**** BEGIN LOGGING AT Mon Mar 02 02:59:58 2015 Mar 02 03:34:37 afaik, yocto has two cross compiler versions..how to select one over another? Mar 02 04:05:30 * nerdboy guesses PREFERRED_VERSION Mar 02 04:08:36 chankit: depends which part, GCCVERSION = "" is the one you are most likely after Mar 02 05:16:35 nrossi: I set GCCVERSION in local.conf am I right? Mar 02 05:26:35 chankit: yep, something like GCCVERSION = "4.9%" Mar 02 06:51:00 can any one help me how to configure ,install and run android on xen domU. Mar 02 07:39:05 * nerdboy shouldn't answer off-shift Mar 02 07:58:49 nrossi: thanks..sorry for the slowpoke reply lol Mar 02 08:11:54 What can cause the following error: Mar 02 08:11:55 ERROR: qttools-plugins not found in the base feeds Mar 02 08:12:00 What did I forget? Mar 02 08:16:12 they simply dont exist Mar 02 08:22:36 and how can I add them? Mar 02 08:22:50 qt-toolchain is looking for them... Mar 02 08:40:03 wv: read qttools/log.do_configure Mar 02 08:56:16 Ja Mar 02 08:56:27 JaMa, it complains about qttools-plugins Mar 02 08:56:34 but there is no recipe as such... Mar 02 08:56:39 why does it look for that? Mar 02 08:56:47 I have a qttools, but no qttools-plugins Mar 02 09:02:20 wv: are you building without opengl ? Mar 02 09:03:02 ALLOW_EMPTY_qttools-plugins = "1" Mar 02 09:03:08 should solve it Mar 02 09:05:58 to be honest, I'm still a bit lost in the wild with this qt thing Mar 02 09:06:17 I used to have qt4, and that was just a matter of adding a few lines to my local.conf Mar 02 09:06:31 now I just added a whole bunch of lines found on the internet to add qt5 Mar 02 09:06:47 My image builds, but I have troubles with my toolchain Mar 02 09:06:57 once I have them up and running both, I 'll see to finetune Mar 02 09:07:16 I think I will need opengl in the end, and I will not need X11/wayland anymore (I think) Mar 02 09:10:19 Hi Mar 02 09:10:21 I am trying to build a dl-dir in a way that I could if necessary build offline. I used the option: 'BB_GENERATE_MIRROR_TARBALLS = "1"'. After running one build like this I commented the 'BB_GENERATE_MIRROR_TARBALLS = "1"' line so it wont be used and added 'SOURCE_MIRROR_URL = "file:///REDO/download"' and 'INHERIT += "own-mirrors"' and 'BB_NO_NETWORK = "1"' to my local.conf. Mar 02 09:10:30 Some of the recipes I use are using the 'SRCREV = "${AUTOREV}"' option and that throws errors in during the parsing process. Mar 02 09:10:41 Am I using autorev wrong ? Mar 02 09:12:04 These recipes also have this configured: 'DEFAULT_PREFERENCE = "-1"' and I don't want them to build any way. There only for development builds. Mar 02 09:13:21 There are also versioned recipes for the same package, and they should be used. Mar 02 09:30:05 wv: qttools-plugins is package created by qttools recipe Mar 02 09:36:53 good morning Mar 02 09:43:37 hm, populate_sdk did the trick where meta-toolchain-qt5 failed Mar 02 09:43:42 dunno if I have everything now Mar 02 09:43:44 will test Mar 02 09:46:15 Anyone know whether it's possible to build a (sub)set of packages as a .tar file or something? Mar 02 09:46:44 I'd like to be able to copy debugging aids on my target if needed, but I don't want them in the main image Mar 02 09:46:49 (think of things like valgrind, which sucks up ~15MB) Mar 02 09:47:15 So what I'd like to do is bundle a few of those tools to a separate archive which I can install on my target as needed Mar 02 09:47:18 anyone any ideas? Mar 02 10:15:03 im editing the source of a downloaded recipe, but changes are not applied. is there any cached source i should be looking for? Mar 02 10:19:37 hi, is there some automated checker tool that I should run before submitting a patch? http://www.openembedded.org/wiki/How_to_submit_a_patch_to_OpenEmbedded Mar 02 10:19:54 I read that, but I cannot find anything that would check the format of a patch, etc. Mar 02 10:44:36 <_4urele_> hi everyone Mar 02 10:46:53 <_4urele_> I'm trying to build 2 images with 2 different distro for 1 machine in the same build directory. I'm using the DISTRO variable in the images itself, but it doesn't seem to be the best idea Mar 02 10:50:15 _4urele_: I don't think it works at all Mar 02 10:50:28 <_4urele_> I explain myself : if I have DISTRO ?