**** BEGIN LOGGING AT Mon Aug 17 02:59:56 2009 Aug 17 05:46:59 kergoth: this local overriding did not work for me Aug 17 05:47:07 I tried it did not honor it Aug 17 06:15:02 well, dunno what to tell you. local is in OVERRIDES. run bitbake -e and see if things are what you think they are Aug 17 06:19:35 could I get some help with a fairly simple question about embedded linux that doesn't really involve oe? Aug 17 06:21:57 only if it's not about kernel modules ( ;) joking ) Aug 17 06:22:59 well Aug 17 06:23:24 it's the same question I asked in elinux, and while I hate to ask in two channels, elinux looks pretty dead Aug 17 06:24:44 the basic problem is that I switched root filesystems and now I'm getting a kernel panic for reasons I don't really understand Aug 17 06:24:53 because I've used this same rootfs image elsewhere Aug 17 06:25:03 with a newer kernel, if that changes anything Aug 17 06:25:18 the boot log is at http://pastebin.com/m188ab733 Aug 17 06:25:27 it's odd because it mounts root and then kernel panics Aug 17 06:25:34 but it doesn't reall give an error message? Aug 17 06:25:40 at least not that I can see Aug 17 06:32:52 I think the last line is quite an error Aug 17 06:33:36 As init is not allowed to die Aug 17 06:34:45 yes Aug 17 06:34:50 but I have no idea why it's dying Aug 17 06:35:07 and 'EXT2-fs warning (device sda1): ext2_fill_super: mounting ext3 filesystem as ext2' doesn't look normal to me (but I might be wrong) Can't you trie to add ext3 drivers to kernel (fixed in kernel, not module) Aug 17 06:35:09 is that a symptom of something I should know about? Aug 17 06:35:16 yea Aug 17 06:35:21 the kernel doesn't have ext3 Aug 17 06:35:24 I'm going to fix that Aug 17 06:35:32 but the old root was on the same filesystem Aug 17 06:35:42 I just deleted all the files and added the new ones Aug 17 06:36:09 but that was used with a different kernel you said, did that kernel have ext3 drivers? Aug 17 06:36:37 sorry, I was kinda confusing there Aug 17 06:36:40 it's the same kernel Aug 17 06:36:52 the kernel is stored in flash memory on the board itself Aug 17 06:37:00 I'm making it use a flash drive as the root filesystem Aug 17 06:37:26 the comment you saw on #elinux was that I didn't copy the kernel modules from the old root to the new root Aug 17 06:37:40 I've done that, and it still panics in the same place Aug 17 06:38:19 so it was mounting an ext3 partition as ext2 before as well Aug 17 06:38:24 the same warning was there Aug 17 06:39:32 o.k. and its a usb flash drive you use for rootfs? Aug 17 06:40:04 yes Aug 17 06:41:55 when you say you switched root filesystems, what exactly do you mean? Aug 17 06:42:37 different file system image? different filesystem on root? different pysical location of root filesystem? Aug 17 06:43:23 well Aug 17 06:43:27 root is on a flash drive Aug 17 06:43:34 so I popped that drive into my laptop Aug 17 06:43:40 moved all the files to a backup directory Aug 17 06:43:47 and then unzipped an ARM distro image on there Aug 17 06:44:02 the image is compiled for ARMv4t Aug 17 06:44:09 unzip does not retain file permissions Aug 17 06:44:14 and my board uses ARM920t Aug 17 06:44:15 that could be an issue Aug 17 06:44:22 well, I untarred it, actually Aug 17 06:44:27 ok. Aug 17 06:44:27 tar -xjf Aug 17 06:44:34 as root? Aug 17 06:44:38 tes Aug 17 06:44:39 yes* Aug 17 06:44:40 good Aug 17 06:45:01 although I'll check that to be sure... Aug 17 06:45:21 i guess that your kernel cannot execute that init binary Aug 17 06:45:26 for whatever reason Aug 17 06:45:45 hrmmm Aug 17 06:45:53 or you might have to wait before the usb-drive is proper initialized Aug 17 06:46:05 it's getting root mounted Aug 17 06:46:17 I have rootdelay=10 