**** BEGIN LOGGING AT Tue May 15 02:59:59 2012 May 15 08:10:20 good morning May 15 08:24:04 gm May 15 08:25:02 what are the variables that determine the search path for bitbake include files? I have tried this, but it is not picked up from my layer: FILESEXTRAPATHS_prepend := "${THISDIR}/${PN}:" May 15 08:55:49 likewise: I'd check three things... May 15 08:55:53 1) The order of the layers in BBLAYERS of /conf/bblayers.conf May 15 08:56:43 2, 3) BBPATH and BBFILES of your conf/layer.conf May 15 08:59:02 Sometimes there are recipes that don't use full qualified paths for their included/required files. In this case the wrong *inc might be pulled in May 15 09:02:21 kenws: Tnx. My layer is latest in the list, but with highest priority. I just wonder if the priority matters for include's. May 15 09:04:49 I've tried add *inc to the BBFILES without affect. May 15 09:04:52 BBPATH := "${BBPATH}:${LAYERDIR}" May 15 09:04:54 BBFILES += "${LAYERDIR}/recipes-*/*/*.bb*" May 15 09:04:55 BBFILES += "${LAYERDIR}/recipes-*/*/*.inc*" May 15 09:04:57 BBFILES += "${LAYERDIR}/images/*.bb" May 15 09:05:57 This is the line that is not picking up my overridden include file: require recipes-kernel/linux/linux-qoriq-sdk.inc May 15 09:06:45 My .bb is pickup up correctly (in the same dir in my layer). May 15 09:06:54 hm, it looks like you are appending ${LAYERDIR} to BBPATH May 15 09:07:04 I'd try: BBPATH .= ":${LAYERDIR}" May 15 09:08:05 kenws: All the layers I have use BBPATH := "${BBPATH}:${LAYERDIR}" May 15 09:08:14 hm, ok May 15 09:08:37 which includes the original layers. I'm quite sure it is the correct way (but I forgot the exact details on why). May 15 09:09:09 For the moment I'll use the .bb file to insert my overrides. May 15 09:09:45 I probably can also use a bbappend, I just want to learn why the include is not picked up from my layer. May 15 09:09:55 I'm not sure I can contribute much but here's my setup: http://pastie.org/private/ok7yxgwjrbg4ggftzgxz0a May 15 09:20:37 kenws: thanks. I think my specific case is that I want to override an include file. I could not find an example in any layer doing this (and maybe it is not recommended). May 15 09:21:15 gm bluelightning May 15 09:21:25 hi likewise May 15 09:21:34 hi all May 15 09:22:40 morning bluelightning May 15 09:23:06 likewise: hm, I thought it's possible to shadow *inc file if there are pulled in using their full qualified path May 15 09:23:15 likewise: if you want your .inc your layer should be listed on the top May 15 09:23:56 seems a bit messy and error-prone to try to do things that way though May 15 09:23:57 s/file if there/files if they/ May 15 09:28:09 likewise: an example is linux.inc, present in many layers May 15 09:33:48 linux.inc which could be shared by all is in meta-oe May 15 09:40:48 JaMa|Off: btw I'm ok for removing it now from meta-handheld May 15 09:41:38 if we're not using it for anything atm then I'm for it too May 15 09:42:49 ant_work: I see it in meta-ti, I really wonder if it works correctly there. May 15 09:42:50 http://git.yoctoproject.org/cgit/cgit.cgi/meta-ti/tree/recipes-kernel/linux/linux_3.1.bb May 15 09:43:58 first line says require linux.inc, but does it pick up the local linux.inc just magically? Not in my case May 15 09:45:57 bluelightning: we ought to remove the recipes needing it or we end up depending on meta-oe layer... May 15 09:46:13 bluelightning: = pls port ipaqs to linux-yocto :) May 15 09:47:02 having a BSP layer depending on meta-oe can lead to horros-stories.... May 15 09:47:23 (udev, systemd, ...) May 15 09:48:49 I'm still of the opinion meta-oe needs to have the disruptive items stripped out rather than just generally considering it as evil May 15 09:51:17 bluelightning: I think that requires a proposal on where the disruptive (possibly future) bits should mature