= 'castel-distro' in my local.conf, and "DISTRO = 'aurele-distro'" in aurele-image.bb in my buildhistory I retrieve "DISTRO = aurele-distro", but the version is poky's version, the 'tmp' directory is not the good one... if I remove the DISTRO ?= 'castel-distro' in my local.con everything is fine for my builds Mar 02 10:51:37 <_4urele_> abelloni, it works more or less but I think it is not expected to be used like this Mar 02 10:52:20 i did a do_populate_sdk and get a .sh to install on my host machine Mar 02 10:52:30 however, there is nog qmake provided...? Mar 02 10:52:32 <_4urele_> abelloni, today i'm using two distro to have a tiny and a normal one. (I'm using the tiny to generate an initramfs) Mar 02 10:53:15 JaMa: I wrote a uthash recipe some 1-2 years ago, and I am throwing it away now as I no longer need it ... but before doing that, I am planning to send it to meta-oe. Mar 02 10:53:30 Which recipes group should I put it into? It is now in our recipes-support. Mar 02 11:30:09 lpapp: or extended Mar 02 12:23:30 Hi! I need to link a Qt application with a shared library (.so) written in C. I have one receipe for the Qt app, and another one for the library. How does the Qt recipe know the location of the shared library on compile time? Mar 02 12:24:02 using DEPENDS/RDEPENDS? Mar 02 12:24:54 and how should I refer to the library in the Qt .pro file when linking Mar 02 12:24:56 ? Mar 02 13:33:08 how do I use the populate_sdk_qt5? Mar 02 13:33:26 I am able to do bitbake -c populate_sdk core-image-minimal Mar 02 13:33:41 but -c populate_sdk_qt5 tells me this does not exist Mar 02 13:33:46 for this image Mar 02 13:33:58 ps: it's not really core-image-minimal, was as exampe Mar 02 13:51:27 wv, you should just need to add `inherit populate_sdk_qt5`to you image. Then you run `bitbake -c populate_sdk`. Mar 02 13:59:48 AndersD, thanks, will try that Mar 02 14:02:39 AndersD, I tryed: but get error Task do_populate-sdk does not exist for target Mar 02 14:20:01 wv, if you copied the error message, it's a typo, it should be an underscore in populate_sdk Mar 02 14:26:25 khem: do you know which recipe installs stdarg.h into ${STAGING_DIR_TARGET}/usr/include/gcc/${TARGET_SYS}//include ? Mar 02 14:43:03 AndersD, it was a typo from console to mirc (did not copy paste) Mar 02 14:46:05 wv, well, if you could run bitbake -c populate_sdk before adding inherit populate_sdk_qt5, it really should work. Could you paste you image recipe somewhere? Mar 02 14:49:17 khem: nevermind. I could sort the issue out. Sorry for the noise. Mar 02 14:52:52 Hi. I am trying to build a recipe that depends on a shared library (.so). but when trying to link it in the main recipe i get this: ../cortexa9hf-vfp-neon-poky-linux-gnueabi/gcc/arm-poky-linux-gnueabi/4.8.1/ld: cannot find -lmylib Mar 02 14:54:06 usually when i am compiling on x86 (ubuntu) i use ldconfig to create/update symlinks needed, could it be something similar that is missing in my recipe for the shared library? Mar 02 14:54:13 anyone having a clue? Mar 02 14:57:43 we set CC, CFLAGS, etc appropriately to find such things. so your buildsystem must not be obeying those variables from the metadta Mar 02 15:07:38 so, no one has any idea how to create, say, a .tar.bz2 in Yocto for a given set of recipes? Mar 02 15:08:00 as in a filesystem image in a tarball? Mar 02 15:08:04 that's just an image in a tarball Mar 02 15:08:24 well, the thing is, currently I build a root fs image Mar 02 15:08:36 add the appropriate type to IMAGE_FSTYPES. Mar 02 15:08:56 http://www.yoctoproject.org/docs/1.7.1/mega-manual/mega-manual.html#var-IMAGE_FSTYPES Mar 02 15:08:58 but I want to be able to add a set of packages to a running system Mar 02 15:09:26 think of stuff like valgrind -- the reason is that it enlarges the rootfs by 20MB which is quite a bit Mar 02 15:09:52 so what I want is to be able to leave my rootfs as-is, have yocto build valgrind+deps as needed and package them up somewhere Mar 02 15:10:02 that would be what's in deploy/ then Mar 02 15:10:18 it generates packages, just use it as a package feed and use opkg/rpm/whatever you picked to install what you want Mar 02 15:10:44 let me check if I understand this :) Mar 02 15:11:10 hmm Mar 02 15:11:11 when you build a recipe, the result is a package. be it a deb or a ipkg or a rpm. Mar 02 15:11:22 an image is constructed