**** BEGIN LOGGING AT Fri Jan 11 02:59:58 2013 Jan 11 09:01:33 Hello. I created two recipes for two packages A and B. The package A depends on some binaries installed in /usr/bin/ so I added RDEPENDS="B" to a recipe... but It seems that B is not automatically installed to my rootfs. Am I missing something ? Jan 11 09:02:59 morning all Jan 11 09:26:20 hi Jan 11 09:26:51 vadmeste: RDEPENDS_${PN} = "b" Jan 11 09:29:05 hrw: thanks, I already fetched other bb files and tested your solution and it works like a charm.. but can you tell me where can I find this kind of information ? in bitbake or open embedded manual ? Jan 11 09:29:38 vadmeste: hard to tell cause I did not read OE manual for ages Jan 11 09:31:19 okay.. because it seems to be easier to read bb files to learn than fetching it in google ^^ Jan 11 09:31:35 s/it// Jan 11 09:32:38 ;D Jan 11 10:11:13 hi all, does anyone know what I have to do to get a recipe to add some files to the sysroot which I want to use in a later recipe? Jan 11 10:25:28 iehhhkss glib-2.0-ptest breaks uclibc builds Jan 11 10:44:49 woglinde: hello there Jan 11 10:55:07 good morning Jan 11 10:57:17 I'm facing to an unexpected behaviour with oe-core. I was used to use oe-classic creating a toolchain with meta-toolchain and using it to build the same kernel oe built with bitbake, but out of oe. Jan 11 10:58:11 now with oe-core this is no longer possible due to compilation errors "arm-oe-linux-gnueabi-ld: unrecognized option '-Wl,-O1'", why? Jan 11 10:59:21 I wonder why oe-core doesn't create a toolchain working like the one it uses inside bitbake process Jan 11 11:01:20 basically oe-classic generated gcc-4.5.3 and oe-core gcc-4.7.2 Jan 11 11:02:37 CROSS_COMPILE=arm-oe-linux-gnueabi-gcc is not able to build kernel 2.6.30 neither 3.4.21 I was testing Jan 11 11:03:18 hm.. -Wl is for gcc to pass options to th elinker...that line is wrong Jan 11 11:04:09 and its CROSS_COMPILE=arm-oe-linux-gnueabi- Jan 11 11:04:28 hi mckoan Jan 11 11:04:32 (remove the trailing gcc) Jan 11 11:04:38 mckoan linker problem gcc vs ld Jan 11 11:05:02 -Wl,-01 is option for gcc to pass down to ld Jan 11 11:05:21 this the log of my test http://pastebin.com/2KJNgV3J Jan 11 11:05:56 bencoh: yes actually I set export CROSS_COMPILE=arm-oe-linux-gnueabi- ;-) Jan 11 11:06:37 check your CFLAGS ? Jan 11 11:07:59 no its LDFLAGS Jan 11 11:08:08 ld is called directly Jan 11 11:08:30 change the make call in recipe to V=1 to see more output Jan 11 11:09:37 hm I would like to know why bitbke ignores my PREFFRED_PROVIDER settings in variuos places Jan 11 11:11:27 woglinde: LDFLAGS are set by meta-toolchain Jan 11 11:11:50 LDFLAGS=-Wl,-O1 -Wl,--hash-style=gnu -Wl,--as-needed Jan 11 11:12:10 than set LD to gcc Jan 11 11:12:40 woops, LDFLAGS, true Jan 11 11:12:46 LD=arm-oe-linux-gnueabi-ld --sysroot=/opt/oecore-x86_64/sysroots/armv5te-oe-linux-gnueabi Jan 11 11:13:09 thats wring than Jan 11 11:13:26 try to set it to gcc Jan 11 11:13:52 woglinde: so is it a bug in meta-toolchain? Jan 11 11:14:37 mckoan hm depends Jan 11 11:15:39 LD=arm-oe-linux-gnueabi-gcc --sysroot=/opt/oecore-x86_64/sysroots/armv5te-oe-linux-gnueabi Jan 11 11:16:01 still the same though :-( arm-oe-linux-gnueabi-ld: unrecognized option '-Wl,-O1' Jan 11 11:16:12 its kernel Jan 11 11:16:21 check the run. file Jan 11 11:16:36 and the kernel Makefile Jan 11 11:17:14 woglinde: but why the kernel build inside oe/bitbake works? Jan 11 11:18:34 mckoan I dont know in my crystal ball who is giving the false setting to the kernel build Jan 11 11:19:07 see even Jan 11 11:19:12 woglinde: I understand :-) Jan 11 11:19:47 so I argue that none of you use a toolchain generated by oe-core to build a kernel Jan 11 11:20:41 in this case, someone should have already faced the same problem Jan 11 11:21:06 mckoan: whiuch branch of OE-core are you using ? Jan 11 11:21:24 it's a shame, because it was a nice feature to oe-classic Jan 11 11:22:01 ericben: master Jan 11 11:23:41 mckoan: LDFLAGS="" CROSS_COMPILE=arm-oe-linux-gnueabi- ARCH=arm