first. May 15 09:51:39 which might be a topic branch, or repo, or May 15 09:52:07 likewise: for my preference, not in the same place where additional recipes that I might need are also maintained... May 15 09:52:22 I have proposed this some months ago (well, proposed the split at least) May 15 09:54:23 updating oe-core's udev would be a first step May 15 16:19:42 what's the easiest way to check why a package is pulled in into an image? May 15 16:20:11 bitbake -g or bitbake -g -u depexp are useful May 15 16:21:04 kergoth`, thanks! will try it out May 15 16:56:18 hi everyone May 15 16:56:46 moin May 15 16:57:15 hey May 15 16:58:02 i'm working with ptxdist all time to generate my distros, but i've had the idea to test OE today :) May 15 16:59:17 but, got stucked in some issues , first the philosophy of both tools are different so i wasn't even able to generate nothing May 15 17:01:08 secondly, in ptxdist, we have to generate rule files and customize them to add extra libraries, but apparently, it's more easier here? May 15 17:05:17 is there a way to override a specific file of a package by another overlay... in this case I want to override defconfig from busybox in oe-core May 15 17:06:05 So I would like to put only defconfig in my own meta-XX bit keep recipes and all that from oe-core/meta May 15 17:29:22 mohtadi_, basically you bitbake an image May 15 17:29:34 then you install the image on the target May 15 17:29:50 and then you bitbake a package for instance or a feed or something like that May 15 17:30:00 and you install that on target with opkg install foo May 15 17:30:12 roric, yes there is May 15 17:30:16 it's called bbapped May 15 17:59:57 hi there May 15 18:00:54 mind if I ask a question, I just can't find an answer through the interwebs May 15 18:01:32 I'm looking for a way to get gdb-cross as an automatic dependency on my build, but haven't found the right way yet May 15 18:02:16 I can't add it to my IMAGE_INSTALL, since it's not technically going on the target image May 15 18:03:44 so, bitbake gdb-cross works fine, but if I try to add it anywhere in a *_DEPENDS or similar, bitbake barfs out with May 15 18:03:45 ERROR: Nothing RPROVIDES 'gdb-cross' (but /home/romain/dev/OE/OpenEmbedded-Archos/recipes/tasks/task-base.bb RDEPENDS on or otherwise requires it) May 15 18:03:46 NOTE: Runtime target 'gdb-cross' is unbuildable, removing... May 15 18:03:46 Missing or unbuildable dependency chain was: ['gdb-cross'] May 15 18:04:56 for the context, I'm using a fairly old stable/2009 branch, pre- oe-core and layering and all that May 15 18:05:38 thx for any tip or advice! May 15 18:15:01 you're mixing up build time and urn time dependencies. *packages* get added to image_install and rdepends, not recipes. May 15 18:16:07 you could create an image recipe, then in the image do something like: do_rootfs[depends] += "gdb-cross:do_build" May 15 18:19:21 kergoth`: ic, will try that th May 15 18:19:22 thx May 15 18:19:50 yeah I figured that I didn't really want a package May 15 18:20:35 actually right now I've just tried adding a PROVIDES="virtual/${TARGET_PREFIX}gdb" to gdb-cross.inc, then add that to DEPENDS on my image recipe May 15 18:20:49 that should work too, right? I'm wondering if it's more correct May 15 18:21:30 I'm thinking that this provides should actually be in there, but is missing in my (old) branch May 15 18:22:04 DEPENDS generally hooks up populate_sysroot to do_configure. if your image runs populate_sysroot, then that'd work, otherwise it won't do much, afaik May 15 18:22:27 yep, fortunately my image does May 15 18:22:45 then yep, should be fine :) May 15 18:22:48 yay :) May 15 18:27:36 yes that worked, woohoo :) May 15 18:27:41 kergoth`: thx again May 15 18:28:26 np May 15 19:34:47 khem: how's the mips64 stuff coming along? May 15 20:21:39 