in the kernel command line Aug 17 06:46:33 that should be enought indeed :) Aug 17 06:47:20 I don't think the kernel was optimized for ARM920t, but if it was, would that make it unable to run ARMv4t binaries? Aug 17 06:47:59 maybe Aug 17 06:48:21 if ARMvt is a subset of ARM920t (i don't know arm very well) Aug 17 06:48:32 well Aug 17 06:48:44 it's basically ARMv4 vs ARMv9 Aug 17 06:48:50 ARM is backwards compatible Aug 17 06:48:57 all the arm versioning confuses me... Aug 17 06:49:01 so an ARMv9 CPU can execute ARMv4 binaries Aug 17 06:49:13 so... ergh Aug 17 06:49:19 I think I just answered my own question XD Aug 17 06:49:41 well, at any rate, I guess I'll try building a new kernel on my other ARM system that uses the same OS image Aug 17 06:49:54 gcc allows you to tune for a processor, but still run on others in the same family Aug 17 06:50:00 -march vs -mtune Aug 17 06:50:09 check the gcc docs Aug 17 06:50:26 well, the issue is that I don't remember how the kernel was initially compiled Aug 17 06:50:33 :) Aug 17 06:50:34 I built it, but it was like a year ago Aug 17 07:26:56 all three networkmanager-applet recipes are broken Aug 17 07:32:41 Should it be possible to create an dir with 700 permissions from an OE receipe? (I tried but the dir in the rootfs ends with 774 Aug 17 07:33:43 And I see only 1 other recipe in the recipes dir wicht uses "install -m 0700 -d ${D}" Aug 17 07:35:38 (I'm now testing if that recipe is able to create a dir with 700 rights) Aug 17 07:37:52 mm, It does, at least in work dir... what are the differences Aug 17 08:08:16 morning Aug 17 08:09:21 does someone knows why does not work this part in the .bb file? ROOTFS_POSTPROCESS_COMMAND += "opkg-cl ${IPKG_ARGS} -force-depends remove ${PACKAGE_REMOVE};" Aug 17 08:09:57 morning Aug 17 09:10:35 good morning Aug 17 09:15:11 Morning florian Aug 17 09:18:07 XorA: are you an angstrom developer? Aug 17 09:18:18 zecke: yes Aug 17 09:18:54 XorA: it seems tht the repo browser is broken Aug 17 09:20:07 zecke: I have no access to that machine, koen does the coding on that Aug 17 09:23:07 Question: What package is loadkeys in? console-tools? (only if someone knows off the top of there head before I grep). Aug 17 09:24:26 Does anyone else actively use the XFCE recipies in OE? Been upgrading them and beating them into some sort of shape but I wondered if there was other users. Aug 17 09:39:39 florian: good morning Aug 17 09:39:40 zecke: good morning Aug 17 09:40:43 pb__: hey Aug 17 09:41:00 pb__: could you take a look ath the glibc patch I sent? Aug 17 09:43:12 zecke: is this the one for ports? Aug 17 09:43:27 pb__: yes Aug 17 09:43:37 okay. I'll try it out this morning. Aug 17 09:43:48 I need to update my tree first, I haven't touched it for about six weeks. Aug 17 09:44:06 pb__: six weeks? that is bad :) Aug 17 09:44:16 heh. well, I have been on vacation :-) Aug 17 09:44:57 not for 6 weeks, admittedly, but the knock-on effects of that have meant I haven't had much time for oe lately. Aug 17 10:19:50 Hi! I have the following problem: there are some packages which have a postinst script which will be run by /etc/rcS/S98configure to setup some systemusers and then run /etc/init.d/populate_volatile.sh update Aug 17 10:20:34 but if there are more then one postinst script which does adduser and populate the first will fail Aug 17 10:21:04 in my system its hal and pulseaudio Aug 17 10:22:33 S98configure runs "opkg-cl configure". in my case this selects pulseaudio first. It adds the pulseaudio user and group and then runs populate volatile Aug 17 10:23:28 but hal is not configured at this time, so the populate-volatile run by pulseaudio-server.postinst will fail and leaves pulseaudio-server in state