**** BEGIN LOGGING AT Fri Mar 08 02:59:59 2013 Mar 08 09:30:33 good morning Mar 08 09:42:49 hello Mar 08 09:44:06 what are the options to givee to cross-compile openssl Mar 08 09:54:16 oe-core_: hi. Do you have poky toolchain installed? Mar 08 09:54:51 no i'm using oe-core toolchain itself Mar 08 09:55:19 but i want openssl to build for qemux86 Mar 08 09:55:37 i'm not cloning the code but using my code Mar 08 09:56:14 Mmm. Are you using Yocto? Mar 08 09:59:25 no oe-core build system Mar 08 10:01:11 panda84kde: do u have any idea? Mar 08 10:03:23 I've got some experience with http://www.yoctoproject.org/docs/1.3/adt-manual/adt-manual.html but don't know at what you're referring to Mar 08 10:03:49 oe-core_: oe-core is part of yocto Mar 08 10:05:14 when i'm building opensssl, first it builds native, there it configures for native os(x86-64-linux mine), during that it gives compilation failure Mar 08 10:05:21 oe-core_: to sum it up: if you want to build outside Yocto build system you should use the sdk. Mar 08 10:05:35 saying ../libcrypto.so: undefined reference to `mod_exp_512' Mar 08 10:05:51 oe-core_: what version of oe-core/bitbake? Mar 08 10:06:46 hm why is it building a native openssl? Mar 08 10:09:37 BBCLASSEXTEND = native option is specified in bb file Mar 08 10:10:12 rpm-native has dependency on openssl-native Mar 08 10:11:07 even if it is building native no issue, but it is giving compilation error Mar 08 10:11:23 ah, rpm is pulling it in, fine. Mar 08 10:11:29 so what version of oe-core? Mar 08 10:13:56 whatersion of oe-core? i did not get u? Mar 08 10:15:13 why it is giving compilation error? Mar 08 10:17:23 yes, what version of oe-core Mar 08 10:17:31 i'm trying to find out what you're building Mar 08 10:19:28 i'm trying to build openssl in oe-core(build system) but i'm not cloning the code Mar 08 10:20:16 whose path is specified in SRC_URI because i want to build with the code which i have Mar 08 10:20:50 there when i'm building i get compilation error as specified above Mar 08 10:22:23 eh? Mar 08 10:22:46 you're swapping the oe-core source for your own openssl source tree? Mar 08 10:22:59 yep Mar 08 10:23:06 are you also ensuring all those patches we apply are being applied? Mar 08 10:24:01 if you've just got a new feature or something, it's best to use a patch Mar 08 10:24:18 but i can't really help you with compilation of sources that are not part of oe, or i can't see Mar 08 10:24:37 even after applying all the patches its failing Mar 08 10:26:26 does the proper source uri build fine? if that does, then it's your changes that are breaking it. Mar 08 10:27:48 k Mar 08 10:28:21 so u tell there might be problem with the changes in the source? Mar 08 10:29:18 i'm saying you should verify that the sources urls as we provide them work before trying your modifications Mar 08 10:29:32 if a pristine oe-core doesn't build, then someone can help Mar 08 10:29:37 ya its building file Mar 08 10:29:43 fine Mar 08 10:30:03 so it's your changes that are breaking the compile in some form Mar 08 10:33:08 k thanks i'll check Mar 08 10:40:49 rburton:Hi Ross, could you please have a look at the following changes in neard recipe ? http://pastebin.com/jLHAwHhx and http://pastebin.com/0E5sZPXJ Does it make sense to you ? Mar 08 10:41:34 OlivierG: yes, that Mar 08 10:41:46 why do you need oe_runmake DESTDIR=${D} libexecdir=${libexecdir} install? Mar 08 10:41:52 surely that's implied Mar 08 10:42:03 as we've already passed libexecdir at configure time Mar 08 10:44:01 because libdirexec in Yocto adds /usr/lib/${BPN} as mine, in neard, refers to /usr/lib... Mar 08 10:45:10 "refers to /usr/lib/nfc" ;) Mar 08 10:47:22 you shoud fix neard Mar 08 10:48:22 ;)... well...we had so many discussion with Marcel and Sameo about this point... Mar 08 10:48:52 