**** BEGIN LOGGING AT Fri Oct 05 03:00:02 2018 Oct 05 04:41:49 New news from stackoverflow: Read and write in ioctl linux device driver Oct 05 08:12:27 New news from stackoverflow: Can I use Google-assistant library on YOCTO linux? Oct 05 08:44:24 how are include paths set when using a library in yocto? e.g. I use glibc (more specific eglibc) in a project but I can't see any include path to glibc directories in the build command yet it seems to find the header files Oct 05 08:47:32 jkliemann: look at the logs of the project, namely the do_configure and do_compile steps. they should be pretty verbose about that. Oct 05 09:49:57 Hi guys. I have added my recipe app, and did: bitbake my-app.. It compiles, BUT: I don't see a release *.ipk package.. I see only my-app-dbg.ipk and my-app-dev.ipg packages.. Why? What is reason? Oct 05 09:50:35 It is Qt app, mased on qmake Oct 05 09:50:42 s/mased/based Oct 05 09:51:01 kuzulis: probably you didn't set FILES_${PN}. it holds the list of files that is meant to go into the package Oct 05 09:51:07 kuzulis: because the app didn't install anything that went into PN Oct 05 09:51:15 the default PN has /usr/bin etc etc Oct 05 09:52:16 rburton: for reasons unknown, "usr bin" just made me think of the trashcan beneath my desk Oct 05 09:55:07 rburton: Hmm.. I did by this wiki: https://wiki.yoctoproject.org/wiki/Building_your_own_recipes_from_first_principles and there are no FILES_${PN} in *.bb file Oct 05 09:55:30 the default is good for stuff which installs to $prefix like it should Oct 05 09:56:09 Hiya! I'm a Yocto noob, but trying my best :-) How could I make Yacto stick to GLibC 2.27: "WARNING: Your host glibc verson (2.28) is newer than that in uninative (2.27). Disabling uninative so that sstate is not corrupted." Oct 05 09:56:31 thats just a warning, ignore it Oct 05 09:57:03 banach-space: this message basically says that the build process need to compile some stuff from scratch for your system and cannot use the prebuilt uninative Oct 05 09:57:11 banach-space: -> ignore it. Oct 05 09:57:47 I'd love to ignore it, but the build fails and a bit of googling revealed that GLibC 2.28 can be a bit problematic. Oct 05 09:58:14 So I wanted to make sure that GLibC 2.28 (the system one) is not used at all. Oct 05 09:58:37 banach-space: thats true but unrelated to uninative, AFAIK. its basically that older rleases don't build on systems with 2.28 Oct 05 09:59:14 banach-space: so either build on a host with an older glibc (can be a docker container too!) or try latest master Oct 05 10:01:13 OK, FYI I'm using Sumo. Will try downgrading GLibC then, thanks! Oct 05 10:01:19 rburton: I have added this: FILES_${PN}-dbg += "${datadir}/${P}/.debug" Oct 05 10:01:19 FILES_${PN} += "${datadir}" Oct 05 10:01:20 to by recipe *.bb file.. But it generates same dbg and dev packages. Oct 05 10:01:36 s/to by/to my Oct 05 10:01:37 kuzulis: don't do that, FILES_PN-dbg is populated magically Oct 05 10:02:23 rburton: Yocto is a full magic ))) Oct 05 10:03:40 rburton: I keep only FILES_${PN} += "${datadir}", but the result same.. (( Oct 05 10:04:18 kuzulis: so what do you expect it to have installed? Oct 05 10:04:31 because the problem is almost certainly your app Oct 05 10:05:35 rburton: I expect that an output IPK files is in /tmp/deploy/ipk directory will be without of dbg suffix.. Oct 05 10:05:49 kuzulis: sure, but what files in the package do you expect to see Oct 05 10:05:58 presuming you can't share the app and recipe Oct 05 10:07:30 rburton: I don't know what I expect... I'm just confused with 'dbg' package suffix Oct 05 10:07:54 kuzulis: surely you know what my-app should be installing Oct 05 10:09:08 rburton: Yes, It just install own 'executable' file. Oct 05 10:09:18 the dbg package is the debug symbols Oct 05 10:09:59 rburton: Yes, I know that is it.. But I don't understand why