**** BEGIN LOGGING AT Tue Aug 27 02:59:58 2013 Aug 27 05:28:10 Jay7: do you know a way how to get the CPPFLAGS settings of a Makefile respected, i.e. not overridden by the bitbak build system? Aug 27 06:25:05 does anyone plan to have a recipe for harfbuzz-ng? Aug 27 06:35:03 JaMa: do you know a way how to get the CPPFLAGS settings of a Makefile respected, i.e. not overridden by the bitbak build system? Aug 27 06:35:23 https://bugzilla.yoctoproject.org/show_bug.cgi?id=5046 Aug 27 06:35:23 Bug 5046: normal, Undecided, ---, scott.m.rifenbark, NEEDINFO , Document that CPPFLAGS override behavior and possible workaround Aug 27 07:07:52 lpapp_: you can remove -e from EXTRA_OEMAKE or unset CPPFLAGS Aug 27 07:08:52 JaMa: how to remove it? Aug 27 07:09:28 JaMa: won't that make the software's CC=gcc make broken as that is probably not appropriate for cross-compilation? Aug 27 07:43:34 In my builds, since I set package_deb, I often see bitbake download apt from debian. Aug 27 07:43:45 I also see opkg get installed. Aug 27 07:43:57 What is the relationship between opkg and apt then? Aug 27 07:44:08 Doesn't apt supply all the needed functionality? Aug 27 07:44:31 Or is apt only partly used and opkg used instead? Aug 27 08:03:49 Hi yocto. Anyway to add my custom files (test apps compiled externally using yocto toolchain) into rootfs image? Aug 27 08:07:30 easiest way would be to make a recipe for your test apps Aug 27 08:07:54 maybe a git recipe with AUTOREV Aug 27 08:18:06 JaMa: sorry, had to disconnect to come to work. Aug 27 08:18:24 JaMa: so, is there a simple to way to preserve CC and CPPFLAGS? Aug 27 08:19:50 Is HOST_ARCH a variable like TARGET_ARCH? The user manual only mentions TARGET_ARCH http://docs.openembedded.org/usermanual/usermanual.html Aug 27 08:26:01 bluelightning: morning, is there a way alternatively to drop the include path in the software, and make it work with Yocto? Aug 27 08:26:09 if dropping the override is not possible simply? Aug 27 08:26:17 Stygia: HOST_ARCH is the architecture you're currently building for (valid when you're building either a native, target or SDK recipe) Aug 27 08:26:36 lpapp: sorry what are we talking about? Aug 27 08:26:50 bluelightning: the issue on Friday evening if you remember. Aug 27 08:26:59 bluelightning: software using CPPFLAGS. Aug 27 08:27:01 friday evening was a long time ago... Aug 27 08:27:30 bluelightning, Alright, just as I hoped, it just wasn't mentioned in the docs. :) And I suppose I could embed python and print the variable, but I'm unsure on how to get to the OE variables from python. Aug 27 08:27:36 bluelightning, Either way, thanks. Aug 27 08:28:21 Stygia: I'll make a note to get Scott to document all of the HOST_* variables, I think that's not the only one we're missing Aug 27 08:28:33 bluelightning: https://bugzilla.yoctoproject.org/show_bug.cgi?id=5046 Aug 27 08:28:34 Bug 5046: normal, Undecided, ---, scott.m.rifenbark, NEEDINFO , Document that CPPFLAGS override behavior and possible workaround Aug 27 08:28:34 bluelightning, No, it isn't, the user guide lacks quite a few things. Aug 27 08:28:49 lpapp: maybe try "unset CPPFLAGS" in do_compile? if not, I don't know Aug 27 08:29:08 Stygia: we definitely take bug reports for the docs Aug 27 08:29:36 actually, I have been told send patches rather than reports. Aug 27 08:29:45 as there is not much doc man power in here. Aug 27 08:29:50 bluelightning, Hmm alright. Which site is the "right" one, openembedded.org, yocto, embedded linux, etc? Aug 27 08:29:57 bluelightning: trying.. Aug 27 08:30:29 bluelightning: before oe_runmake ? Aug 27 08:30:43 Stygia: bugreport is the same for those... Aug 27 08:31:10 I know poky's the distro, openembedded the project, and yocto the... newer incarnation of the project? Aug 27 08:31:32 no, Yocto is the project. Aug 27 08:32:49 bluelightning: bitbake foo -c clean && bitbake foo should be enough? Aug 27 08:33:01 (just in case) Aug 27 08:33:30 because I am getting include path issues for that, still. Aug 27 08:34:13 how have you determined that it is CPPFLAGS being cleared that is causing invalid include paths to be used? Aug 27 08:34:15 ndec: does that work for you? Aug 27 08:34:29 bluelightning: it is not passed to the compiler.. that is the first symptom. Aug 27 08:34:39 bluelightning: then, ndec told me to print it out in the Makefile. Aug 27 08:34:43 bluelightning: it got never printed out for me. Aug 27 08:34:50 bluelightning: but it got for him, and he told me it is cleared. Aug 27 08:35:18 I may have additional issues as well... for starter, echo ${CPPFLAGS} does not work under the "all:" target. Aug 27 08:36:08 does not work means, it does not get printed even if I prefix it with a "TEST" string. Aug 27 08:36:26 added to where? Aug 27 08:36:28 so as if it was either not executed, or stdout blocker in there... Aug 27 08:36:42 bluelightning: Makefile Aug 27 08:36:55 workdir Aug 27 08:37:04 there's no stdout "blocker" Aug 27 08:37:11 building the archive with that information in. Aug 27 08:37:21 if that statement is being executed the result will appear in the task log Aug 27 08:37:45 bluelightning: I have no clue why it is not printed here. Aug 27 08:37:55 nor have I Aug 27 08:37:59 bluelightning: but ndec could reproduce anyway. Aug 27 08:38:13 bluelightning: I think you would be able to with the first example of the Makefile tutorial Aug 27 08:38:19 and running bitbake. Aug 27 08:38:29 I will actually probably build such a skeleton to experiment with. Aug 27 08:39:53 bluelightning: there'ya go, http://www.cs.colby.edu/maxwell/courses/tutorials/maketutor/makefile.2 Aug 27 08:39:59 yes, we saw the other day that CPPFLAGS is 'cleared'. e.g. it's an env variable set by bitbake , so with make -e, the makefile can't override Aug 27 08:40:19 yes I understand that part Aug 27 08:40:42 ah. Aug 27 08:40:46 bluelightning: I will even simplify that example into one object file. Aug 27 08:40:51 so lpapp do you need anything from me? Aug 27 08:40:51 and attach it to the bugreport. Aug 27 08:40:59 ndec: any workaround is welcome, yes. Aug 27 08:41:04 this is my current task Aug 27 08:41:09 so I need to work the solution out. Aug 27 08:41:28 unset does not seem to work for me. Aug 27 08:41:50 i think i told you to set CPPFLAGS in the .bb file Aug 27 08:42:21 well, I tried before you told. Aug 27 08:42:23 but it did not work Aug 27 08:42:46 well, given that it worked for me, you must have been doing something else ;-) Aug 27 08:42:55 also, looking at bitbake.conf: Aug 27 08:42:55 export CPPFLAGS = "${TARGET_CPPFLAGS}" Aug 27 08:42:55 export TARGET_CPPFLAGS = "" Aug 27 08:43:06 you should be able to set TARGET_CPPFLAGS Aug 27 08:43:29 right, that is not the way how I tried. Aug 27 08:43:37 I tried as kergoth mentioned. Aug 27 08:43:45 that is why I asked for clarification ... Aug 27 08:43:55 what did you try? Aug 27 08:44:35 ndec: for instance exorcism. :} Aug 27 08:45:29 ndec: 18:43 < lpapp> ndec: I tried this: EXTRA_OEMAKE = "'CFLAGS=${CFLAGS}' 'CPPFLAGS=${CPPFLAGS}'" Aug 27 08:46:11 ah, right. and iirc, i told you why this wouldn't work, and told you another method which works . Aug 27 08:46:19 bluelightning: btw, can you confirm that bitbake foo -c clean && bitbake foo is ok? Aug 27 08:46:30 ndec: not that I can remember. Aug 27 08:47:03 or something that I could understand apparently anyhow. Aug 27 08:47:03 lpapp: that will work yes; but note it will restore from sstate if no task signatures have changed Aug 27 08:47:24 bluelightning: right, so -c cleanall is safer, then. Aug 27 08:47:42 lpapp: cleanall is further than necessary, that will delete the downloads as well Aug 27 08:47:45 Hmm I'm trying to file a bug in the documentation. It wants me to either choose a manual, or "general", meaning no manual. However, none of the options are "user