**** BEGIN LOGGING AT Wed Jul 13 02:59:58 2016 Jul 13 05:46:17 Hello, I am having a problem compiling code, here is the problem < https://gist.github.com/Technicus/84a51de81e460e07d6a0d33cda29acec >, how do I get this to compile? Jul 13 05:54:11 hi, I tried to compile a 3rd party kernel module but somewhat /usr/src/kernel/scripts/mod/elfconfig.h is absent so I can't compile that. any idea? Jul 13 06:44:51 parrot: did you compile it on target ? Jul 13 07:11:31 Hi, I have two out of tree kernel modules where ond depends on the other. This means module B requires some headers provided by module A. Where would I put these headers inside inside the OE build system to make it available for module B so that everything works properly with sstate etc.? Jul 13 08:11:45 hi Jul 13 08:12:05 s.a Jul 13 08:12:37 help Jul 13 08:12:46 !help Jul 13 08:14:13 can I ask a question Jul 13 08:14:14 ?? Jul 13 08:14:24 yes Jul 13 08:15:31 I started to use yocto linux on sabre automative board i.mx6q of freescale Jul 13 08:15:54 is there any download or update commands at yocto terminal Jul 13 08:17:07 to update your board? Jul 13 08:17:28 to update an application Jul 13 08:17:55 or download an application Jul 13 08:19:01 does the OS on the board features a package management tool? (dpkg, opkg, or rpm?) Jul 13 08:20:27 The OS features rpm Jul 13 08:21:12 I'm not aware of any yocto specific tool for it, but you can either ship your new package and install it locally, or setup an ftp mirror and fetch it with your package manager Jul 13 08:22:52 can I make opkg or dpkg usable on my OS? Jul 13 08:23:14 khem: I did. Anyway, I ran make config and make prepare on the kernel source directory and it got better. Jul 13 08:25:26 parrot: did you inherit module? that helps a lot Jul 13 08:27:00 akoc: the package format should be defined in your loca.conf file Jul 13 08:27:05 local.conf* Jul 13 08:27:06 akoc: you can find some information online: for example: http://wiki.hioproject.org/index.php?title=Yocto:_Opkg_Package_Management_System_Set_Up Jul 13 09:18:53 Hi, I have two out of tree kernel modules where ond depends on the other. This means module B requires some headers provided by module A. Where would I put these headers inside inside the OE build system to make it available for module B so that everything works properly with sstate etc.? I tought about the kernel-source or kernel-build artifacts idrectory in work-shared but I'm not sure this is the right place. Jul 13 10:23:05 What's up with Yoctos fetishism for using names like 'tmp' and 'temp'? ;) Jul 13 10:28:54 yeah the logs in temp/ should really be in a logs/ directory Jul 13 10:35:19 Is there a way to get the sysroot path from within a recipe? I've found PKG_CONFIG_SYSROOT_DIR as the closest match. Jul 13 10:35:31 hello Jul 13 10:36:14 hello can: Jul 13 10:36:17 can: hello Jul 13 10:37:13 Is anybody here worked on sabre automative boards? Jul 13 10:37:31 The problem is upstream opus, which in its /usr/include/opus/opusfile.h, include , and as such I need to add $SYSROOT/usr/include/opus to gcc's search path. When we compiled our Qt app using the Yocto SDK, we got a ton of env-vars that this could be extracted from, such as SDKTARGETSYSROOT. Jul 13 10:37:57 or any similarboard from nxp Jul 13 10:38:12 How is SDKTARGETSYSROOT thought to be replaced when running as a normal recipe? Jul 13 10:38:23 can: what is your problem ? Jul 13 10:39:15 I need some help on building distro and sending it with mfg tool? Jul 13 10:40:53 I have a smilar problem.. Jul 13 10:55:25 at least can someone tell me how to reach local.conf to disable x11 Jul 13 10:55:30 ? Jul 13 10:57:05 can: DISTRO_FEATURES_remove = " x11" in your local.conf Jul 13 11:02:06 since the sysroot is cumulative, is there a QA check that verifies that its not linked against libs that happens to be forgotten i DEPENDS and RDEPEDS? Jul 13 11:07:50 sveinse: yes Jul 13 11:08:01 anyone aware of any problems with sourceforge? getting quite a few checksum failures in poky recipes Jul 13 11:08:16 CTtpollard: master or a branch? Jul 13 11:08:32 jethro Jul 13 11:08:52 quite a few package are failing checksums from sourforge and switching to to other sources Jul 13 11:09:14 http://sourceforge.net/error-404.html Jul 13 11:09:19 might be stemming from this? Jul 13 11:10:55 luckily not had a package that has yet to find another upstream source as of yet Jul 13 11:11:17 Hi there Jul 13 11:11:48 this is more a bitabke than a yocto question but... Jul 13 11:11:57 is there a