**** BEGIN LOGGING AT Tue Oct 06 02:59:58 2015 Oct 06 03:14:32 Hm anyone know of any good examples of writing both custom do_unpack, and do_unfetch Oct 06 03:15:02 i have a working setup already, but it does seem to interfere with dependency checking somehow. Oct 06 03:15:51 In case you're wondering why it's because the recipe has to use google's depot tools. Which kind of throws SRC_URI and all of that out the window sadly. Oct 06 03:18:33 could always write a custom fetcher so you can use your own url scheme to use it Oct 06 03:18:41 bitbake fetcher, that is Oct 06 03:33:08 kergoth: Hmm.. how involved is that? Oct 06 10:34:01 Im having a small problem with recompiling the kernel with a patch Oct 06 10:34:45 ive added gator to the kernel/drivers changed the Makefile and Kconfig file accordingly but I'm unable to see a "Gator" option when i run menuconfig Oct 06 10:34:52 anyone can point me in the right direction ?? Oct 06 10:36:53 sighack: is that inside some kind of poky environment, devshell? or just a pure kernel menuconfig, e.g. you are actually having problems with the manual process? Oct 06 10:37:17 its in poky dizzy branch Oct 06 10:38:03 sighack: um, so you are sure that you have properly added that patch into the build process and that it gets certainly applied? Oct 06 10:38:41 or did you do that manually and are now trying to tinker the kernel thtough bitbake menuconfig Oct 06 10:39:24 Followed the instructions at : http://infocenter.arm.com/help/index.jsp?topic=/com.arm.doc.dui0482t/index.html Oct 06 10:39:42 I've manually added a folder "gator" in the kernel/drivers folder Oct 06 10:40:15 sighack: so your build process will know nothing about it and probably overwrite any changes at will Oct 06 10:40:42 LetoThe2nd: post that I've added "obj-$(CONFIG_GATOR) += gator/ " to the Makefile Oct 06 10:40:48 sighack: those instructions target the manual building of a kernel binary or module, *NOT* the integration into a larger build process. Oct 06 10:41:00 sighack: so you first need to sort out the way you want to go. Oct 06 10:41:16 aahh.. that makes sense now, so how do I integrate it into my process? Oct 06 10:41:22 ie, using poky Oct 06 10:41:41 sighack: see the kernel development manual as provided on the yocto project website. Oct 06 10:41:51 LetoThe2nd: alright! thanks Oct 06 10:41:53 sighack: it should give plenty of information on that. Oct 06 11:57:08 hello everybody Oct 06 11:58:05 i have an issue with the crosstoolchain, when i run cmake (after calling source enironment-setup script) i get : Fatal python error : Py_initialize unable to get the local encoding Oct 06 11:58:16 ImùportError : No module named encoding Oct 06 11:58:36 i understand that some python modules are missing but i dont know how to install them Oct 06 11:59:18 some posts sugggest adding TOOCHAIN_HOST_TASK_append natievsdk-python-modules but this does not sole my problem Oct 06 11:59:22 any idea please , Oct 06 12:23:12 any idea please ? Oct 06 14:58:32 eft the room. Oct 06 14:58:32 (7:58:05 AM) sjolley: YPTM: Ready-Access Number: 8007302996 Access Code: 2705751 Oct 06 14:58:57 YPTM: Stephen Joined Oct 06 15:01:41 YPTM: Randy joined Oct 06 15:01:51 YPTM: Sona joined Oct 06 15:03:29 YPTM: Jussi Kukkonen on call Oct 06 15:07:06 YPTM is over. Oct 06 15:07:14 :( Oct 06 15:07:24 I had the idea that it was going to be by IRC Oct 06 15:07:38 vmrod25: there was a conf call - it was short. Oct 06 15:08:03 see sjolley's 800 number above. Oct 06 15:08:32 lolz Oct 06 15:08:33 ok Oct 06 15:08:53 np Oct 06 15:10:32 I have in a bbclass file a anonymous python function which does the following, d.appendVar("INHERIT", " deploy") however I can't see the full deploy.bbclass being imported to the recipe, i.e. DEPLOYDIR is not being set Oct 06 15:23:24 i activated a PR server and build history for my package feeds Oct 06 15:24:54 for a large number of packages, i am noticing a switch in dependencies between busybox and bash Oct 06 15:26:05 it happens occasionally, and it seems like something due to a non-deterministic dependency - possibly on the shell Oct 06 15:28:49 dguthrie: is your bbclass inherited