes/cairo/cairo_1.8.10.bb: cairo: add 1.8.10 Feb 22 08:58:32 03Koen Kooi  07org.openembedded.dev * r6986f36e79 10openembedded.git/recipes/xorg-driver/ (2 files in 2 dirs): xf86-video-displaylink 0.3: add gentoo patch to fix build Feb 22 08:58:33 03Koen Kooi  07org.openembedded.dev * r0659293092 10openembedded.git/recipes/libdlo/kernel-module-udlfb_git.bb: Feb 22 08:58:34 kernel-module-udlfb: add git version Feb 22 08:58:34 * the recipe has hints for the needed kernel config Feb 22 08:58:34 * doesn't require special Xserver, fbdev works Feb 22 08:58:34 03Koen Kooi  07org.openembedded.dev * rb9efaa7bf4 10openembedded.git/recipes/dbus/ (6 files in 2 dirs): dbus: add 1.2.20, convert to new style staging Feb 22 08:58:39 03Koen Kooi  07org.openembedded.dev * ra8616e05f6 10openembedded.git/recipes/udev/udev_151.bb: udev 151: bump PR Feb 22 08:58:39 03Koen Kooi  07org.openembedded.dev * r4c664de181 10openembedded.git/recipes/networkmanager/networkmanager_0.7.999.bb: networkmanager 0.7.999: bump PR Feb 22 08:58:39 03Koen Kooi  07org.openembedded.dev * rc478cd48cc 10openembedded.git/conf/distro/include/angstrom-2008-preferred-versions.inc: angstrom: finally fix CVE-2008-4311 Feb 22 08:58:40 03Koen Kooi  07org.openembedded.dev * r99f01db28b 10openembedded.git/recipes/gnome/gnome-bluetooth_git.bb: gnome-bluetooth git: raise default pref for angstrom Feb 22 09:28:13 why is default FULL_OPTIMIZATION -Os in sane-toolchain-eglibc.inc and -O2 in glibc.inc? at least qt4-x11-free generates wrong code with -Os (checking what exactly ATM) but maybe it's more wide problem Feb 22 09:28:26 khem: you would probably know ^ Feb 22 09:47:42 03Koen Kooi  07org.openembedded.dev * rfe179a8cbf 10openembedded.git/recipes/powervr-drivers/ (3 files in 2 dirs): libgles-omap3: depend on proper fbset, set rgba correctly for 32bpp and try to enable tripple buffering Feb 22 09:47:53 03Koen Kooi  07org.openembedded.dev * r0452a571a0 10openembedded.git/contrib/angstrom/sort.sh: angstrom feed sorter: qemux86 changed arch Feb 22 09:58:25 morning Feb 22 09:58:37 hi hrw Feb 22 09:58:50 gm Feb 22 10:11:59 gource running against OE.dev repo makes nice thing to watch Feb 22 10:14:58 hrw: do you have it rendered somewhere? :) Feb 22 10:15:21 JaMa: it generated 3GB movie and did not made half of history... Feb 22 10:17:48 yup :/ I remember reading about how long and how big gentoo code swarm generation has taken http://www.gentooexperimental.org/~patrick/weblog/archives/2009-09.html#e2009-09-24T03_09_24.txt Feb 22 10:25:48 thx Feb 22 10:28:47 ? Feb 22 10:29:48 nice read it was Feb 22 10:32:03 did you give up, wainting for OE.dev movie? I hope not :) Feb 22 10:33:27 I am trying other options Feb 22 10:35:32 lets check how much time is needed for 10% of time Feb 22 10:44:04 hi Feb 22 10:44:28 i need to build clutter 1.0 Feb 22 10:45:15 by bb files contains http://paste.debian.net/60913/ Feb 22 10:45:24 anybody can tel me is it ok or not? Feb 22 10:45:50 cose am getting error while bitbaking Feb 22 10:47:15 the error you get might be more useful Feb 22 10:47:37 ok wil copy it Feb 22 10:48:03 ERROR: function staging_helper failed Feb 22 10:48:03 ERROR: see log in /home/siji/OE/angstrom-dev/work/beagleboard-angstrom-linux-gnueabi/clutter-1.0+git${SRCPV}-r3/temp/log.staging_helper.13729 Feb 22 10:48:03 NOTE: Task failed: /home/siji/OE/angstrom-dev/work/beagleboard-angstrom-linux-gnueabi/clutter-1.0+git${SRCPV}-r3/temp/log.staging_helper.13729 Feb 22 10:48:03 ERROR: TaskFailed event exception, aborting Feb 22 10:48:04 ERROR: Build of /home/siji/OE/openembedded/recipes/clutter/clutter_1.0_git.bb do_setscene failed Feb 22 10:48:06 ERROR: Task 1 (/home/siji/OE/openembedded/recipes/clutter/clutter_1.0_git.bb, do_setscene) failed Feb 22 10:48:08 NOTE: Tasks Summary: Attempted 2259 tasks of which 2259 didn't need to be rerun and 1 failed. Feb 22 10:48:10 ERROR: '/home/siji/OE/openembedded/recipes/clutter/clutter_1.0_git.bb' failed Feb 22 10:48:23 wil apste the log error @debian paste Feb 22 10:48:27 *paste Feb 22 10:50:00 log file contains only one line Feb 22 10:50:02 sh: /home/siji/OE/angstrom-dev/work/beagleboard-angstrom-linux-gnueabi/clutter-1.0+git-r3/temp/run.staging_helper.13729: No such file or directory Feb 22 10:50:47 huh scary Feb 22 10:51:12 * florian votes for cleaning this package first Feb 22 10:51:47 u mean bitbake -c clean ? Feb 22 10:51:47 siji: do you have SRCREV specified somewhere? Feb 22 10:53:01 siji: I would also clean all traces from pkgdata find tmp/pkgdata/ -name clutter\* -exec rm -rf {} \; Feb 22 10:53:42 siji: works for me, when I mess some recipe or arch is changed and staging_helper is looking for index file in wrong directory Feb 22 10:53:53 oh ok Feb 22 10:54:05 let me try Feb 22 10:54:45 siji: and having two _ in recipe name seems wrong too Feb 22 10:55:37 siji: use clutter-1.0_git.bb (as is used for 0.9) or clutter_1.0+git.bb as used in 0.8 Feb 22 10:55:50 ok Feb 22 10:55:53 siji: and note that there is extra 'r' in +gitr${SRCREV} Feb 22 10:56:05 ok Feb 22 10:57:19 I mean there should be extra 'r' clutter-0.9_git.bb has +git${ too Feb 22 10:58:09 ok Feb 22 10:58:36 do i have to modify clutter_svn.bb too? Feb 22 10:59:08 http://git.pokylinux.org/cgit.cgi/poky/tree/meta/packages/clutter/clutter-1.0_git.bb Feb 22 10:59:17 is this bb file is OK? Feb 22 11:09:15 Jama, wht's SRCREV Feb 22 11:09:21 is it md5sum? Feb 22 11:22:36 siji: SRCREV