make -j9 works here Jan 11 11:23:52 same here Jan 11 11:24:38 ericben: LDFLAGS="" solved the issue here too, thanks Jan 11 11:25:02 mckoan: the SDK is targetting application development so defaults flags are provided for application and can create troubles with the kernel as you can see here ;-) Jan 11 11:25:42 ericben: thank you for this clarification too, it is really precious :-) Jan 11 11:28:24 meta/classes/kernel.bbclass:87 : unset CFLAGS CPPFLAGS CXXFLAGS LDFLAGS MACHINE Jan 11 11:28:52 (the rest of this file might interest you) Jan 11 11:29:23 CXXFLAGS is funny Jan 11 11:29:31 yeah :) Jan 11 11:30:28 our hardware switch crashed an hour ago.. i was just asking: hi all, does anyone know what I have to do to get a recipe to add some files to the sysroot which I want to use in a later recipe? So, package A creates some file F, and package B needs F in its sysroot. At the moment, I added DEPENDS = "A" in the recipe of B, but the file F does not appear in the sysroot -- how do I mark the files to be added to sysroot? Jan 11 11:31:02 bencoh: interesting line ;-) thank you Jan 11 11:31:56 hi everybody! I've got a strange issue on booting and i.MX6 board. I have bootloader, kernel and filesystem, but i'm not able to get a serial console. Jan 11 11:32:13 this is boot log: http://pastebin.com/CcsdCHEf Jan 11 11:32:26 (the hwclock error is not fatal) Jan 11 11:33:16 I've checked my /etc/inittab and set checked the presence of this line: "S:2345:respawn:/sbin/getty 115200 ttymxc1" Jan 11 11:33:53 it is actually present but I still fail to get a serial console login. Anything that anyone can suggest? Jan 11 11:33:57 panda84kde check thje topic Jan 11 11:34:25 ok, woglinde. sorry. moving to #yocto Jan 11 12:09:35 someone has finally discovered what's happening with patchwork for oe? Jan 11 12:14:41 hi Jan 11 13:09:56 khem ping when you are awake Jan 11 13:11:04 hi likewise and hi ensc Jan 11 13:14:23 woglinde: while you're there, do you think I can safely add the linker --fix-v4bx patch to linux-uclibceabi builds as for gnueabi? Here it is for klibc recipes. Jan 11 13:14:53 ant I saw your patch Jan 11 13:15:07 I think so Jan 11 13:15:18 just try it with qemu build Jan 11 13:17:28 yes, better to test, even if objdump is enough in this case Jan 11 13:58:37 isnt pkg_prerm() not called while updating? Jan 11 14:36:30 devzero_: from what I know it's not due to missing implementation in opkg Jan 11 14:36:57 devzero_: https://code.google.com/p/opkg/source/browse/trunk/libopkg/opkg_install.c#541 Jan 11 14:38:10 ok thx, i wanna stop a binary before upgrading - whats the best way to do? Jan 11 17:39:24 moin Jan 11 18:35:31 ok, I went with the latest fad Jan 11 18:35:35 moar ram Jan 11 18:36:20 ? Jan 11 18:42:01 fast ram is cheap Jan 11 18:42:13 older DIMMs not so cheap... Jan 11 18:53:34 does PACKAGE_FEATURES exists? Jan 11 18:53:57 nope Jan 11 18:55:25 me and some of the other guys that use gnuradio cant really agree on figure optios and depends :) Jan 11 18:56:22 hm berlios git still down Jan 11 18:58:55 Crofton|work: there's PACKAGECONFIG, however Jan 11 18:59:09 which is basically USE w/ automatic autoconf args and deps/rdeps Jan 11 19:06:28 So I'm looking at sdcard_image.bbclass in meta-ti and I'd like to do something similar for my own board, but I'm unfamiliar with the image type stuff. I copied and modified the class into my own meta-layer and added the new type to IMAGE_FSTYPES in my local.conf, but I just get a symlink to nothing in my images folder. Any ideas? Jan 11 19:06:35 kergoth: is there OE and/or BitBake artwork anywhere? I'd like to put something on the updated BitBake User Manual's title page. Jan 11 19:12:31 jawilson, they are talking about it on the meta-ti list atm Jan 11 19:12:44 wmat, we haev artwork for the oe logo on the wiki Jan 11 19:13:32 I need to look at buying some stickers based on it also Jan 11 19:14:09 jawilson: it's a bad example to go from. it requires /etc/fstab changes on the build machine, etc. look at meta-fsl-arm's sdcard class