**** BEGIN LOGGING AT Thu Jun 29 03:00:01 2017 Jun 29 07:29:27 good morning Jun 29 09:30:48 Good morning! I am working on a recipe for octave. One of the products of the build is a host binary (mkoctfile) that is needed to compile the extensions Jun 29 09:31:30 I used to install it in ${STAGING_BINDIR_NATIVE}/mkoctfile but this does not work anymore with the sysroots-componets approach Jun 29 09:31:52 where do I have to install the binary so I can DEPEND it on other packages? Jun 29 09:39:48 yeah that's really nasty don't do that Jun 29 09:41:03 so you have other recipes that need that tool to build? Jun 29 09:41:24 yep octave-image-acquisition Jun 29 09:41:33 the standard practise would be to make an octave-native recipe (be it a new recipe, or bbclassextend) that installs that binary to ${D} like normal. Jun 29 09:41:45 it can be a cut down recipe that *only* installs that tool Jun 29 09:42:21 eg gtk-icon-utils-native.bb is basically gtk-native but it only builds a single binary Jun 29 09:43:23 but the tools is already built and tested on the target recipe. fetch, patch, configure, build and install from 0. seems overkill Jun 29 09:44:11 tough. a recipe can't be target and native at the same time. your trick of throwing directly into STAGING_BINDIR_NATIVE wouldn't have worked if octave was installed from sstate. Jun 29 09:45:32 and SYSROOT_PREPROCESS_FUNCS_append and install into "${SYSROOT_DESTDIR}/${STAGING_DIR_NATIVE}" Jun 29 09:45:40 will that work? Jun 29 09:48:30 just have a native form that just builds that single binary Jun 29 09:49:31 ok, thanks! Jun 29 12:38:45 Hi, I need a litle bit more help with recipe-sysroot Jun 29 12:39:39 The recipe RDEPEND on libecj-bootstrap-native, which has the fllowing file: sysroots-components/./x86_64/libecj-bootstrap-native/usr/share/java/ecj-bootstrap.jar Jun 29 12:40:01 but the file is not present on the reice-sysroot-native. Any ideas how to debug it? Jun 29 13:00:58 ribalda_: RDEPENDS - shouldn't that be DEPENDS? Jun 29 13:04:15 bluelightning, it depends! Jun 29 13:04:24 heh Jun 29 13:04:38 sorry, I'll shut up now Jun 29 13:09:44 bluelightning: sorry, it is in DEPENDS. And I can see that it is being considered in do_prepare_recie_susroot Jun 29 13:09:50 Adding dependency on libecj-bootstrap-native Jun 29 13:11:05 ribalda_: that's odd then, it should be in the sysroot... Jun 29 13:11:46 maybe because the file is a .jar? Jun 29 13:17:05 ribalda_: I don't think we filter based on file type, at least any such filtering would be before it made it into sysroots-components Jun 29 13:23:25 bluelightning: im having some issues with bitbake giving me AST / parsing errors of my python code. Do you know if its allowed to nest functions in a bitbake python function? like: python asdf() { \ def nested_func(..): pass \ .. print(nested_func(..)) } Jun 29 13:26:26 yes, the core classes do that all the time Jun 29 13:26:43 pastebin? probably a genuine typo Jun 29 13:29:17 mago1: I believe that's allowed yes... I think we do that in a few places Jun 29 13:30:06 bluelightning: I think I found it, on java-library.bbclass there is a task, removebinaires that removes jars and classes Jun 29 13:30:11 :) Jun 29 13:30:36 ribalda_: ah, I see :) Jun 29 13:30:45 glad you figured it out though Jun 29 13:36:51 rburton: you're probably right, lemme just recheck. is there a better way to get parsing error output? the oneliner outputted by bitbake is very difficult to read Jun 29 13:37:41 only if you file a bug to get the ast producing better errors... Jun 29 14:27:15 tlwoerner, ERROR: /home/balister/src/oe-core/oe-core/../meta-raspberrypi/recipes-core/images/rpi-hwup-image.bb: Deprecated variable(s) found: "IMAGE_DEPENDS_rpi-sdimg". Use do_image_[depends] += ":" instead Jun 29 14:27:58 Crofton|work: yea... every darn BSP is "broken", i didn't have a singe successful build last night :-( Jun 29 14:28:12 bother Jun 29 14:28:55 it's a fairly easy fix, however Jun 29 14:29:52 although the fix doesn't seem to like dashes (?) so i'm still trying to figure that one out Jun 29 14:30:55 I'd help, but no cycles until tomorrow at earliest Jun 29 14:32:02 i think it's a case of every BSP maintainer looking at everyone else's BSP, copying what they see, but unfortunately the first one used this syntax (which is now deprecated) so now they're all bad :-) Jun 29 14:33:24 i haven't done rpi yet (assuming someone else is going to jump all over that one) i've got 3 out of the 8 i care about working and ready to push Jun 29 14:33:44 ...and working on the rest Jun 29 14:35:07 how can I create an ipk package which contains a single file '/.foo'? Classes use glob('*') which ignores dotfiles Jun 29 14:35:46 oops, misspoke, 2 out of my 8 did succeed, have fixes for 3, leaving 3 more to go Jun 29 14:59:40 ensc|w: glob doesn't ignore dotfiles, it just doesn't match them if it follows a / Jun 29 14:59:52 ensc|w: just FILES_${PN} += "/.*" instead of /* or / Jun 29 15:00:00 for example Jun 29 15:17:27 kergoth: package_ipk.bblcass contains Jun 29 15:17:28 g = glob('*') Jun 29 15:17:28 if not g and localdata.getVar('ALLOW_EMPTY', False) != "1": Jun 29 15:17:51 'g' is empty in my case Jun 29 15:19:11 ALLOW_EMPTY_${PN} = "1" does not seem to work; perhaps there is a similar statement below Jun 29 15:19:14 either patch package_ipk.bbclass to check for both, or set ALLOW_EMPTY even though it's not empty Jun 29 15:19:17 huh Jun 29 15:22:18 ... or wait... ALLOW_EMPTY works. when I tried it I got probably a broken package from sstate Jun 29 15:42:02 allow_empty is generally an outcome of something else Jun 29 15:47:35 Crofton: https://github.com/agherzan/meta-raspberrypi/pull/90 Jun 29 18:20:35 Crofton: ...and they discovered the dash/underscore issue too :-) Jun 29 18:49:24 tlwoerner, good thing I have real work to keep me busy then Jun 29 18:49:33 damn it, wanted to attack the qwt strip issue though Jun 29 18:51:08 Crofton|work: i'm sure the BSP trolls will have everything fixed up by the time your "real work" is done ;-) Jun 29 19:14:03 lol, Trained Bootloader Monkeys! Jun 30 00:29:01 Has anyone out there successfully built oe/poky using Windows Subsystem for Linux? I got the build to start, but I had to build in /mnt/e since I didn't have enough space in my C drive, so I ran afoul of filenames with ":" in them. Jun 30 02:28:02 MrKevin: I would suggest to use a linux VM on windows Jun 30 02:28:31 We do support SDKs on mingw Jun 30 02:28:37 would be nice to support the windows subsystem for linux in general, my main concern would be filesystem capabilities, but i doubt anyone has messed with it much yet Jun 30 02:28:43 a vm is definitely a good way to go **** ENDING LOGGING AT Fri Jun 30 03:00:03 2017