**** BEGIN LOGGING AT Fri Oct 30 02:59:59 2015 Oct 30 06:45:27 morning all Oct 30 07:08:35 ohai Oct 30 07:15:50 fray: thanks Oct 30 07:16:33 i was complaining about ebuilds that pass --llvm-config to packages and the output is stupid Oct 30 07:17:00 i don't want to hack every ebuild that uses clang Oct 30 07:17:08 that would also be stupid Oct 30 07:18:13 the facilities i have for setting per-package env flags are completely ignored Oct 30 07:18:32 apparently you're better at gentoo than we are Oct 30 08:12:30 morning all Oct 30 08:14:48 morning bluelightning :) Oct 30 08:15:07 bluelightning: can you help me with a python build issue? Oct 30 08:16:39 bluelightning: I was able to build 2.7.3 cleanly for my target but 2.7.9 is failing and interestingly only for my machine Oct 30 08:16:41 :( Oct 30 08:17:19 bluelightning: if you have a qemu build setup just add TUNE_CCARGS_pn-python=" -m64 -march=bdver4 -msse3 -mfpmath=sse" to your local.conf Oct 30 08:17:27 and do a bitbake python Oct 30 08:17:34 the do_install fails.... Oct 30 08:17:51 but I am pretty sure it is a build issue not install Oct 30 08:24:32 abelal: bin/sh: line 6: 26046 Illegal instruction (core dumped) - is that what you see? Oct 30 08:43:00 hi guys Oct 30 08:45:14 hi lazao Oct 30 08:45:33 hi bluelightning :) Oct 30 08:48:24 do you know how can i specify a root password for dropbear ssh ? atm i'm using the ssh-server-dropbear EXTRA_IMAGE_FEATURES Oct 30 08:54:41 wouldn't that jsut be the general root password in the end? Oct 30 08:57:13 lazao: remove debug-tweaks from features if it's there, then add Oct 30 08:57:15 inherit extrausers Oct 30 08:57:17 EXTRA_USERS_PARAMS = "usermod -P mypassword root;" Oct 30 08:57:24 it would be, i guess Oct 30 08:57:36 obviously i didn't read the faq: https://wiki.yoctoproject.org/wiki/FAQ:How_do_I_set_or_change_the_root_password Oct 30 08:57:39 thanks jku_ Oct 30 08:59:24 Oct 30 08:59:44 Oct 30 09:00:40 sure, i'm just doing some testings, i'll disable the root password, at least for ssh Oct 30 09:02:44 bluelightning: ping Oct 30 09:03:01 parrot1: pong Oct 30 09:05:17 bluelightning: The app that I'm writing recipe for is ignoring CMAKE_INSTALL_LIBDIR variable so it's installing the lib files to /usr/lib instead of /usr/lib64. So how do I manually "move" those libraries files to /usr/lib64 instead? Oct 30 09:05:39 and the fact that a lot of apps are ignoring those variables frustrates me :-( Oct 30 09:11:43 parrot1: seems to me you'd need to figure out why it's not using that path and fix it... Oct 30 09:11:51 parrot1: I'm not a cmake expert unfortunately Oct 30 09:18:48 bluelightning: neither am I. I'm also not questioning why the dev ignores it as some of them simply have no way to be contacted. I'm just wondering if it's possible to hack do_install or do_package method to copy the libfiles from the original /usr/lib to /usr/lib64 Oct 30 09:19:57 jku_: this has to be done in my image recipe? its not working :( Oct 30 09:22:00 parrot1: certainly you can do that, but if the software bakes the path into the binaries then you may still have problems Oct 30 09:22:41 parrot1: however to do that you would just add a do_install_append() and mv the files from ${D}/usr/lib64 to ${D}${libdir} Oct 30 09:27:10 bluelightning: Will certainly try that although it's not the most elegant way. Thanks Oct 30 09:31:32 image recipe sounds fine... can also do it in configuration but the inherit functionality is slightly different there Oct 30 09:31:37 lazao: ^ Oct 30 09:32:36 lazao: you aren't trying to bbappend image recipes, right? Oct 30 09:36:50 actually i have a personnal custom-image.class, and i have 3 images recipes that inherit from custom-image Oct 30 09:37:26 i only want to have the root password updated in one of the 3 images recipes, so