kergoth`: pretty good May 15 20:21:52 kergoth`: I have a branch which you can try out May 15 20:22:25 kergoth`: I am able to boot console-image as well as sato image using xorg xserver on qemu May 15 20:23:08 git.openembedded.org/openembedded-core-contrib kraj/misc May 15 20:23:41 I would like someone to try it out but you need to use linux-yocto-dev kernel May 15 20:23:58 which has recipes in poky-extra May 15 20:32:08 ah, k, will try May 15 20:34:20 kergoth`: cool thx May 15 20:34:51 kergoth`: eventually we will add multilib to it but for now its pure 64bit abi (n64) only May 15 20:35:20 kergoth`: and also look at the patches and any feedback you would provide will be very valuable May 15 20:35:33 since I am planning on proposing it for OE-Core soon May 15 20:43:23 k. sounds good May 15 21:14:28 GNUtoo-desktop, thanks for the bbappend hint :) May 15 21:15:05 np May 15 21:15:32 GNUtoo-desktop: you mentiones you had issues likining firefox with gold May 15 21:15:40 do you still have a way to repoduce that May 15 21:15:49 or affirm it somehow ? May 15 21:16:01 GNUtoo-desktop: I am willing to understand and fix it if need be May 15 21:16:09 khem, yes it still segfault May 15 21:16:15 khem, wow thanks a lot May 15 21:16:29 khem, I bugreported on the meta-firefox May 15 21:16:38 GNUtoo-desktop: I read that May 15 21:16:46 but it did not give me enough info May 15 21:16:55 so I need a lot more info May 15 21:17:02 e.g. how was it compiled May 15 21:17:06 what kind of infos? May 15 21:17:12 ah, using SHR May 15 21:17:13 was it using thumb1 oor thumb2 or arm ISA May 15 21:17:15 let me find a log May 15 21:17:20 ok shr May 15 21:17:25 what was the device May 15 21:17:31 all armv7 May 15 21:17:34 gta04 May 15 21:17:39 n900 May 15 21:17:39 ok armv7 May 15 21:17:50 and was it using t2 or t1 or arm isa May 15 21:17:56 how to check? May 15 21:18:03 I use rm_old_work May 15 21:18:08 so I've full build kept May 15 21:19:24 ok so if you can provide access to a running box where segfault is seen I can do live debugging May 15 21:19:44 or you can post core dumps May 15 21:19:51 and the binaries involved May 15 21:20:23 arm-oe-linux-gnueabi-g++ -march=armv7-a -fno-tree-vectorize -mthumb-interwork -mfloat-abi=softfp -mfpu=neon -mtune=cortex-a8 May 15 21:20:32 do you have an armv7 device? May 15 21:20:43 else I can provide ssh to an armv7 device if you want May 15 21:21:09 let me look if I've images for the gta04 May 15 21:21:54 crespo runs replicant too so not a good idea since it has some personal infos but with gta04 no issues and it has 512M of ram May 15 21:22:52 ssh to an armv7 phone is enough May 15 21:22:59 or do you need a box attached to it too? May 15 21:23:32 because I can forward you ssh May 15 21:24:31 JaMa|Off, hi May 15 21:26:09 GNUtoo-desktop: ssh to device is better May 15 21:26:20 I do have armv7 devices but they are busy doing stuff May 15 21:26:21 device beeing armv7 May 15 21:26:27 ok May 15 21:26:35 and it takes a while to reproduce May 15 21:26:37 because I've no issues at all giving ssh to the gta04 May 15 21:26:56 just ondevice is good May 15 21:27:00 ok May 15 21:27:03 if image is having debug info May 15 21:27:08 and tools on it May 15 21:27:10 ok May 15 21:27:16 I'll prepare an image May 15 21:27:23 I'll take an official SHR image May 15 21:27:27 so add tools-sdk to the image and also debug stuff May 15 21:27:29 k May 15 21:27:37 ah sdk? May 15 21:27:48 shr-feeds is not enough? May 15 21:27:55 sdk should run on target ? May 15 21:28:00 or on build machine? May 15 21:28:12 because there are stuff like gcc-symlinks which runs on target May 15 21:28:17 or even binutils May 15 21:29:01 on target May 15 21:29:18 if you add tools-sdk to target it should do that May 15 21:30:34 ok May 15 21:39:42 GNUtoo-desktop, another bbappend-question, can it be used