from those packages Mar 02 15:11:35 and you can use the packages on an image to add more stuff afterwards Mar 02 15:11:44 Yes, so I was wondering, is it possible to generate a second image? Mar 02 15:11:56 with a given set of packages Mar 02 15:12:05 if you want to have a "small" image and a "bigger" image then just make a second image Mar 02 15:12:16 I was considering that Mar 02 15:12:19 i thought you wanted to add stuff as required Mar 02 15:12:29 but the main issue is that it's nontrivial to update the rootfs in our system Mar 02 15:12:59 So the idea is to just untar additional software as needed - of course, this won't work well with init scripts and such but I don't care about that :) Mar 02 15:13:14 use the package manager and it will work well Mar 02 15:13:18 hmm Mar 02 15:13:28 I guess I could add rpm to the image, yeah Mar 02 15:13:31 its exactly like installing software using eg rpm on fedora Mar 02 15:13:41 * rink_ always stayed far from rpm :) Mar 02 15:13:51 the RedHat 5.0 memories still haunt me Mar 02 15:13:52 s/rpm on fedora/dpkg on debian/ Mar 02 15:14:09 * rink_ wonders if he dares to change the package format. Mar 02 15:14:13 oh its trivial Mar 02 15:14:15 well, is the runtime of rpm smaller than dpkg ? Mar 02 15:14:23 well don't use dpkg. use opkg or rpm. Mar 02 15:14:37 the dpkg backend is the least-tested Mar 02 15:14:44 That's a good point Mar 02 15:17:18 rburton: thanks! Mar 02 15:43:31 otavio: https://autobuilder.yoctoproject.org/main/builders/nightly-fsl-arm/builds/211/steps/BuildImages/logs/stdio Mar 02 15:44:09 otavio: master fails in your uboot recipe Mar 02 16:03:35 rburton: yeah I hit that one today Mar 02 16:17:27 hello Mar 02 16:18:03 how can I prevent a recipe from being built ? (at a higher level), is there something like DO_NOT_BUILD_RECIPE = recipe1,recipe2, etc Mar 02 16:18:06 thanks Mar 02 16:18:48 <_4urele_> acidfu, did you try with "BBMASK" Mar 02 16:18:59 no I didn't, I'll try it thank you ! Mar 02 16:20:01 <_4urele_> I use it to mask "bbapend" but check in the ref manual if it works for recipes Mar 02 16:20:22 it's worked since before bbappends existed Mar 02 16:20:38 but it's an extremely blunt instrument, prevents parsing entirely. not recommended unless absolutely necessary Mar 02 16:20:48 of course, if something depends on what you mask out, your build will just immediately fail Mar 02 16:20:53 that is, something you want to build Mar 02 16:50:46 hello, Mar 02 16:51:14 i'm facing a bug on dizzy with gcc-4.9 and gstreamer-1.0-plugins-base on imx platform Mar 02 16:51:26 a problem with optim flag O2 Mar 02 16:51:40 where should i post the issue ? Mar 02 16:52:37 sce: I would start with meta-freescale Mar 02 16:59:01 acidfu: check PNBLACKLIST as well Mar 02 17:02:07 angolini: meta-nxp now eh :) Mar 02 17:02:43 sce: what is the issue Mar 02 17:03:12 angolini, khem, its in gstvolume.c Mar 02 17:03:15 Hi khem` Mar 02 17:03:26 you can also file a bug report, at https://bugzilla.yoctoproject.org/ Mar 02 17:03:42 ulf`: hi Mar 02 17:04:13 angolini, khem, I can not change the volume in gstreamer pipeline with volume element Mar 02 17:04:35 khem, if i remove the -O2 its working Mar 02 17:05:15 maybe I should create meta-nxp, I have a few lpc3250 board left at home ;) Mar 02 17:05:19 sce: you need to get better info Mar 02 17:05:39 abelloni: do they run linux Mar 02 17:06:11 yeah, I even committed the buildroot support Mar 02 17:06:22 khem, what do you mean ? I guess yes I know where the issue in gstvolume.c. Its when i affect a variable Mar 02 17:07:57 sce: ok that better Mar 02 17:08:06 can you get more details Mar 02 17:08:48 i can give the line Mar 02 17:09:07 i'm trying to build gstreamer with -02 but i dont know how to do that Mar 02 17:09:07 sce: if this is only gstreamer and toolchain, maybe the better place is gstreamer ML, not meta-freescale. Mar 02 17:10:41 sce: i remember it was already replied on a past imx.community question (the bbappend needed to -O2) have you searched? Mar 02 17:11:20 no you mean freescale community or yocto ? Mar 02 17:12:11 I mean freescale community Mar 02 17:12:19 I'm searching the link..... Mar 02 17:13:20 angolini, i'm searching too ... Mar 02 17:15:37 khem, its gst/volume/gstvolume.c +251 in gstreamer1.0-plugins-base 1.4.1 Mar 02 