i just added the EXTRA_USER_PARAMS in one of them Oct 30 09:38:53 my bad, i put EXTRA_USERS_PARAMS instead of EXTRA_USER_PARAMS Oct 30 09:38:56 should be okay now Oct 30 09:44:30 parrot1: if the project uses some other variable instead, you can also use something like 'EXTRA_OECMAKE = "-DLIB_INSTALL_DIR=${libdir}"' Oct 30 09:45:06 parrot1: that's probably a safer choice than moving files in do_install_append() Oct 30 10:18:26 bluelightning: yup Oct 30 10:19:25 bluelightning: when the setup.py script tries to import optparse Oct 30 10:19:39 bluelightning: optparse then imports strop Oct 30 10:19:46 and then this happens Oct 30 10:20:18 bluelightning: should the target side py modules be loaded when building sharedmods or host? Oct 30 10:21:05 bluelightning: because in the current scenario it is trying to load the target side ones Oct 30 10:21:26 bluelightning: and I am surprised why it isnt failing for any other -march Oct 30 10:21:32 :( Oct 30 10:24:06 hi - is there a quickstart anywhere showing a sample devshell session, when one attempts to understand what gets wrong in a thirdparty rule ? Oct 30 10:25:05 s/rule/recipe/ Oct 30 10:25:06 abelal: could you please file a bug report on bugzilla.yoctoproject.org? should go under oe-core / dev tools/toolchain Oct 30 10:26:00 yann|work: not really, but having opened the devshell if you cd .. (usually) you'll be in the workdir for the recipe, in which you'll find a "temp" directory containing all the logs and run files for the tasks Oct 30 10:27:07 bluelightning: I will but can you shed some light on this on your part? Oct 30 10:27:18 bluelightning: this doesnt happend with py 2.7.3 Oct 30 10:27:33 so I was thinking this might be a problem there Oct 30 10:27:38 abelal: I honestly don't know enough about the details, I would have to actually start debugging it and I'm running short of time today Oct 30 10:28:06 bluelightning: ok thanks :) I wont hold you up Oct 30 10:28:13 if you can file a bug I will ensure it gets assigned to the right person to look into it Oct 30 10:28:20 bluelightning: ok, I'm there already, but wondering if there was no more direct way - the run. Oct 30 10:28:22 oops Oct 30 10:28:26 the run. Oct 30 10:28:31 bluelightning: is there anyone else here who might be helpful at the moment Oct 30 10:28:42 not that I know of Oct 30 10:28:51 hmmm ok Oct 30 10:28:53 but if so they should please speak up ;) Oct 30 10:29:01 the run.* scripts include both env/funcs settings, and the actual rules Oct 30 10:29:36 (please excuse my silly keyboard with "*" located on the RETURN key) Oct 30 10:30:16 it would be great to have those more separated Oct 30 10:32:04 yann|work: I'm not sure that would make sense, what it's showing to you is exactly what gets run and that includes both the environment and the functions Oct 30 10:32:25 yann|work: if you just want to examine the environment for the recipe you can do so with bitbake -e recipename | less Oct 30 10:32:39 like, a devshell that would get a shell just instead of running do_ Oct 30 10:33:44 erm, but that is what happens when you run bitbake -c devshell recipename Oct 30 10:33:49 it gives you a shell Oct 30 10:34:07 in my case, I'm investigating a configure failure in a linux recipe, which is a bit more convoluted than the average recipe - something like "-c devshell-configure" would help Oct 30 10:34:34 hmm, I guess I see what you're driving at Oct 30 10:34:46 not precisely, since the run.do_configure script sets up a lot of stuff before running do_configure Oct 30 10:35:24 so I'm editing a copy of the script to get it to run bash instead of do_configure, but having a more comfortable way to get there would be great Oct 30 10:36:20 hm, except I get a child shell which does not inherit all those variables I need :) Oct 30 