to modify .inc-files in some way? May 15 21:39:57 no idea about .inc May 15 21:40:10 but since .inc are included.... May 15 21:40:20 you modify the recipe itself and it should work May 15 21:41:29 GNUtoo-desktop, but the problem for me is that the .inc's does a lot of things and extends DEPENDS for instance May 15 21:41:41 (It's Qt4 I want to make smaller) May 15 21:42:56 roric, do you need to target multiples recipes? May 15 21:43:54 GNUtoo-desktop, I build Qt4 embedded, which in several steps requires some .incs, which adds on dependencies which I don't need May 15 21:44:16 so I would like to extend the different .inc's to tune what they do May 15 21:49:59 Time to leave. GNUtoo-desktop, thanks for the help! May 15 21:58:21 khem, tools-sdk is not in the feeds, is it a meta-package? May 15 21:58:33 because we have compilers etc.... May 15 22:02:29 GNUtoo-desktop: oh yes May 15 22:02:37 if you are going by feeds May 15 22:02:49 then install gcc-symlinks and binutils-symlinks May 15 22:02:53 and also gdb May 15 22:02:56 ok May 15 22:03:04 I'm installing firefox-dbg right now May 15 22:03:05 and all -dbg packagaes May 15 22:03:10 ok May 15 22:03:16 that should be enough May 15 22:03:19 to start with May 15 22:07:41 also I guess you'll also need gdbserver May 15 22:07:53 because 512M of ram is not enough.... May 15 22:08:02 for such huge debugging symbols May 15 22:08:20 even 1G is not enough May 15 22:14:58 GNUtoo-desktop: hmmm if I run gdbserver then I will need a host May 15 22:15:08 ah ok May 15 22:15:09 so I will try to run it locally on device May 15 22:15:16 and see if I get to see something May 15 22:15:26 if not then we might need to have some more solution May 15 22:15:37 else an imx qsb with 1GB would be better? May 15 22:16:13 anyway you don't need *all* the debug symbols May 15 22:16:17 GNUtoo-desktop: yep better machine May 15 22:16:21 is always better May 15 22:16:23 :) May 15 22:16:37 the problem with qsb is that there are no feeds May 15 22:16:44 GNUtoo-desktop: mostly I need firefox and may be libc May 15 22:16:45 thatsd it May 15 22:16:49 ok May 15 22:17:02 hmmm May 15 22:17:04 * opkg_download: Failed to download http://build.shr-project.org/shr-core-staging/056/ipk/armv7a-vfp-neon/firefox-dbg_10.0.2-r2_armv7a-vfp-neon.ipk, wget returned 1. May 15 22:17:36 let me try on my desktop May 15 22:58:01 khem, with gdb it OOMs May 15 22:58:10 :( May 15 22:58:10 only with firefox-dbg May 15 22:58:26 did you try epiphanu May 15 22:58:30 epiphany May 15 22:58:45 now there is midori that works again May 15 22:58:51 k May 15 22:58:57 but I would really loved to see firefox fixed May 15 22:59:03 it's a really good browser May 15 22:59:09 I think firefox mobile is prolly better suited for our space than regular ff May 15 22:59:26 regular firefox works so well.... May 15 22:59:31 it's not like some years ago May 15 22:59:38 my problem is none of my boards have displays May 15 22:59:43 it fits well the screen etc... May 15 22:59:44 ah ok May 15 22:59:49 so running a GUI app needs more setup on my side May 15 23:00:02 ssh -Y? May 15 23:00:09 secondly if the setup is there its easy and quicker to diagnose problem May 15 23:00:40 khem, in how much time can you diagnose the problem? May 15 23:00:43 I can try that but images are not running gui at all May 15 23:00:47 thats why I wanted something to look at May 15 23:01:01 ok May 15 23:01:10 GNUtoo-desktop: hard to say :) it can be a day May 15 23:01:30 because I can else give you access to a machine + a target May 15 23:01:54 and the machine is in my room currently May 15 23:02:04 so when I go to sleep.....it has to be off May 15 23:02:04 ok. May 15 23:02:12 oh May 15 23:02:28 may be tomorrow May 15 23:02:30 I can move it tough May 15 23:02:37 yes May 15 23:02:42 else there is another way May 15 23:02:51 I