manual". (http://docs.openembedded.org/usermanual/usermanual.html) Which category should If ile it in? Aug 27 08:47:59 lpapp: use cleansstate if you really feel the need to clear out shared state as well Aug 27 08:48:20 Stygia: do not worry that much ... triagers will help... just do your best. Aug 27 08:48:22 lpapp: note that this shouldn't be necessary, changing variable values will force tasks to re-execute without doing that Aug 27 08:48:58 [20:02:37] [19:40:03] lpapp: temp workaround set CPPFLAGS in the .bb file Aug 27 08:48:58 [20:05:50] lpapp: look. i have tested this workaround before giving it to you. if you set the variable in the .bb it will work. you must be doing something wrong. Aug 27 08:49:02 lpapp: ^ Aug 27 08:49:02 ndec: so you claimed you made the change. Let me ask again: can you paste the line to avoid any differences? Aug 27 08:49:09 ndec: that is totally unclear. Aug 27 08:49:14 as I replied after that. Aug 27 08:49:22 but you were not willing to provide the only line. Aug 27 08:49:40 'set CPPFLAGS in the .bb file'. not sure what is not clear here? Aug 27 08:49:42 My options are: ADT docs, BSP docs, development, general docs ("No associated manual"), kernel development, mega manual, profile manual, quick start and reference. Aug 27 08:49:51 ndec: there are thousand ways to set that ... Aug 27 08:49:53 here is the line: Aug 27 08:49:54 +CPPFLAGS="-Ifoobar" Aug 27 08:50:12 now, you are saying something different again Aug 27 08:50:18 i just tried with TARGET_CPPFLAGS and it works too, and seems nicer. Aug 27 08:50:26 See? That is why I have been asking for clarification since Friday. Aug 27 08:50:33 started with set it through EXTRA_OEMAKE Aug 27 08:50:38 then you say TARGET_CPPFLAGS above Aug 27 08:50:45 now, you claim a different version CPPFLAGS Aug 27 08:50:55 clarifying is very important here. Aug 27 08:51:13 rather than throwing continuously: "you still do not get it?" Aug 27 08:51:47 but neither of those works after all, for me. Aug 27 08:52:12 not to mention, it is reasonable those do not work. Aug 27 08:52:19 again, check the log. i told 'using EXTRA_OEMAKE won't work, and explained why (e.g. because the variable is cleared before calling make anyways)', and i told 'set CPPFLAGS in the recipe'. i can't do better, i believe. I think you could have 'tried' that method since Friday.. Aug 27 08:52:21 because "./" is set inside the workdir somewhere in a folder. Aug 27 08:52:32 so the path is only known at assembly time. Aug 27 08:52:38 you cannot really predefine that. Aug 27 08:52:46 so trying to set what you do not know will not obviously work. Aug 27 08:53:12 yeah, you told *after* trying. Aug 27 08:53:17 iirc, you had -I./ in your makefile, right? Aug 27 08:53:30 and ./ is relative. Aug 27 08:53:37 and it might change to something else in the future Aug 27 08:53:39 etc Aug 27 08:54:01 fine. then do CPPFLAGS="-I./" the flags will be used in your makefile, so you will get the same command line as you have otherwise. Aug 27 08:54:05 it would be very fragile to try to set from outside. Aug 27 08:54:31 are you sure it is not expanded at processing time? Aug 27 08:54:46 also, your line is still unclear. ;-) Aug 27 08:54:58 I have no idea if it matters where I put into the file. Aug 27 08:55:21 I will put into do_compile now before oe_runmake Aug 27 08:55:39 lpapp: ok... so i have no idea what you want. we told you that bitbake set CPPFLAGS (to avoid nasty cross compilation issues), so whatever is in your makefile will *never* be used. but there is a workaround which is to set it in your recipe. if you don't want to set CPPFLAGS in the recipe, then there is no solution. Aug 27 08:55:52 also, I am afraid that CPPFLAGS will be passed to everything... Aug 27 08:56:00 which could cause potential clashes. Aug 27 08:56:17 to everything in your recipe, right. Aug 27 08:56:19 actually, that is how you will get nasty compilation issues imho. Aug 27 08:56:29 no, not hte recipe what I meant. Aug 27 08:56:33 I meant the makefiles recursively, etc. Aug 27 08:56:37 the* Aug 27 08:56:46 yes, indeed. that's true. Aug 27 08:57:17 it might be that there are similar files, and since you would modify the include path for each subproject, they might get collided. Aug 27 08:57:25 and the whole will collapse. Aug 27 08:57:38 yes. true. that's why it was called a 'workaround'. Aug 27 08:58:13 well, it is not a workaround if it can blow up for me. Aug 27 08:58:26 i think someone suggested to unset CPPFLAGS (another workaround), if you do that before calling make, you should then be able to control CPPFLAGS from your makefile. Aug 27 08:58:49 I already replied to that above Aug 27 08:58:54 if it had worked, we would not discuss it now... Aug 27 09:06:51 can I add the Makefile to SRC_URI? Aug 27 09:06:57 or usually people will package those? Aug 27 09:07:13 so a test case would only contain a main.cpp and Makefile. Aug 27 09:10:45 is it possible to switch the LIC_FILES_CHKSUM off when doing a skeleton? Aug 27 09:13:26 ok, copied one over from skeleton.. Aug 27 09:16:57 ndec: bluelightning http://paste.kde.org/~lpapp/p83c389f0/ Aug 27 09:17:06 make and make clean work manually. Aug 27 09:17:39 er, I think you may not have set S correctly... Aug 27 09:17:53 or it wouldn't be complaining about not finding a makefile Aug 27 09:18:26 bluelightning: have no any idea I am afraid ... Aug 27 09:18:55 bluelightning: http://paste.kde.org/~lpapp/pdc63ff52/ Aug 27 09:20:02 lpapp: yep, you'll need S = "${WORKDIR}" to match your SRC_URI Aug 27 09:20:16 bluelightning: where? Aug 27 09:20:21 in the recipe Aug 27 09:20:30 since you're not extracting an archive that contains a directory with the source in it Aug 27 09:20:44 bluelightning: I am still not getting it where. Aug 27 09:21:03 just otu side anything? Aug 27 09:21:06 outside* Aug 27 09:21:17 yep, it's not critical where it appears Aug 27 09:21:56 bluelightning: bitbake testcase -c clean && bitbake testcase -> still fails. Aug 27 09:22:04 lpapp: there is commit in meta-qt5 which removes -e from EXTRA_OEMAKE, but calling unset is easier if you care about only 2 variables Aug 27 09:22:54 hmm, I think I will need this, hellomake: hellomake.c hellofunc.c gcc -o hellomake hellomake.c hellofunc.c -I. Aug 27 09:24:10 bluelightning: I do not understand why it is this hard to get a skeleton done Aug 27 09:24:14 is there any documentation about it? Aug 27 09:24:26 it should be the "Hello Makefile bitbake recipe" stuff. Aug 27 09:24:35 I guess it is useful for people when debugging. Aug 27 09:24:38 there is one in the doc, iirc. Aug 27 09:24:50 right, we have an example of a makefile-only recipe in the manual Aug 27 09:25:04 http://www.yoctoproject.org/docs/current/dev-manual/dev-manual.html#usingpoky-extend-addpkg-singlec Aug 27 09:25:41 I do not see how it is the same. Aug 27 09:25:49 it is not even using explicit Makefiles. Aug 27 09:25:56 next sample does... Aug 27 09:26:36 I do not see any example in there including a Makefile. Aug 27 09:27:23 they seem to use the compiler either directly or upstream build systems. Aug 27 09:27:31 those are not good for makefile related issue debugging. Aug 27 09:27:35 with a minimal test skeleton. Aug 27 09:27:53 lpapp: http://www.yoctoproject.org/docs/current/dev-manual/dev-manual.html#usingpoky-extend-addpkg-makefile ??? Aug 27 09:28:29 Hmm. Aug 27 09:28:33 Guys, I have a rather complex issue. Aug 27 09:28:41 I need to sed on a Makefile.PM Aug 27 09:28:59 Since I cannot pass the arguments I need, but I can pass them to the ./configure call within the script. Aug 27 09:29:07 bluelightning: http://paste.kde.org/~lpapp/pbf1e9fab/ Aug 27 09:29:13 However... ${D} expands in such a way that it overlaps with sed metacharacters. Aug 27 09:29:18 So, I tried this: clean_D = sed -i 's/\//\\\//' ${D} Aug 27 09:29:32 ndec: ??? Aug 27 09:29:37 But when I used ${clean_D} in the second sed (on the makefile), I'm told that clean_D does not exist. Aug 27 09:29:37 Where does that include any Makefile ??? Aug 27 09:29:44 Any suggestions for how I could proceed? Aug 27 09:29:48 Stygia: use some other separator character e.g. : or ! Aug 27 09:29:50 it is fetching a full blown upstream repository. Aug 27 09:30:03 bluelightning, ... I always forget you can even do that. Aug 27 09:30:07 bluelightning, But for future reference. Aug 27 09:30:16 bluelightning: why complaints from ld? Aug 27 09:30:16 bluelightning, How would I assign a variable? Is that possible, if I needed it? Aug 27 09:31:03 lpapp: the command being run looks invalid to me... Aug 27 09:31:21 bluelightning: why is yocto running an invalid command for a valid Makefile? Aug 27 09:31:48 ok, unset CPPFLAGS makes run it through. Aug 27 09:31:55 I presume I need to check the log for verbose make. Aug 27 09:32:08 or is make running with VERBOSE=1 by default? Aug 27 09:32:19 Stygia: in shell script I think you would actually want something like: varname=`echo "${D}" | sed 's/\//\\\//'` Aug 27 09:33:01 bluelightning, Hmm alright, but it is possible to assign my own variables and expand them with ${VAR} somewhere, then. Cool. :) Aug 27 09:33:09 hmm, how can I assume which the latest log is in temp? Aug 27 09:34:17 there is a symlink to the latest one. Aug 27 09:35:04 Stygia: if you mean outside of a shell function, you can't really execute an external command as shell like that in order to assign a value to a variable, but you can call into python code with VARNAME = "${@...}" Aug 27 09:35:43 bluelightning, Hmm. I suppose there's some comprehensive docs for this somewhere? Aug 27 09:37:06 Stygia: there's the bitbake manual; it's in need of some updates though (work is ongoing in that area) Aug 27 09:37:08 bluelightning, I'll go read the megamanual sometime. It's alright, I'm gradually getting this silly recipe to work. :) Aug 27 09:42:58 ndec: you mean log.do_compile? Aug 27 09:44:27 hmm, indeed, unsetting works for the skeleton! Aug 27 09:45:12 JaMa: hmm, why is -e removed in meta-qt5? Aug 27 09:45:23 JaMa: because it should be dropped at most places anyway? Aug 27 09:45:48 bluelightning: should my skeleton be proposed to meta-skeleton? Aug 27 09:45:52 bluelightning: or just blog post? Aug 27 09:47:14 Hey, I'm struggling with a bunch of error messages I don't quite understand, and that I haven't had much luck googling for. Aug 27 09:47:19 This is the recipe and error messages: http://pastebin.com/JAfGtbsc Aug 27 09:47:45 The more profilic ones seem to be "No GNU_HASH in the ELF binary" and "non-debug package contains .debug" Aug 27 09:48:02 If I don't have a FILES_${PN} listing, then it's fine, but obviously also worthless. Aug 27 09:48:39 Stygia: the trick is careful setting of FILES_* for the latter Aug 27 09:48:56 Stygia: the former I've never had to fix although I have seen it Aug 27 09:49:17 bluelightning: posted it in there, https://bugzilla.yoctoproject.org/show_bug.cgi?id=5046 Aug 27 09:49:18 Bug 5046: normal, Undecided, ---, scott.m.rifenbark, NEEDINFO , Document that CPPFLAGS override behavior and possible workaround Aug 27 09:49:41 bluelightning, So I should set FILES_* so as to avoid including the .debug file... alright. Aug 27 09:49:52 bluelightning, I'll try that first, see if it fixes the other one Aug 27 09:51:20 Stygia: set the FILES_${PN}-dbg first to contain the .debug dirs Aug 27 09:51:41 tf, Ah, then I can set FILES_${PN} to whatever without getting complaints? Aug 27 09:51:46 yes Aug 27 09:51:57 tf, Awesome. :) Thanks. Aug 27 09:53:08 tf: the order used to set the FILES_** vars matter? Aug 27 09:53:13 Sorry for asking so many newb questions, there may be some good docs for all this somewhere, I just apparently suck at setting them. Aug 27 09:53:37 *finding Aug 27 09:54:23 lpapp: because it provides more control about the environment, e.g. LD was incorrectly set and had to be unset in many tasks Aug 27 09:54:40 ndec: no, but the order of packaging does Aug 27 09:54:41 JaMa: fair enough Aug 27 09:55:04 And hmm. I fixed the critical error about the .debug file, but I still get "No GNU_HASH in the elf binary". Aug 27 09:55:19 Does this have to do with faulty/mis-targeted cross-compilation perhaps? Aug 27 09:55:25 tf: ah.. i see... i got it wrong. because you said 'set it first' i thought the 'order' was important and was surprised Aug 27 09:55:27 ;-) Aug 27 09:55:45 ndec: I realize I forumulated that wrong Aug 27 09:56:29 the order of packaging is given by the order of the PACKAGES variable Aug 27 09:56:39 i spent a couple of hours debugging FILES_** and i would be ashamed if i didn't notice the order was important ;-) Aug 27 09:56:48 which has sensible default Aug 27 09:57:11 but if you add your own packages, care needs to be taken Aug 27 10:02:37 Stygia: smthg wrong with LDFLAGS Aug 27 10:08:13 bluelightning: strange, I can print it from the skeleton. Aug 27 10:14:34 lpapp: maybe I miss something but you can add include paths i.e. CFLAGS in EXTRA_OEMAKE Aug 27 10:14:37 EXTRA_OEMAKE = "'CC=${CC}' 'RANLIB=${RANLIB}' 'AR=${AR}' 'CFLAGS=${CFLAGS} -I${S}include -DWITHOUT_XATTR' 'BUILDDIR=${S}'" Aug 27 10:15:04 bluelightning: finally got it working! :) Aug 27 10:15:25 ant_work: unfortunately, that does not work for the previously explained reason. Aug 27 10:16:02 bluelightning: arm-foo-linux-gnueabi-gcc: error: unrecognized command line option '-m32' Aug 27 10:16:07 getting this now, but the include path present. Aug 27 10:16:11 why is -m32 invalid for gcc? Aug 27 10:17:11 perhaps because it is an old version? Aug 27 10:34:50 ant_work, Hmm quite possible. Sorry for the delayed response btw, lunch. Aug 27 10:50:55 Heya. I found a recipe for libdevice-serialport-perl, which is just what I need, but it fails complaining that it can't find "oenote". What's this and how do I get it in my OE install? Aug 27 10:51:05 This recipe here BTW: http://cgit.openembedded.org/openembedded/tree/recipes/perl/libdevice-serialport-perl_1.04.bb Aug 27 10:53:52 It looks like a sort of debugging thing... a fance "print" if you will. Can I safely remove it? (My current MO) Aug 27 10:54:29 Stygia: oenote has been removed in 2003 Aug 27 10:55:32 ant_work, Ah... Well that repo is presumably far out of date, then. No worries though, I deleted it and the rest of the recipe was, well, more or less fine. Aug 27 10:55:35 Either way, my recipe works now. :) Aug 27 10:55:47 - oenote Executing intltoolize --copy --force --automake Aug 27 10:55:47 + bbnote Executing intltoolize --copy --force --automake Aug 27 10:56:24 Hello all Aug 27 10:57:03 I have one issue and want to check with you guys for a fix. Aug 27 10:57:27 agherzan: go ahead Aug 27 10:57:41 I want to base my distro on poky distro. Now, poky adds x11 as distro feature but i want to have that switchable to fb, directfb or wayland. Aug 27 10:58:36 Am i able to get out all x11, directfb, wayland etc from DISTRO_FEATURES and add my preferred backend? Aug 27 10:59:04 it seems i can't use a python function for this in a distro conf Aug 27 10:59:09 agherzan: perhaps you wanna use core-image-minimal? Aug 27 10:59:13 or some image on top of that? Aug 27 10:59:23 I mean surely, an image with your additions. Aug 27 10:59:24 ant_work, Hmm alright. It's just a "print" type statement, though, isn't it? I'd leave that out personally. Aug 27 10:59:38 lpapp: image? i'm talking about distro feature Aug 27 10:59:51 distro conf sorry Aug 27 10:59:59 agherzan: you can add any package to your distro Aug 27 11:00:20 you didn't understand. i want to remove something from distro feature in a distro conf Aug 27 11:00:37 Stygia: yes, is in logging.bbclass Aug 27 11:01:03 ant_work, Alright then, unless nobody complains I'll leave it out, then (And you should complain if you want, I'm gonna be pushing all these recipes up). Aug 27 11:01:09 Stygia: in a python task you could use bb.note( Aug 27 11:01:25 agherzan: that does not make much sense to me at least. Aug 27 11:01:31 agherzan: to inherit something to drop it then. Aug 27 11:01:37 why not just not inherit in the first place? Aug 27 11:01:43 i.e. building your distro on top of oe? Aug 27 11:01:50 poky is only a very thin layer after all. Aug 27 11:01:53 ant_work, Hmm alright? Good to know, but it's a do_configure_prepend Aug 27 11:02:12 ant_work, I'm not sure what python tasks have to do with this? :) Aug 27 11:02:20 http://hambedded.org/blog/2012/11/24/from-bitbake-hello-world-to-an-image/ Aug 27 11:02:40 he he , I kept it to show most print() stuff ;) Aug 27 11:03:02 agherzan, I would think so, yup. if you pass the proper flags to X11 and so on. Aug 27 11:03:19 agherzan, You _should_ be able to create or bbappend the existing recipes to get where you want to be. Aug 27 11:03:23 agherzan: you can run phython bits in assignment to variables in distro.conf the usual way Aug 27 11:03:32 agherzan, Generally speaking, if you can do it manually, you can do it with bb Aug 27 11:03:36 yeah, that too. Aug 27 11:03:36 tf that was what i was looking for i think Aug 27 11:03:37 ant_work, Oh fancy. Thanks. :) Aug 27 11:04:02 tf: python __anonymous() { doesn not work.... Aug 27 11:04:07 agherzan: what is the use case btw? Aug 27 11:04:44 agherzan: no, I mean in assignments, like https://github.com/Guacamayo/meta-guacamayo/blob/master/meta-guacamayo/conf/distro/guacamayo.conf Aug 27 11:05:28 agherzan: you should be able to process the distro features that way, I think Aug 27 11:07:22 tf: is there any product based on this distribution? Aug 27 11:15:40 tf: but i can't define a function in the .conf file? a python function Aug 27 11:22:42 So i understand that there is no support for unning python code in a distro conf file right? Aug 27 11:23:04 tf: i actually need a function or so to get out some features fron distro_features... Aug 27 11:24:36 agherzan: you can inherit a bbclass where you can put the function Aug 27 11:24:53 I have wondered if we should relax the rules about conf files... Aug 27 11:25:23 RP: what is wrong about bbclasses Aug 27 11:26:36 RP: So in the distro itself i can't define python code right? Aug 27 11:26:58 agherzan: correct, .conf files don't support function syntax Aug 27 11:27:09 RP: thanks. Aug 27 11:28:29 RP: unparsed line: 'inherit backend' Aug 27 11:28:58 agherzan: INHERIT from config context Aug 27 11:29:09 and you need to append to it Aug 27 11:29:24 INHERIT += "backend" Aug 27 11:30:15 RP: yes. got it now Aug 27 11:30:32 agherzan: https://git.yoctoproject.org/cgit/cgit.cgi/poky/tree/meta-yocto/conf/distro/poky.conf#n101 Aug 27 11:30:39 just for a real world example. ^ Aug 27 11:31:39 hi all, I have build a Poky image, but ny rootfs does not heve ipkg manager included (not dpkg). How to instruct Youcto to include this in the rootfs build? Aug 27 11:32:12 drasko_: the default is rpm. Aug 27 11:32:15 sec... hold on. Aug 27 11:32:34 lpapp, I have changed this in conf/local.conf Aug 27 11:32:47 drasko_: you need to add package_manager Aug 27 11:32:48 I am using ipk on my embedded device Aug 27 11:32:58 agherzan, exactly. How to do this? Aug 27 11:33:00 drasko_: then it should be fine. Aug 27 11:33:01 drasko_: http://www.yoctoproject.org/docs/current/mega-manual/mega-manual.html#setting-up-runtime-package-management Aug 27 11:33:42 drasko_: its called opkg fwiw Aug 27 11:33:51 yes, opkg Aug 27 11:34:03 but how to instruct Yocto to include it in image? Aug 27 11:34:13 drasko_: see the link above Aug 27 11:34:20 It should be present in rootfs Aug 27 11:34:33 drasko_: regular poky image? Aug 27 11:34:40 drasko_: it depends whether you configure your image to include a package manager or not Aug 27 11:35:03 RP, I am looking for a way to do this configuration. Aug 27 11:35:06 How is this done? Aug 27 11:35:09 drasko_: have you set up package_ipk? Aug 27 11:35:23 lpapp, no. I do not know how to do this Aug 27 11:35:34 this is exactly what I am asking Aug 27 11:35:40 drasko_: ok, so your locan conf is not set up properly then. :) Aug 27 11:35:56 no, this I have set Aug 27 11:36:09 PACKAGE_CLASSES ?= "package_ipk" Aug 27 11:36:16 drasko_: see the link above, http://www.yoctoproject.org/docs/current/mega-manual/mega-manual.html#setting-up-runtime-package-management Aug 27 11:36:17 no probelm there, packages are produced Aug 27 11:36:27 drasko_: please read the link ndec pasted, this is documented in the manual Aug 27 11:36:33 my problem is that when I transfer package to the board Aug 27 11:36:37 I can not install them Aug 27 11:36:42 drasko_: yes, we understood Aug 27 11:36:45 because opkg is missing from the rootfs Aug 27 11:37:15 drasko_: "When you build your image, you select to use the appropriate package manager by setting the PACKAGE_CLASSES variable." Aug 27 11:37:21 "You enable package management on the target by listing "package-management" in the IMAGE_FEATURES variable." Aug 27 11:37:40 drasko_: all you need is: IMAGE_FEATURES += " package-management" Aug 27 11:38:01 agherzan: + PACKAGE_CLASSES at least. Aug 27 11:38:11 drasko_: http://www.yoctoproject.org/docs/current/mega-manual/mega-manual.html#var-PACKAGE_CLASSES Aug 27 11:38:55 lpapp, I have set PACKAGE_CLASSES correctly Aug 27 11:39:04 I am missing maybe IMAGE_FEATURES Aug 27 11:39:16 drasko_: I do not follow anymore. :) Aug 27 11:39:36 but make sure you have both. :) Aug 27 11:39:44 thanks! Aug 27 12:23:19 JaMa: I'm curious if the hangs/Ctrl+C behaviour are looking any better in your general usage? Aug 27 12:29:25 is there some trick for yocto execute makefile recursion? Aug 27 12:29:29 to* Aug 27 13:18:50 HI all. In order to add package manager to the image I have set IMAGE_FEATURES += "package-management" Aug 27 13:19:03 However, only rpm package manager was built in Aug 27 13:19:18 How to add other package manager, i.e. opkg? Aug 27 13:19:48 the package manager used is the one specified in the package class value in the local.conf Aug 27 13:20:03 if you have multiples listed there, I believe it's the last one... (but there should be a comment to indicate which) Aug 27 13:20:13 the package-management system installed is the one which was used to generate the image Aug 27 13:20:17 fray, PACKAGE_CLASSES ?= "package_ipk" Aug 27 13:20:21 yes Aug 27 13:20:24 this is only thing in my local.conf Aug 27 13:20:52 if RPM was installed, and that is your configuration -- then something you built specifically asked for 'rpm'.. Likely an LSB component Aug 27 13:20:59 (rpm is required for LSB compliance) Aug 27 13:21:11 do I have to do some "clean" of previous build for this to be taken into account? Aug 27 13:21:19 no, you should not Aug 27 13:21:26 I can see Yocto building many ipk packages Aug 27 13:21:44 but on the image it pu rpm manager Aug 27 13:22:28 actually I think it should be 'EXTRA_IMAGE_FEATURES += "package-management"' Aug 27 13:22:35 but that's from emmory, I might be wrong Aug 27 13:22:50 https://wiki.yoctoproject.org/wiki/How_do_I Aug 27 13:22:52 yes.. likely you built/installed something that enabled LSB compliance components... 'rpm' is one of those components Aug 27 13:23:46 how to diasble this and force opkg? Aug 27 13:24:03 try EXTRA_IMAGE_FEAUTRES += ... instead Aug 27 13:24:21 if that works, update the wiki -- or ask someone with permission to. Aug 27 13:24:43 (I'm working off master, not any of the older branches.. it's very possible something has changed in the last 6 months or so..) Aug 27 13:24:53 and leave IMAGE_FEATURES unpopulated? Aug 27 13:24:57 yes Aug 27 13:25:04 OK, let me try rebuilding Aug 27 13:26:11 EXTRA_IMAGE_FEATURES ??