way to remove a flag from a variable Jul 13 11:12:03 sveinse: STAGING_DIR_HOST? Jul 13 11:12:09 rburton: interestingly all of them fail with the same sum, maybe it's offering a temp placeholder Jul 13 11:12:14 Anyone knows if it is possible to use kernel 4.5 in Yocto? If so, how can that be managed? Jul 13 11:12:19 i.e : if somewhere there is var[name] = "1" Jul 13 11:12:24 how do I unset it ? Jul 13 11:12:43 in python there is a delVarFlag but is there a non-python way to do it ? Jul 13 11:15:40 @can Jul 13 11:17:43 maxin: Where is local.conf? Is it in already working system or in rootfs? Jul 13 11:18:01 can: in your build folder in the conf subfolder Jul 13 11:21:21 neverpanic: I don't build OS. There is a program called mfgtool for freescale boards which has all the OS/document needed for board. That program also setup the OS to board. Can I take build folder from there? Jul 13 11:27:14 can: in that case, you probably has the final rootfs, I guess.. Jul 13 11:27:54 any .bb for kernel 4.5? Jul 13 11:30:21 can: it might be worth reading the i.MX User's Guide documentation for MFGTool Jul 13 11:33:11 neverpanic: No, that env var is not available when running the recipe apparently Jul 13 11:43:00 neverpanic: I'm using PKG_CONFIG_SYSROOT_DIR, it seems to work Jul 13 11:43:56 BTW, isn't bitbake slightly inconsistent: bitbake -C configure sp:do_configure <-- 'configure' in one instance, 'do_configure' in another? Jul 13 12:49:23 sveinse: it might not be an environment variable, but it's certainly a bitbake variable, because it's set in meta/conf/bitbake.conf Jul 13 12:49:53 also, yes, bitbake is inconsistent; e.g. addtask $name_without_do before $name_with_do after $name_with_do Jul 13 12:49:59 there are multiple places like this Jul 13 12:50:02 neverpanic: Is there a bb directive to export it? Jul 13 12:50:35 That being said, I'm happy with using PKG_CONFIG_SYSROOT_DIR unless that changes at some point Jul 13 12:50:49 export; but you could also pass it to your build in EXTRA_QMAKEVARS_SOMETHING? Jul 13 12:52:18 neverpanic: Yeah, probably. I just have to dig into the variable system of qmake (which can be, uhm, eccentric at times) Jul 13 13:00:58 The qt5 defines qt.conf (store in the build tmp work dir), which defines the paths to the system. https://bpaste.net/show/acf89f7de0b7 Jul 13 13:01:52 However, build fails, as HostBinaries is not prefixed with the sysroot, as the other Host* vars are. This is probably a bug, right? Jul 13 13:02:06 Because it causes our build to fail Jul 13 13:02:26 Where should I start looking for a fix to this? Jul 13 13:14:31 Oh joy, it seems this /is/ a larger discussion between OE and Qt when it comes how to cross build :( Jul 13 13:15:07 sveinse: JaMa might have some helpful advice, he works on the layer a lot Jul 13 13:16:15 Yes, I'm reading a IRC disussion between him and lpapp regarding this exact topic. Just trying to figure what to do next. I've found the code, but I'm reluctant to fix it in case it is deliberate and breaks a lot of stuff... Jul 13 13:24:43 ziggo: o/ Jul 13 13:27:17 JaMa: Is it deliberate that "HostBinaries=" in qt.conf in a qt5 app recipe contains "/usr/bin/qt5" while the other Host paths contains the sysroot prefix? I notice that the qt.conf included in the SDK, does not. It contains the sysroot path. Bug or intentional? Jul 13 13:45:41 Does anybody have experiance with building external kernel modules that depend on each other? Jul 13 13:47:02 :) Jul 13 13:48:26 bachp: so how do you expect thats going to work? Jul 13 13:49:12 I would like to put a DEPENDS="kernel-module-a" into kernel module b Jul 13 13:50:06 Inside kernel module a I tried to put the header file that is reqired by b into STAGING_KERNEL_DIR so that it is available to module b Jul 13 13:50:22 howver this breaks down when I clear tmp and rebuild from sstate Jul 13 13:51:07 rburton: So I was wondering if there is another way? Jul 13 13:51:19 I think you would need to depend on a specific task Jul 13 13:51:39 DEPENDS="kernel-module-a:do_" Jul 13 13:51:51 not sure of the something... the task that does the staging, I guess Jul 13 13:52:02 bachp: don't just write directly to STAGING_* as sstate won't capture it Jul 13 13:52:03 boucman_work: I put the header there in the module_do_install task Jul 13 13:52:22 i'm more wondering how modules that depend on each other can be built Jul 13 13:53:47 rburton: what do you mean how they can built? Outside of OE module a s Makefile just puts the header into the kernel source and then