unpacked Aug 17 10:24:52 my quickhack is to replace the "/etc/init.d/populate_volatile.sh update" line with /e/i.d/p_v.sh update || /bin/true" but this is just a quick hack Aug 17 10:25:02 is there any better solution? Aug 17 10:27:05 I guess you should teach populate_volatile.sh to work on a single package at a time. Aug 17 10:28:34 That'd be desirable anyway, since there's no point re-creating the data for every package whenever a single one is updated. Aug 17 10:31:10 pb__: ok, this sounds reasonable Aug 17 10:40:51 Jeff__: I would love you if you fixed that problem, its been annoying me for years, but I never got round to it Aug 17 10:50:14 hi @all Aug 17 10:51:42 did anyone tried the mini2440 build in the stable branch recently? Aug 17 11:14:13 I have installed bash in my image but /bin/sh is still symlinked to busybox, what do I need to do to stop that from happening? Aug 17 11:15:05 install bash as a higher-priority alternative for sh, I guess. Aug 17 11:15:18 I thought it was Aug 17 11:15:29 what priority is it using? Aug 17 11:15:50 looks like busybox is 50, if I'm reading the bb file right Aug 17 11:16:06 I think it only installs as bash by default Aug 17 11:16:22 it doesn't have any alternatives magic in its bb Aug 17 11:16:42 dash does, however, so I guess I can copy/paste that Aug 17 11:16:46 ah, well then. that'd certainly be an issue. Aug 17 11:16:52 yeah, you should be able to copy it from dash. Aug 17 11:19:16 ok I'll try with priority 60 Aug 17 11:19:54 I guess it should be higher than busybox by default, since if you install bash it's probably because you detest the busybox sh Aug 17 11:21:16 mm, I dunno. there's a bit of a conflict there between /bin/sh as used for scripts (for which you might want busybox, since it's faster) and /bin/sh as used as a login shell (for which I could understand wanting bash since it sucks less) Aug 17 11:21:40 also, you might want to install bash just to satisfy some script which requires #!/bin/bash, without necessarily making everything else use bash all of a sudden. Aug 17 11:23:07 so, it's not completely clear to me that either one of them is necessarily "better" as /bin/sh. I think both of them do implement all the stuff that POSIX mandates for /bin/sh (and hence all the stuff that scripts are entitled to rely on). Aug 17 11:23:41 it's also interesting to note that other distros are tending to move away from installing bash as /bin/sh, replacing it with things like dash instead. Aug 17 11:23:42 mmm Aug 17 11:23:58 you got a point, then I guess the solution would be for /etc/passwd to use /bin/bash as login shell by default Aug 17 11:24:08 or /bin/interactive-shell which could be alternatives-driven Aug 17 11:24:12 if your main complaint is just that busybox sh sucks for interactive use then you might be better off just changing your login shell selection in /etc/passwd Aug 17 11:24:15 right, yeah, exactly Aug 17 11:24:46 what's the advantage of dash over bash, just a size thing? Aug 17 11:24:56 size and allegedly speed Aug 17 11:25:03 while being more bash like than other alternatives Aug 17 11:25:11 not that I'd use it unless I had to Aug 17 11:25:17 I put bash back since I had an app that complained bitterly about it, but I can't recall now what exactly it was Aug 17 11:25:27 ubuntu uses it by default Aug 17 11:25:31 the shell gets loaded hundreds of times in linux initscripts, smaller means faster Aug 17 11:25:42 as for some reason linux doesnt cache that sort of thing well Aug 17 11:25:48 XorA: not necessarily, there's only one copy in memory at any given time Aug 17 11:25:54 I think it's partly speed, partly memory usage, and partly a desire to identify (and eliminate) scripts which use