= "" Aug 27 13:26:12 IMAGE_FEATURES += "${EXTRA_IMAGE_FEATURES}" Aug 27 13:26:25 that is being done within the bitbake.conf file. Due to ordering that -should- cause it to load 'later' Aug 27 13:26:39 doing IMAGE_FEATURES += ... should ahve worked, but orderign may be affected.. Aug 27 13:26:48 if this still doesn't work, I'm not sure what the problem is Aug 27 13:27:10 I was confused because of the comments in local.conf template Aug 27 13:27:24 stating available options for EXTRA_IMAGE_FEATURES Aug 27 13:27:35 if you look at meta/classes/image.bbclass -- you should see 'package-management' listed as one of the 'IMAGE_FEATURES[validitems] .. if you don't maybe it's not supported in the version you are using? Aug 27 13:27:40 and package-management is not there Aug 27 13:27:46 ya, it looks like someone missed adding it there.. it should be listed Aug 27 13:27:56 along w/ 'read-only-rootfs'.. Aug 27 13:28:26 those two were added somewhat recently, at least within the last year Aug 27 13:28:56 bitbake is on course Aug 27 13:28:59 ;) Aug 27 13:29:56 the other thing you can do is run: bitbake -e -- and then check the value of 'IMAGE_FEATURES'. It -should- show built-ins and the ones you added.. the debug information should show it inherited the stuff from EXTRA_IMAGE_FEATURES Aug 27 13:34:50 the kconf_check script from kern-tools checks for non-hardware-specified kernel configuration options. for example, CONFIG_EXT3_FS & CONFIG_INET Aug 27 13:35:34 however, the yocto-bsp layer creation script creates a layer that contains several of these options Aug 27 13:36:51 which is more correct, kconf_check, or yocto-bsp? Aug 27 13:43:01 j8 -- I'd defer to zedd if he ever wakes up.. Aug 27 13:43:30 but the check is really a way to verify that what you are doing is reasonable.. if it works in the end it's a warning.. if it doesn't, it might point you to the source of a problem.. Aug 27 13:44:12 if this is the check I'm thinking of, it's telling you that a configuration fragment for a piece of hardware is also enabling generic support, for EXT3 or whatever.. which would make it difficult to disable it if you wanted to do that Aug 27 13:45:05 fray, putting EXTRA_IMAGE_FEATURES += "package-management" does not work Aug 27 13:45:15 now even rpm was not present on the rootfs Aug 27 13:45:29 ok.. then you need to first verify that package-management is supported.. look at meta/classes/image.bbclass, grep for package-management.. Aug 27 13:45:42 (what version of YP code are you using?) Aug 27 13:45:53 fray, a cusom one from Altera Aug 27 13:46:31 ok.. last one I knew of from them was Danny.. but I don't know if thats what you are using.. Aug 27 13:46:42 best to check that file quickly.. Aug 27 13:46:53 ROOTFS_BOOTSTRAP_INSTALL = "${@base_contains("IMAGE_FEATURES", "package-management", "", "${ROOTFS_PKGMANAGE_BOOTSTRAP}",d)}" Aug 27 13:46:58 * lpapp suspects -c clean actually does not make a proper clean Aug 27 13:47:21 ok.. next try bitbake -e then look at the output for 'IMAGE_FEATURES' Aug 27 13:47:50 see if it contains the package-management in it.. Aug 27 13:48:05 fray, but I added package-management to EXTRA_IMAGE_FEATURES... Aug 27 13:48:14 anyway... Aug 27 13:48:17 correct, but IMAGE_FEATURES should inherit the values of EXTRA_IMAGE_FEATURES Aug 27 13:48:30 they do in the current version, but this may have been a change made -after- the version of YP you are using Aug 27 13:48:34 http://pastebin.com/zAGWBtgL Aug 27 13:49:19 looks like it is not inherited Aug 27 13:49:25 ya, I'd agree.. Aug 27 13:49:33 so what you had the first time was likely right.. Aug 27 13:49:39 yes... Aug 27 13:49:46 but it created only rpm manager Aug 27 13:49:51 change EXTRA_IMAGE_FEATURES = "debug-tweaks package-management" back to just IMAGE_FEATURES Aug 27 13:50:00 thats the part I don't understand.. it shouldn't have.. Aug 27 13:50:23 you might want to do the same thing (bitbake -e), and check the value of of the PACKAGE_CLASSES Aug 27 13:50:34 maybe something is overriding or adding to it? Aug 27 13:50:49 The altera stuff I worked on was using 'rpm', not ipk... some maybe something is broken in the integration Aug 27 13:51:17 yes... Aug 27 13:51:45 I will try and if not I will stick with rpm for the moment and contact directly Altera for support Aug 27 13:51:46 thanks Aug 27 13:52:07 no problem.. let me know if you figure it out.. I'm curious as to what is wrong Aug 27 13:55:12 hmmm... I just falled back to first version of my local.conf change : IMAGE_FEATURES += "package-management". And when I check output, I have : http://pastebin.com/QC8SriaR Aug 27 13:55:19 so this does not look good neither Aug 27 13:56:03 it looks to me like altera-image may be clearingthe image_features.. Aug 27 13:56:18 I think that I have to add change here : ./meta-altera/recipes-core/images/altera-image.inc Aug 27 13:56:22 if you look at the comment lines around the # IMAGE_FEATURES= comment it should show you Aug 27 13:56:42 ya, if they are hard coding the image_features there, that is the problem Aug 27 13:56:47 here : http://www.yoctoproject.org/docs/current/mega-manual/mega-manual.html#var-IMAGE_FEATURES Aug 27 13:56:54 Although you can use this variable from your local.conf file, which is found in the Build Directory, best practices dictate that you do not. Aug 27 13:57:05 you can do: IMAGE_FEATURES_append = " package-management" (note the additional space) Aug 27 13:57:05 Typically, you configure this variable in an image recipe. Aug 27 13:57:37 ya.. that is why EXTRA_IMAGE_FEATURES was added.. to allow base features to be defined.. but the EXTRA... were still supposed to be applied.. Aug 27 13:57:41 OK, let me try first modufying image recipe Aug 27 13:57:57 usually an image will have something like: Aug 27 13:58:16 IMAGE_FEATURES = "my_feature my_other_feature ${EXTRA_IMAGE_FEATURES}" Aug 27 13:58:31 then you set the extra_image_features like I had yout ry earlier Aug 27 13:58:37 I see, this makes sense Aug 27 13:58:47 alternatively in your local.conf you can do: Aug 27 13:58:57 IMAGE_FEATURES_append = " ${EXTRA_IMAGE_FEATURES" Aug 27 13:59:03 'er.. make that Aug 27 13:59:09 IMAGE_FEATURES_append = " ${EXTRA_IMAGE_FEATURES}" Aug 27 13:59:22 EXTRA_IMAGE_FEATURES = "debug-tweaks package-management" Aug 27 13:59:32 the _append will cause the feature to be added unconditionally Aug 27 14:00:20 I can see now opkg compiling... Aug 27 14:00:27 excellent Aug 27 14:02:26 Looks better : http://pastebin.com/mtXMTe9x ;) Aug 27 14:02:37 I will test it, but it should work now Aug 27 14:02:39 thanks a lot Aug 27 14:02:43 ya.. Aug 27 14:03:01 you may want to add the debug-tweaks to that.. otherwise you won't be able to log-in as root.. :) Aug 27 14:03:20 according to your grep, debug-tweaks is already in EXTRA_IMAGE_FEATURES Aug 27 14:03:37 EXTRA_IMAGE_FEATURES="debug-tweaks" is already there... Aug 27 14:04:00 ya.. but it's not showing up in IMAGE_FEATURES (for the same reason that package-management wasn't.. so it's not going to do anything) Aug 27 14:14:52 zibri: did you ever get to the bottom of the syslinux -malign-labels build failure? Aug 27 14:15:04 tf: nope :( Aug 27 14:16:07 I am seeing it as well Aug 27 14:16:43 fray, actually ./ipk/armv7ahf-vfp-neon/opkg_0.1.8+svnr633-r10.7_armv7ahf-vfp-neon.ipk was produced Aug 27 14:16:52 but it is not installed on rootfs Aug 27 14:16:55 good.. that sounds right.. Aug 27 14:16:59 I have to install it manually Aug 27 14:17:10 I'm not sure why that would be.. it should be automatic Aug 27 14:17:12 but can not, as it is ipk :( Aug 27 14:17:58 it seems to me like they hae something wrong in their image recipe, which is causing other things to ignore values or do the wrong thing Aug 27 14:19:15 tf: if you learn anything, please let me know. i was looking at the syslinux build system, but i'm still a bit confused Aug 27 14:21:07 zibri: I think the malign-labels option is only valid on FRV? Aug 27 14:22:36 it's added as a fallback to -falign-labels=0 Aug 27 14:22:40 GCCOPT += $(call gcc_ok,-falign-jumps=0,-malign-jumps=0) Aug 27 14:23:00 where gcc_ok tries to compile a dummy.c with the given flag. Aug 27 14:24:47 yes, but that's a bug in syslinux, since that option is not available Aug 27 14:25:52 yes, sure. but i'm not sure why -falign-labels isn't available. Aug 27 14:26:07 indeed Aug 27 14:26:27 or rather, why gcc_ok don't think it is Aug 27 14:27:51 zibri: something to do with the host compiler Aug 27 14:28:34 at least here it's syslinux-native that's failing Aug 27 14:28:43 same for me Aug 27 14:28:59 doh, you running gcc 4.8? Aug 27 14:29:44 indeed Aug 27 14:30:29 you figured out something? :) Aug 27 14:41:49 zibri: looking at the log, all the gcc_ok tests fail Aug 27 14:42:07 the intall log Aug 27 14:52:54 tf: ah, it fails because of the other issue i have (in dylan). "../../dummy.c:1:0: error: CPU you selected does not support x86-64 instruction set" Aug 27 14:59:21 YPTM: We are about to start the Yocto Project Techincal Meeting (YPTM), this call is open to all Aug 27 14:59:27 Dial-in number: 1.972.995.7777 Aug 27 14:59:32 Participant passcode: 42001078 Aug 27 14:59:47 YPTM: AlexG online Aug 27 14:59:56 YPTM: Björn is on the call Aug 27 15:00:04 YPTM: Micheal on the call. Aug 27 15:00:08 YPTM: MatthewW on the call Aug 27 15:00:10 If you have any issues feel free to join the call, this does not prevent other discussion here Aug 27 15:00:22 YPTM: Saul is on Aug 27 15:00:54 YPTM: Bruce is on Aug 27 15:00:58 YPTM: Denys is here Aug 27 15:01:49 * zeddii hears nothing Aug 27 15:02:04 YPTM: Nitin is on the bridge Aug 27 15:02:14 YPTM: davest here Aug 27 15:02:30 YPTM: Paul Eggleton is on Aug 27 15:02:53 YPTM: Tom Z on the call Aug 27 15:02:53 YPTM: Polk is here Aug 27 15:02:55 YPTM: Welcome to the technical team meeting. Please let me know who's on the bridge. Thanks! Aug 27 15:03:10 YPTM: Belen is on the call Aug 27 15:03:10 Song_Liu: Richard is on the call Aug 27 15:03:54 YPTM: Beth is on Aug 27 15:03:56 YPTM: HAve an open regarding 4 tc's: https://bugzilla.yoctoproject.org/tr_show_case.cgi?case_id=290 https://bugzilla.yoctoproject.org/tr_show_case.cgi?case_id=291 https://bugzilla.yoctoproject.org/tr_show_case.cgi?case_id=292, and https://bugzilla.yoctoproject.org/tr_show_case.cgi?case_id=169 Aug 27 15:04:04 YPTM: any opens? Aug 27 15:04:15 Song_Liu: automated testing on real hardware Aug 27 15:04:18 YPTM: Darren is on Aug 27 15:04:24 YPTM: jzhang on the call Aug 27 15:04:41 Song_Liu: autobuilder during M5. Aug 27 15:05:21 Song_Liu: M4 and 1.4.2 Aug 27 15:05:33 AlexG: I just looked at those test cases and I think its fine to drop them Aug 27 15:05:47 RP: thanks. Aug 27 15:06:34 YPTM: THe full pass report for 1.4.2 is available: https://wiki.yoctoproject.org/wiki/Full_Pass_Test_Report_for_Yocto_1.4.2_RC1_2013-08-18_build Aug 27 15:06:41 AlexG: They're all covered in the autobuilder now Aug 27 15:06:47 great! :) Aug 27 15:07:29 AlexG: these are one of the kinds of tests I wanted to automate away Aug 27 15:09:44 https://www.google.com/calendar/embed?height=600&wkst=1&bgcolor=%23FFFFFF&src=mg0m150m30gs891eqqhtfu5jgg%40group.calendar.google.com&color=%238C500B&src=sc7rov4ck552t2f0pav120t7r0%40group.calendar.google.com&color=%23528800&src=theyoctoproject%40gmail.com&color=%23182C57&ctz=America%2FLos_Angeles Aug 27 15:10:54 https://wiki.yoctoproject.org/charts/combo.html Aug 27 15:13:14 where do I find the 1.4.2 schedule? Aug 27 15:17:30 https://www.google.com/calendar/embed?src=3a0927ctps51u8rs2jv1uligdc%40group.calendar.google.com&ctz=America/Los_Angeles Aug 27 15:27:01 thread starts here FYI: https://lists.yoctoproject.org/pipermail/yocto/2013-August/017716.html Aug 27 15:30:26 does anybody know where /usr/share/X11/xorg.conf.d/10-evdev.conf is coming from Aug 27 15:31:05 I would say meta/recipes-graphics/xorg-driver/xf86-input-evdev_2.8.1.bb but it doesn't seem to be the case Aug 27 15:32:40 Can anyone help with this? $ arm-angstrom-linux-gnueabi-gcc --version Aug 27 15:32:41 arm-angstrom-linux-gnueabi-gcc: error while loading shared libraries: __vdso_time: invalid mode for dlopen(): Invalid argument Aug 27 15:32:41 I used MACHINE=beaglebone ./oebb.sh bitbake meta-toolchain to build the tool chain and use the generated sh script from the SDK to install it. Aug 27 15:32:41 The generated shell script was named "angstrom-eglibc-x86_64-cortexa8hf-vfp-neon-v2013.06-toolchain.sh" Aug 27 15:32:41 My development machine is x86_64 Ubuntu 12.04.2 Aug 27 15:32:43 and I did "$ source /usr/local/oecore-x86_64/environment-setup-cortexa8hf-vfp-neon-angstrom-linux-gnueabi" prior to trying the arm-angstrom-linux-gnueabi-gcc --version Aug 27 15:32:48 Any insights to what I'm doing wrong would be greatly appreciated. Thanks Aug 27 15:36:45 YPTM: thank you all for joining the meeting, you all have a nice day/evening Aug 27 15:37:15 tf: i got it working, i think. by installing the debian package libc6-dev-i386 on the build host and adding -m32 to GCCOPT :-(... so not a pretty solution. Aug 27 15:37:56 i'm not sure if syslinux has to be built for i386, and how to solve it properly in that case. Aug 27 15:38:26 zibri: hmm Aug 27 15:40:38 zibri: right, there is a -m32 parameter already, but that too uses the wretched gcc_ok snippet Aug 27 15:40:49 so that's getting lost Aug 27 15:41:21 ah, of course. so the addition of libc6-dev-i386 should have been enough Aug 27 15:41:37 not sure Aug 27 15:42:11 one way to find out, but I got to go now Aug 27 15:43:18 it's a bit funny that all of the building takes place in do_install as well... Aug 27 15:43:29 "all" Aug 27 15:46:47 zibri: that's usually an indication that something is very wrong :/ Aug 27 15:59:39 http://paste.kde.org/~lpapp/pf44b6a56/ -> what could cause this error error message when building an own software/ Aug 27 15:59:42 ? Aug 27 15:59:44 Does anyone know the syntax for a PREMIRROR line that includes more than one destinations to search? All the examples I've seen only have a single search destination for one protocol (e.g. http). Aug 27 16:00:32 Glenn: sorry? The documentation has several exampes. Aug 27 16:00:34 examples* Aug 27 16:00:35 Glenn: see meta/classes/mirrors.bbclass Aug 27 16:00:37 https://www.yoctoproject.org/docs/current/poky-ref-manual/poky-ref-manual.html Aug 27 16:00:42 Glenn: see the variable definition Aug 27 16:00:46 (MIRRORS and PREMIRRORS are the same syntax) Aug 27 16:01:00 Glenn: PREMIRRORS_prepend = "\ Aug 27 16:01:00 git://.*/.* http://www.yoctoproject.org/sources/ \n \ Aug 27 16:01:00 ftp://.*/.* http://www.yoctoproject.org/sources/ \n \ Aug 27 16:01:00 http://.*/.* http://www.yoctoproject.org/sources/ \n \ Aug 27 16:01:01 https://.*/.* http://www.yoctoproject.org/sources/ \n" Aug 27 16:01:30 Glenn: you could also look at the own-mirrors.bbclass which shows a multi protocol list and allows you to set it with one URI Aug 27 16:01:49 That's not what I want. I want more than one destination for a protocol (e.g ftp://.*/.