ha Mar 08 10:49:22 sameo: sounds like you're doing it wrong :) Mar 08 10:50:00 what's wrong with just installing into libexecdir, which defaults to /usr/libexec i think still, but increasingly is being re-set to $libdir/$packagename Mar 08 10:50:19 he is not here (back in few minutes)... but we discussed about this issue this morning... Mar 08 10:50:49 OlivierG: is he coming to osts? Mar 08 10:50:54 yes Mar 08 10:51:11 i'll throw him into the hot tub clothed unless he sorts this out. Mar 08 10:51:34 good luck ;) Mar 08 10:51:37 yeah i may need a team to assist me Mar 08 10:54:01 btw, the idea behind all these changes (according to Marcel) is to have the daemons path as /usr/lib/{package} ... Mar 08 10:54:42 that's a packaging choice Mar 08 10:54:46 so he's Wrong. Mar 08 10:55:02 i won't throw marcel in the hot tub though Mar 08 10:55:18 O:-) so am i.... Mar 08 10:57:05 rburton: Hey...So what's wrong with @libexecdir@/nfc ? Mar 08 10:57:38 well, most sane distributions are moving to --libexecdir=/usr/lib/@package@ Mar 08 10:57:43 so you'll be double-nesting Mar 08 10:59:00 right, which is why we're overriding poky's install for neard. Mar 08 11:02:11 morning all Mar 08 11:03:17 rburton: According to the GNU standards: The definition of ‘libexecdir’ is the same for all packages, so you should install your data in a subdirectory thereof." Mar 08 11:04:28 So we just do that from the neard Makefile directly. Mar 08 11:09:03 sameo: iirc, FHS disagrees with you Mar 08 11:10:42 http://www.pathname.com/fhs/pub/fhs-2.3.html#USRLIBLIBRARIESFORPROGRAMMINGANDPA Mar 08 11:10:57 well, FHS says don't use libexec Mar 08 11:11:19 rburton: I think they do: http://www.linuxbase.org/betaspecs/fhs/fhs/ch04s07.html Mar 08 11:11:43 rburton: But basically FHS says you can add a subdir to libexec for your program, it doesn't say it must be pkgname Mar 08 11:12:22 ha Mar 08 11:12:47 so as more and more distros move away from libexec, FHS approves it. epic. Mar 08 11:12:59 heh :) Mar 08 11:14:47 although more and more distros are moving to /usr/lib/@package@ for libexecdir, this is not something we can take for granted according to the GNU and FHS folks. Mar 08 11:16:53 I like osts hot tub btw. Mar 08 11:17:04 Just not clothed though... Mar 08 11:50:49 hi I was using tcpdump recipe from oe on github but the url mention there seems to be unavailable so I replaced it with another but now its not able to parse Mar 08 12:02:06 its done thanks anyways :) Mar 08 13:35:45 any help for yocto-bsp script around ? Mar 08 14:05:42 howdy folks, i have set "PREFERRED_PROVIDER_virtual/kernel = "linux-emac" in my machine configuration file but yocto still gives me linux-dummy whenever I build using 'bitbake -c build virtual/kernel' (this is also after running 'cleanall' for the same package) Mar 08 14:06:41 we have a custom kernel version driven by the requirement that it be compatible with Xenomai i-pipe patch Mar 08 14:06:47 waynr: perhaps that value is being overridden somewhere else? Mar 08 14:07:13 what does this report: bitbake -e | grep PREFERRED_PROVIDER_virtual/kernel Mar 08 14:08:46 PREFERRED_PROVIDER_virtual/kernel=linux-emac Mar 08 14:09:15 hmm if something were wrong with the recipe that i wrote or if the recipe file weren't found shouldn't that cause an error? Mar 08 14:11:36 waynr: if it was skipped I think it might (e.g. if COMPATIBLE_MACHINE / COMPATIBLE_HOST were not set correctly in the linux-emac recipe) Mar 08 14:11:53 if the recipe wasn't available at all you would get a warning Mar 08 14:12:20 hmm the build attempt does indicate 4 warnings, only 3 of which stuck on the bitbake console output Mar 08 14:14:24 ah nevermind, the fourth warning is "linux-dummy: No generic license file exists..." Mar 08 14:16:33 hmm, I don't see COMPATIBLE_HOST in the yocto