**** BEGIN LOGGING AT Mon Mar 30 03:00:00 2020 Mar 30 05:21:34 yohboy: last I checked cross compiling any BLAS library was really painfull, I got Scipy working properly with ATLAS on some x86 architectures but not on ARM, this was years ago btw Mar 30 07:13:44 mcfrisk: kudos for writing that answer! Mar 30 07:25:35 New news from stackoverflow: what is a reference distribution? Mar 30 07:53:40 LetoThe2nd: thanks! Mar 30 07:55:41 mcfrisk: i just am pretty much convinced the asker will be dissatisfied because it doesn't involve direct copy-pasteable instructions. but the effort shall be appreciated, hence, thank you for doing it! Mar 30 08:10:49 LetoThe2nd: But why can't you just tell him the exact steps required?! :-) Mar 30 08:11:17 erbo: yeah i know. its all my bad, because i'm an arrogant elitist. Mar 30 08:14:01 Not arrogant elitist, just "not that much helpful" :p Mar 30 08:15:08 hehe Mar 30 08:16:07 erbo: its just the usual for me, the one less-than-average-gifted guy who ruins it for many. Mar 30 08:53:08 hi guys, I have a problem with my bitbake server, when I run bitbake commands it says: NOTE: Retrying server connection (#1)... Mar 30 08:53:59 I'm not using containers, and the only hints I found on google are containers related, anybody has some tips? Mar 30 09:01:38 stefanOffred123: since a particular event, maybe? Mar 30 09:06:17 thanks, Leto, no seems not a particular event. I simply add a custom recipe with devtool to Poky, I have previously run bitbake core-image-minimal, and bitbake without problems. Mar 30 09:06:56 can I do something as clear bitbake cache or something similar? Mar 30 09:07:48 stefanOffred123: my gut feeling is that some background process is stuck or such. are you intentionally using a memory resident bitbake server or such? Mar 30 09:10:16 good morning Mar 30 09:11:04 mcfrisk: howdy! Mar 30 09:16:13 stefanOffred123: did you by any chance Ctrl+C bitbake multiple times to make it stop :)? Mar 30 09:16:51 stefanOffred123: otherwise, ps aux | grep bitbake and see if there's one running somewhere (without you calling one) Mar 30 09:17:09 I'm using a remote server on LAN that has enought power, but I can't reboot this machine to restart things from scratch. Mar 30 09:17:40 stefanOffred123: that does not answer any of our questions :) Mar 30 09:18:08 ps aux | grep bitbake show no bitbake running Mar 30 09:20:12 stefanOffred123: wait, are you saying bitbake server is on a remote server, different from the one you have Yocto on? (I don't know if this is possible honestly) Mar 30 09:20:52 no, all is on a remote server, I would only explayn why I cant reboot the system Mar 30 09:21:10 stefanOffred123: did any of your SSH connection failed while bitbake was running? Mar 30 09:21:15 fail* Mar 30 09:21:20 I'm doing obviusly ps aux on that server :-) Mar 30 09:21:45 did you by any chance Ctrl+C bitbake multiple times to make it stop :)? -> could be.. Mar 30 09:22:04 stefanOffred123: first, I'd suggest using screen/tmux on the server when running bitbake. Or use mosh (you should, mosh is the future) Mar 30 09:22:11 or a combination of both Mar 30 09:22:14 yes could be that some SSH connection fail Mar 30 09:22:42 qschulz: are you just trying to start a mosh pit? :P Mar 30 09:22:43 @qschulz this is a good hint! Mar 30 09:22:46 stefanOffred123: then bitbake was stopped and it can end in a incoherent state :) Mar 30 09:23:05 stefanOffred123: go into your build dir, rm bitbake.lock maybe even bitbake.sock if there's one Mar 30 09:23:19 I didn't know about mosh, I'll check for it Mar 30 09:23:21 DO NOT do it if there is a bitbake running, because that's supposed to happen Mar 30 09:23:36 (to not have two bitbake instances fro the same project running at the some time) Mar 30 09:24:34 yes, I understood this topic, so maybe the problem is some ssh connection failed bad? Mar 30 09:24:38 and don't ever do anything without a