globally by INHERIT? I'm guessing not. INHERIT is parsed at the end of the config file parsing, it'll never be used if set for an individual recipe Oct 06 15:29:38 dguthrie: if you want to conditionally inherit a class in a recipe, you'll have to do something like inherit ${SOMEVAR} and then have SOMEVAR use inline python, returning an empty string when you don't want it inherited Oct 06 16:08:47 hi all, hope everyone is having a peachy october Oct 06 16:09:36 I've got a problem with depmod and out-of-tree modules and an SDK Oct 06 16:10:30 My environment is Ubuntu 14.04, with the depmod shipped with the distro. Oct 06 16:11:22 The problem is the depmod shipped with ubuntu 14.04 is a little old, and doesn't pick up new files (like *.bin) Oct 06 16:12:18 So when I run depmod with a kernel module build inside the SDK environment, it isn't too happy. Oct 06 16:12:46 I'd like to include kmod (and the depmod symlink) in the SDK as a nativesdk package Oct 06 16:13:31 That way the depmod/kmod in the SDK is the same as the one the kernel was built with Oct 06 16:14:58 Any ideas on how I could construct a .bbappend file for kmod to do this? Oct 06 16:15:44 given you aren't going to be buliding a new rootfs with the sdk anyway, i don't really see why you wouldn't just run depmod on target, but yes, you could add kmod to the sdk if you add BBCLASSEXTEND += "nativesdk" to the kmod recipe first, then add nativesdk-kmod to TOOLCHAIN_HOST_TASK Oct 06 16:16:45 kergoth: I'd like to run depmod during build time before deploying the module to target, so I can see any warnings/errors Oct 06 16:17:17 kergoth: I will attempt to modify the kmod recipe itself, I didn't try that yet, thanks! Oct 06 16:17:26 ? Oct 06 16:17:34 just use a bbappend. that's what they're for Oct 06 16:17:47 a bbappend modifies a recipe by definition. i dind't mean the recipe *file* Oct 06 16:18:21 OK, I tried a .bbappend before. there's several kmod recipes... kmod-native and kmod_git Oct 06 16:18:26 no, there are two Oct 06 16:18:40 one of which is native, which is a build tool run during the build process to run on the machine running bitbake Oct 06 16:18:45 not of any use to you here Oct 06 16:18:52 agreed Oct 06 16:20:24 I get a build failure with a kmod_git.bbappend in the install step. It can't find the testsuite directory. It looks like it never got copied to the build directory. Oct 06 16:21:08 I suppose I can add a do_install_prepend and try copying... Oct 06 16:21:12 hello? Oct 06 16:21:12 bluelightning: are you there? Oct 06 16:21:34 raykinsella78: hi yes Oct 06 16:21:58 I'm still at the YP booth at ELCE Oct 06 16:23:25 I guess I'm confused why adding the BBCLASSEXTEND += "nativesdk" makes the install step fail, but the native and vanilla kmod all build/install fine. Oct 06 16:24:01 darthduck: what is the failure? Oct 06 16:24:47 Making install in libkmod/docs | make[2]: Nothing to be done for `install-exec-am'. | make[2]: Nothing to be done for `install-data-am'. | find: `testsuite': No such file or directory | WARNING: exit code 1 from a shell command. Oct 06 16:25:15 bluelightning: it's in the do_install step Oct 06 16:25:28 that sounds like it's being run in the wrong directory Oct 06 16:25:45 not sure why that would happen only at do_install or why it would be specific to nativesdk Oct 06 16:26:04 I don't see any uses of PN that should be BPN, doesn't seem like that should be failing Oct 06 16:26:10 * kergoth yawns Oct 06 16:27:53 Hmm, we should think about prepending to PS1 in our setup scripts Oct 06 16:28:43 there's a ptest.patch file that messes with the testsuite dir, wonder if that has anything to do with it Oct 06 16:29:52 <3 yocto-layer create+bitbake-layers add-layer+recipetool. so nice for quickly appending something for testing Oct 06 16:38:21 it looks like the tip of kmod meta/recipes-kernel/kmod/kmod.inc doesn't include ptest.patch anymore. The fido branch has it tho. Oct 06 16:53:19 bluelightning, kergoth: I think the problem may be in a check in the Fido branch of kmod_git.bb. It isn't present in the git tip/master. There's a DISTRO_FEATURES check in the Fido branch that calls