instead as a starting point, that uses parted instead of losetup, etc Jan 11 19:14:39 beyond that, it sounds like you didn't add it to IMAGE_CLASSES Jan 11 19:14:44 it wont be automagically inherited Jan 11 19:15:08 also, the sector 63 stuff is no longer relevant Jan 11 19:15:41 kergoth, I was wondering how it'd be automagically inherited, but nothing complained so I went full speed ahead. where would I do the IMAGE_CLASSES stuff? Jan 11 19:17:07 Crofton|work, well, I'm copying the existing method for setting up the card Jan 11 19:17:10 grep around. Jan 11 19:17:13 from iveia Jan 11 19:18:24 https://groups.google.com/forum/?fromgroups=#!topic/beagleboard/ro5k5r4Cuq4 Jan 11 19:20:02 Crofton|work: thx, I'll check it out Jan 11 19:20:41 wmat use the new logo Jan 11 19:20:42 Crofton|work: ah, interesting Jan 11 19:20:56 yeah Jan 11 19:21:10 older mlo's needed it Jan 11 19:21:21 but at this point they are several years old Jan 11 19:21:45 I always break fat/root at about 64M to be on an erase block boundary Jan 11 19:21:54 start rootfs on 131072 sector Jan 11 19:22:23 http://git.yoctoproject.org/cgit/cgit.cgi/meta-mentor/tree/classes/image_types_sdcard.bbclass is basically the meta-fsl-arm/meta-ti bits but without the actual per-machine functions to do the work, then http://git.yoctoproject.org/cgit/cgit.cgi/meta-mentor/tree/classes/image_types_mentor.bbclass was my naive attempt that currently results in filesystems too large to fit in the partitions it creates Jan 11 19:22:33 interesting Jan 11 19:22:41 * kergoth will have to revisit this in a future sprint Jan 11 19:24:03 kergoth, ah, well thanks for the reference Jan 11 19:24:44 kergoth, grepping meta-* and oe-core for IMAGE_CLASSES yields no results... Jan 11 19:24:46 seems like there are quite a few of us all playing with similar stuff right now :) i'm looking forward to the image/deployment rework Jan 11 19:25:00 jawilson, read the meta-ti list archiove Jan 11 19:25:35 jawilson: grep fail. it's in image.bbclass in oe-core. Jan 11 19:25:36 multi-tasking will be the end of me, I was headed that way earlier Jan 11 19:25:50 kergoth, I'm an idiot Jan 11 19:26:01 you'd want to define it to something like IMAGE_CLASSES = "image_types my_image_types_class" Jan 11 19:26:50 I understand now Jan 11 19:26:55 thanks Jan 11 19:27:17 np Jan 11 19:30:30 Crofton|work, I find it eerily ironic that they happen to be talking about this on the meta-ti list just as I finally decide to get into it. I've been looking at it for months... Jan 11 19:30:42 heh Jan 11 19:30:44 yeah Jan 11 19:31:05 although I suspect everyone starting out the year with good intentions Jan 11 19:39:58 excellent, now when I test gnuradio builds, I can still surf pr0n Jan 11 19:40:08 20G of ram lets you do that Jan 11 19:40:47 ok, need to start booking flight for ELC and FOSDEAM Jan 11 19:41:55 me will tell croftons wife Jan 11 19:42:29 1MB for gnuradio builds, right? Jan 11 19:43:18 rofl Jan 11 19:43:53 desktop performance became sluggish during compile Jan 11 19:46:34 they are almsot giving ram away now Jan 11 19:48:14 yes Jan 11 19:48:31 but prices will start to climb again soon they claimed toda Jan 11 19:48:32 y Jan 11 20:06:34 Crofton|work: where on the wiki? I can't find any artwork there. Jan 11 20:10:00 hmmm Jan 11 20:11:10 i bet it is in git Jan 11 20:11:53 http://git.openembedded.org/openembedded/tree/contrib/artwork Jan 11 20:28:29 wmat, oe.svg in that directory is what you want Jan 11 21:19:17 Crofton|work: shouldn't that pic open with Inkscape? It doesn't on my machine. Jan 11 21:22:20 it does on mine Jan 11 21:22:28 make sure you do not downoad the html Jan 11 21:23:02 * Crofton|work made that mistake Jan 11 21:31:59 heh, I did Jan 11 21:32:52 ok, I need to find a source for OE stickers Jan 11 21:33:46 bring some to ELC, please ;) Jan 11 21:46:41 will do Jan 11 21:46:48 need to get them on order fast for FOSDEM Jan 11 21:48:52 flight for fosdem booked Jan 11 22:10:56 Crofton|work: ping **** ENDING LOGGING AT Sat Jan 12 02:59:58 2013