bash-only features by mistake. Aug 17 11:25:58 bss and data would be instantiated for each of course Aug 17 11:26:05 but there's only one .text area around Aug 17 11:26:26 tzanger: ubuntu benchmarks pretty much speak for themselves, whole seconds saved on boot time Aug 17 11:26:30 ... unless you're nommu... in which case yes, smaller is very much better :-) Aug 17 11:26:48 XorA: I doubt very much dash has anything to do with that and more things like upstart Aug 17 11:26:58 * tzanger curses the gpio gods this morning Aug 17 11:27:06 tzanger: except they run sysvinit under upstart Aug 17 11:27:12 tzanger: so there is no different there Aug 17 11:28:01 I'd have to benchmark it myself to be sure Aug 17 11:28:13 but right now I'm lusting after this bdi2000 and the 3000 across from me Aug 17 11:28:45 tzanger, OpenEmbedded manual specifies that dash doesn't cut it if it's symlinked to /bin/sh, for building OE Aug 17 11:28:59 maybe that's where you saw it? Aug 17 11:29:01 tzanger: the bss/data/heap requirements for bash are certainly not negligible, there seems to be several hundred kilobytes there judging from my machine. Aug 17 11:29:20 th1: heh, yeah, that is a longstanding piece of oe suckage. Aug 17 11:29:33 rather like the various weird restrictions on where you can put your build tree and what you can call it. Aug 17 11:29:33 th1_: coudl be yes Aug 17 11:30:34 pb__, well compared to Buildroot it's pretty unrestricted Aug 17 11:30:39 pb__: the most fun one is putting it in /usr/src :-) Aug 17 11:32:53 what do I have to put in local.conf, to have my dev pc act as a package repository for deployed images? e.g. http://mymachine.intra/oe/ Aug 17 11:34:52 th1_: see recipes/meta/distro-feed-configs.bb? Aug 17 11:35:14 nice thx Aug 17 11:39:31 hello Aug 17 11:41:55 can someone please let me know how to compile a specific application with debugging information. I'm building the default console-image but would like to enable the -g switch in the "ls" utility. Aug 17 11:45:22 this seems like a simple task but i cant find it documented anywhere Aug 17 11:57:09 delayen: i think you just have to set DEBUG_BUILD = "1" in your conf/local.conf, then rebuild your pkg Aug 17 12:07:08 thanks msmith. I just checked that the poky handbook also mentions the same method. Aug 17 12:20:34 delayen, I'd guess you could do DEBUG_BUILD_packagename = "1" but I'm not sure if it would work Aug 17 12:32:37 03Felix Domke  07org.openembedded.dreambox * r066d7cb132 10openembedded.git/packages/enigma2/enigma2.bb: enigma2.bb: create meta- for plugin metadata Aug 17 12:40:58 /join #ocaml Aug 17 12:51:04 I'd like to build the python bindings for farsight2 and telepathy-farsight but they're packaged in the main farsight2 and telepathy-farsight releases (and the recipies currently add --disable-python to them), can anyone point me at an existing package that works in the way and produces a separate package for the python bindings? Aug 17 12:51:21 Or is the only option to --enable-python and have the bindings bundled in with the main packages? Aug 17 12:51:43 s/the way/that way/ Aug 17 13:01:19 I'm compiling a qt application using openembedded. I've created a .bb file and just bitbaking it and it generates .ipk file. How to generate .tar.bz2 file too? Aug 17 13:01:39 What should I add to my .bb file to do this? Aug 17 13:02:40 sx: nothing to your .bb file, tar.bz2 would be another packaging option and I think it is currently broken... Aug 17 13:02:55 sx: INHERIT += package_tar in your local.conf should do it... Aug 17 13:04:14 zecke, thanks, I was trying to remember how to do that :) Aug 17 13:07:12 zecke, thanks, I'll try it now Aug 17 13:11:12 will a patch which adds interactive-shell to alternatives, uses