is git revision used, usually defined in sane-srcrevs.inc Feb 22 11:22:44 ok Feb 22 11:23:35 and in oe.dev is not allowed to use SRCPV in git recipes... poky has probably it allowed already Feb 22 11:23:46 oh ok Feb 22 11:24:43 in 0.8 it's mentioned as CLUTTER_REV Feb 22 11:36:15 JaMa: Poky has, yes :/ Feb 22 11:45:38 RP: I know you're already concentrating only on bitbake master, but would be great to release one more 1.8 bitbake version with LOCALCOUNT override support Feb 22 11:46:40 RP: at least for case when someone needs to build image in newer oe.dev (hopefully with SRCPVs) with older bitbake when hunting some bitbake bug or something like that Feb 22 11:47:55 JaMa Sorry for bothering u I did some modificaiton in my bb file Feb 22 11:48:31 with that release in sanity check, and override enabled by default, I hope I could be allowed to push SRCPV changes to oe.dev Feb 22 11:48:44 can u hve a look on it @ http://paste.debian.net/60923/ Feb 22 11:50:34 siji: I dont understand why cannot you just put SRCREV_pn-clutter-1.0 ?= "whatever_revision_you_want" in sane-srcrevs.inc or even better just SRCREV = "whatever_revision_you_want" in that recipe (and remove rev= from SRC_URI) Feb 22 11:51:20 siji: also now you have extra ' ' in CLUTTER_REV which won't make fetcher happy I guess Feb 22 11:55:09 JaMa i just followed the 0.8 bb file convention ( ' ' that i removed trying now) Feb 22 11:56:50 YES I think the error is now with "file://enable_tests-1.0.patch;patch=1 " Feb 22 12:03:46 JaMa, I think I did it Feb 22 12:03:56 got .ipk package now Feb 22 12:04:00 let me try it now Feb 22 12:20:47 JaMa: I really want to get people using bitbake master tbh Feb 22 12:29:13 03Koen Kooi  07org.openembedded.dev * r1e454b8556 10openembedded.git/conf/machine/hawkboard.conf: hawkboard: add ubifs support, number of LEBs might change due pending kernel partition change Feb 22 12:37:46 RP: ok, btw have you seen somewhere broken tinderclient.bbclass with master? It doesn't see depends right http://pastebin.ca/1805660 Feb 22 12:43:16 RP: and if you're tracking dependencies for bitbake releases than between 1.8 and master is new dependency on setuptools because of pkg_resources usage in parse_py Feb 22 13:22:55 JaMa: Hmm, that isn't good :/ Feb 22 13:23:06 We should probably make that an optional import Feb 22 13:23:58 JaMa: Thanks for the tinderclient bug report, we need to fix that regression Feb 22 13:26:17 hi mckoan Feb 22 13:36:04 recalcati: hi Feb 22 13:36:30 sed -i -e "s/compliment/complement/g" docs/usermanual/chapters/recipes.xml ;# I would have said, no? Feb 22 13:37:34 i mean, compliments are nice and all but.. :) Feb 22 13:50:43 can i somehow access a previously setVar from a shell function and if so, how? Intended flow would be like so: recipe A does setVar('THING', 'xyz', d). recipe B uses THING in a shell function Feb 22 13:54:34 (or rather recipe B inherits A.bbclass which provides a shell-function foo that uses THING and that foo is addtask'ed for something in B, but that shouldn't matter i think) Feb 22 13:56:06 mckoan: I sent you an e-mail for a compilation workstation. Feb 22 14:13:15 anybody knows who is the maintainer (if there is one) of the python recipes ? Feb 22 14:14:24 mlip2: mickeyl is Feb 22 14:14:46 hrw: thx Feb 22 14:21:24 blindvt: don't understand the question. any variable in the datastore can be referenced with ${}. Feb 22 14:24:13 kergoth, that's what i assumed but it somehow didn't work for me. Let me retry from scratch.. Feb 22 14:31:15 blindvt: there is no sharing of data between recipes. you can't set a var in one recipe and then use it from another. Feb 22 14:31:18 kergoth: good morning Feb 22 14:31:35 oh, right, misread his question :) Feb 22 14:31:38 hey pb_ Feb 22 14:31:47 * pb_ hands kergoth a vat of coffee Feb 22 14:31:53 woo Feb 22 14:44:34 Can some kind and generous soul attempt to build inetutils and see if it fails do_configure for them as well? I'd like to submit a patch for it if it's broken for everyone. Feb 22 14:44:47 * mwester needs more coffee too. Feb 22 14:45:08 pb_, that's a pity. thanks Feb 22 14:48:15 so what's in that persist-database. would it be legal to just store stuff in there? *blink* :) Feb 22 14:49:54 mwester: building now Feb 22 14:50:09 blindvt: what exactly are you trying to accomplish? Feb 22 14:51:19 mwester: | configure.ac:122: required file `build-aux/config.rpath' not found Feb 22 14:51:28 mwester: http://tinderbox.openembedded.org/builds/61130/ Feb 22 14:51:43 gettext issue, iirc Feb 22 14:51:44 heh Feb 22 14:51:52 i hate waking up with a headache Feb 22 14:53:16 kergoth, i have a package without an install target (gnulib) that is ready to use from a git-checkout. And i have to somehow remember ${S}/gnulib-tool to call it from the gnulib.bbclass that get's inherited from affected packaged. I try to avoid copying that ${S} dir to a known path that i cwould have to hardwire (like STAGING_DATADIR/gnulib) Feb 22 14:54:07 s/packaged/packages/ Feb 22 14:56:01 hrw, yes, that's the error I get now. It used to work; configure would copy in a config.rpath but it does so no longer. :( Debian has a patch for this problem, which is to add the config.rpath -- I'm sure that is patching the symptom, and not really fixing it, but I cannot find the source of the problem and I don't think many people really care about inetutils anymore either. Feb 22 14:56:13 kergoth, i