10:37:32 that would be very interesting yann, it can help a lot debugging Oct 30 10:42:46 not sure how that can be done - switching a shell into interactive mode is not something I think is implemented in any shell. But starting a child shell inheriting the parents' vars could be a way Oct 30 10:46:46 we do set variables in the shell's environment already, so it's not going to be that hard Oct 30 10:49:07 yes, can be as easy as exporting functions Oct 30 10:53:55 btw, in case someone involved in that would be around: I'm trying to get the bananapi support to work Oct 30 11:00:26 changing the run.* script to use bash, and run "export -f" on all defined funcs before launching a shell works nearly nice indeed - bbfatal causes exit from the shell, which is not that nice, and not exiting prevents the error to be caught :) Oct 30 11:00:37 but that still likes promising Oct 30 11:06:02 looks like the sunxi linux recipe has been developped by copying an old version of linux.inc... Oct 30 11:10:44 I think I'm going to cry ... devtool deleted my kernel :'( Oct 30 11:11:26 when is the 2.0 release ? Oct 30 11:12:03 kbingham: erm, what? its absolutely not supposed to do that... is this with 1.8? Oct 30 11:12:16 bluelightning: Yes ... 1.8 ... Oct 30 11:13:08 the answer to your second question is, it's imminent - RC3 is being pulled together at the moment, if all goes well, sometime next week or early the following week Oct 30 11:13:58 I think we did talk about devtool and the kernel not playing well together previously although I would not have expected any deletions Oct 30 11:14:20 presumably it's externalsrc that actually deleted it as the result of a clean / version change, rather than devtool Oct 30 11:14:32 though perhaps the distinction is moot Oct 30 11:14:46 bluelightning: I ran a bitbake - and I 'think' what happened is it tried to re populate my kernel source ... even though I had done a devtool modify -x virtual/kernel src/kernel ... so now I'm left with a symlink to an empty directory in src/kernel Oct 30 11:15:23 is this a linux-yocto kernel? Oct 30 11:15:29 bluelightning: sounds feasible ... I haven't dug into the internals to see what happened yet ... I just ran a build and my tree vanished. Oct 30 11:15:43 bluelightning: - no freescale tree. Oct 30 11:16:46 I'll just remove the symlink, and put my own tree in src/kernel to try to recover for now. Oct 30 11:16:53 lost a couple of patches but nothing too critical. Oct 30 11:17:04 ok, my apologies for the inconvenience Oct 30 11:17:25 it would be interesting to try the same thing with 2.0, I'd wager it would not happen Oct 30 11:17:36 bluelightning: hehe it's ok .. I knew devtool was still beta ... I should have pushed my work back to the server more frequently! Oct 30 11:17:49 if you feel adventurous you can checkout the jethro branch and try it Oct 30 11:19:06 bluelightning: That is tempting ... but I'll have to save that for another day ... need to get this driver integration done for the boss :) Oct 30 11:22:04 bluelightning: Were you at the Yocto dev day in dublin? I can't put a face to your name ... Oct 30 11:22:41 I was yes, though I was in the beginner's room Oct 30 11:23:50 bluelightning: Ahh ok ... I stayed in the advanced room. which for my second week with yocto may have been brave ... but it worked out ok :) Oct 30 11:33:11 Phew - I had pushed the commits ... so I haven' t lost anything except for a change in my .config file :) Oct 30 11:35:04 kbingham: ok, that's good Oct 30 11:35:25 kbingham: I wonder if the issue is reproducible Oct 30 11:36:01 I'll try to pay attention to what I do / what happens and let you know. Oct 30 11:36:08 ok, thanks Oct 30 11:36:40 bluelightning: It's odd though - because I had done several builds successfully ... so I'm not sure what