17:16:10 self->current_volume stays always the same with -O2 Mar 02 17:19:40 sce: is this using hard-float Mar 02 17:20:18 I wonder if its failing in some vectorization Mar 02 17:20:48 sce: the link i was talking about is https://community.freescale.com/message/391614#391614 but now I'm not sure it can help you any more.... Mar 02 17:20:51 can u try to compile with -O2 -fno-tree-vectorize Mar 02 17:21:51 how can i do that ? is there a configure flags where i could add this ? Mar 02 17:24:57 khem, yes looks like its hard Mar 02 17:26:04 CFLAGS += ? Mar 02 17:33:24 khem, -fno-tree-vectorize is not working either Mar 02 17:34:15 khem, i need to go now.. We can discuss tommorrow i hope Mar 02 17:47:19 build fails with perl: symbol lookup error: /usr/lib/libgudev-1.0.so.0: undefined symbol: g_type_class_adjust_private_offset Mar 02 17:47:39 nm says that symbol is undefined in libgudev, but it's exported in libgobject Mar 02 17:50:03 hmm, libgobject inside the sysroot doesn't have this symbol however Mar 02 17:53:59 I see, it's using system libgudev, which links to a newer libgobject, which has this symbol Mar 02 17:59:17 abelloni: hmm, do you understand how ntp works? Mar 02 17:59:50 abelloni: I launched the server instance with busybox on one board (192.168.0.1) and the client on the other with the peer, but the client does not seem to get adjusted even though the client can ping 192.168.0.1, the server. Mar 02 18:02:23 I'd first have a look at ntpd.conf then try to know if every server understand in which stratum they are Mar 02 18:03:32 abelloni: there is no ntpd.conf. Mar 02 18:06:17 right, busybox only uses the command line Mar 02 18:06:38 oh, that's ntp.conf Mar 02 18:07:07 http://git.busybox.net/busybox/tree/networking/ntpd.c#n2161 Mar 02 18:07:24 question is what peers are configured Mar 02 18:07:34 -p 192.168.0.1 on the client Mar 02 18:11:03 abelloni: I thought that would be enough. Mar 02 18:11:44 should be Mar 02 18:11:57 you can start your server with -n to get some logs Mar 02 18:12:11 abelloni: heh, /etc/ntp.conf support was added after I raised this issue Mar 02 18:12:38 when I submitted my hackish approach to oe-core... then I also brought it up on the busybox mailing list. Mar 02 18:13:45 honestly, I wouldn't use the busybox ntp server Mar 02 18:15:26 why is that Mar 02 18:16:52 abelloni: this is how the server is running now, /sbin/ntpd -p 127.0.0.1 -l -n Mar 02 18:17:13 I don't like the idea of busybox being internet / network facing Mar 02 18:17:56 ok, well, that is a personal call. Mar 02 18:17:57 I'm not sure -p 127.0.0.1 is supposed to work Mar 02 18:18:11 well, I cannot set up anything else, really. Mar 02 18:18:37 the board cannot connect now to the external world Mar 02 18:18:49 then don't give any peers Mar 02 18:19:08 http://git.busybox.net/busybox/tree/networking/ntpd.c#n2177 Mar 02 18:19:18 -l and no peer => stratum 1 Mar 02 18:19:28 I guess that is what you want to achieve Mar 02 18:19:49 I can try that. Mar 02 18:20:10 still no syncing Mar 02 18:20:28 the client is 17 secs off the server. Mar 02 18:22:18 what is tcpdump showing ? :) Mar 02 18:23:04 oh, it is now sync'd, interesting. Mar 02 18:23:45 but will it also work with real peers? Mar 02 18:24:08 like pool.ntp.org? Mar 02 18:24:14 Unfortunately, I cannot test that scenario rightn ow. Mar 02 18:24:16 right now* Mar 02 18:24:35 but it takes a lot of time to sync up. Mar 02 18:25:42 it is certainly more than five minutes. Mar 02 18:26:05 I wonder if I had just been impatient when trying 127.0.0.1... Mar 02 18:28:09 I wonder how the algorithm is made up when exactly to sync, etc. Mar 02 18:28:25 whether it also synchronizes for 10 secs drift, etc, which is what I currently have. Mar 02 18:30:28 it adjust maxim 1s every 32s Mar 02 18:30:55 abelloni: I checked every minute now Mar 02 18:31:02 it was the same drift at around 4 minutes Mar 02 18:31:09 I checked it again around 6 minutes, then it was in sync. Mar 02 18:31:19 my gut feeling is that it syncs up after 5 minutes or so. Mar 02 18:46:17 Hi! I built core-image-base for beaglebone and it all works nice but I am unable to run any command which requires superuser privilages (I'm logged in as root). Has anyone experienced a similar issue and was able to solve it? Mar 02 18:52:09 what