see that everybody else hardwires that stuff (to pick just one: palmtopdir) but i don't particularly like that so i was looking how to get around it differently (and failed miserably so far) Feb 22 14:56:32 * mwester hands kergoth an aspirin and some coffee to go with it. Feb 22 14:58:53 mwester: tried inetutils 1.7? Feb 22 14:59:03 1.5 is 3y old Feb 22 14:59:58 gettext-0.17 will also need this hunk, btw: busybox.net/~aldot/oe/m4_copy.patch Feb 22 15:01:00 SlugOS is growing rather old; I have not tried upgrading inetutils for fear of requiring upgrades to automake and all the rest of the (very frightening) dev tools. But it's worth a try, I'll do that and see if it is better. :) Feb 22 15:02:19 (and yes, it also needs updated mbrtowc handling but all that stuff is imho better handled by just using gnulib-tool --update in the first place, instead of hunting down individual breakage and fixing of fallout from ancient m4 and lib hunks) Feb 22 15:05:14 blindvt: if gnulib-tool is an executable, can you not just install it into ${STAGING_BINDIR_NATIVE}? Feb 22 15:05:35 how and where does one set the linux kernel version in oe? I need to specify the 2.6.30 kernel Feb 22 15:07:18 bch132: check recipes/linux/linux_2.6.30.bb Feb 22 15:07:25 pb_, it's an executable but it relies on all m4 and lib hunks to be in-place (relative to the binary -- resp. shell script). Just copying the binary is not sufficient. And copying it around is just ugly and a waste of space (yea, i know. but still) Feb 22 15:08:50 blindvt: you might consider it ugly, but it is a fairly deep-rooted principle in oe that stuff to be used by other packages goes into staging. If you try to leave it in ${S} and refer to it there then you will lose in any configuration where rm_work is used, and it will also go wrong with packaged staging. Feb 22 15:10:18 pb_, yup. I'm giving up on that one. Feb 22 15:13:04 pb_, i take it that you've seen that i mentioned yesterday that (for me) AUTOREV with bitbake master is broken, i.e. throwing 'FetchData' object has no attribute 'basecmd' while parsing Feb 22 15:13:10 hrw: using gumstix, so I'm attempting to "bitbake omap3-console-image" - what should I be changing in linux_2.6.30.bb? In overo.conf, I see PREFERRED_PROVIDER_virtual/kernel = "linux-omap3" - can I change that to a version of the kernel? Feb 22 15:14:07 ah. Feb 22 15:14:19 so you need to check linux-omap3_2.6.30.bb recipe Feb 22 15:14:38 blindvt: I saw that, but I don't have anything useful to say about it. RP would be your resource for issues with bitbake master. Feb 22 15:15:32 hrw: ok, how do I specify to use that recipe? Feb 22 15:15:34 * blindvt nods pb_ Feb 22 15:16:39 bch132: you can specify inside:DEFAULT_PREFERENCE_overo = "1" Feb 22 15:17:07 running gource on OE git makes it sluggish as hell Feb 22 15:21:11 hrw: ok, thanks, so my understanding then is when bitbake is processing all the files since the prefered provider for the kernel is "linux-omap3" it looks through all the linux-omap3 recipes, sees that 2.6.30 has the highest preference once I put that line in there and uses that version - is that accurate? Feb 22 15:24:02 yes Feb 22 15:26:19 great, thanks! really trying to understand how OE works Feb 22 15:28:51 then if I have a patch that needs to be applied, I can just add it in that recipe under SRC_URI_append_overo? such as file://mypatch.patch;patch=1 assuming the command line to apply the patch is patch -p0 -i mypatch.patch? Feb 22 15:35:30 OE assume p1 Feb 22 15:36:49 so it should be ;patch=0 if it's -p0 on the cmd line? Feb 22 15:37:00 no, patch=1;pnum=0 Feb 22 15:37:35 ahh, ok Feb 22 15:39:32 thanks! do I need to bump the PR up anywhere afterwards? Feb 22 15:40:44 conf/machine/include/omap3.inc perhaps? Feb 22 15:43:31 or MACHINE_KERNEL_PR maybe? Feb 22 15:44:02 no neeeed Feb 22 15:45:29 cool, thanks for all your help Feb 22 15:47:13 Finally found a channel that isn't dead, related to OE (coming from angstrom). I'm wondering if there's a nice way to deal with a kernel module's name being changed (and potentially breaking encryption for a number of wireless cards) Feb 22 15:59:51 blindvt: can you summarize the issue with bitbake master, if you would? Feb 22 16:09:01 wow is GStreamer bad... my desktop is 40% busy to play a mp3... my ARM Shark only used 28% :( Feb 22 16:10:22 ouch Feb 22 16:13:53 what's arm shark ? Feb 22 16:14:17 how can i compile a specific kernel module from oe ? Feb 22 16:14:18 whatnick: dnard Feb 22 16:15:18 pb_ to compile a module ? Feb 22 16:15:34 no, the shark Feb 22 16:15:58 compiling your kernel module should be as easy as "bitbake " but it is hard to give a specific answer to a generic question Feb 22 16:18:09 03Denys Dmytriyenko  07org.openembedded.dev * r26c064da16 10openembedded.git/recipes/meta/external-toolchain-csl/SUPPORTED: external-toolchain-csl: add the list of supported locales Feb 22 16:18:55 speaking of kernel modules - how does one enable support for say v4l2 or specific video capture adapters in the kernel with oe? Feb 22 16:23:01 there isn't anything very oe specific that you need, it's just a case of providing an appropriate .config file for the kernel. Feb 22 16:26:25 pb: thanks Feb 22 16:30:51 pb_: here are better specifics -> http://docwiki.gumstix.org/index.php/Template:FAQ/Questions/USB-Webcam describes how to do it in buildroot Feb 22 16:30:59 i would like to do the same in oe Feb 22 