bitbake compiled it as 'debug' target Oct 05 10:10:07 it didn't Oct 05 10:10:21 it did a normal build, extracted the debug symbols, and put them into the PN-dbg package Oct 05 10:10:35 the PN package contains the stripped binaries Oct 05 10:11:44 rburton: Ah..So, so, how then I can say to bitbake to build my app in 'release'? Oct 05 10:12:31 rburton: Do I need .bbappend file for this? Oct 05 10:12:49 this build vs release thing is generally the mindset of someone who has used visual studio Oct 05 10:13:04 there is no build/release toggle in yocto Oct 05 10:13:13 erm debug/release Oct 05 10:13:30 a normal build has debug symbols, these are extracted from teh binaries and put in a separate package Oct 05 10:13:42 so the normal package is the binaries, and the dbg package is *just* the debug symbols Oct 05 10:14:08 ok, but how then I can create a 'normal' binaries package? )) Oct 05 10:14:27 I don't understand, why it does not creates Oct 05 10:14:37 so again, what file do you expect it to have built? Oct 05 10:14:43 /usr/bin/my-app? Oct 05 10:15:02 Hmm.. lett assume, that yes Oct 05 10:15:24 then assuming that your app knows how to install properly, then it would have been installed and you'll have a my-app package Oct 05 10:15:39 if that didn't work then my-app is broken Oct 05 10:16:21 read the recipe, look at how it does install, look at the my-app rules for install, debug why you don't end up with /usr/bin/my-app installed under ${D} during do_install Oct 05 10:16:50 Hmm.. my recipe has not any do_install() scripts.. Oct 05 10:16:59 does it inherit any classes? Oct 05 10:17:30 I thought everything would be done automatically. Oct 05 10:18:01 It inherit only qmake5 Oct 05 10:18:07 in .bb file Oct 05 10:18:26 right, and that class has a do_install Oct 05 10:18:41 so now you get to figure out why your my-app's qmake file doesn't install properly Oct 05 10:19:05 congrats you've moved the problem out of the recipe and into my-app itself Oct 05 10:19:50 the qmake5 class will configure/build/install correctly *assuming* the qmake file actually works Oct 05 10:19:57 So, do I need in do_install() ? Oct 05 10:20:09 no Oct 05 10:20:14 the qmake class has a do_install Oct 05 10:20:38 but if that is failing then you know the problem is in my-app itself Oct 05 10:21:00 Do you mean, that it failing in compile-time? Oct 05 10:21:09 well install time Oct 05 10:21:40 Ah.. ok... many thanks for your time :) Oct 05 10:21:45 the class just asks the app to install Oct 05 10:21:52 from your description, that is broken Oct 05 10:22:05 I will check, thanks )) Oct 05 10:23:16 Is in yocto any logs to see what happens ? e.g. in 'install' time? Oct 05 10:23:28 yes in the work dir Oct 05 10:23:40 under your recipe's temp/ there's log.do_install Oct 05 10:29:24 thanks Oct 05 10:38:10 I'm bbappending to the dhcp_4.3.6.bb recipe, but that recipe already has a do_install_append().. can I have another do_install_append() in my .bbappend? Oct 05 10:38:15 yes Oct 05 10:38:38 So it appends to the append(ix) or does it override it? Oct 05 10:39:00 appends Oct 05 10:39:12 Nice! Thanks :) Oct 05 10:43:11 rburton: A new question, what does this mean: Files/directories were installed but not shipped in any package ? Oct 05 10:43:34 e.g. my app has install path as /home/root/ Oct 05 10:44:26 so that warning was telling you that files were installed (by do_install) but not put into any packages Oct 05 10:44:41 ideally, don't install apps to /home/root Oct 05 10:44:54 if you must add that to FILES_${PN} Oct 05 10:45:12 apps should be in /usr/bin... Oct 05 10:46:39 yes, I know about /usb/bin, but I need in custom location.. Oct 05 10:46:50 then you have to deal with the fallout of that Oct 05 10:47:07 which starts with setting FILES_${PN} as appropriae Oct 05 10:47:58 WOW, it works: FILES_${PN} += "/home/root/" .. many thanks ) Oct 05 