does 'id' say Mar 02 18:52:40 rink_: uid=0(root) gid=0(root) groups=0(root) Mar 02 18:53:06 hmm thatshould be enough Mar 02 18:53:10 rink_: 'whoami' shows I am indeed logged in as root Mar 02 18:53:11 which kind of command do you use ? Mar 02 18:53:24 shutdown -h now is the command I use Mar 02 18:53:41 and that just says: permission denied ? Mar 02 18:54:10 Yes.shutdown: warning: cannot open /var/run/shutdown.pid shutdown: /dev/initctl: Permission denied init: must be superuser. Mar 02 18:54:13 sure there isn't something like SELinux or other access control stuff active? Mar 02 18:54:35 Not sure how to check that Mar 02 18:55:12 * rink_ neither. Mar 02 18:55:28 this channel is less inactive during midday, fwiw Mar 02 18:55:36 perhaps soemeone has a better hint than I do Mar 02 18:55:49 I did make one change to the core-image-base.bb. I followed this: https://wiki.yoctoproject.org/wiki/FAQ:How_do_I_set_or_change_the_root_password Mar 02 18:56:08 well, assuming you could log in ... that should be fine Mar 02 18:56:38 Prior to doing what is suggested on that web page I could not log in. Default username (root) and blank password didn't work Mar 02 19:00:23 guys, what do you know what is the difference between ${base_bindir} and ${bindir} ? Mar 02 19:00:24 rink: sestatus ? Mar 02 19:00:36 s/what// Mar 02 19:02:09 rink_: even ^^ Mar 02 19:02:38 vmeson: sestatus "command not found". Mar 02 19:02:38 guess Casper_ should try that :) Mar 02 19:02:54 i doubt anything SELinux related is running Mar 02 19:03:11 Casper_: maybe it's a silly question but what are the permissions on /dev/initctl Mar 02 19:03:54 :-) "prw------- 1 root root 0 Jan 1 2000 initctl" Mar 02 19:04:41 I would strace shutdown and see if I could figure out the call that was making him think he doesn't have permission Mar 02 19:08:34 I did chmod 700 on the initctl to get prwx------ permissions but still i get "permission denied" when shutting down. Mar 02 19:09:21 rewitt: Not sure how to use strace. Typing 'strace' gives a "command not found" Mar 02 19:09:57 realBigfoot: bindir ~= /usr/bin, base_bindir ~= /bin Mar 02 19:11:27 kergoth: PREMIRRORS, do they fetch git clones form premirrors too ? or just tarballs of git repos Mar 02 19:11:56 just tarballs IIRC Mar 02 19:12:01 it depends on the mirror configuration. afaik either should work, but i've never actually tried to set the mirror replacement url to a git url Mar 02 19:12:10 the latter definitely works Mar 02 19:12:54 JaMa: yeah tarballs is whats working Mar 02 19:13:08 but a complete clone it still goes back to SRC_URI Mar 02 19:13:24 its not documented that way though Mar 02 19:14:20 Casper_: I forgot they don't put strace in the image by default. :( You would have to add it. IMAGE_INSTALL_append = " strace" Mar 02 19:15:36 rewitt: Allrighty. Thanks. Back to building. Mar 02 19:16:13 Casper_: I'm off to lunch. If you get the strace output and want us to look at it just put it on pastebin. Mar 02 19:16:16 * rewitt heads to lunch Mar 02 19:16:33 khem: you changed the rights ide of the replacement to a git:// url? Mar 02 19:16:46 rewitt: OK. Thanks. Mar 02 19:19:05 kergoth, thanks! Mar 02 19:20:35 realBigfoot: i recommend reading meta/conf/bitbake.conf Mar 02 19:20:46 it defines all those basic variables like that Mar 02 19:21:04 kergoth, wow nice this will make my life much easier Mar 02 19:21:22 it's that file which includes the distro and machine config, and local.conf, actually. they wouldn't be parsed otherwise Mar 02 19:39:05 https://plus.google.com/101339419642360856354/posts/7fpMfQzJAf8 Mar 02 19:39:12 Does this run Yocto Linux? Mar 02 19:45:41 kergoth: yes Mar 02 19:45:49 kergoth: I used own-mirrors class Mar 02 19:45:54 which does that too Mar 02 19:46:57 own-mirrors just sets the replacement to SOURCE_MIRROR_URL Mar 02 19:47:07 rewitt: Here is the strace of running 'shutdown -h now': http://pastebin.com/fKUGHNsC I'd really apreciate it if you, or anyone else, could have a look. Thanks. Mar 02 19:47:08 unless SOURCE_MIRROR_URL is a git:// url, that's not going to do what you think Mar 02 20:04:54 erm, I'm trying to build my BSP with the linaro tool chain, but everytime I'm getting problem when it tries to compile poky/meta/recipes-devtools/libtool/libtool-cross_2.4.2.bb Mar 02 20:05:18 in