16:51:41 Hi, I always failed to bitbake images like /home/tang/Project/OE/build/tmp/cross/armv7a/arm-angstrom-linux-gnueabi /bin/as:unrecognized option '-Qy' | make[2]: *** [dummy-checksum.o] Error 1 | make[2]: *** Waiting for unfinished jobs.... Feb 22 16:58:08 are RDEPENDS for libraries handled automatically? Feb 22 16:58:19 yes Feb 22 16:59:26 so, if there is lib in DEPENDS, which is not actually linked, it won't be in RDEPENDS? Feb 22 16:59:33 correct Feb 22 16:59:55 and because of that, the lib will be built, staged, but not packaged. right? Feb 22 17:00:03 yes that is possible Feb 22 17:00:36 ok, that's explains why I don't have a package for a built and staged lib... :) Feb 22 17:01:01 have a nice rest of day Feb 22 17:01:07 hrw: gn Feb 22 17:13:15 bye Feb 22 17:35:08 JaMa: to answer your question about Optimization settings IIRC at one time glibc was not buildable using -Os that could be the reason. However eglibc is always compilable with -Os Feb 22 18:02:48 RP: we're done with contrib-rday and 1.8-parser-rework, correct? Feb 22 18:05:49 What are the rules re: tagging a point on the dev branch? May I do so, or should I just create a (long-lived) branch for a particular distro? (I wish to mark a baseline before I make significant restructuring sorts of changes) Feb 22 18:06:22 Or should I just write down the date, and leave it at that? :D Feb 22 18:09:05 kergoth, i can't reproduce the exception with AUTOREV. Very embarrassing, i'm sorry for the noise :/ Feb 22 18:09:13 hmm, odd. np Feb 22 18:13:05 kergoth, i added default params["name"]=pn yesterday but wiped it again since i wasn't sure what to do if somebody adds a second {ht,f}tp:// for that would suddenly require adding a name so make them unambiguous i thought. If it's always the first file that gets the name=PN then all would be jazz but i that will not always be the case Feb 22 18:13:56 right, but it would be easy to override the default. just don't add it if the name param is already set on a uri. Feb 22 18:16:15 could some native speaker NAK or apply: sed -i -e "s/compliment/complement/g" docs/usermanual/chapters/recipes.xml Feb 22 18:25:45 khem: interesting is that glibc.in sets O2 as system default and also has FULL_OPTIMIZATION_pn-glibc with O2 and without frename-registers, on the other side eglibc sets -Os as default and without -pn-eglibc Feb 22 18:27:24 khem: so if I found what exactly is broken in qt4-x11-free, then I would propose to add FULL_OPTIMIZATION_pn-qt4-x11-free to eglibc.inc as it's probably problem for all distros using -Os Feb 22 18:27:58 JaMa: qt4-x11-free can not use Os ? Feb 22 18:28:57 khem: no.. it produces faulty code.. running diff between -O2 and -Os build right now.. Feb 22 18:30:25 JaMa: a gcc bug ? Feb 22 18:30:30 khem: in this thread is description and also testcase I'm using http://lists.shr-project.org/pipermail/shr-devel/2010-February/002169.html Feb 22 18:31:16 khem: happens with 4.4.2, 4.4.3, I'm waiting for 4.3.3 -Os build to finish Feb 22 18:32:48 JaMa: It would be nice to understand the real cause of the problem if possible in given time sometimes changing opt levels smear over the problems Feb 22 18:34:11 khem: that's why I refused to blindly add those FULL_OPTIMIZATION_pn-qt4-x11-free :) Feb 22 18:34:26 JaMa: very good Feb 22 18:34:44 khem: first I was wrong thinking that it's problem only on armv4t.. because Angstrom on armv5te worked ok Feb 22 18:34:46 I know its hard to spot such issues but if we could it would help Feb 22 18:35:32 khem: but I didn't noticed that it's because angstrom is using glibc and in angstrom-glibc.inc is that -O2 (and also -Os in angstrom-eglibc.inc) Feb 22 18:36:14 khem: yes and maybe if we find the real cause then maybe other apps will behave better too Feb 22 18:36:56 always the case Feb 22 19:06:32 JaMa, so couldn't we as well use plain [xyzsum] for the first file? That would nullify "archive" vs. ${PN} arguments too Feb 22 19:08:38 JaMa, busybox uses "tarball" and i was about to use PN everywhere. Doesn't really matter much though Feb 22 19:11:18 blindvt: I don't understand what do you mean by "plain [xyzsum]" ? Feb 22 19:15:03 JaMa, SRC_URI[md5sum] = "0xf00" Feb 22 19:17:12 blindvt: I agree that there is need to use name= ie in kernel x.y.z-rcN are 2 checksums in one recipe, there I used SRC_URI[kernel.md5sum] and SRC_URI[patch.md5sum] Feb 22 19:18:18 blindvt: but having different name in every recipe IMHO doesn't improve anything.. but I can be wrong and need to check .inc is no big deal.. so I don't care (much) Feb 22 19:19:27 JaMa, another possibility would be to use positional numbers, like 1.md5sum 2.md5sum for the first, second file. Just an idea for the case where there's more than 1 "not file" (i.e. not file://; they're usually not checksumed, right?) Feb 22 19:19:35 blindvt: and first I was thinking also about tarball, but using name=tarball and name=zip is even more strange then $PN :) Feb 22 19:20:03 JaMa, indeed Feb 22 19:20:20 blindvt: but that would be really difficult to track ie in mesa where you have about 3 levels of .inc files.. Feb 22 19:21:21 blindvt: and in case someone need to add second fetched source to top level .inc, he would have to update all positional sums in all recipes (maybe even outside main tree) do definitely no for positional :) Feb 22 19:22:16 JaMa, yes, the positional numbers are not a good idea for everything but the most