it in /etc/passwd, and adds busybox and bash in increasing priority order be accepted? Aug 17 13:12:06 sx: remove the IMAGE_PKGTYPE line from classes/package_tar.bbclass Aug 17 13:12:36 th1_: so long as it's arranged in such a way as to not break existing installations, and with the appropriate dependency constraints between the different packages, yes, I can't see why not. Aug 17 13:12:47 ok :) Aug 17 13:15:29 zecke, it is INHERIT += package_tar or INHERIT += " package_tar " ? which one? Aug 17 13:15:55 1/24 Aug 17 13:15:56 oops Aug 17 13:16:44 sx: haeh? INHERIT+= "package_tae" in the local.conf and remove the IMAGE_PKGTYPE line in classes/package.bbclass Aug 17 13:18:27 Crofton|work: acked-by: you for removing IMAGE_PKGTYPE ?= "tar" from classes/package.bbclass, as rootfs_tar.bbclass does not exist and inherit image.bbclass will fail.. Aug 17 13:27:35 morning Aug 17 13:33:06 03Roman I Khimov  07org.openembedded.dev * r2dd3dd7304 10openembedded.git/site/ix86-common: Aug 17 13:33:06 site/ix86-common: fix grave bug in ac_cv_sizeof_unsigned_char Aug 17 13:33:06 ac_cv_sizeof_unsigned_char definition accidentally redefines Aug 17 13:33:06 ac_cv_sizeof_unsigned_int to 1. Luckily, not many packages are using that Aug 17 13:33:06 but for example it completely breakes gnutls. Aug 17 13:33:08 Fix that. Aug 17 13:33:10 Acked-by: Holger Freyther Aug 17 13:43:19 morning Aug 17 13:44:30 to do a rebuild I delete the tmp/ dir but this rebuilds the cross compiler everytime, is there a way to do a complete rebuild without having to rebuild the cross compiler? Aug 17 13:46:38 mpoullet: using bitbake -cclean on the files you don't want? Aug 17 13:47:28 mpoullet: you could write a cleanall target for base.bbclass that will recursively clean but stop at the toolchain Aug 17 13:47:29 hi zecke, no sometimes I'd like to rebuild everything but the cross compiler Aug 17 13:48:21 zecke: interesting, how can I determine which package is the toolchain, aren't there multiple packages for it? Aug 17 13:54:46 Hi All. What is the best way to go about submitting a large set of recipes etc for a new board? Split it up into lots of small patches over several days or post big set to mailing list in one go? Aug 17 13:56:50 mpoullet: mostly not... virtual/libc and virtual/gcc-for-${TA.. is the toolchain (see base.bbclass were these depdencies get injected) Aug 17 13:58:41 zecke: ok, thanks for the infos Aug 17 14:01:25 jic23, in a developer branch and post a link to the list? Aug 17 14:03:35 jic23: if you have patches for existing files, post them to the list individually. Aug 17 14:03:49 if you're just adding new files, you can probably post them in a big batch or a separate git branch. Aug 17 14:07:03 does bitbake have an equivelant to debian's .install files, so I can list what files I want included in what packages? Aug 17 14:08:42 Thanks, I'll probably try and get the new packages out the way first in seperate posts then do the board specific stuff later. Aug 17 14:18:15 Elleo: Look at FILES_* Aug 17 14:18:28 Elleo: http://docs.openembedded.org Aug 17 14:18:39 The manual should have the information you need Aug 17 14:20:19 Laibsch: yeah, found it thanks Aug 17 14:20:27 dtnrg seems to be a good example of what I want to do Aug 17 14:26:45 can you please update git repository ? Aug 17 14:26:46 http://www.angstrom-distribution.org/unstable/sources/git_git.denx.de.u-boot.git.tar.gz Aug 17 14:26:53 this link is invalid Aug 17 14:31:42 03John Willis  07org.openembedded.dev * r48011a79e2 10openembedded.git/recipes/gamin/ (files/configure_ucred.patch gamin_0.1.8.bb): Aug 17 14:31:42 gamin-0.1.8: Fix build issue with recent GCC/GLibC combos. Aug 17 14:31:42 Update gamin_0.1.8.bb to build