when module b is buitl the header is already there. Jul 13 13:54:08 so they don't depend on each other Jul 13 13:54:56 i'd see if installing the header to the usual location in the standard sysroot works Jul 13 13:55:01 rburton: they kinda do, if module b needs a header provided bymodule a Jul 13 13:55:42 THe header provides some symbols that are used by b to talk to a Jul 13 13:55:54 but I'm not to deep into how exactly the work Jul 13 13:56:02 I just try to get them working with OE :) Jul 13 13:56:44 try installing the header into the sysroot like you normally would, ie ${D} not STAGING_KERNEL_DIR Jul 13 13:58:10 That might be one way Jul 13 13:58:31 But is there also a way to properly make the recipe write to STAGING_KERNEL_DIR? Jul 13 13:59:02 Hello, how do I write a systemd unit? I am writing a program that I want to start everytime the system starts. The program will need to check if /dev/input/event2 is present. If it is not present the program fails. I could probably modify the program to wait until /dev/input/event2 is present and pause if it goes away, that would probably be best. The systemd unit for the program should be able to start the program at boot and r Jul 13 13:59:09 How do I create that? Jul 13 13:59:20 bachp: not afaik, that's a magic location used by the kernel source package Jul 13 13:59:38 Technicus: i suggest trying #systemd (or reading the systemd documentation) Jul 13 14:00:16 right on :) Jul 13 14:03:10 rburton: It would be ok if the module was rebuilt like the kernel if you do a do_shared_workdir Jul 13 14:04:12 I have samba running on an intel edison. Every time after reboot inorder to connect to the server from another machine, I must reset the password with `smbpasswd -a root`. How do I make this persistent? Jul 13 14:05:21 I also just asked that same question on #samba Jul 13 14:20:16 rburton: Puting the header to normal sysroot doesn't seem to work, I guess kbuild does only look for includes inside KERNEL_SRC resp. STAGING_KERNEL_DIR :( Jul 13 14:20:49 bachp: i suggest asking zeddii / zeddii_home Jul 13 14:25:44 rburton: is it worth to bring this up on the mailing list? Jul 13 14:26:14 bachp: try zedd here first, but yeah why not Jul 13 14:26:54 Ok I will try irc first then thx Jul 13 14:27:10 * zeddii reads Jul 13 14:31:29 zeddi: Did you get our conversation or should I repeat the question? Jul 13 14:36:35 I think I have it. This is something that pops up off and on, but normally about module A needing symbols from module B, which we've addressed in krogoth/master. But yah, most of them are just looking in the common kernel source dir. Jul 13 14:39:15 What is the preferred method to reach the maintainers of meta-qt5? Which email list? Jul 13 14:41:40 openembedded-devel iirc Jul 13 14:43:39 rburton: What's the address/subcription page to it? I cant seem to find it listed on the yoctoproject.org site... Jul 13 14:45:54 sveinse: http://lists.openembedded.org/mailman/listinfo/openembedded-devel Jul 13 14:48:24 ooi, I've just read quite a bit in the mega manual. I get the impression that deb is lesser used than rpm and ipkg. Is that correct? Jul 13 14:48:50 (Because I'm coming from a debian background so I tend to default to debs...) Jul 13 14:49:06 ipk is pretty similar to deb, but opkg is faster than dpkg Jul 13 14:53:03 sveinse: since not many folks use the deb support, there's no guarantees on how well it behaves. but yes, it's lesser used. opkg is a smaller lighter dpkg+apt-get, so is more useful for very small systems, but admittedly nowadays "embedded" doesn't always mean very resource constrained, so.. :) Jul 13 14:56:03 kergoth: no it does not. The system we're replacing with yocto runs Ubuntu today actually. But on-target dpkg upgrades is an exercise in slow, so we don't use that. Jul 13 16:01:11 kergoth: SRCREV_FORMAT has a bug it seems when I have two repos and I say in SRC_URI name=libcxx and name=libcxxabi and say SRCREV_FORMAT = "libcxx_libcxxabi" in recipe then I get the SRCREV_FORMAT as (sha of libcxx)_(sha of libcxx)abi Jul 13 16:01:57 it seems we have to specify either define the limitations saying the name= in SRC_URI can not be substrings of one another Jul 13 16:02:06 or fix the parser Jul 13 16:06:40 see https://github.com/kraj/bitbake/blob/master/lib/bb/fetch2/__init__.py#L772 Jul 13 16:07:21 we repace name with SHA which will replace libcxx substr in libcxx as well as libcxxabi Jul 13 16:10:06 we should be using regexp instead of replace() Jul 13 16:16:11 how do we find the