proper screen multiplexer. Mar 30 09:24:39 LetoThe2nd: I'd start a mosh cult for sure. That thing is one of the first thing I install when installing servers/personal Mar 30 09:24:51 computers Mar 30 09:25:13 LetoThe2nd: I use byobu BTW, but because I've no knowledge of any other :) Mar 30 09:25:20 ok, on your live session I have seen you're using tmux, right? Mar 30 09:25:21 stefanOffred123: honestly, could be Mar 30 09:25:35 i'm a tmux guy, yes. Mar 30 09:25:37 stefanOffred123: I've read tmux is for the cool kids Mar 30 09:25:58 ok, and do you nknow how to escape from this state, without rebooting? Mar 30 09:26:18 escaping from a mosh pit? drinking until reboot. Mar 30 09:27:25 hehe, I used in past, but I honestly hate the scroll down that sometimes hang up and clear the stdout.. but I presume that with som tweaking it can be resolved Mar 30 09:28:02 ok so I will drink until the end of days :-D Mar 30 09:28:16 stefanOffred123: what do you mean which state? Mar 30 09:28:45 the main reason why i'm using tmux is that it doesn't, by default, clash with the screen keybindings. so i can use screen as a serial terminal inside a screen session. Mar 30 09:28:59 admittedly, less and less needed these days. Mar 30 09:30:07 hum intresting.. Mar 30 09:30:24 stefanOffred123: 11:23 qschulz| stefanOffred123: go into your build dir, rm bitbake.lock maybe even bitbake.sock if there's one? Mar 30 09:32:35 qschulz yes there is one, I'll remove both Mar 30 09:33:41 Hi my dearest Yocto Community, I refer to https://www.yoctoproject.org/irc/%23yocto.2020-03-27.log.html#t2020-03-27T10:42:19Well, a full wipe of sstate did not solve the problem I have with https://github.com/chrusel/meta-chruselcrypto/blob/master/recipes-crypto/cryptopp/libcryptopp_6.1.0.bb and dunfell release. I've figured out, the evil part is Mar 30 09:33:42 the symlink called libcryptopp.so.6 in the generated lib package. My question today: Why is that symlink dangerous here? Do I really have a issue or does bitbake trow a false exception? Mar 30 09:34:45 qschulz after removing .lock and .sock still the same problem (Retrying server connection (#1)...), and lock and sock has been recreated. Mar 30 09:36:12 qschulz there is also a hashserve.sock, maybe should be removed as well? Mar 30 09:40:13 stefanOffred123: there is probably a process left behind in memory? Mar 30 09:40:51 Chrusel: what is the exact command for the symlink creation that you think is an issue? Mar 30 09:41:15 qschulz: I'd recommend never manually deleting those files. Its almost certainly something left in memory that is the issue Mar 30 09:43:53 RP how can I find it? ps aux|grep bitbake has shown nothing and I can't find strange prosess running.. Mar 30 09:44:24 mckoan: got your mail. will kick off a build in a second and then report Mar 30 09:44:36 RP: I am putting my new threadripper box to good use :) http://git.yoctoproject.org/cgit/cgit.cgi/poky-contrib/log/?h=akanavin/package-version-updates-later Mar 30 09:44:44 mckoan: what poky revision are you on? Mar 30 09:44:44 (patches are held until late april obviously) Mar 30 09:45:28 If can be useful, In other directories other projects can run bitbake correctly on the same server Mar 30 09:47:09 qschulz: ln -sf --> https://github.com/weidai11/cryptopp/blob/master/GNUmakefile-cross fefined in line #33 Mar 30 09:48:01 RP: how do you suggest fixing the double Ctrl+C which results in bitbake not able to run again? (this fixed it, but that's just by empirism that I found those files, so eager to learn how one\s supposed to recover from this state? Mar 30 09:48:35 (which could technically not be what stefanOffred123 is experiencnig, I understand) Mar 30 09:49:54 LetoThe2nd: I forgot to mention that, sorry. It is zeus Mar 30 09:50:12 mckoan: np. ok, i'll check Mar 30 09:52:37 qschulz: kill the process left in memory Mar 30 09:53:04 qschulz: if the process is left in memory