**** BEGIN LOGGING AT Fri May 08 02:59:59 2020 May 08 04:35:19 I have known how to use wic to create a image with my build target May 08 04:35:48 but the problem is I just want to create a ext4 image for the rootfs and a tar package for debugging rootfs May 08 04:36:03 can wic do that ? Or I should do it in the old way ? May 08 04:44:16 wic doesn't have anyting to do with that May 08 04:44:27 you can use wic as as a normal image fstype as you would any other image fstype May 08 04:44:36 set WKS_FILE and add wic to IMAGE_FSTYPES May 08 04:44:55 whether you also build a tarball of it, or use the debug rootfs support is independent of that May 08 05:26:06 kergoth, but I think it is a common usage, it is better to create wks files for those purpose May 08 05:26:17 that doesn't make sense. May 08 05:26:26 the wks controls the creation of the .wic disk image only May 08 05:26:36 whether you're also building other things has nothing to do with what wic is doing May 08 05:27:08 if I submit those patches, will you accept it May 08 05:27:08 and wic works independent of oe, whereas debug image construction is a specific oe feature May 08 05:27:47 you see, many developer would like to only flash its rootfs while keeping its data partition May 08 05:27:53 and debug its system May 08 05:28:54 oe already has support for generating a parallel debug rootfs. whether you use wic to create the rootfs image doesn't matter May 08 05:49:50 is it possible to have npm install dependency from private repository? I get host key verification failed, although it is in same repository than the git sources in the same recipe that use the same ssh key May 08 06:05:27 stuom1: ssh key != key verification May 08 06:06:20 stuom1: in most cases, its rather that the host serving the code through https/gits is having a self-signed or ortherwise invalid certificate May 08 06:10:26 it says also npm ERR! Could not create directory '/home/root/.ssh'. May 08 06:14:01 basically i can do "npm install" on host but not in recipe do_install anymore because of this new package in private repository, and need to figure out why May 08 09:10:02 hi, where could i find info on how to setup SRCREV_FORMAT ? May 08 09:10:56 on the manual i see "We need to document AUTOREV and SRCREV_FORMAT here." but maybe i am lookign the wrong place May 08 09:12:30 kergoth: you mean add IMAGE_GEN_DEBUGFS=1 in local.conf May 08 09:12:31 ok, found it, sry May 08 09:12:43 is it recommended way ? May 08 09:37:12 I hope this is the right place to ask this question, I'm trying to build a native SDK including protobuf compiler (protoc) and the grpc plugin (which is just an executable) for C++ (both in meta-oe/recipes-devtools). Now, I do see the protoc in the SDK, and also the grpc libraries, but not the plugins for code generation, which would be essential May 08 09:37:13 for the SDK including grpc. What could be the cause of this? Thanks in advance. May 08 09:38:00 I have to say in advance, I'm using petalinux, so, it is not pure Yocto. But they couldn't help me there. May 08 09:47:15 Plinnie: maybe the plugins are in a second package or recipe? no idea, jsut guessing - you'll have to look at the recipes. May 08 09:56:06 Thank Leto. not as far as I can see. Then again, I don't see a separate recipe for the protobuf-compiler either. i don't know how bitbake determines which files end up in the sdk. Here is the protobuf recipe: https://git.openembedded.org/meta-openembedded/tree/meta-oe/recipes-devtools/protobuf/protobuf_3.11.4.bb and here the grpc recipe May 08 09:56:06 https://git.openembedded.org/meta-openembedded/tree/meta-oe/recipes-devtools/grpc/grpc_1.24.3.bb Both extend the nativesdk class May 08 09:59:22 well, using SRC_URI_append is bbappend to add a git repo from where getting some additional files (overlay), but seems this repo is totally replacing the .bb one. What may be wrong here ? May 08 09:59:46 s/is/in May 08 10:00:04 Plinnie: was just giving ideas, i have never used this, sorry. May 08 10:00:22 ok, that is fine. Thanks for your pointers May 08 10:01:51 Would #oe be a more appropriate place for these questions? May 08 10:02:59 Plinnie: both are fine, but as you've already asked here probably everybody in #oe already knows anyways. :) May 08 10:03:44 