10:57:57 rburton: A last question (maybe). So, what a last step to add my app recipe to an image? Is it just add it to IMAGE_INSTALL_append of my image.bb file? Oct 05 10:58:03 yes Oct 05 10:58:11 thx Oct 05 10:58:35 LetoThe2nd: sorry for the late answer, i looked into the do_compile logs and in the whole command only the projects own include paths are set, how does it find the libc? my original problem is that it indeed finds the libc headers but some defines from signal.h are missing yet all signal.h files I could find have them so I have no idea where the include comes from and why it doesnt define some symbols Oct 05 10:59:24 jkliemann: no problem. have you looked into do_configure too, and the environment it sets? specifically the sysroot? Oct 05 11:03:21 do_configure just shows that there are no errors and everything seems to be fine, this seems to be okay since the original project doesnt need any further configure steps. which sysroot exactly do you mean? i noticed that each library seems to have its own (sometimes?) to build, what still confuses me is that other headers such as stdio.h work fine Oct 05 11:03:44 jkliemann: each recipe gets its own sysroot, indeed Oct 05 11:04:35 should it be under tmp/work///sysroot ? Oct 05 11:06:27 i don't have a suffiently new workdir handy to check, sorry. Oct 05 11:08:45 jkliemann: also look at the run.* files of the stages. those should contain what bitbake feeds into the repective tasks Oct 05 11:12:29 FWIW, bitbake -c devshell is what I normally use for this kind of debugging Oct 05 11:12:45 jkliemann: FWIW, bitbake -c devshell is what I normally use for this kind of debugging Oct 05 11:18:18 LetoThe2nd: it seems to use the correct sysroot, it has the same sysroot options as all other packages and if I add an #error it fails there. some of the missing defines seem to be missing due to another missing define yet when I add a -D__USE_POSIX it doesnt seem to have any effect Oct 05 11:19:15 jkliemann: hm, no idea right now (and i gotta run, sorry) Oct 05 11:20:21 jofr: when I call make in the devshell it fails to link but the initial error is gone Oct 05 11:20:43 LetoThe2nd: no problem, still thanks, you pointed me into some good directions Oct 05 11:32:30 got it! i defined -std=c99 instead of -std=gnu99 which lead to __USE_POSIX being undefined and therefor not including some definitions Oct 05 11:34:41 rburton: Why my app which I install to custom /home/root/foo directory has not an executable rights? Oct 05 11:36:48 rburton: But in /tmp/work/.../image/../ it has executable rights Oct 05 11:39:18 Did you do an install -m in your do_install? Oct 05 11:39:54 rburton: i.e. it has not execurable rights after I flashed an image to target HW.. But if I mount same image via 'mount' utility, then I see an executable rights.. Oct 05 11:40:44 jofr: No, I did nothing.. I have not an do_install() script in my .bb file Oct 05 11:41:39 kuzulis: is /home mounted with executable mode allowed? Oct 05 11:43:11 New news from stackoverflow: Can I use Google-assistant library on YOCTO linux? [on hold] Oct 05 11:44:00 sorry, guys.. all ok.. it is my inattention.. Oct 05 12:25:31 can I print a warning from just parsing a file? Oct 05 12:26:06 this is a .conf file actually, I guess that's different from recipes Oct 05 13:04:14 Dear all, Oct 05 13:04:26 What is the scope of PACKAGE_EXCLUDE when used inside a recipe (to build image) Oct 05 13:04:34 do I need to add PACKAGE_EXCLUDE_pn-${PN} = "foo" to not pollute the name space? Oct 05 13:04:40 when I do want to exclude "foo" ? Oct 05 13:27:33 lukma: image-scope, just set it in the image recipe Oct 05 13:32:17 rburton: Is this somewhere written? Oct 05 13:32:48 _pn-foo is for when you're doing a recipe-specific change from a global config such as local.conf Oct 05 13:33:09 so PACKAGE_EXCLUDE_pn-core-image-base = "foo" in