trivial recipe. So what about just using plain [md5sum] if there is exactly one resource that is _not_ a file:// and require explicit names for everything else? Feb 22 19:23:29 JaMa, where docs would suggest to use $you_name_it as explicit name Feb 22 19:23:34 Tartarus: ping Feb 22 19:23:59 blindvt: which will fail to parse when someone add another fetched source somewhere? Feb 22 19:25:37 JaMa, exactly. This was exactly what made me back out that params["name"] = pn default yesterday evening :) But i would think that this is expected behaviour. If you complicate stuff then you can't rely on the simple automatism to take care for you, so that's perfectly fine IMHO Feb 22 19:26:01 blindvt: I'm probably not the right person to discuss checksums :) I like the idea to move them to recipe and kill too-often-conflict-change file checksums.ini, but from my just user POV, having different name in .inc files means that I usually have to check name used before I generate that checksum file in tmp Feb 22 19:26:31 03Michael 'Mickey' Lauer  07org.openembedded.dev * r4476d7a11a 10openembedded.git/recipes/mbuffer/mbuffer_20091227.bb: mbuffer: new recipe; measuring buffer Feb 22 19:26:37 03Michael 'Mickey' Lauer  07org.openembedded.dev * r62ad3249b2 10openembedded.git/recipes/tasks/task-cli-tools.bb: task-cli-tools: add mdbus2 and mbuffer Feb 22 19:39:30 as long as the recipe can override the default names, it'd be fine. that sort of automation can't be expected to work for every case, only the common case Feb 22 19:53:32 kergoth, yes. I had something like this in mind, fwiw: http://paste.debian.net/61000/ Feb 22 19:56:36 (modulo testing, typos and style -- dunno if you prefer if len("") or if None) Feb 22 19:57:40 "if len(name)" isn't very pythonic. objects have a protocol for evaluation in boolean context Feb 22 19:57:42 "if name" Feb 22 19:57:44 should do Feb 22 20:07:08 Anyone, may help me with PSPLASH, please? I installed psplash but the progress bar doesn't work, i looked at google but I didn't found anything about how to fix the progress bar. Feb 22 20:38:19 eFfeM: don't you have DEPLOY_DIR outside of TMPDIR? Feb 22 20:55:46 darn. It helps if one renames all variables uniformly. Feb 22 20:55:56 kergoth, JaMa, http://paste.debian.net/61017/ Feb 22 20:56:12 better? (that seems to actually run now) Feb 22 20:57:09 modulo the last hunk dealing with USE_LFS. Whole different typo :) Feb 22 20:58:59 i hate to do that, but history|grep 'could some' Feb 22 20:58:59 could some native speaker NAK or apply: sed -i -e "s/compliment/complement/g" docs/usermanual/chapters/recipes.xml Feb 22 20:59:08 or should i send a patch? Feb 22 20:59:54 khem, *blink* Feb 22 21:13:06 hmmm I will queue it along with some other patches I have Feb 22 21:20:49 khem, how do i build docs ? Feb 22 21:21:48 khem, oh nm. found the README Feb 22 21:25:02 n8 Feb 22 21:26:21 no Feb 22 21:33:29 denix, was afk, i don't have deploy dir outside tmpdir, but I do have staging dir outside of tmpdir Feb 22 21:35:18 eFfeM: ah, so you don't suffer from http://cgit.openembedded.org/cgit.cgi/openembedded/commit/?id=8dc84badb1d772f5953492c4022c1f644c4fe278 Feb 22 21:41:07 JaMa, see ML Feb 22 21:50:19 denix, i did not have that problem recently, got some issues a while ago in that area though Feb 22 21:50:47 denix, do you get the problem when restoring from packaged staging Feb 22 21:51:06 eFfeM: I'm still a little bit lost about the reason for that change... Feb 22 21:51:28 denix, if so, my hunch is that there is something odd with timestamps, if you are restoring from pacakged staging do_deploy should not be needed Feb 22 21:51:51 denix wrt the patch: I seem to recall it came from Tartarus Feb 22 21:51:54 eFfeM: actually, I had to disable pstage due to that change - it doesn't work with pstage AND deploy out of tmp... Feb 22 21:52:22 yeah, I tried to ping him... Feb 22 21:53:19 at some point i made a local patch to add a dir or so, but didn't commit that because that was not the root cause and just mimicking hthe problem Feb 22 21:53:36 and after the patch frm Tartarus i don't think i saw my problem again Feb 22 21:53:57 what is going wrong for you, is do_deploy failing when restoring from pstaging ? Feb 22 21:56:17 eFfeM: no, as I said, I can't even use pstage for now... Feb 22 21:57:11 eFfeM: it fails in do_deploy when building, if deploy is outside of tmp Feb 22 21:57:40 eFfeM: need to talk to ant - I think he was also using it this way... Feb 22 21:57:47 crumbs, a patch with documentation. unconventional! Feb 22 22:01:07 :) Feb 22 22:01:28 * denix thinks uri1/uri2 can be confusing... Feb 22 22:02:57 denix, odd Feb 22 22:03:01 no idea Feb 22 22:03:18 pb___, heh, call me wimp since i even checked with lynx ;) Feb 22 22:04:22 heh Feb 22 22:04:41 pb___, but serious. is md5sum= deprecated by now? see note hunk in docs Feb 22 22:04:58 yeah, looks like a reasonable patch to me. might want to kill that stray bb.note(), but I think the basic idea is fine. Feb 22 22:05:23 pb___, if it's not sha256sum should be documented and implemented if it isn't already -- didn't check Feb 22 22:05:41 blindvt: no sha256sum Feb 22 22:05:49 RP wanted to add one to bb Feb 22 22:06:06 pb___, the bb.note is a debug-leftover, yep. sorry for that Feb 22 22:07:03 khem, btw.. I cought me adding USE_LFS since i want to be able to choose if i want it, per distro. Feb 22 22:07:54 blindvt: afaik, md5sum