documentation, what is that for? Mar 08 14:18:22 maybe my bb file patterns are incorrect for this layer Mar 08 14:27:03 waynr: COMPATIBLE_HOST is similar, it just matches on the full host string e.g. i686-linux Mar 08 14:31:04 okay. looks like my bsp layer's BBFILES was set incorrectly...fixed now, but still not catching that recipe Mar 08 14:31:56 so does bitbake linux-emac work? Mar 08 14:32:33 nothing PROVIDES linux-emac Mar 08 14:32:46 might indicate my problem Mar 08 14:37:32 right, so it's not even seeing the recipe Mar 08 14:37:38 so you still have a problem with your BBFILES Mar 08 14:38:27 well I do see the directory in which this bitbake recipe is found in the 'bitbake -e | grep BBFILES' Mar 08 14:39:08 (I fixed the layer.conf for this bsp (maybe), previously its recipes were not found when running that command) Mar 08 14:41:03 just to double check, what is the BBFILES line in the layer's layer.conf, and what is the path to the recipe? Mar 08 14:43:23 okay BBFILES had ${LAYERDIR}/common/recipes-*/*/*.bb but the recipe was in ${LAYERDIR}/recipes-kernel/linux/linux-emac.bb Mar 08 14:43:46 ah right yes that would do it Mar 08 14:43:48 starting to get messages useful for debugging the actual kernel recipe now Mar 08 14:44:21 thanks for the help bluelightning :) Mar 08 14:44:26 no worries Mar 08 15:09:59 bluelightning: does exixt a doc explaining how does Yocto branch releases works? (1.3_M5, 1.4_M3, ...) Mar 08 15:13:40 mckoan: those are interim release milestones when working towards a release Mar 08 15:14:08 mckoan: e.g. 1.4_M4 is milestone 4 towards 1.4 release Mar 08 15:14:32 something equivalent to kernel RC1... Mar 08 15:15:05 theres lot of stuff with the schedule and milestone plans on the wiki Mar 08 15:15:07 systemd 198 is booting here ... Mar 08 15:15:17 rburton: hi Mar 08 15:15:55 rburton: have hashed out the systemd plan to address the meta-systemd concerns yet ? Mar 08 15:16:34 rburton: parts of the wiki are quite out of date though :( Mar 08 15:18:03 khem: I have this doubt because usually I use oe-core with origin/master but yocto with danny Mar 08 15:18:30 mckoan: do you mean poky with danny? Mar 08 15:18:30 khem: makes it sense to work with yocto origin/master too? Mar 08 15:19:03 khem: going to send a rfc series today, i've got hybrid builds working. annoyingly didn't get to finish it before i go to a conference for a week Mar 08 15:19:36 mckoan: I see why not Mar 08 15:19:52 mckoan: I work with origin/master on yocto/poky Mar 08 15:20:53 mckoan: poky is just oe-core+meta-yocto+otherbits. using poky with one branch and oe-core with another branch, when one encompasses the other, doesn't make a great deal of sense :) Mar 08 15:21:42 khem: hmmm actually I don't understand why I am afraid of this :-) I'll switch to that too Mar 08 15:23:06 kergoth: BTW I'm using two separate trees the Poky(yocto) way and the raw oe-core way Mar 08 15:23:17 with 2 different builds Mar 08 15:23:44 I figured, but I'm confused about the issue. You're basically already building poky master if you're bulding oe-core master, all that's missing is meta-yocto & friends Mar 08 15:23:53 so there's no real reason for worry Mar 08 15:24:29 kergoth: in fact I was interested to experiment master meta-yocto & friends Mar 08 15:24:39 * khem off to a F2F Mar 08 15:25:00 khem: thx Mar 08 15:25:33 I'd be more worried about using bleeding edge core classes than a machine .conf ;) Mar 08 15:25:35 but fair enough Mar 08 15:30:48 khem: basically with this branch if you activate both systemd and sysvinit features you'll be able to pick the init manager at image time. Mar 08 15:31:10 your sysvinit one will likely pull in libsystemd-login, but that's a tiny dep. Mar 08 15:33:13 I run into the following when my image recipe does do_rootfs: Mar 08 15:33:15 | Computing transaction...error: Can't install seemore-image-plugin-meta-0.1-r0@armv7a_vfp_neon: no package provides seemore-image-plugin-glib Mar 08 15:34:25 It seems to pick a random gstreamer plugin to quit on each time i re-run it. Has anyone else run into something like this? Mar 08 15:35:17 i wouldn't be worried about the ordering of the errors Mar 08 16:30:10 (good afternoon|goede middag) Mar 08 16:32:00 when I try to compile a package I get this error: libtool: link: cannot determine absolute directory name of `=/usr/lib64' Mar 08 16:32:24 the intarwebs has little if any usefull info. Any ideas Mar 08 16:32:25 ? Mar 08 17:01:30 hello helpfull yocto-people! Mar 08 17:03:52 I added a new layer including conf/distro/.conf, then changed "DISTRO", "DISTRO_NAME" etc. but when I boot the built image, it always tells me the old name (from the distro.conf I copied and changed)... What do I have to change so my image uses the new name? Mar 08 17:39:29 why are there many yocto kernel git repo instead of one with difference branches? Mar 08 17:41:37 xxiao, check the kernel architecture docs, but it's due to transient sets of features over time. just like linux-next, linux-stable, etc, but with more topic branches. you can't maintain a clean set of series that transition to mainline and encompass hundreds of commits over time, unless you can drop them completely and regerate/stack them clean each time. Mar 08 17:46:00 somehow i feel yocto devel is much more complicated than debian... Mar 08 17:46:45 * zeddii shrugs Mar 08 17:47:00 they are different use cases. Mar 08 17:48:17 xxiao: there are other options for building the kernel if you feel that linux-yocto is more than you need Mar 08 17:48:48 exactly. people can use what they want. linux-yocto-custom, completely custom, etc. Mar 08 17:50:10 need add a few recipes, kernel patches, external modules to yocto, the document is overwheleming to say the least :) Mar 08 17:50:38 which one ? don't read the kernel architecture document as a start. Mar 08 17:50:50 xxiao: we're happy to take suggestions if you have concrete ones Mar 08 17:50:52 there's a new kernel one for 1.4, hands on labs, quickstart guides that are simple enough. Mar 08 17:51:15 right, the labs get good feedback as far as I've seen Mar 08 17:51:35 at the basic level, exactly what you know about patching any package with bitbake/oe-core/yocto works exactly the same way for the kernel. Mar 08 17:52:04 i need add scst to yocto, which itself is a few modules, scripts, and also need patch the kernel Mar 08 17:52:11 it can just grow to suit advanced needs and the chaos that can be vendor patches, upstream releases, out of tree serieris, etc. Mar 08 17:52:19 and some user spaces program to admin it Mar 08 17:52:41 xxiao: do you know how to do the userspace parts already ? and the startup scritps ? if so, you have most of the knowledge that you need. Mar 08 17:53:18 https://www.yoctoproject.org/sites/yoctoproject.org/files/elc2013-kernel-lab.pdf Mar 08 17:53:19 https://www.yoctoproject.org/sites/yoctoproject.org/files/devday-kernel-tzanussi-elc-2013.pdf Mar 08 17:53:26 i have it working with yocto, just not the 'right' way, i.e. not built with recipes etc, it's done in a hackish way and now i want to clean it up Mar 08 17:54:21 http://www.yoctoproject.org/docs/1.4/kernel-dev/kernel-dev.html Mar 08 17:54:25 that's all kernel. Mar 08 17:54:47 but you also need to read the other yocto docs for userspace/recipes/packaging, the concepts are the same. Mar 08 17:55:29 thanks, yeah...quite some reading to do Mar 08 17:56:36 it gets easier, once the basic concepts become clear. Mar 08 18:20:17 mhh... I wonder if anybody answered my question after my PC crashed... ? Mar 08 18:26:06 hsychla: where exactly are you seeing the old $DISTRO