local.conf would exclude foo from core-image-base Oct 05 13:33:18 but if its your image, just put PACAKGE_EXCLUDE=foo in it Oct 05 13:36:43 Hello folks Oct 05 13:36:53 I've got a question about SDK. Oct 05 13:38:09 Is it possible to embed source code of all applications installed to a distro to eSDK, so it would be used for offline builds? Oct 05 13:39:21 I wanna have an autonomous SDK, which would need an internet to fetch sources when image rebuild is requested by the SDK user. Oct 05 13:39:36 * wouldn't need an internet Oct 05 13:40:22 Is it possible to create something like this based on for instance eSDK? Oct 05 13:45:55 exit Oct 05 13:47:20 ok I just created a dummy.bb recipe with an __anonymous function Oct 05 13:50:26 is it possible to append a string to a variable only when it is set? Oct 05 14:34:32 RP, rburton I started a build. you can kill it if need be Oct 05 14:37:15 my son's birthday dinner shortly, so i'm outta here. However, I do have an almost-working nfs-ganesha recipe in case anyone fancies continuing it and eventually moving runqemu to using that instead. Oct 05 15:09:24 rburton, have fun.. Oct 05 15:09:35 rburton, i have a fancy Oct 05 15:09:52 where can I find it? Oct 05 15:13:54 New news from stackoverflow: Install Tensorflow on yocto linux for R-CAR H3- ARM -SOC Oct 05 15:44:01 New news from stackoverflow: How to enable tc command when building a kernel using Yocto recipes Oct 05 16:40:25 Hi, I modified a recipe yesterday to use some complicated python macros to set PN. Basically, if you are building a release the PN is a bare version string like 0.4.5 pulled out of an include file. If you are building out of the git develop branch, add a suffix with the git commit ID. This part worked with other recipes. But this recipe has a SRC_URI += "file://somefile". Now when I run that recipe I get Oct 05 16:40:31 an error that bitbake can't calculate the checksum on somefile. Any ideas? Oct 05 16:41:21 it can't find the file at all, most likely. if you put it in a bbappend, amke sure you also adjusted FILESEXTRAPATHS Oct 05 16:41:29 or put the recipe in the right place on disk Oct 05 16:41:33 s/recipe/file/ Oct 05 16:43:50 It has no problem finding the file in do_fetch. It is only when it goes to checksum that it no longer finds it Oct 05 16:44:27 files are checksummed at parse time Oct 05 16:44:32 long before tasks are run Oct 05 16:44:56 I was thinking it might be specifically related to the logic to checksum with PV Oct 05 16:45:01 https://patches.openembedded.org/patch/117673/ Oct 05 16:45:43 sorry. I meant to say I modify PV, not PN Oct 05 16:45:49 where exactly is the file? if it's in a path that includes pv, of course changing pv would make it no longer find it.. Oct 05 16:46:04 recipes-foo/foo/files/somefile Oct 05 16:46:24 where the recipe is recipes-foo/foo/foo_%.bb Oct 05 16:47:06 I also tried FILESEXTRAPATHS and that did not change anything. Oct 05 16:47:33 _% makes no sense, i hope that isn't the actual path on disk Oct 05 16:47:49 first i'd suggest postoing the actual error message you're seeing, and when it's happening, rather than summarizing Oct 05 16:48:27 I am including a version number centrally from an include file, so yes, I am actually using _% (and later replacing PV. Oct 05 16:50:02 i'd just use foo.bb in th at case, rather than using the same character used in bbappend wildcards, just to avoid confusion Oct 05 16:50:12 but i doubt that's the issue, just worth noting Oct 05 16:50:17 ok. i will give that a try. Thanks Oct 05 16:55:10 Hey guys, I'm new to yocto, and I'm looking for some info about LSB and Yocto. Oct 05 16:55:33 I found this wiki page about LSB and Yocto, but it looks dated. -> https://wiki.yoctoproject.org/wiki/LSB_Result Oct 05 16:56:17 Is there somewhere I can find out about LSB compliance by version on Yocto? Oct 05 16:58:19 @kergoth, same error message: Oct 05 16:58:25 WARNING: /home/rohan/plt/fsl-release-bsp/sources/meta-plantronics-marmot/recipes-tools/foo/foo.bb: Unable to get checksum for foo SRC_URI entry VERSION: file could not be found Oct 05 16:58:28 WARNING: /home/rohan/plt/fsl-release-bsp/sources/meta-plantronics-marmot/recipes-tools/foo/foo.bb: Unable to get checksum for foo SRC_URI entry VERSION: file could not be found Oct 05 16:58:42 ERROR: An uncaught exception occurred in runqueue########## | ETA: 0:00:00 Oct 05 16:58:45 Traceback (most recent call last): Oct 05 16:58:48 File "/home/rohan/plt/fsl-release-bsp/sources/poky/bitbake/lib/bb/runqueue.py", line 948, in RunQueueData.prepare(): Oct 05 16:58:51 (mc, fn, taskname, taskfn) = split_tid_mcfn(tid) Oct 05 16:58:53 > self.runtaskentries[tid].hash = bb.parse.siggen.get_taskhash(taskfn, taskname, procdep, self.dataCaches[mc]) Oct 05 16:58:56 task = self.runtaskentries[tid].task Oct 05 16:58:59 File "/home/rohan/plt/fsl-release-bsp/sources/poky/meta/lib/oe/sstatesig.py", line 139, in SignatureGeneratorOEBasicHash.get_taskhash(fn='/home/rohan/plt/fsl-release-bsp/sources/meta-plantronics-marmot/recipes-tools/foo/foo.bb', task='do_fetch', deps=[], dataCache=): Oct 05 16:59:04 def get_taskhash(self, fn, task, deps, dataCache): Oct 05 16:59:06 > h = super(bb.siggen.SignatureGeneratorBasicHash, self).get_taskhash(fn, task, deps, dataCach Oct 05 16:59:12 that should have failed at parse time, not runqueue generation time, unless you have something which changes from one parse to the next Oct 05 17:02:45 that's what i thought too if there was an error Oct 05 17:04:08 RP: interesting, https://www.python.org/dev/peps/pep-0540/ Oct 05 17:06:56 kergoth: should I paste the recipe? Oct 05 17:08:18 LICENSE = "CLOSED" Oct 05 17:08:19 S = "${WORKDIR}" Oct 05 17:08:19 SRC_URI += "file://somefile" Oct 05 17:08:19 do_install() { Oct 05 17:08:19 echo ${PIKA_VERSION} ${PV} | cat - somefile >>VERSION Oct 05 17:08:21 install -d ${D}/home/root Oct 05 17:08:23 install -m 0644 VERSION ${D}/home/root Oct 05 17:08:26 } Oct 05 17:08:28 FILES_${PN} += "/home/root/VERSION" Oct 05 17:08:31 require recipes-tools/version/pika-version.inc Oct 05 17:08:33 # this just has one line: PIKA_VERSION = "0.4.5" Oct 05 17:08:36 PIKA_BUILD_SOURCE ??= "git" Oct 05 17:08:38 LOCAL_PV = "${PIKA_VERSION}-local" Oct 05 17:08:41 GIT_PV = "${PIKA_VERSION}+git${SRCPV}" Oct 05 17:08:43 RELEASE_PV = "${PIKA_VERSION}" Oct 05 17:08:46 SRCREV = "${@bb.data.getVar('AUTOREV',d,1) if bb.data.getVar('PIKA_BUILD_SOURCE',d,1)=='git' else bb.data.getVar('PIKA_VERSION',d,1)}" Oct 05 17:08:49 PV = "${@bb.data.getVar('LOCAL_PV',d,1) if bb.data.getVar('PIKA_BUILD_SOURCE',d,1)=='local' else bb.data.getVar('GIT_PV',d,1) if bb.data.getVar('PIKA_BUILD_SOURCE',d,1)=='git' else bb.data.getVar('RELEASE_PV',d,1)}" Oct 05 17:11:12 RP: https://www.python.org/dev/peps/pep-0538/ too Oct 05 17:12:51 kergoth: Any suggestions about something I could try next? Oct 05 17:27:04 the first thing worth noting is oe-core already provides a mechanism to do what you want, so you don't have to do this yourself at all Oct 05 17:27:48 https://github.com/openembedded/openembedded-core/blob/master/meta/classes/devupstream.bbclass — if you follow the steps in the comments, you'll have two versions of your recipe available to be built, and you can select which using PREFERRED_VERSION as though you had two recipe files for the two versions Oct 05 17:28:50 although, i think both would be parsed, and if upstream isn't available for every user, maybe that wouldn't be ideal.. hmm Oct 05 17:39:45 IRC check. Is this received? Oct 05 17:44:12 Hello all. IRC check. Is this received? Oct 05 17:44:51 test failed Oct 05 17:44:57 lol Oct 05 17:45:21 The new authentication makes IRC login more cumbersome. Ugh :( Oct 05 17:46:34 new authentication? Oct 05 17:46:56 if