know how to setup gdb+gdbserver May 15 23:03:01 I am in west coast so there is time difference involved too May 15 23:03:07 you just tell me what to do then May 15 23:03:07 ok May 15 23:03:38 let me see. May 15 23:03:56 I'll fetch my laptop May 15 23:03:59 I am going to build an angstrom image which includes firefox May 15 23:04:11 with gold on armv5te + qemu May 15 23:04:18 and see if I can reproduce it on qemu May 15 23:04:26 ok May 15 23:04:47 if you have time now, I fetch my laptop, setup gdb+gdbserver May 15 23:12:45 khem, 0xb6fd8d10 in _start () from /home/gnutoo/sshfs/lib/ld-linux.so.3 May 15 23:14:07 should I try to look at the bugreport May 15 23:14:15 and then go right before it crash May 15 23:14:41 GNUtoo: ok is it running under gdb ? May 15 23:14:47 yes May 15 23:14:51 gdb+gdbserver May 15 23:14:57 GNUtoo: ok let it run and wait till it dies May 15 23:16:18 ok May 15 23:16:57 ok it died May 15 23:17:29 http://www.pastie.org/private/iijigtlerfmwzifpjgjn6g May 15 23:18:46 khem, do you need bt or info threads May 15 23:18:46 ? May 15 23:18:52 GNUtoo: no May 15 23:18:54 ok May 15 23:19:03 GNUtoo: I need to see the registers May 15 23:19:03 disass something I guess May 15 23:19:05 so info reg May 15 23:19:05 ok May 15 23:19:25 and disassemble around the illegal intruction May 15 23:19:26 http://www.pastie.org/private/hk1ap58utgyoqi7uvyajq May 15 23:19:49 what disass command should I use exactly? May 15 23:20:16 disassemble $pc-20, $pc+20 May 15 23:20:18 e.g. May 15 23:20:19 last time we went trough it we found out that it was a way to switch to thumb which was not correct if I remember well May 15 23:20:20 ok May 15 23:20:38 yes but why is it switching into thumb is my question May 15 23:20:42 http://www.pastie.org/private/oufzoieq9b7m9ndscnrqw May 15 23:20:45 ok May 15 23:21:01 unless there are bits in firefox which are always compiled in thumb this should not happen May 15 23:21:11 ok May 15 23:21:29 part of the trampoline for javascript is in assembly if I remember well May 15 23:21:30 hmmm May 15 23:21:39 and that's exactly what we are dealing with May 15 23:21:55 that code seems to be bogus for arm mode May 15 23:22:03 ok do one thing May 15 23:22:24 cherry pick this patch http://git.openembedded.org/openembedded-core-contrib/commit/?h=kraj/gold&id=b72f3238ad32ab420306a0226d8c2e57c52ddf45 May 15 23:22:45 also I found out recently that firefox can work in safe mode if I disable every extensions and javascript included May 15 23:22:46 this will need your binutils-cross to be recompiled May 15 23:22:52 ok May 15 23:23:01 and then rebuild/relink firefox with new gold May 15 23:23:06 and see if it helps May 15 23:23:13 ok I do it right now May 15 23:23:24 only binutils cross and firefox then May 15 23:23:25 nice May 15 23:23:28 since this seems to me the problem is in handling thumb trampolines May 15 23:23:33 yes May 15 23:23:47 I will drive home now May 15 23:23:54 but will checkback in 1 hr or so May 15 23:23:58 nice May 15 23:24:00 thanks a lot May 15 23:24:10 so if you have some info for me dump it here May 15 23:24:18 I will read it later May 15 23:24:20 ok May 15 23:24:23 if it worked or not May 15 23:24:32 thanks a lot!!!! May 15 23:24:42 thank me if it works :) May 15 23:25:03 and if it does then I will propose it to be pulled into OE-core May 15 23:25:06 yes, but thanks for the help even if it doesn't work yet May 15 23:25:13 ok nice May 15 23:25:14 np May 16 01:14:36 khem, hi May 16 01:15:32 khem, http://www.pastie.org/private/m6pic0222et9gn8fct3hzw May 16 01:15:44 and it seem right at the beginning May 16 01:15:51 even before the previous issue May 16 01:15:59 (which was trampoline related) May 16 01:16:46 for instance: May 16 01:16:48 ~# firefox --help May 16 01:16:48 Segmentation fault **** ENDING LOGGING AT Wed May 16 03:00:01 2012