and you start more, who knows what will happen to cache files etc Mar 30 09:55:43 RP: same as stefanOffred123 never had a bitbake in the running processes, so how do I know which one's keeping the lock? Can't it just be bitbake no properly cleaning up its files because it was interrupted? Anyway, that didn't seem to fix stefanOffred123 error:/ Mar 30 09:56:42 qschulz: bitbake recovers from the lock files fine. It will only show an issue if there is something left in memory causing a problem Mar 30 09:56:55 lsof on the lockfile might show it? Mar 30 09:58:22 Chrusel: DESTDIR should be set to ${D}? Otherwise, try to print the exact command run to create the symlink (echo in the makefile and check in the WORKDIR/temp/log.do_install). Mar 30 09:58:47 Chrusel: also, ll on the symlink to show what it is set to. Might just be you're symlinking wrongly for Yocto Mar 30 09:58:59 RP: indeed, too late for stefanOffred123 since I made him delete the files :/ Mar 30 09:59:07 RP lsof bitbake.lock show noting Mar 30 09:59:34 RP files has been recreating each time I run bitbake.. Mar 30 10:01:23 stefanOffred123: I'd have a look through "ps ax" and see if you can see the processes that are causing it Mar 30 10:03:12 RP I presume it could be not necessary "bitbake", right? Mar 30 10:18:16 qschulz: DESTDIR=${D} in the install task throws tons of package QA [installed-vs-shipped] issues. Mar 30 10:18:41 then its time to properly sort out FILES_... Mar 30 10:19:47 qschulz: The exact ln command comes in a few minutes Mar 30 10:22:49 ls -als tmp/work/core2-64-miipoky-linux/libcryptopp/6.1.0-r0/git/libcryptopp.so.60 lrwxrwxrwx 1 g041840 t_adp_cc_user 20 Mar 30 12:17 tmp/work/core2-64-miipoky-linux/libcryptopp/6.1.0-r0/git/libcryptopp.so.6 -> libcryptopp.so.6.1.0 Mar 30 10:22:56 Chrusel: I'd say this DESTDIR is actually what you want. Because this means the files are installed in the correct directoy but you're missing paths/files in the appropriate FILES_ variable in your recipe Mar 30 10:23:36 Chrusel: not this one, the one in package or packages-split for the ls -lah ;) Mar 30 10:24:04 qschulz: ok, I'll check that Mar 30 10:24:55 i wonder if we need a 'debug packaging' command Mar 30 10:25:21 'these files were installed to do_install, the packages were split into these groups' etc Mar 30 10:25:49 rburton: isn't it something you could get with oe-pkg-data? Mar 30 10:25:59 -util Mar 30 10:27:12 rburton: but yes, we need tutorials/more tools to help debug :) Mar 30 10:27:29 qschulz: oe-pkgdata is only useful once the package has been built Mar 30 10:27:59 a 'its broken and i don't know why' tool could look at $D to tell you what do_install actually installed, then $PKGD to tell you what got packaged where Mar 30 10:28:07 instead of rooting around temp yourself Mar 30 10:28:34 would actually only be a few lines of bash Mar 30 10:45:45 rburton: good stuff then! Mar 30 10:54:41 qschulz: the libcryptopp package contains the library libcryptopp.so, a symlink to that library called libcryptopp.so.6.1.0 and another symlink with the name libcryptopp.so.6. Could that make problems to bitbake, having two symlinks pointing to the same library? Mar 30 10:56:08 Chrusel: ls -lah of those libs please? Mar 30 10:56:50 but no, that is standard. However, so and so.6 should be part of the -dev package IIUC Mar 30 11:07:59 stefanOffred123: I'd grep for python Mar 30 11:13:31 ok i'm starting to feel crazy... is there anything done between do_install and do_package for binaries? the md5sum is different in image and packages-split.... (and a .debug appeared) Mar 30 11:13:49 (i have update-rc.d and bin_package inherited) Mar 30 11:15:46 qschulz: yes, stripping for a start Mar 30 11:16:19 do_install drops unstripped binaries into image/ and do_package will strip on the way to packages-split Mar 30 11:16:36 symbols go to .debug/ in PN-dbg Mar 30 11:18:20 rburton: alright, thanks, that