triggered the source change/ update here. Oct 30 11:42:52 bluelightning: did the slides from the devday make it up on the web anywhere? Oct 30 11:43:07 that's a good question, I suspect not Oct 30 11:43:12 I'll ping some people next week Oct 30 12:17:00 Hi, is there a sanity check for runtime dependency which is not mentioned in build dependencies? Oct 30 12:17:43 We have way too many build errors from here and there which are clearly sysroot race conditions and missing build dependencies. Oct 30 12:19:09 Hi! Is the oprofile recipe dependent on kernel headers or can I use oprofile on top of different kernel versions without rebuilding ? Oct 30 12:25:44 Marex: afaik, oprofile and other userspace component do not depend the kernel version, of course some features with in might but they rely on ABI, not kernel versions. Oct 30 12:26:45 mcfrisk: perf and oprofile are sitting in meta/recipes-kernel , which confuses me Oct 30 12:27:15 mcfrisk: I'm aware that regular system components can deal with changing kernel underneath without much problem, yep Oct 30 12:28:49 yes, but they are also 'just' userspace apps. They rely on stable (hah :) Linux kernel ABI to function. They must behave the same on other kernels too. Oct 30 12:40:54 mcfrisk: I would have hoped so, I think perf doesn't for example ;-) Oct 30 12:52:38 Marex: I know pref is trickier, original question was about oprofile Oct 30 13:08:01 mcfrisk: I am still looking into that, looks like oprofile is using perf headers :/ Oct 30 13:08:32 the question now is, are those headers part of kernel abi Oct 30 13:08:34 probably yes Oct 30 13:09:56 Marex: in kernel tree all in include/uapi are uapi, and available in userspace from /usr/include/linux etc. Oct 30 13:10:46 but there is much more to uapi than just those headers which in quite many places incomplete and even broken (my fixes for them are in review for over a year now..) Oct 30 13:12:39 mcfrisk: yup Oct 30 13:49:50 anyone saw on linux do_install, something like that? : No rule to make target '.../build/tmp/work/bananapi-poky-linux-gnueabi/linux/3.4.90-r1/image/lib/firmware/./', needed by '.../build/tmp/work/bananapi-poky-linux-gnueabi/linux/3.4.90-r1/image/lib/firmware/ti_3410.fw'. ? Oct 30 13:51:14 if I run a devshell, get a shell with exported functions as described earlier, and run the "make modules_install" by hand - or even do_install itself, BTW, then the install goes as expected Oct 30 13:53:01 even just launching run.do_install in the devshell succeeds, while running a plain bitbake does not Oct 30 14:15:51 wow, using "oe_runmake -d" to run modules_install does prevent the problem! Oct 30 14:16:15 sounds like a gnumake bug, in fact... Oct 30 14:17:23 hm... not that easy :) Oct 30 14:43:55 the problem rather looks like a concurrency problem, it works flawlessly when I disable PARALLEL_MAKE - is the latter disabled when running in devshell ? Oct 30 15:24:29 2.0 fails pretty hard on atmel platforms :) Oct 30 15:27:16 hi. anyone got RIoTboard up and running on Yocto? for some reason I can't even see u-boot on boot... using yocto fido, meta-fsl-arm and meta-fsl-arm-extra Oct 30 15:29:49 Hi, I built a toolchain and I'm trying to compile a big makefile project with it, I made the recommended changes, and my makefiles are using the right compiler and flags, however, I keep running into undefined references to to __aeabi_uidiv etc, is there some additional library I need to link against? I have -lc -lm Oct 30 15:30:58 xp190_: there is probably an issue with the floating point settings of your toolchain Oct 30 15:31:42 abelloni: where would I look to see what the flating point settings are and how would I change them? Oct 30 15:33:06 I'm crosscompiling from ubuntu, simple sample files work fine, but my project fails to link