* should should search more than one location). Aug 27 16:02:09 Glenn: and how is that different in theory? Aug 27 16:02:19 you can prepend more lines. Aug 27 16:02:21 no? Aug 27 16:02:37 Don't thin so Aug 27 16:02:46 sure, you can Aug 27 16:02:54 if you check the example classes provided above, you will see. Aug 27 16:03:23 you can also use the += syntax. Aug 27 16:03:30 Glenn: just add more lines Aug 27 16:03:40 it will try each matching line Aug 27 16:05:50 here is my full error message: http://paste.kde.org/~lpapp/p7e52f101/ Aug 27 16:06:28 sgw_: would fltk be acceptable for oe-core? Aug 27 16:07:22 I probably need to inherit autotools? Aug 27 16:08:15 basically I have a small software using raw GNU Makefile. Aug 27 16:08:29 and that invokes autoconf, configure, and then make for a nested subfolder's makefile. Aug 27 16:08:40 do I need to inherit autotools in such a scenario? Aug 27 16:15:05 how will autotools know whether to run autoreconf or ./boot.sh? Aug 27 16:22:20 I believe --enable-maintainer-mode or some other option will make automake generate a makefile that will re-run autoreconf Aug 27 16:22:31 but normally you're supposed to run it explicitly Aug 27 16:22:41 ./boot.sh is not a standard script afaik Aug 27 16:24:10 I tried appending more lines but it didn't appear to work. I was just looking for a concrete example. If you tell me a PREMIRRORS_append += will continue to add more search destinations for a given protocol I will trust you and try again. Aug 27 16:26:39 Glenn: have you checked the examples? Aug 27 16:26:41 do they work? Aug 27 16:27:18 Which one precisely. I looked at several of them already. Looking at mirrors.bbclass right now. Aug 27 16:27:30 that should have been the first, yeah. Aug 27 16:29:32 What mirrors.bbclass seems to indicate that the same protocol can be listed more than once with each line having a different search path. Is that a safe assessment? Aug 27 16:30:28 Glenn: why do you think it is restricted to protols or urls? Aug 27 16:30:33 they can be mixed as you wish, no? Aug 27 16:32:00 I was trying to specify more than one search location on the same protocol line, e.g. PREMIRRORS_append = "git://.*/.* file://toplevel/premirror ftp://ftp.acme.com/pre_mirror/foo \n" Aug 27 16:32:14 Glenn: why/ Aug 27 16:32:15 ? Aug 27 16:32:16 Using a space as a separator. Aug 27 16:32:27 Glenn: why haven't you tried in separate lines for starter? Aug 27 16:33:13 yes, I'm doing that now and getting hit/miss results. Aug 27 16:33:17 Still investigating Aug 27 16:35:48 Glenn: if you want a second mirror url, then add additional lines. \n is the separator between mirror definitions. Aug 27 16:37:37 But you still need to list the protocol part again, right, eg. PREMIRRORS_append = "git://.*/.* file://toplevel/premirror \n git://.*/.* ftp://ftp.acme.com/pre_mirror/foo \n " Aug 27 16:38:01 Glenn: yes, so? Aug 27 16:38:22 there is no groupping feature... Aug 27 16:38:51 It's not a problem. I just wanted to understand the syntax. I mistakenly thought you could somehow group them. Aug 27 16:40:03 Right now I'm seeing intermittent behavior. It works correctly on some runs and then on others it fails. Could be a problem with the reliability of my FTP destination. Sometimes it works, sometimes it doesn't. Aug 27 16:41:30 Glenn: ok, good. Aug 27 16:43:32 any idea what host should I pass to configure? | configure: error: cannot run C compiled programs. Aug 27 16:43:35 | If you meant to cross compile, use `--host'. Aug 27 16:49:24 Glenn: the syntax is for each entry, newline separated. that's it, no shortcuts Aug 27 16:51:14 Glenn: you could report an FR if you wished, though. Aug 27 16:52:27 Thanks for the info. I think I'm fighting a crappy link. Occasionally it will time out which causes it to fail. No need yet for any FR. Aug 27 16:52:28 is there an image type that partitions the hdd or flash disk into to (boot, home) ? Aug 27 16:52:34 currently I am testing the image with liveimage Aug 27 16:52:44 the changes in the /etc directory is temporary Aug 27 16:53:03 eren: there is a raspberry pi BSP on github that has one Aug 27 16:53:42 eren: it makes an SD card image with an fat32 boot partition (the hardware needs it) followed by an ext3 root Aug 27 16:54:08 http://git.yoctoproject.org/cgit/cgit.cgi/meta-raspberrypi/tree/classes/sdcard_image-rpi.bbclass Aug 27 16:54:17 yup Aug 27 16:54:43 I guess I will need to change a bit Aug 27 16:54:44 lpapp: either $HOST or $TARGET should work... Aug 27 16:54:51 because the platform I use it i586 Aug 27 16:54:55 regular pc with bios Aug 27 16:55:01 you mean via extra_oeconf? Aug 27 16:55:03 I mean, not that regular :) Aug 27 16:55:18 nerdboy: actually not in this case. Aug 27 16:55:35 nerdboy: it is inside a Makefile Aug 27 16:56:56 BCMM: oh, and I will need syslinux for booting :\ Aug 27 16:57:09 I will probably need to use something like HOSTCC Aug 27 16:57:17 $HOSTCC -dumpmachine Aug 27 16:57:22 eren: btw i'm not saying there isn't a better/more official example, just that that bbclass exists Aug 27 16:57:32 oh I see Aug 27 17:03:25 ffs why doesn't the kernel build see my cfg file Aug 27 17:10:30 Does anyone here actually develop anything using the Angstrom tool chains generated by OE/Yocto? Aug 27 17:10:30 From the angstrom-v2013.06-yocto1.4 git branch that is? Love to know how you got it working if you do. Aug 27 17:21:33 hello Aug 27 17:21:49 anybody here ? Aug 27 17:23:06 no Aug 27 17:23:10 jsut ask Aug 27 17:30:41 Dennis: sure, what is the question? Aug 27 17:32:22 http://pastebin.com/n057BH88 Aug 27 17:34:11 meta-toolchain-qte linking against /lib64/ld-linux-x86-64.so.2 using the angstrom-v2013.06-yocto1.4 git branch Aug 27 17:37:41 Crofton|work: what is the meaning of life? Aug 27 17:38:23 And, please don't ask/tell me to post it in #angstrom, it's already there! Aug 27 17:38:56 mranostay, http://www.youtube.com/watch?v=6PQ6335puOc Aug 27 17:39:26 cfo215: silly question but you did run the env setup script before using the binaries the SDK contains, right? Aug 27 17:39:36 mranostay, its a movie by "Monty Python's Flying Circus" Aug 27 17:40:35 . Aug 27 17:40:56 bluelightning, why of course! I did: source /usr/local/oecore-x86_64/environment-setup-cortexa8hf-vfp-neon-angstrom-linux-gnueabi Aug 27 17:41:08 cfo215: ok, just checking Aug 27 17:41:17 $ which qmake2 Aug 27 17:41:17 /usr/local/oecore-x86_64/sysroots/x86_64-angstromsdk-linux/usr/bin/qmake2 Aug 27 17:41:28 bluelightning, np Aug 27 17:42:00 cfo215, is getting a little fed up with the whole OE/Yocto/Angstrom doesn't do anything as documented circle.... Aug 27 17:42:15 I'm back and want to ask something about meta-freescale. Aug 27 17:42:28 Dennis, please just ask! Aug 27 17:42:42 ok thx, Aug 27 17:43:06 Dennis, everyone here is asking or answering something. You don't have to ask to ask... It's ok... we're here to help! Aug 27 17:45:23 there is master branch whcih has new 3.0.8 libfslcodec Aug 27 17:46:24 but otavio salvador commented AAC OGG codecs were removed. so i want to know why. Aug 27 17:46:33 so based on the last line of my pastebin (http://pastebin.com/n057BH88) it looks like qmake2 was linked against /lib64/ld-linux-x86-64.so.2 (0x00007f3e84b88000) if I'm understanding correctly. Aug 27 17:46:40 and how to put them back Aug 27 17:49:46 this is what I get when trying the qmake2 against the sample program: :~/qte-example$ qmake2 Aug 27 17:49:46 qmake2: error while loading shared libraries: __vdso_time: invalid mode for dlopen(): Invalid argument Aug 27 17:58:58 cfo215, I ahve an sdk, but no qt supoprt in it Aug 27 18:01:35 Crofton|work, I'm trying to build a custom sdk. one that includes libphidget21.so and Qt 4.8. I can build the system image just fine with qt4 and the phidgets library (which i use for RFID). I just can't get the sdk image to work to support that combination. Aug 27 18:01:56 using -c populate_sdk? Aug 27 18:02:19 Crofton|work, yes that really blows up. Aug 27 18:02:26 ? Aug 27 18:02:31 all sorts of errors. I'll pastebin for you. Aug 27 18:15:56 Crofton|work, OK so it's one error... but still one too many: | DEBUG: Python function do_populate_sdk finished Aug 27 18:15:56 | ERROR: Function failed: populate_sdk_image (see /media/toshiba-usb3/work/setup-scripts/build/tmp-angstrom_v2013_06-eglibc/work/beaglebone-angstrom-linux-gnueabi/console-image/1.0-r0/temp/log.do_populate_sdk.5886 for further information) Aug 27 18:15:56 ERROR: Task 10 (/media/toshiba-usb3/work/setup-scripts/sources/meta-angstrom/recipes-images/angstrom/console-image.bb, do_populate_sdk) failed