with recent GCC and GLibC 2.8 > via addition Aug 17 14:31:42 of configure.in patch (use AC_USE_SYSTEM_EXTENSIONS instead of Aug 17 14:31:44 AC_ISC_POSIX). Aug 17 14:31:46 Signed-off-by: David-John Willis Aug 17 14:31:48 Signed-off-by: Holger Freyther Aug 17 14:31:50 03Holger Hans Peter Freyther  07org.openembedded.dev * r0907be2124 10openembedded.git/site/ (mipsel-linux mipsel-linux-uclibc): Aug 17 14:31:53 site: Add ac value for mipsel-linux and mipsel-linux-uclibc Aug 17 14:31:55 The results were found using google and applying Aug 17 14:31:57 common sense. Aug 17 15:32:22 how can I export my packages from deploy into a repository that opkg on the target expects? I use the angstrom distro and it's looking for e.g. "ipkg/glibc/i686/{base|debug|gstreamer|perl|python}/Packages" etc. but my deploy dir only has all packages in one level, not split like that Aug 17 15:32:49 or must I provide a new angstrom-feeds-config.bb to override that behaviour? Aug 17 16:16:58 03Koen Kooi  07org.openembedded.dev * rc661fd43a2 10openembedded.git/ (6 files in 5 dirs): linux-omap* 2.6.29: add patch that should solve the memhole problem Aug 17 16:18:36 hi,is there a good calculator that is like shr's calc Aug 17 16:18:41 it's for testing xorg Aug 17 16:18:48 maybe xcalc Aug 17 16:45:20 03Khem Raj  07org.openembedded.dev * r3de2a694e2 10openembedded.git/recipes/python/python-sip_4.8.2.bb: Aug 17 16:45:20 python-sip_4.8.2.bb: set DEFAULT_PREFERENCE = -1 Aug 17 16:45:20 Signed-off-by: Khem Raj Aug 17 16:59:29 khem: I haven't tested yet, but I'm kind of suspicious the problem will be fixed when -1 is set Aug 17 18:09:24 anyone had any issues building openssl-native on x86_64? Aug 17 18:14:50 when does it build? Aug 17 18:17:33 git-native depends on it... only for IMAPS, I believe, which is overkill Aug 17 18:18:23 either way, it builds fine on x86 32bit, but throws some weird register-related issues on x86_64... Aug 17 18:21:23 looks like it builds on F11 which has gcc4.4 Aug 17 18:22:39 ok, thanks Aug 17 18:25:10 denix, it builds no problem on ubuntu 9.04 x86_64 Aug 17 18:26:39 th1_: thanks! I guess it maybe the case of an old distro/gcc - need to check that with the customer who's reported this issue... Aug 17 18:47:17 denix, damn customers :) Aug 17 18:47:25 :) Aug 17 19:02:11 denix, seriously, customers feeding back problems are great for us Aug 17 19:16:10 Crofton|work: unfortunately, their feedback is usually not detailed - need to "interrogate" them for every piece of information... :) Aug 17 19:18:59 yeah, I am familiar with customers :) Aug 17 20:34:25 03Fraxinas  07org.openembedded.dreambox * rd2fd5f61e6 10openembedded.git/packages/gstreamer/ (files/playsink-textsink.diff gst-plugins-base_0.10.24.bb): gstreamer: add patch to fix bug #591664 (srt subtitle resync issue after seeking) Aug 17 21:12:22 ant__: Hi Aug 17 21:12:30 I have some kexecboot questions Aug 17 21:12:35 #kexecboot? Aug 17 21:12:41 hey Laibsch Aug 17 21:13:14 np Aug 17 21:14:43 ant__: do you build on a 32bit or 64bit host? Aug 17 21:31:57 03Felix Domke  07org.openembedded.dreambox * r25c8bb45fd 10openembedded.git/: Merge branch 'org.openembedded.dreambox' of git://git.opendreambox.org/git/ghost/opendreambox-1.5 into org.openembedded.dreambox Aug 17 21:32:05 03ghost  07org.openembedded.dreambox * r6ebd5d5704 10openembedded.git/packages/enigma/ (enigma/fix_xml_defaults.diff enigma_cvs.bb): enigma_cvs.bb: backport needed fix for current terrestrial.xml from tuxbox-cvs Aug 17 23:30:34 Is anyone successfully using connman-gnome? If so what versions of connman and connman-gnome have you chosen? Aug 18 00:32:19 bitbake wpa-gui pulls in postgres and mysql! wtf? **** ENDING LOGGING AT Tue Aug 18 02:59:56 2009