value? Mar 08 18:57:47 evanp: when booting the qemu image. but I got it now, was a stupid mistake after all. did not change the distro parameter in local.conf... but thanks for taking your time! Mar 08 19:26:57 I have a recipe that under certain circumstances may want to add its multilib variant into the final image, I've been trying to do this by trying to append to RRECOMMENDS_${BPN} in my __anonymous() function but it doesn't work. what's the recommended way of doing this? Mar 08 19:34:38 off late I am seeing hangs when building the final image on poky master Mar 08 19:34:42 it uses systemd Mar 08 19:34:54 but I am just seeing it since couple of days Mar 08 19:37:28 paw: rather than appending, I'd probably stick a variable reference directly in RRECOMMENDS_${BPN} and then arrange for that variables value to be empty or the list of additional packages Mar 08 19:38:15 paw: a lot of the task- (v1.3) / packagegroup- (v1.4) recipes do that sort of thing. e.g., look at how FEATURES works Mar 08 19:41:19 i am getting "ERROR: ExpansionError during parsing /opt/....linux-emac.bb: Failure expanding variable do_patch: ExpansionError: Failure expanding variable do_patch, expression was cd /opt/.... Mar 08 19:42:55 is do_patch a template string variable that gets expanded with certain other variables made available somewhere long the recipe? and if so, does this error indicate that I am missing a variable necessary to complete the expansion? Mar 08 19:45:24 waynr: why are you overriding do_patch? Mar 08 19:45:37 it's rare to need to do that Mar 08 19:45:59 i'm not doing it on purpose Mar 08 19:46:11 where's "cd /opt" coming from then? Mar 08 19:46:32 that's part of the error output Mar 08 19:46:37 do_patch will apply the patches listed in src_uri Mar 08 19:46:48 so maybe you've got a messed up src_uri statement Mar 08 19:46:54 that is possible Mar 08 19:47:09 i don't have any patches lsted in src_uri Mar 08 19:47:20 just a defconfig and a git repository Mar 08 19:47:30 waynr: sounds like you need to share your recipe ;) Mar 08 19:47:48 (yes, the functions are for some time plain text, which are then parsed as python) Mar 08 19:50:37 rburton: that is probably a good idea...I am kinda just working in the dark copying existing recipes and making changes necessary to pull source from a git repository that i run Mar 08 20:42:27 can anyone remember when openzaurus was formed? Mar 08 20:42:47 trying to put a date on "key components of yocto are X years old" Mar 08 20:47:06 hmm, Mar 08 20:47:20 I think we have been using 2004 Mar 08 20:48:02 Domain Name:OPENZAURUS.ORG Mar 08 20:48:02 Created On:05-Feb-2002 17:07:13 UTC Mar 08 20:48:15 that's a good idea :) Mar 08 20:48:23 Domain Name:OPENEMBEDDED.ORG Mar 08 20:48:24 Created On:21-Jan-2003 23:58:22 UTC Mar 08 20:48:45 10 years old is good number Mar 08 20:48:56 thanks :) Mar 08 20:49:13 someone at ELC made the remark The Yocto Project hadn't been around long enough to judge Mar 08 20:49:24 I found that "amusing" Mar 08 20:51:18 ha Mar 08 21:05:04 I have eclipse making an arm binary of hell world with the toolchain from populate sdk Mar 08 21:05:22 any hints on how to run this binary on the target hw from eclipse? Mar 08 21:26:42 Crofton|work: you are almost there Mar 08 21:26:57 Crofton|work: if you can compile an app half the battle is won Mar 08 21:35:35 hello,world ... Mar 08 21:51:18 I think I ran it. There are way too many things to chose frmo. Seriously confusing Mar 09 01:12:24 rburton: somewhere I have the irc logs from when we first started oe and were arguing about names, and the wiki pages where we did the original design and requirements Mar 09 01:12:28 i can't remember where i put it, though Mar 09 01:12:40 was trying to put together a timeline **** ENDING LOGGING AT Sat Mar 09 02:59:59 2013