anchors it going to be problem Jul 13 16:16:34 I dont think we flag name=foo_bar and name=zoo Jul 13 16:17:00 do we disllow _ in name= param of SRC_URI ? Jul 13 16:22:53 hi - where can I find out from within a booted yocto the build timestamp string we get in image filenames ? running strings on the iso shows it is present in something looking like a bootloader commandline, but I can't find it in the filesystems Jul 13 16:51:44 khem: huh, yeah, that should definitely be fixed Jul 13 16:52:06 khem: i'd say we split by _ and replace matching elements, rather than doing string search/replace of any kind Jul 13 16:58:37 i hate changes that require me to run world builds to make sure they didn't break anything Jul 13 16:58:40 takes forever to make progress Jul 13 16:58:56 actually going to need to do 3-4 world builds to test incremental changes to this.. Jul 13 17:29:38 hey all Jul 13 17:29:48 is there a way to unset a variable attribute ? Jul 13 17:30:09 I am trying to reactivate a task that was marked as noexec, and setting it to 0 doesn't seem to help... Jul 13 17:30:17 setting it to "" neither Jul 13 17:32:42 strip it in an anonymous python function? Jul 13 17:38:27 boucman_work, neverpanic: yeah, i think you'll ahve to use delVarFlag() in anonymous python, unfortunately. we need to fix that Jul 13 17:39:43 RP: any chance on being able to look at shallow one of these days? :) just want to make sure it's still on the radar Jul 13 17:40:34 shallow clone patches ? I think that should be very helpful as more and more stuff become git'ified Jul 13 17:41:02 shallow mirror tarball patches, actually. clone isn't viable unless you use AUTOREV, since we don't know the depth between branch HEAD and SRCREV at fetch time Jul 13 17:41:07 kergoth: for now I just changed the names of SRC_URIs to avert the issue but it took a while to understand it Jul 13 17:41:26 yeah, understandable, I had no idea that was an issue. should probably open a bug for tracking purposes Jul 13 17:41:37 yeah let me do that Jul 13 17:43:00 kergoth: fyi i am trying to change a file in a particular image (image, not package) so I am trying to reenable the fetch target to have my file in the image's builddir Jul 13 17:43:29 ah, yeah, i've done that from time to time Jul 13 17:45:52 khem: if you're curious, https://github.com/openembedded/bitbake/compare/master...kergoth:shallow-simplify -- the third commit describes it - https://github.com/openembedded/bitbake/commit/d3fc8d7 Jul 13 17:49:30 Side notes: The shallow mirror tarballs are never unpacked into DL_DIR, they're unpacked directly from DL_DIR into WORKDIR, and a current clone is preferred to either tarball, a shallow tarball is preferred to an out of date clone, and a missing clone will use either tarball. Jul 13 17:52:33 doesn't do what I want... i'm investigating Jul 13 17:53:14 kergoth: shallow clones would be awesome... nowdays a git clone of the kernel is ~3Go Jul 13 17:54:19 a shallow git kernel tarball is around 150mb instead of >1gb for a full git mirror tarball. of coruse if you fetch enough updates the advantage would be lost, but i think there's value Jul 13 17:57:42 people still using modems out there? :) Jul 13 18:00:13 zeddi: sorry for the delayed response, what do you mean this was adressed in krogoth/master? Jul 13 18:01:21 * paulg dials up the BBS to download zeddii 's latest kernel Jul 13 18:01:55 paulg: your broadband provider may put max data limits too :) Jul 13 18:02:35 so you may rather want to stream a 4K video instead of downloading OE packages again and again Jul 13 18:03:15 or update the laptop while at work. :) Jul 13 18:03:32 ok, that "helps" the file ends up in downloads, but not in ${WORKDIR} i'm probably missing something obvious, is there another task responsible for copying from download to workdir that would also be noexec ? Jul 13 18:03:34 or at Starbucks. :) Jul 13 18:09:19 boucman_work: yes, do_fetch downloads to DL_DIR, do_unpack copies/unpacks from DL_DIR to WORKDIR Jul 13 18:13:58 thx, i'll look if it' noexec'd too Jul 13 18:15:40 it is Jul 13 19:01:48 hello - if i've deleted a device tree file in my build folder, is there an easy way to recover it? Jul 13 19:01:57 without deleting the build folder and re-running everything Jul 13 19:02:06 i've tried bitbake -c fetchall Jul 13 19:02:10 and bitbake -c clean Jul 13 19:02:35 i also tried deleting the entire /arch/arm/boot/dts folder, hoping that it would trigger downloading it from source **** ENDING LOGGING AT Thu Jul 14 02:59:59 2016