Oct 30 15:35:24 it depends on how you built your toolchain Oct 30 15:36:14 I didn't do anything special, just bitbake meta-toolchain Oct 30 15:36:34 I changed the machine to my board (sabrelite) Oct 30 15:36:54 I'm using poky 1.8.1 Oct 30 17:12:30 * armpit really could use that Yocto cricket bat Oct 30 17:29:02 armpit: you mean like this? http://imgur.com/HaSHBAa Oct 30 17:30:30 fredcadete, nah.. thats too soft what I want to do. I need a real english cricket bat ; ) Oct 30 18:01:54 hi! I'm trying to use fido to build an armv6 qemu image, with kernel 2.6.x and glibc 2.9 Oct 30 18:02:10 I can't seem to find working recipes for the last two aspects Oct 30 18:02:49 I tried checking out an older release (daisy), but that has python errors on recent Ubuntu Oct 30 18:04:04 (I'm trying to make a VM I can test my Kindle homebrew mods in) Oct 30 19:15:51 why is it called "native"sdk? native to what? what would a non-native sdk be? Oct 30 19:16:58 since you can build SDKs for a different architecture than the host's, it can't be "native" in the sense of being compiled for the host either Oct 30 20:23:07 Hi, I built a toolchain and was able to compile a project, however, after compiling the binary won't run, when I use file to see what I built, I see the interpreter listed incorrectly compared to binaries that run on the target, how can I correct this? Oct 30 20:23:25 I can do a sym link and the binary runs, but this is not acceptable in my case Oct 30 20:23:45 how are you compiling your binary? are you using the build system, a generated SDK, or? Oct 30 20:25:06 I have makefiles which use the environment variables setup when I source the environment from the toolchain Oct 30 20:25:55 When I manually use $CC, a test binary builds properly and the interpreter is setup correctly, but in my project, this does not happen, i can't figure out why this is Oct 30 22:07:11 now that my core-image-minimal of 1.6 for bananaPI is working, i'm trying to build a core-image-x11 - however, I have a funny configure failure in qemu native, where the test for SDL fails because of seemingly-not-found libdbus versioned symbols from libpulse Oct 30 22:07:46 yann|work: which host distro is this on? Oct 30 22:07:58 (building on Debian/testing x86_64 with no known problem like that) Oct 30 22:08:42 hmm, we saw a similar issue on Ubuntu 15.10 Oct 30 22:08:59 https://bugzilla.yoctoproject.org/show_bug.cgi?id=8553 Oct 30 22:09:00 Bug 8553: normal, Medium+, 2.0, eduard.bartosh, IN PROGRESS DESIGN , qemu-native to build on Ubuntu 15.10 Oct 30 22:11:18 ah, yes, that's it Oct 30 22:12:43 hi! Oct 30 22:17:37 and the BUILD_LDFLAGS_prepend_pn-qemu-native does seem to help Oct 30 23:43:23 When is a driver compatible with a linux version? Example I have a driver for linux version 3.16 should it be still compatible with 3.18 or does the second number mean there changes in the interface and it probably won't work ? Oct 30 23:45:38 I'm not a kernel dev, but I'm fairly sure it'll entirely depend on what the driver is and what APIs it uses Oct 30 23:47:16 Hmm ok ... Oct 30 23:49:12 bluelightning: do you know if I can set the kernel version which the minimal image should use? Could I set 3.16 somewhere and it would use this version? (for example the LINUX_VERSION tag) Oct 30 23:49:58 you can create a recipe for a kernel version to build and then set PREFERRED_PROVIDER_virtual/kernel to point to that Oct 30 23:50:05 see our kernel / BSP manuals for coverage on that Oct 30 23:52:49 ah ok so there is not just a switch for that ... ok I see ... thanks Oct 31 01:08:51 I know I need a bbappend file for this but where should I place it and how should the content look like: linux-yocto-rt was skipped: incompatible with machine mymachine **** ENDING LOGGING AT Sat Oct 31 02:59:59 2015