you're just talking about nickserv, use a decent irc client and it'll auto-login Oct 05 17:47:07 msg NickServ identify blahblah Oct 05 17:57:42 All: I deployed an embedded application using jethro a few years ago. I got krogoth compiling project, and am now jumping to morty. What are benefits of jumping to pyro->rocko->sumo? Oct 05 17:58:07 I hold the kernel fixed due to hardware. Oct 05 18:07:06 mrk377: pyro added recipe specific sysroots, which are reallty nice Oct 05 18:08:50 also, like updated versios of stuff Oct 05 18:09:01 some CVE's fixed, new ones introduced Oct 05 18:09:18 also, people will laugh at you for using ancient versions Oct 05 18:09:43 older kernels need patches to compile with newer gcc Oct 05 18:13:00 JPEW/Crofton - The CVE's are important for the Information Assurance junkies. Having updated packages is also a requirement too. My recipes have required small tweaks for Morty. Oct 05 18:13:33 I wonder if leap frogging to pyro->rocko->sumo is best, or jump straight to sumo? Is one way better for sanity. Oct 05 18:17:33 old kernels can hav eissues also Oct 05 18:17:40 I'd go all the way Oct 05 18:17:54 and try to track new releases Oct 05 18:28:34 thx crofton. I will try sumo next. Oct 05 18:38:20 All - Is there a way to override checking QAPATHTEST[host-user-contaminated] = "package_qa_check_host_user" in insane.bbclass. Our docker swarm runs as uid 0, and creates a ton of "is owned by uid 0... contamination" Oct 05 19:10:04 mrk377: I wouldn't recommend it, but you can do WARN_QA_remove = "host-user-contaminated" Oct 05 19:11:09 There are legitamate and serious problems that QA warning can find :) Oct 05 19:11:21 thx JPEW Oct 05 19:11:47 It would be *much* better to run as any user other than root :) Oct 05 19:11:52 What are impacts from your experience? Oct 05 19:12:47 That particular QA warning can find cases where recipes are written incorrectly, and the files end up on the rootfs with the UID of the user running bitbake instead of the correct UID (which can be very bad) Oct 05 19:13:23 Oh, that is important. Oct 05 19:13:54 Running bitbake as root confuses it because there are *lots* of files on the rootfs owned by root (on purpose), so it cant distinguish between ones that are supposed to be owned by root and those that are accidentally owned by root Oct 05 19:16:51 I appreciate your time/input. It is very helpful for me. Oct 05 19:52:30 Is there any package that provides the rest of sysvinit functionality like "service"? Oct 05 20:03:21 tgoodwin: service is a ubuntu-specific thing iirc Oct 05 20:03:54 rburton: I was going off this: https://fedoraproject.org/wiki/SysVinit_to_Systemd_Cheatsheet Oct 05 20:04:06 ah, fedora ship it too Oct 05 20:04:11 either way its not standard sysvinit Oct 05 20:04:20 alright, thanks Oct 05 20:06:30 armpit: poky-contrib:ross/mut. their cmake files are nonsense, and it fails to link natively right now Oct 05 20:07:42 rburton, thanks Oct 05 20:51:10 I didn't get this error with Krogoth, but now it shows on Morty **QA Issue: core-image-project: Recipe file fetches files and does not have license file information (LIC_FILES_CHKSUM) [license-checksum]. I have about 19 IMAGE_INSTALL packages in my core-image-project.bb that I don't control (packages from internet). Must I create LIC_FILES_CHKSUM for the culprits?? Oct 05 20:56:40 Kick me. I had a blank (i.e../recipes-connectivity/remserial/remserial_1.4.bb:LIC_FILES_CHKSUM = "") Oct 05 21:10:35 Folks, one more time the question regarding SDK. Oct 05 21:10:35 Is it possible to embed source code of all installed to the image applications into eSDK, so the SDK would be used for offline builds/rebuilds? Oct 05 21:10:35 I wanna have an autonomous SDK, which wouldn't need the Internet to fetch sources when image rebuild is requested by the SDK user. **** ENDING LOGGING AT Sat Oct 06 03:00:01 2018