the find command that looks like it's failing. Oct 06 17:24:07 hi. is it possible to build the same image for several machines at the same time? Oct 06 17:41:17 sa2ajj: not in one bitbake command, no. you can do it with multiple in a single builddir by either modifying local.conf or exporting MACHINE in the shell environment Oct 06 17:43:50 kergoth: right Oct 06 17:44:54 i have two [rather] different target platforms [different CPUs], however the set of packages is _almost_ the same Oct 06 17:45:30 i hoped to use `bitbake image-platform-1 image-platform-2` to get both at the same time Oct 06 17:45:35 nope Oct 06 17:45:53 MACHINE=platform1 bitbake image; MACHINE=platform2 bitbake anotherimage Oct 06 17:45:54 the reason why i'd like them to be produced at the same time is build history feature... Oct 06 17:45:56 would be as close as you can get Oct 06 17:46:15 right Oct 06 17:46:52 so if i'd want to the output information to end up in a build history of sort, i'd have to manage the latter myself. Oct 06 17:46:57 kergoth: thank you! Oct 06 17:47:05 buildhistory distinguishes between archs and platforms, it can handle multiple machines just fine. Oct 06 17:47:12 so not sure what exactly you're looking for there Oct 06 17:47:58 what i found about build history is that it creates a separate commit for _each_ file produced... Oct 06 17:48:22 and what i wanted is a _single_ commit with the result Oct 06 17:48:30 no, there's one commit for each bitbake command, one for packages, one for the image Oct 06 17:49:02 and _then_ i wanted to be able to produce images for several targets and still to put their information in the same commit :) Oct 06 17:50:11 according to buildhistory.bbclass it's one commit per file produced Oct 06 17:50:41 no idea what your'e reading, but i'm talking about what actually happens Oct 06 17:51:09 did you test it? Oct 06 17:51:33 i was reading buildhistory.bbclass for fido... lemme check what happens in master... Oct 06 17:51:53 fido didn't behave that way either Oct 06 17:52:06 i've been using it since paul submitted it to oe-core, and have never seen it commit once per file Oct 06 17:52:40 yes, in master it's the same thing Oct 06 17:52:49 (though coded a bit differently) Oct 06 17:53:01 on sec Oct 06 17:53:30 again, i'm talking about what it actually does, not your potentially flawed reading of the code Oct 06 17:53:39 i have a buildhistory repo sitting right in front of me Oct 06 17:54:10 do i read it wrong: http://git.yoctoproject.org/cgit/cgit.cgi/poky/tree/meta/classes/buildhistory.bbclass#n661 ? Oct 06 17:54:43 actually, 661-663 Oct 06 17:56:27 my reading: for each file produced, record it in a separate commit Oct 06 17:57:41 nope Oct 06 17:57:49 it only takes the leading part of the path (images or packages) Oct 06 17:57:53 awk -F/ '{print $1}' Oct 06 17:58:42 * sa2ajj goes to check what -F/ means Oct 06 17:59:01 field separator. instead of whitespace, use / Oct 06 17:59:05 effectively the same as cut -d/ -f1 Oct 06 17:59:25 right. my bad. sorry Oct 06 18:01:04 probably a comment or two would help those who come after me :) Oct 06 18:11:23 before i go to bed: kergoth: thank you very much for you help: the thing with different target machines was not obvious Oct 06 18:11:48 s/you/your/ Oct 06 18:12:13 not a problem Oct 06 18:57:28 strange, I'm cross compiling "gpsd", but it doesn't use the proper "ranlib" binary, but it compile with the right gcc Oct 06 19:11:23 hrm... good news: kmod gets farther when I update the bitbake recipes to use version 20 of kmod. bad news: I get a python exception in do_package. Oct 06 19:12:06 Exception: OSError: [Errno 2] no such directory /sandboxes2/yocto_1.8/poky/build/tmp/work/x86_64-nativesdk-pokysdk-linux/nativesdk-kmod/20+gitAUTOINC+d9c7175859-r0/package/opt/poky/1.8/sysroots/x86_64-pokysdk-linux/opt Oct 06 19:36:25 any idea why would the package Im trying to build (gpsd) would use the wrong ranlib ? Oct 06 19:36:46 it is using the host one, instead of from the cross compiler toolchain :" Oct 06 19:57:43 dangit... bitbake kmod works fine. bitbake nativesdk-kmod fails. Oct 06 20:17:31 The failure I'm