isn't deprecated as such, though its use probably shouldn't be encouraged. Feb 22 22:08:14 calling it a day, cya Feb 22 22:08:22 03Graeme Gregory  07org.openembedded.dev * ra9484a8291 10openembedded.git/recipes/gtk+/gtk+_2.18.6.bb: Feb 22 22:08:22 gtk+_2.18.6.bb : fix EXTRA_OECONF for native builds. Feb 22 22:08:22 We dont have a native libjpeg so make sure its disabled. Feb 22 22:08:23 03Graeme Gregory  07org.openembedded.dev * r5aa5974c1f 10openembedded.git/recipes/maemo4/ (libgpsbt/configure-fix-version.patch libgpsbt_0.1.bb): Feb 22 22:08:24 libgpsbt_0.1.bb : apply patch to remove debian requirement, update staging. Feb 22 22:08:24 http://wiki.navit-project.org/index.php/Navit_on_OpenEmbedded_for_n810 has Feb 22 22:08:24 a patch that is required on non debian systems. Feb 22 22:08:24 03Graeme Gregory  07org.openembedded.dev * rd88aa38f6b 10openembedded.git/recipes/maemo4/ (libgpsmgr/configure-fix-version.patch libgpsmgr_0.1.bb): Feb 22 22:08:25 libgpsmgr_0.1.bb : apply patch to remove debian requirement, update staging. Feb 22 22:08:25 http://wiki.navit-project.org/index.php/Navit_on_OpenEmbedded_for_n810 has Feb 22 22:08:25 if the recipe declares an md5sum for a file then it'll be checked. if you don't declare an md5sum then that's fine too. Feb 22 22:08:26 a patch that is required on non debian systems. Feb 22 22:09:05 it's perfectly legitimate to only declare the sha256sum and I guess that might make sense for new recipes. Feb 22 22:09:23 khem, but this gets me into wanting to add more (there's already NLS in there). Now i wonder if there is some more generic way. In my buildroot, i export the whole uClibc.config hierarchy as UCLIBC_CONFIG__knobs and can configure/depend/rdepend packages accordingly Feb 22 22:09:42 khem, ideas on how i can transplant that idea to oe? Feb 22 22:09:56 pb___: is there sha256sum checker in bb? Feb 22 22:10:49 denix: in bb? no, but none of this logic is in bitbake anyway Feb 22 22:10:50 pb___, i ment md5sum as in SRC_URI="uri://foo;md5sum=4711". Is that notation deprecated in favour of SRC_URI[md5sum]= ? Feb 22 22:11:13 pb___: see blindvt msg :) Feb 22 22:11:34 blindvt: oh, right, sorry, I misunderstood what you meant. yeah, in terms of oe it probably is deprecated. Feb 22 22:11:45 * denix followed blindvt's line of thought... :) Feb 22 22:12:03 I guess it's up to the bitbake dudes to decide whether they want to go on supporting that for non-oe uses. Feb 22 22:12:30 pb___, if it's not then 1) let's deprecate it or 2) add sha256sum= -- assuming that notation is not there. And if the ;sha256sum= notation in SRC_URIs is there, why were SRC_URI[{md5,sha256}sum] added in the first place? :P Feb 22 22:12:35 blindvt: you mean to make uclibc .config visible to other packages right ? Feb 22 22:12:50 khem, yes, exactly. Feb 22 22:13:21 blindvt: the answer to the latter question is that, with the checksum as a URI parameter, it is more tedious to have a generic .inc file define the URI and provide the checksums in version-specific bb files. Feb 22 22:13:21 pb___, blindvt: actually, as I said earlier, RP went on record saying he'd like to get sha256sum added to bitbake Feb 22 22:14:00 blindvt: obviously you could write SRC_URI = "http://x/y;md5sum=${MD5SUM}" and then define ${MD5SUM} independently, but that idea was not very popular. Feb 22 22:14:10 blindvt: A lot should already be exported in uclibc headers Feb 22 22:14:21 khem, it's a chicken and egg problem though. Depends on if you want to rule over packages per (certain) libc config or vice versa. Not sure. For some libc knobs it's fine but others, like locale/LFS/WCHAR/ distros/users should decide. Feb 22 22:14:34 blindvt: in OE I think distro rules Feb 22 22:14:46 denix: well, that's fine, he's welcome to do that. Feb 22 22:15:01 so essentially a .config is a property of a distro Feb 22 22:15:27 pb___: but now since you mention it, looks like using ;md5sum=${MD5SUM} is not popular in OE... right? Feb 22 22:15:44 right Feb 22 22:15:53 I can't actually think of any recipes that are doing that, offhand Feb 22 22:16:09 routine=${INSANELY_LONG_HASH} is never going to look nice Feb 22 22:18:45 khem, but which knobs should be decided by distro (and how?). Take LFS. For a distro to set that (now, at least AFAIK) you have to add blobs to base.bbclass that is 0) overridable by distro 1) honored by libc 2) passed down to configury of packages. Feb 22 22:19:03 The code in OE is just stopgap until we get that sorted out - http://thread.gmane.org/gmane.comp.sysutils.bitbake.devel/1089/focus=1115 Feb 22 22:20:13 blindvt: probably distro should export equivalent variables into metadata Feb 22 22:20:41 blindvt: and now how do we define the new variables is as hard as posix groups Feb 22 22:20:48 I guess it's still confusing what's considered recommended way from OE and bitbake perpective... Feb 22 22:21:30 khem, same holds true for stuff like malloc_0: has to be passed down to packages: Feb 22 22:21:32 $ grep malloc_ ../../buildroot.git.m68k/package/Makefile.in Feb 22 22:21:32 BR2_AC_CV_FUNC_MALLOC_0_NONNULL=ac_cv_func_malloc_0_nonnull=yes \ Feb 22 22:21:32 gl_cv_func_malloc_0_nonnull=yes \ Feb 22 22:21:32 BR2_AC_CV_FUNC_MALLOC_0_NONNULL=ac_cv_func_malloc_0_nonnull=no \ Feb 22 22:21:32 gl_cv_func_malloc_0_nonnull=no \ Feb 22 22:22:23 blindvt: well we have site/ files Feb 22 22:22:28 khem, though a distro might not really be concerned about