build/local.conf I added: TARGET_SYS = "arm-linux-gnueabihf" , TCMODE = "external-linaro", EXTERNAL_TOOLCHAIN = "/work/gcc-linaro/" Mar 02 20:05:24 anyone has an idea? thanks :) Mar 02 20:09:23 does linaro have examples on using their external toolchain? Mar 02 20:09:34 I've never used it.. but the TARGET_SYS change seems suspect to me.. Mar 02 20:09:53 no, I don't find anything :| Mar 02 20:10:05 I'm trying trial and error Mar 02 20:10:12 I suspect it's not supported without someone giving you instructions.. Mar 02 20:10:22 Usually the instructions are included in the implemnetation of the TCMODE Mar 02 20:10:47 conf/distro/include/tcmode-${TCMODE}.inc Mar 02 20:10:52 search you layers for that.. Mar 02 20:10:59 the one I know often exists in the 'external-sourcery' Mar 02 20:11:09 that's for the Mentor Sourcery G++ Mar 02 20:11:36 yes I have this file here: meta-linaro/meta-linaro-toolchain/conf/distro/include/tcmode-external-linaro.inc Mar 02 20:11:56 look in there or other places along that path and see if there are instructions on usage Mar 02 20:12:00 fray, and do you usually specify TCLIBC too ? Mar 02 20:12:22 I use glibc, which is the default.. but the only external toolchains I use are the ones I build myself.. Mar 02 20:12:31 so I don't use the regular external-* configs Mar 02 20:12:57 the thing is just that i'm trying to build the whole thing with gnuabihf instead of gnuabi Mar 02 20:13:06 Casper_: What are the user/group/permissions on the shutdown binary? Mar 02 20:13:11 then you are going about it the wrong way.. Mar 02 20:13:24 fray, anything else to suggest ? Mar 02 20:13:27 ignoring the external toolchain, you select the ABI, processor optimizations and related options using the "TUNE" parameters.. Mar 02 20:13:59 specifically "DEFAULTTUNE" in the machine (BSP), or overriden in your local.conf Mar 02 20:14:25 this is mine: DEFAULTTUNE ?= "cortexa15thf-neon" Mar 02 20:14:27 see oe-core/meta/conf/machine/include/README for in place documentation Mar 02 20:14:33 ok Mar 02 20:14:42 the ?= says if not previously defined.. so assume thats not an issue.. Mar 02 20:15:05 the configured environment will be a cortex a15 w/ VFP hard float ABI and neon instructions Mar 02 20:15:08 I should probaly do = directly Mar 02 20:15:18 and not ?= Mar 02 20:15:22 the ABI is ARM EABI HF variant Mar 02 20:15:30 depends where it's implemented.. Mar 02 20:15:34 ?= is processed in the order in which events occur.. Mar 02 20:15:42 local.conf is processed before the machine files.. Mar 02 20:15:54 we recommend that the tune be included early in the amchine config.. Mar 02 20:15:57 so: Mar 02 20:15:59 it's defined here conf/machine/device_name.conf Mar 02 20:16:07 DEFAULTTUNE ?= "coretexa15hf-neon" Mar 02 20:16:26 include conf/machine/include/tune-cortexa15.inc Mar 02 20:16:28 .... Mar 02 20:16:36 is very different then doing those two lines in the other order Mar 02 20:16:37 yup that is done Mar 02 20:17:04 so defining defaulttune, will automatically will choose the right compiler ? Mar 02 20:17:05 you can veriy settings by doing bitbake -e , and inspecting the output for "DEFAULTTUNE=" Mar 02 20:17:14 it will (baring external toolchains) Mar 02 20:17:16 Casper_: the strace looks fine until geteuid() returns 1000 -- can you grep root /etc/passwd -- should be 0. Mar 02 20:17:39 when you introduce external binary toolchains, that is when you have to follow that providers requirements.. which they should tell you what ABIs are supported and how to use them Mar 02 20:18:29 but I always have this: TARGET_SYS = "arm-poky-linux-gnueabi" Mar 02 20:18:34 (I tend not use external binary toolchains for my oe-core development, because often vendor/semi toolchains are optimized at the 'expense' of other architectures.. I.e. it produces correct code for their ONE chip, but you deviate at all and now you are getting bd code elsewhere.. Mar 02 20:18:42 but I would expect to see gnuabihf Mar 02 20:19:06 rewitt: shutdown -> /sbin/shutdown.sysvinit lrwxrwxrwx ; shutdown.sysvinit -rwsr-xr-- Mar 02 20:19:15 acidfu, why? Mar 02 20:19:39 those values are only used by configure.. and as far as I know gnueabihf has never been 'blessed' my GNU.. Mar 02 20:19:42 fray, why am I expecting to see gnuabihf ? because I want to make sure I compile with