seeing is inside update-alternatives.bbclass... is "inherit update-alternatives" compatible with a nativesdk package? Oct 06 20:27:26 darthduck: I think inheriting it is not a problem itself, trying to use it might be Oct 06 20:28:06 Depends on what you want to do; I know that I had to patch some stuff to get it to work in our model of an SDK Oct 06 20:38:54 neverpanic: thanks. I'm attempting to write a stripped-down version of the kmod recipe right now. It looks like a .bbappend will be difficult at best (how does one uninherit?) :) Oct 06 20:46:18 darthduck: one does not. You can hack the internals of the inherit command to make the inherit u-a a no-op though Oct 06 20:46:57 there's a list of bbclasses that were already included; if you can get u-a in there without actually inheriting it, the inherit becomes a no-op Oct 06 20:51:59 neverpanic: hrm... I'll look into that... do you know the approx. location of the python file where the inherit command is defined? Oct 06 20:52:30 $(dirname $(which bitbake))/../lib, I think Oct 06 20:52:47 obviously that's a hack, though Oct 06 20:53:33 neverpanic: true. I suppose the best thing would be to fix the kmod recipe directly and submit a patch. :) Oct 06 20:54:10 that, or you could wrap the u-a bbclass in a separate layer you include before poky/meta and add apropriate changes there Oct 06 20:54:18 certainly less hacky Oct 06 20:57:23 there's already an anonymous function in update-alternatives.bbclass that has a comment "Update Alternatives only works on target packages...", and proceeds to check if it's native, cross, or crosssdk. Oct 06 20:58:16 I'll try adding a check for nativesdk there... maybe that'll work. Oct 06 21:02:06 dangit, not that easy. Oct 06 21:57:50 neverpanic: Wow! Found this little gem in recipes-extended/tar/tar.inc: ALTERNATIVE_${BPN}_class-nativesdk = "" Oct 06 21:58:05 That worked like a charm. Oct 06 21:58:13 Yeah, could have told you... Oct 06 21:58:19 just wasn't sure that's what you needed. Oct 06 21:58:43 Sorry, should have mentioned that Oct 06 21:59:07 neverpanic: heh, I'm still not sure nativesdk is what I need... I have no intention of running depmod on the target, I just need to run it for out-of-tree module builds. Oct 06 21:59:37 If you build those modules against the sdk, that may be what you need. Oct 06 21:59:44 yup Oct 06 22:01:17 Coolbeans... regardless, thanks for the help, it is much appreciated! Oct 06 22:25:17 that's precisely what nativesdk *is*, binaries that run on the machine the sdk is installed on Oct 06 22:25:27 target stuff can be placed in the sdk as is, don't need to be built special Oct 07 01:54:08 Hey guys, I have a layer I have found that I want to add to my project, and it says to add it to sources/layers.txt Oct 07 01:54:24 but I don't see that referenced in the developer documentation anywhere Oct 07 01:54:32 Can someone point me in the right direction? Oct 07 02:01:18 sources/layers.txt isn't associated with oe-core mechanisms, most likely it's one of the custom setup script setups. are you using a chip vendor sdk or something? Oct 07 02:01:33 you could always clone the layer you want and bitbake-layers add-layer /path/to/the/layer Oct 07 02:03:31 https://github.com/imyller/meta-nodejs Oct 07 02:03:41 That's the reference Oct 07 02:04:29 I'm building a recipe that requires that I 'npm install' some package dependencies, since it's a nodejs application Oct 07 02:05:06 from a quick google, it sounds like layers.txt is an angstrom-specific file used by its setup scripts. Oct 07 02:05:34 just clone it yourself instead of relying on a script to do it, and add thel ayer to BBLAYERS in conf/bblayers.conf or use the aforementioned bitbake-layers command Oct 07 02:05:48 ahh dang I googled the heck out of that jazz Oct 07 02:06:00 sorry, I guess i could have figured that out on my own Oct 07 02:32:33 If I want to build a back-rev of a recipe, where is that typically specified? Oct 07 02:38:47 sorry, I guess i could have figured that out on my own Oct 07 02:39:10 ls Oct 07 02:39:24 gah sorry, new irc client Oct 07 02:42:27 np **** ENDING LOGGING AT Wed Oct 07 02:59:58 2015