it Feb 22 22:22:39 looks under site/ Feb 22 22:22:56 yeah site files do exactly that Feb 22 22:23:07 khem, mhm. Looking.. Feb 22 22:23:09 you have an arch and a libc and and os Feb 22 22:23:38 and some stuff is linux specific some stuff arch specific some stuff libc specific Feb 22 22:23:47 and some combo of both Feb 22 22:23:58 hi kehm and blinder Feb 22 22:24:04 woglinde: hello Feb 22 22:24:09 woglinde, heya Feb 22 22:24:53 hms Feb 22 22:24:59 woglinde, oh, i've seen that there is a real blinder, but i take it you know that we're distinct beings :) Feb 22 22:25:19 no rfkill driver for viliv ex79 Feb 22 22:25:22 ups ex70 Feb 22 22:26:44 khem: I guess the issue with malloc-0 is that it depends on how you configured uClibc, and that in turn is a distro decision. so there is no single right answer for a given arch, or a given TARGET_OS. Feb 22 22:27:27 khem, the site thing doesn't help me, i think Feb 22 22:27:37 oh yes Feb 22 22:27:48 of course, you could argue that setting it to be anything other than the same behaviour as glibc is madness, and distros that do that can pick up the pieces for themselves, but the underlying issue does have some validity. Feb 22 22:27:55 hi pb Feb 22 22:28:01 pb___: I think we need thoughtful distro vars Feb 22 22:28:03 hi woglinde Feb 22 22:28:20 khem: well, that's what DISTRO_FEATURES is for. Feb 22 22:28:27 pb___: yeah Feb 22 22:28:34 khem, it's libc config dependant (and that was what i was aiming at -- how to propagate what to whom, decided by whom^2) :) Feb 22 22:28:34 I would not be in favour of some massive proliferation of new variables. Feb 22 22:29:06 me neither. That's why i'm asking Feb 22 22:29:07 I have to enhance those for eglibc as I will soon add configurablility for eglibc which uses another set of config options Feb 22 22:29:36 bbiab, gotta go put the chickens to bed Feb 22 22:30:20 blindvt: libc config that changes a system wide feature is not libc contained Feb 22 22:30:41 it has to have global visibility and distro seems to be the right choice Feb 22 22:31:12 I think making distro's a bit configurable is going to help Feb 22 22:31:58 khem, agree. But that means we need a knob that 1) comes from deity (user/distro) 2) is honored by libc 3) is consequently (potentially) passed to packages. This is currently -- AFAICS -- done via base.bbclass, please correct me if i'm wrong Feb 22 22:32:53 DISTRO_FEATURES Feb 22 22:32:55 is the way Feb 22 22:32:57 khem, and putting all that stuff into base doesn't stike me as a sensible thing to do but i could of course be wrong Feb 22 22:33:11 khem: hi Feb 22 22:33:17 ant_home: hello Feb 22 22:33:42 khem: we need your help again: klibc vs. headers Feb 22 22:33:56 with 2.6.26 it fails on do stage Feb 22 22:33:57 blindvt: from practical POV adding knob for every uclibc option wud be an overkill impo Feb 22 22:34:05 ant_home: hmm ok Feb 22 22:34:10 whats the error Feb 22 22:34:14 ant_home: hi Feb 22 22:34:17 mom Feb 22 22:34:33 http://fr.pastebin.ca/1804701 Feb 22 22:34:40 I read smthg in patchwork Feb 22 22:34:47 hi ant Feb 22 22:34:51 hello Feb 22 22:35:22 http://paste.pocoo.org/show/180726/ Feb 22 22:35:25 khem, fair enough. DISTRO_FEATURES sounds promising. Why is NLS not handled there? :) Feb 22 22:35:50 ant_home: hey, is your DEPLOY_DIR out of TMPDIR? Feb 22 22:35:58 not yet ... Feb 22 22:36:19 khem: this is with 2.6.3x http://paste.pocoo.org/show/180726/ Feb 22 22:36:33 ant_home: I was asking eFfeM if he had issues with http://cgit.openembedded.org/cgit.cgi/openembedded/commit/?id=8dc84badb1d772f5953492c4022c1f644c4fe278 Feb 22 22:36:52 blindvt: it should be put there too Feb 22 22:37:02 denix: this commit actually solved my issue Feb 22 22:37:28 ant_home: strange. what was the issue? Feb 22 22:37:45 ant_home: as it actually breaks my setup now :) Feb 22 22:37:50 iirc cp cannot find zImage Feb 22 22:37:58 let see if I kept some traces Feb 22 22:38:59 ant_home: hmmm it seems you need linux_libc_headers Feb 22 22:39:11 khem, i see. Noted. Feb 22 22:39:18 * blindvt calls it a day Feb 22 22:39:22 g'night * Feb 22 22:40:04 nite blinder Feb 22 22:40:20 ant_home: after this change, I have "cp cannot copy file into itself" Feb 22 22:40:26 khem: this new version adds the install_header patch... Feb 22 22:40:38 hm Feb 22 22:41:07 denix: you patched pacjkage_stagefile_shell? Feb 22 22:41:48 ant_home: I had it patched to have deploy dir out of tmp Feb 22 22:42:10 ok. then it seem sth eissue just apply in that case Feb 22 22:42:23 oh..my Space key... Feb 22 22:42:34 ant_home: now it cannot be fixed one way or another Feb 22 22:42:40 he he Feb 22 22:42:50 ant_home: right, that's why I asked if you use this setup :) Feb 22 22:43:08 incredibly I can rebuild from pstage, initramfs and second kernel included Feb 22 22:43:53 without that patch it was failing Feb 22 22:44:36 ant_home: I see. but can you confirm if you have deploy dir inside tmp dir or out? :) Feb 22 22:44:44 inside Feb 22 22:44:51 thanks Feb 22 22:44:53 'angstrom vanilla' Feb 22 22:45:33 right Feb 22 22:45:48 iirc I could build/rebuild minimal too Feb 22 22:46:56 khem: I'm a bit lost with that klibc patches Feb 22 22:47:27 we have some conditionals in signal-cleanup.patch e.g. Feb 22 22:47:37 +#if LINUX_VERSION_CODE >= KERNEL_VERSION(2,6,31) Feb 22 22:48:37 I suspect that the impact of th e'install_header' patch and the refactor of the other staging patches could have