ah, ok! :-) May 08 12:43:25 Hi, i want to use D-Bus in yocto. Is these somebody who can recommend a C++ open source library? May 08 12:43:49 https://www.freedesktop.org/wiki/Software/DBusBindings/ This list looks very old :) May 08 12:47:35 Or should i switch to systemd and use the sdbus-c++ binding? May 08 12:51:54 PatrickE, use the plain C API ? May 08 12:53:52 kroon yes. But that would mean that i would have to write the C++ abstraction. I would like to avoid using a mixture of C and C++ in the upper layers of my software May 08 12:55:10 I don't see why one would need to add a C++ abstraction May 08 12:56:10 I've had good results with the QtDbus API in the past. May 08 12:57:09 mfe555 yes it looked really mature. But i think the license would be an issue :) May 08 13:06:39 ./meta-openembedded/meta-oe/recipes-core/dbus/libdbus-c++_0.9.0.bb May 08 13:09:51 hey good morning !/? , anyone ever had issue with pyro rootfs full host poisoning ? I'm building within a docker container running ubuntu1604. Using a image that inherit from core-image and the distro is "somewhat" inspire from poky-tiny. the final image is expected to be a tar.gz file. All file on the rootfs are own by my user in docker. May 08 13:27:46 aleblanc: in those kind of cases with poky based distros the QA checks would fail the build early. Sounds like pseudo or something else is not working at all which can be because of pseudo missing some new syscalls for example. I would try to reproduce with a minimal image debug the steps.. May 08 13:32:17 what's the process for submitting a recipe upstream to oe? May 08 13:32:24 mcfrisk: I looked into the check rootfs_check_host_user_contaminated , it look specifically for a /home/USER, as /home/root exist , the test pass. I'm not aware of any other test. (feel free to tell if yo know any) . As for pseudo, running a "bitbake some package -c devshell" follow by doig a touch FILE show that the file does endup with user root. But will try with a minimal image an debug for May 08 13:32:26 sure May 08 13:37:04 aleblanc: you need to run the task scripts. in devshell temp/run.do_install for example. the tasks marked with fakeroot will be run with pseudo and should result in files with root. You cal also see file permissions from buildhistory btw. a custom image class may do mayhem as well. May 08 13:37:34 mcfrisk: ah ! ok will look into right away May 08 13:38:22 mcfrisk: btw , thank for the support May 08 13:38:28 / help May 08 14:17:48 mcfrisk: hum interesting, still investigating but , when in devshell for my image, the rootfs show the correct ownership, when not , it goes back to my owner May 08 14:24:13 aleblanc: that's pseudo doing it's job May 08 14:24:48 the tar ball needs to be generated as task inside bitbake with fakeroot, then the permissions will be correct. May 08 14:32:16 mcfrisk: ah ! , weird, I would expect the tar ball recipe from image_types.bbclass to this by default ... May 08 14:33:45 when using rpm, should pkg_prerm be getting run via the preuninstall scriptlet during a package upgrade? in my environment it doesn't seem to be :( May 08 14:53:40 aleblanc: I've been using rootfs tarballs generated by bitbake and yocto bbclasses, haven't seen any problems. root permissions are there. Rember to unpack as root or "fakeroot bash" to see the owners correctly. May 08 15:26:33 buildbeef: much like other open source projects: https://www.openembedded.org/wiki/How_to_submit_a_patch_to_OpenEmbedded May 08 15:44:11 mcfrisk: finally found my issue, pseudo does not work outside for file outside the devshell environment, which is why I need to use sudo and not fakeroot to unpack my tar. Many taks for you help. May 08 15:46:44 tanks** May 08 15:52:16 Hello everybody May 08 15:57:18 I'm trying to add cmake-format (a simple python app) to my SDK, so i generated the recipe with devtool, and added nativesdk-cmake-format to TOOLCHAIN_HOST_TASK. That almost worked, but when I run cmake-format, interpreter is not able to find six (a dependency of the tool). Looking to the recipe generated by devtool I see that python3-six is listed in RDEPENDS. How can I make it available in SDK? May 08 16:00:03 afaik it should be already, but you can always explicitly set RDEPENDS_${PN}_class-nativesdk = "nativesdk-python3-six" or whatever May 08 16:00:07 to override the non-nativesdk one May 08 16:01:48 kergoth: So, I'll first check if maybe the dependency is already there (and there is some other kind of problem), and the I'll try that trick. Thanks! May 08 16:05:55 seems like package_rpm.bbclass specifically wraps (wrap_uninstall) the prerm script with a test to only run if all versions were removed (so no run on upgrade) :( May 08 16:30:41 RP: I have send some armv8 tuning additions also suitable for dunfell May 08 16:43:03 khem: in the future if you could tag those with [master][dunfell] when you send them it will help me to avoid missing them! May 08 16:44:14 In general I'll wait for Richard to pull them into master before I'll take them (usually after 2 or 3 days of aging in master) May 08 17:03:20 sakoman: Yes, I should do that, however I wanted to wait on dunfell pull for sometime thats why I did not May 08 18:08:22 hmmm, who let sakoman in May 08 18:09:44 Crofton|road: be nice! May 08 18:10:09 How are you doing? May 08 18:10:55 Doing well all things considered :-) May 08 18:11:33 Hawaii hasn't been hit too bad by covid May 08 18:27:19 sakoman: you should put a live webcam pointing towards beaches for people too remotely enjoy vacations in Hawaii May 08 18:29:06 Crofton|road: you've been on the road for past 3 months if not longer - be nice to sakoman, we all depend on him! :) May 08 18:31:31 I take it you don't plan on hosting an on site developer meeting there :( May 08 18:31:44 I need to adjust the nick on this account May 08 18:40:50 Crofton|road: that long on road you must be "tire"d May 08 18:41:04 urg May 08 18:43:05 better? May 08 18:44:52 uh-oh, Crofton is in/on the cloud now... May 08 18:45:18 this plague is making you all loopy May 08 18:45:20 the singularity is here May 08 18:48:13 khem: I don't have any sandy beaches nearby. Its more like 200 foot cliffs on this side of the island May 08 18:49:20 Crofton|cloud: yeah, I'll pass on hosting any on site meetings ;-) May 08 18:49:54 Post vaccine though, we need to talk May 08 19:03:35 vaccines will make my autism flare up May 08 19:05:44 post vaccine will be 4 years+ as thats what is best we have done May 08 19:30:44 that makes my head explode May 08 19:31:19 We are schduling an oe happy hour with a pacific friendly time May 08 19:39:11 Crofton|cloud: what is your definition of Pacific? The time zone, or those of us in the middle of the similarly named ocean? ;-) May 08 19:39:29 middle of the ocean May 08 19:39:35 trying to get into NZ May 08 19:39:53 people all sleeping at different times is inconvenient May 08 19:39:59 Excellent! I hate these 4:30am meetings I have every week May 08 19:56:43 @Crofton: postfuncs[vaccine]+="real_meeting" May 08 20:34:20 RP: Are you waiting for something on https://lists.openembedded.org/g/openembedded-core/message/137946 or did it just get lost? May 08 23:35:49 halstead: i'd like to confirm that the YP has download mirrors in their infrastructure so if a download fails (https://oldproject.org/oldproject-1.2.tar.gz) it should be found on a mirror (https://yp-infrastructure/oldproject-1.2.tar.gz)? May 08 23:37:34 tlwoerner, For most files yes. Which are you looking for? May 08 23:44:26 tlwoerner, Like the files at http://downloads.yoctoproject.org/mirror/sources/ May 09 02:43:05 halstead: thanks for the link, that's what i was looking for May 09 02:43:46 You're welcome tlwoerner :) May 09 02:44:11 halstead: the XOrg/freedesktop project is thinking of reorganizing their infrastructure and, apparently, "yocto" downloads are so prevalent it came up as part of the discussion :-) May 09 02:45:07 halstead: but with the YP's mirror, i told them it's safe to reorganize. old branches can find stuff on the mirror, and newer stuff can be updated to point to the new location(s) May 09 02:46:46 halstead: i've asked for a list, and if anything's not on the mirror i'll poke you about getting it transferred? May 09 02:47:18 tlwoerner, I think that should work well. Then we can see how many are building using un-maintained branches based on mirror fetches. ;) May 09 02:47:36 nice :-) May 09 02:47:54 That is very helpful. Thanks for doing the double check tlwoerner! **** ENDING LOGGING AT Sat May 09 02:59:58 2020