with exit code '1' Aug 27 18:15:57 NOTE: Tasks Summary: Attempted 3992 tasks of which 3991 didn't need to be rerun and 1 failed. Aug 27 18:15:59 Summary: 1 task failed: Aug 27 18:16:01 /media/toshiba-usb3/work/setup-scripts/sources/meta-angstrom/recipes-images/angstrom/console-image.bb, do_populate_sdk Aug 27 18:16:04 Summary: There was 1 WARNING message shown. Aug 27 18:16:08 Summary: There was 1 ERROR message shown, returning a non-zero exit code. Aug 27 18:16:10 c Aug 27 18:16:46 you'll need to look in the logfile and poke arond Aug 27 18:17:01 and use pastebin, don't paste more then a few short lines in the channel Aug 27 18:21:51 Crofton|work, sure thing. Unfortunately most of it is Greek to me. The "collected errors" are at http://pastebin.com/kHhvEKf4 Aug 27 18:22:17 are you asking for dropbear and/or openssh? Aug 27 18:22:18 Crofton|work, if you like I can paste the whole log. Aug 27 18:22:41 I wonder why console-image is trying to install both? Aug 27 18:23:02 I'm not asking for anything... I'm trying $ MACHINE=beaglebone bitbake -c populate_sdk console-image Aug 27 18:23:35 or is that not correct? Aug 27 18:23:41 that should work Aug 27 18:23:43 at a guess, this is dev/dbg packages for openssh-sftp + dropbear interacting painfully Aug 27 18:23:49 yeah Aug 27 18:24:42 is there an eay way to get console image to avoid dropbear? Aug 27 18:25:34 dropbera vs openssh is a package group / image feature selection Aug 27 18:31:07 OK guys. I have a $1000 USD budget for anyone who can take a day and show me how to get a WORKING linux distro on which I can run Qt 4.x on a Beaglebone Black with attached LCD7 (rev A3) cape and touchscreen. I will gladly come to your location or we can do a webex or equivelent. Aug 27 18:32:39 I'm really tired of getting nowhere with OE/Poky/Angstrom/bitbake/etc. It's all crap to me if it doesn't work "out of the box". Aug 27 18:33:07 I've lost a month trying to learn this stuff. Aug 27 18:37:42 Sorry, I can't help out in that way. I haven't got a complete OS built for BBB myself yet, either. My main problem was running my builds on a distro not supported by Angstrom, though. Aug 27 18:39:46 cfo215, you are using the angstrom build instructinos? Aug 27 18:41:15 Crofton|work, yes. The only difference is I'm using the v2013.06 git branch. Aug 27 18:41:40 Crofton|work, not *master* Aug 27 18:42:51 I'll fire off a build, but it will take ages unless I can get the dl dir shared :) Aug 27 18:43:06 someone is looking at an issue for me, so I can look a little athis Aug 27 18:43:15 core-image contains qt? Aug 27 18:44:17 excuse me, console-image? Aug 27 18:44:24 I've tryied it with "MACHINE=your_machine ./oebb.sh bitbake console-image" using the beaglebone for "your_machine". I've also tried it with . ./envron...;MACHINE=... bitbake... Aug 27 18:45:13 I am trying to figure out how you inserted qt into the image? Aug 27 18:45:13 Crofton|work, afaik it does. Aug 27 18:46:13 IMAGE_INSTALL_append += "qt4-embedded" Aug 27 18:46:16 sorry Aug 27 18:46:22 :) Aug 27 18:46:47 I am no qt guru Aug 27 18:47:02 did you get any qt demos to run on your hw? Aug 27 18:47:04 nor I. but I can't use it if it's not there.... Aug 27 18:47:40 kind-a-sort-of... they didn't play well with the touch screen. Aug 27 18:48:11 levi, I use Ubuntu 12.04.2 for my development machine. Aug 27 18:50:41 building Aug 27 18:50:56 I'll see if I can track down the dropbear/opensdsh issue Aug 27 18:51:36 Crofton|work. I really need the Touchscreen to work. Which is why I picked Angstrom. It's gnome image works with the touch screen. I just don't need the overhead of X11 on my BBB. Thanks for trying to diagnose this... Aug 27 18:52:04 you have to ask around about the touchscreen Aug 27 18:52:52 I've been tryin g to leave on a bike ride for the last 3 hours Aug 27 18:54:33 Sorry to hear that. You should go riding. There's no time like the present! I'll survive. Aug 27 18:58:09 how can i find out the value of a variable used by bitbake? specifically, i want to know the IMAGE_CLASSES it is building Aug 27 18:58:19 well, someone is looked at something for me Aug 27 18:58:44 BCMM, try bitbake -e thing | grep ^IMAGE_CLASSES Aug 27 18:59:07 I just started baking "systemd-gnome-image" with the 'IMAGE_INSTALL_append += "qt4-embedded"' local.conf. Assuming that goes well, I'll try it with -c populate_sdk and see what happens. Aug 27 19:02:36 YAY, my TNC USB KISS modem is working with AX25 support Aug 27 19:02:46 it's good to see your BSP and packages in action :P Aug 27 19:03:26 * eren deserved a wiessbier Aug 27 19:03:32 see you tomorrow folks! Aug 27 19:12:20 First Build stuck at Aug 27 19:12:22 0: linux-yocto-3.8.13+gitAUTOINC+e7f42a416a_aa76cc2840-r4.2 do_fetch (pid 10937) Aug 27 19:12:43 the GIT clone in this case is Painstakingly slow... Aug 27 19:13:00 is this a problem with the servers or my build environment Aug 27 19:13:33 the size of build/downloads/git2 is increasing but the speed is extremely slow... Aug 27 19:14:01 PS: The internet connection is clean and has been supporting other downloads at good speeds Aug 27 19:14:13 this however seems to be stuck at 3-4KB/s Aug 27 19:18:30 cfo215: I don't think Angstrom supports much in the way of Ubuntu. I think they use some sort of Redhat-based system on their build servers. That may not be your problem at all, though. What kind of errors are you seeing? Aug 27 19:19:02 levi, Do you have a few days... ;) Aug 27 19:19:41 problem building SDK mainly. Trying to do MACHINE=beaglebone ./oebb.sh bitbake -c populate_sdk console-image Aug 27 19:21:26 end goal. Getting ANY distro working on Beaglebone Black, which supports the LCD7 cape for beaglebone, associated touchscreen, and Qt 4.x. as well as some misc libraries which compile for ARM. Aug 27 19:22:19 I started with Angstrom because it shipped with the BBB. Aug 27 19:23:28 My development PC is running Ubuntu... my target machine is the BBB running Angstrom Aug 27 19:29:05 anybody facing problem.... with this Aug 27 19:29:13 0: linux-yocto-3.8.13+gitAUTOINC+e7f42a416a_aa76cc2840-r4.2 do_fetch (pid 10937) Aug 27 19:31:02 Bhawna: it's probably cloning the git repo from yoctoproject.org and the traffic at this time might be somewhat higher Aug 27 19:31:36 Bhawna: my fetches for the kernel have been slow. the PREMIRROR has been the culprit Aug 27 19:35:01 thanks, good to know that its not machine specific Aug 27 19:35:13 any idea, how long this takes Aug 27 19:35:18 i am pretty new to all this Aug 27 19:40:21 a while! the kernel git archive it downloaded for me is 700 MB. eventually, i created my own mirror Aug 27 20:15:54 RP no more hangs after many hours of the stress test at this point. Aug 27 20:17:31 At least that is one less problem to mess with. It did make me wonder if we had an easier way to invoke bitbake with the python-dbg, since that is how I ultimately diagnosed the problem. Aug 27 20:22:06 I just took the approach of swapping python for python-dbg, but I figure there is probably a better way. Aug 27 20:58:37 JaMa: hmm, not sure why openflow is in meta-virtualism rather than meta-networking? Aug 27 21:25:34 Crofton|work: thanks Aug 27 23:15:15 ERROR: Can NOT get PRAUTO, exception timed out Aug 27 23:15:18 that can't be good.. Aug 27 23:16:22 i like exceptions... Aug 27 23:16:56 bug 5051 has been filed for that it seems Aug 27 23:16:57 Bug https://bugzilla.yoctoproject.org/show_bug.cgi?id=5051 critical, High, 1.5 M5, richard.purdie, NEW , ERROR: Can NOT get PRAUTO, exception timed out Aug 27 23:17:19 ah, right, thanks, didn't see that Aug 27 23:22:08 night all Aug 27 23:22:42 damn, i didn't get to leave a mint on the pillow... Aug 27 23:38:05 jwessel: good to hear, I'll take the patch :) Aug 28 01:12:20 RP: seems multiple issues with bitbake recently... **** ENDING LOGGING AT Wed Aug 28 02:59:58 2013