hard float Mar 02 20:19:54 vmeson: root:x:0:0:root:/home/root:/bin/sh Mar 02 20:20:23 that won't do anything but pass that value to configure, which will happily ignore it Mar 02 20:20:32 you verify it by building something and inspecting the executable.. Mar 02 20:20:50 if you want to do it before building, inspect the toolchain arguments, inspect target_CC= (might have that name wrong) Mar 02 20:21:11 ok i'll try it, and see if my stuff is compiled with hard-float even tho the target sys is soft-float Mar 02 20:21:21 target_sys is NOT saying soft-float.. Mar 02 20:21:23 Casper_: so what does ls -l show for /sbin/shutdown.sysvinit? Because it has the setuid bit set, so my guess is that it's dropping privileges Mar 02 20:21:33 it's saying arm-...-linux-gnueabi which is what it is Mar 02 20:21:42 ok! Mar 02 20:21:43 don't get confused.. HF is a varient of gnueabi.. Mar 02 20:22:02 what says hard and soft float as an indicator is the tune settings, and ELF attributes Mar 02 20:22:12 ah ok great Mar 02 20:22:33 that part was confusing me, i'll recompile with the default setting and the right tune then Mar 02 20:22:44 seriously.. GNU configure ignores pretty much everything after the 'OS' field in a system setting.. there are a few scripts that for ARM will look for 'gnu' vs 'gnueabi' to tell if it's OABI or EABI Mar 02 20:22:58 but hard float is an eabi variant.. there is nothing else to be learned from the system.. Mar 02 20:23:03 rewitt: It gives -rwsr-xr-- Mar 02 20:23:06 rewitt: -rwsr-xr-- 1 1000 1000 18448 Mar 1 23:45 shutdown.sysvinit Mar 02 20:23:08 (for customer configure scripts, adding 'hf' might even break the configure) Mar 02 20:23:17 'er.. customer -> custom.. i.e. written in perl or python) Mar 02 20:23:31 Casper_: Yeah the uid/gid is 1000, does that user/group even exist? Mar 02 20:23:50 for an ARM system, you can inspect TUNE_PKGARCH, and you'll see 'hf' if the system belives it's building for hard float Mar 02 20:24:21 (that name can be overriden, but it's discouraged).. the package name is just human readable text.. the system doesn't actually care) Mar 02 20:24:25 Casper_: It's definitely not root. So what's happening as vmeson pointed out is that when the the binary is ran, it's changing the effective uid to be 1000. That uid does not have the correct permissions. Mar 02 20:24:39 fray, by doing bitbake -e, I see this: export CC="arm-poky-linux-gnueabi-gcc -march=armv7-a -marm -mthumb-interwork -mfloat-abi=hard -mfpu=neon -mtune=cortex-a15 Mar 02 20:24:47 which clearly confirm what you are trying to explain to me ;) Mar 02 20:24:55 yes, that looks correct Mar 02 20:25:40 rewitt: cat /etc/group | grep 1000 does not output anything Mar 02 20:25:56 if it's returning '1000'. that says that it doesn't have any name to translate Mar 02 20:25:57 rewitt: Am I looking in the right place? Mar 02 20:26:00 so it's just using what you said.. Mar 02 20:26:43 based on it being 1000, I'd say the problem is with either the creation of the package (the wrong owner/group was selected), or you are NFS serving your filesystem, and either didn't extract teh filesystem properly -- or pseudo didn't run in the NFS server Mar 02 20:31:23 Casper_: my system says: Mar 02 20:31:25 sh-4.3# ls -l /sbin/shutdown.sysvinit Mar 02 20:31:26 -rwxr-xr--. 1 root shutdown 29112 Feb 27 15:07 /sbin/shutdown.sysvinit Mar 02 20:31:28 sh-4.3# grep shutdown /etc/group Mar 02 20:31:29 shutdown:x:70: Mar 02 20:32:15 fyi: it's oe-core-1.7 based... Mar 02 20:32:56 fray, thank you for those explications ! Mar 02 20:33:25 no problem.. I hope it helps Mar 02 20:33:30 yes it does ! Mar 02 20:34:16 vmeson: I get the same. @beaglebone:# grep shutdown /etc/group Mar 02 20:34:17 @beaglebone:/# shutdown:x:70: Mar 02 20:34:58 Casper_: Are you serving the filesystem across nfs as fray mentioned? Mar 02 20:36:26 rewitt: No, I extracted the rotfs tar.bz in my deploy directory to my SD card Mar 02 20:36:29 fray: Sorry. I missed your comments. Mar 02 20:37:19 how did you extract it? Mar 02 20:37:28 most likely you did it as a user and it corrupted all of the file permissions.. Mar 02 20:37:52 if you deploy to an SD card (a tar ball vs a constructed filesystem) you likely will need to upgrade you permissions to root Mar 02 20:38:10 fray: Im using Ubuntu so