broken the thing :/ Feb 22 22:51:40 is there an oe recipe that will enable v4l2 support in the kernel? Feb 22 22:53:09 or can I just change the option to y in kernel-config? Feb 22 22:54:15 ant_home: Does klibc expect linux src tree to be available during build ? Feb 22 22:55:03 do_configure () { Feb 22 22:55:03 ln -sf ${STAGING_KERNEL_DIR} linux Feb 22 22:55:03 } Feb 22 22:57:00 khem: about 2.6.33 http://lists.linuxtogo.org/pipermail/openembedded-commits/2010-February/042611.html Feb 22 22:58:41 ant_home: I think you need to create a symlink Feb 22 22:58:46 in the klibc recipe Feb 22 22:59:12 we do as above Feb 22 22:59:29 in klibc-common.inc Feb 22 23:01:48 anyway, for 2.6.26, I understand that the new sanitized headers installed by klibc using 2.6.26 are defining getline, and that's the error Feb 22 23:02:11 for 2.6.33 the issue is the 'asm' directory, I suppose Feb 22 23:03:06 ah, the full log is here http://tinderbox.openembedded.org/public/logs/task/5094363.txt Feb 22 23:03:59 hm.. Feb 22 23:04:00 usr/klibc/../include/sys/types.h:15:31e error: osix_types.h: No such file or dierror: llinux/posix_types.h: No such file or director Feb 22 23:04:08 ^^this seems very strange... Feb 22 23:04:30 osix? Feb 22 23:09:22 ant_home: so which kernels are failing for you Feb 22 23:09:29 for me 2.6.26 Feb 22 23:09:34 for JaMa 2.6.33 Feb 22 23:09:47 with same error as you gave pointers too ? Feb 22 23:10:01 and klibc_1.5.15 fails against 2.6.33 too Feb 22 23:10:12 do you have posix_types.h in your staged kernel sources Feb 22 23:10:17 under linux/ Feb 22 23:11:13 yes, posix_types.h 1270n Feb 22 23:11:20 ok Feb 22 23:13:01 khem: the point is klibc_1.5.15+1.5.16 now fails for both: old and new :D Feb 22 23:13:08 and if you create the above symlink say like this { cd usr/include; ln -sf ${STAGING_KERNEL_DIR}/include/linux usr/include/; ln -sf ${STAGING_KERNEL_DIR}/include/asm usr/include/; ln -sf ${STAGING_KERNEL_DIR}/include/asm-generic usr/include/} Feb 22 23:13:14 does that help Feb 22 23:13:30 don't we have apatch for this? Feb 22 23:13:30 oh get rid of that cd usr/include Feb 22 23:16:10 ant_home: hmmm usually having symlink to top of lnx src tree is enough Feb 22 23:16:24 what do you have under the staged linux tree Feb 22 23:16:27 03Denys Dmytriyenko  07org.openembedded.dev * rc079ec93e2 10openembedded.git/recipes/eggdbus/eggdbus_0.6.bb: Feb 22 23:16:27 eggdbus: disable parallel make due to eggdbusenumtypes.h gen/use Feb 22 23:16:27 Quick workaround to make it build again. Can probably be fixed properly in Feb 22 23:16:27 the Makefile... No PR bump. Feb 22 23:17:14 khem: I have 2.6.26 staged now. I'd like to have klibc_1.5.16 able to build on toip of it, then move to the new 2.6.33 kernels/headers Feb 22 23:18:18 it seems all comes by the sanitized headers from headers_install Feb 22 23:18:34 these redefinr e.g. getline Feb 22 23:19:22 # Feb 22 23:19:22 | scripts/unifdef.c:209: error: conflicting types for 'getline' Feb 22 23:19:22 # Feb 22 23:19:22 | /usr/include/stdio.h:651: error: previous declaration of 'getline' was here Feb 22 23:20:37 ant_home hm that should be only in older kernels Feb 22 23:20:57 yep..2.6.26 is old enough Feb 22 23:21:01 mostfixes renames getline to get_linet in unifdef.c Feb 22 23:21:05 ant jupp Feb 22 23:21:42 I had similar issues with kexec-tools and was just hacking/purging the headers Feb 22 23:22:07 but I asked khem if he sees a 'preferred' method to stage the damned asm headers :) Feb 22 23:23:06 ant_home: As I see klibc seems to use linux src tree and not sanitized headers things may be weird Feb 22 23:23:21 exactly Feb 22 23:23:35 but finally they added a patch Feb 22 23:23:51 ant_home: oh to klibc ? Feb 22 23:24:06 ant_home: pointer ? Feb 22 23:24:19 mom Feb 22 23:24:49 http://git.kernel.org/?p=libs/klibc/klibc.git;a=commit;h=dc5aff1731bc4708d3391311074f026690212f8d Feb 22 23:25:32 this was interfering with the staging.patch Feb 22 23:25:39 we have in OE Feb 22 23:26:46 khem: I added all pending patches to the 1.5.15+1.5.16 recipes Feb 22 23:28:00 klibc_1.5.15 compiles just fine and does not have that patch, just the staging.patch Feb 22 23:34:19 khem: I have to sleep now...thx for your attention Feb 22 23:34:43 woglinde: ebenfalls...feel free to help khem :) Feb 22 23:34:48 good night Feb 22 23:39:43 ant_home: I understand the problem a bit now Feb 22 23:40:28 I think it expects a raw linux src tree but OE's staged tree is not same Feb 22 23:40:39 so make header_install wont work as it should Feb 22 23:40:57 because we already create those headers syslinks in kernel.bbclass Feb 23 00:42:24 hmm, wonder if anyone's looked at adding the toolchain for the Go language to oe.. though i guess gccgo is just a matte rof using a branch and --enable-languages.. Feb 23 00:42:26 * kergoth thinks Feb 23 00:45:28 Why go there, kergoth? Feb 23 00:45:42 heh Feb 23 00:45:45 * Crofton grumbles Feb 23 00:46:03 git is not built with ssh auth support in OE Feb 23 00:46:40 hmmm Feb 23 00:47:11 we depend on openssl Feb 23 00:49:14 what do you mean? pretty sure git just runs ssh, it doesn't have an internal ssh implementation.. Feb 23 00:53:19 looks like it does not work with dropbear Feb 23 00:53:29 replaced that with open ssh and beating some more Feb 23 00:53:46 yeah, I am trying to clone a got repo onto the beagle Feb 23 00:53:58 that needs a key to login **** ENDING LOGGING AT Tue Feb 23 02:59:58 2010