I just opened nautilus, double-clicked the archive and used Ubuntu-s archive manager to extract onto the mounted SD card partition Mar 02 20:38:20 yup, that corrupted it. Mar 02 20:38:28 it extracted the files as your local user in order to protect you. Mar 02 20:38:48 extracting files for a tar archive as root is dangerous, as it can set files with permissions that may permit an attacker to do something harmful.. Mar 02 20:38:55 Casper_: put the mouse away. :) Mar 02 20:39:13 Since you created the tar, it should be safe to do this.. so you will need to change your uid to root, and extract with full permissions.. (and the CLI is the best way to do this..) Mar 02 20:39:20 cd Mar 02 20:39:33 sudo tar xvf... Mar 02 20:39:37 cd ~ Mar 02 20:39:46 fray: :-) That's what I get for beeing lazy. The first two tries I did it as root and then it seemed too much hastle and so I just gave user permissions to the SD card partition. Mar 02 20:39:46 (eject SD card).. try again Mar 02 20:40:00 Thanks. I'll try it now. And put my mouse away. :) Mar 02 20:40:05 :) Mar 02 20:41:24 I didn't even think about the tarball extraction, I'm always booting images directly in qemu :) Mar 02 20:45:21 One other question while I'm doing this. I also have to extract the kernel modules, correct? Mar 02 20:50:58 fray: Works like a charm now. Thanks a million for your help. And rewitt, rink and vmeson. By the way, how to I address more people on IRC? "member: blablabla" seemes to address one. Any chance of addressing more? Mar 02 20:53:43 it's convention.. ther eis no formal way.. Mar 02 20:53:58 I usually just do, name1, name2, ... nameN: if addressing more then one Mar 02 20:54:09 (my client highlights any message with 'fray' anywhere in it) Mar 02 20:57:48 fray: I see. Thanks. Now on to the next challange. Trying to add the vsftpd package to my image. I'll chew on it myself for now but I'll probably end up here with questions in a few days. :) Thank you again for todays help. Mar 02 21:26:21 Casper_: np, glad it's working for you. Avoid the mouse! Mar 02 21:29:54 eh rburton :) Mar 02 21:34:59 anyone know if thereis a recipe for dhrystone ? thanks Mar 02 21:41:13 is that still relevant ? Mar 02 21:45:04 I mean it ouch, the buildroot recipe is ugly: http://git.buildroot.net/buildroot/tree/package/dhrystone Mar 02 21:45:20 s/I mean it // Mar 02 23:11:23 I'm getting a python error on relocatable_binaries_preprocess(). Mar 02 23:11:31 It says File not found Mar 02 23:12:13 It's working in Ubuntu 14.04.1, but not 14.04.2 Mar 03 01:48:42 hi guys, anybody here? Mar 03 01:49:13 I am but that might not be that useful :) Mar 03 01:49:25 I have minor dylan tweaks, was wondering who do I talk about patches Mar 03 01:49:49 I'm talking meta-oe Mar 03 01:51:08 openembedded-devel@lists.openembedded.org Mar 03 01:51:36 So you can subscribe following the link at http://www.openembedded.org/wiki/Mailing_lists Mar 03 01:53:14 rewitt: I sent my patch there and no response + no commits Mar 03 01:53:24 rewitt: I thought maybe dylan has died already? Mar 03 01:53:41 Xz: how long ago did you send said patch(s)? Mar 03 01:54:04 nrossi: 19 feb 2015 Mar 03 01:54:14 nrossi: it's opencv backport Mar 03 01:55:58 Xz: did you remember to prefix your change with [meta-oe]? i cant see it in the meta-oe patchwork Mar 03 01:57:00 nrossi: did you try just looking for 'opencv' on the list? I put a subject: Mar 03 01:57:03 [oe] [meta-oe][dylan]opencv: master->dylan backport Mar 03 02:01:20 Xz: i see your email, its not a patch though? you are just requesting the backport of a commit. I cant see from the list but did you send it with to: Paul? Mar 03 02:03:04 Xz: it might be just building up in Paul's email and he has not see it yet. Also it has not yet been 2 weeks, but if you are after quicker action you can ping that email and see if you get a response. Mar 03 02:05:56 nrossi, sounds like zynq SPL might want u-boot-dtb.img now Mar 03 02:17:33 how do I add kernel source code to a toolchain SDK? is there a package name for kernel source code? Mar 03 02:26:13 nrossi: so should I just wait? Mar 03 02:26:27 nrossi: I could send a patch if that's necessary instead of request Mar 03 02:26:42 nrossi: it just made sense for me to do it this way **** ENDING LOGGING AT Tue Mar 03 02:59:58 2015