makes sense. Just to be sure... stripping cannot have any side-effect at runtime right? Mar 30 11:18:36 correct Mar 30 11:19:50 rburton: thanks! Mar 30 11:22:34 qschulz: since do_package fails withWARNING: libcryptopp-6.1.0-r0 do_package: KeyError in ./package/usr/lib/libcryptopp.so.6ERROR: libcryptopp-6.1.0-r0 do_package: Error executing a python function in exec_python_func() autogenerated:(...)Exception: KeyError: 'getpwuid(): uid not found: 1107'I can't ll on packages_split data. Mar 30 11:24:26 qschulz: I have to dig into project's makefile ... Mar 30 11:25:14 Chrusel: something's messing up the owner/group of the files (chown or wrong user when running the script)? Mar 30 11:27:45 qschulz: I run bitbake as normal user on a regular Ubuntu 18.04 host ... I will take a look into the cryptic cryptopp makefile ... Mar 30 11:29:52 qschulz: "git grep chown" shows no hits -.- Mar 30 11:48:17 Hi, is it possible to ship the rootfs.manifest in the image file? Does yocto have a generic command for this, or do i need to write my own? Mar 30 11:51:10 pharaon2502: does it have to be that exact file? Mar 30 11:51:31 RP unfortunately no python istance running.. I fear I must make another poky folder and rerun bitbake in a different folder project. Mar 30 11:51:33 pharaon2502: in that case it is probably more like a postprocess command then. Mar 30 11:51:50 if i don't need the exact file? Mar 30 11:51:59 pharaon2502: then there is https://git.yoctoproject.org/cgit/cgit.cgi/poky/tree/meta/classes/image-buildinfo.bbclass Mar 30 11:52:30 pharaon2502: generally speaking this sounds very much like a xy question anyways. so, what is it that you *ACTUALLY* want to archieve? Mar 30 11:53:16 stefanOffred123: ps ax | grep python and ps ax | grep bitbake don't show anything? That is unusual for it not to start like that :( Mar 30 11:53:36 LetoThe2nd: cca 20 recipe Names + srcrev so it would be the easies to copy the whole manifest, and then just parse my content on the web Mar 30 11:54:01 pharaon2502: that sentence does not parse, sorry. Mar 30 11:54:38 LetoThe2nd: image-buildinfo.bbclass just creates which metas i used, and which is the DISTRO, or am I missing something? Mar 30 11:55:32 pharaon2502: and offers customization through IMAGE_BUILDINFO_VARS Mar 30 11:55:41 RP no, sorry. I presume that some ssh disconnection while running bitbake in the past has lesft something wrong in the build folder. so even if no python or bitbake process is actually running, when I run new bitbake command it's not able to recover Mar 30 11:56:10 pharaon2502: but of course it might not fitting. yet, thats what we have in-tree. if you need somethign else, its up to you. Mar 30 11:57:16 LetoThe2nd: Do you maybe have the list of avaliable variables? Mar 30 11:58:59 pharaon2502: should be anything that is visible in the scope where the class' tasks are being executed. Mar 30 11:59:44 like, whatever parameter which i can display with bitbake -e, that can be added here? Mar 30 12:00:09 pharaon2502: go ahead and find out. Mar 30 12:00:19 LetoThe2nd: Thanks!! Mar 30 12:03:38 pharaon2502: have fun. and additionally, you can always use the class as a blueprint for a custom version. Mar 30 12:28:22 stefanOffred123: what does bitbake-cookerdaemon.log say at the end? Mar 30 12:42:46 RP File "/usr/lib/python3.5/asyncio/unix_events.py", line 264, in create_unix_server raise OSError(errno.EADDRINUSE, msg) from None Mar 30 12:43:34 stefanOffred123: so something is holding the socket it wants. There has to be some process left in memory Mar 30 12:44:37 yes, right. I try with netstat looking for sockets Mar 30 12:48:59 RP strange, I guess I would find a bitbake.sock.. but nope Mar 30 12:54:17 RP thanks a lot, i've found the problem. Was hashserve.sock. I simply remove it. Mar 30 12:54:39 so all the .sock files must be removed from build/ Mar 30 12:54:50 many thanks guys Mar 30 12:56:08 stefanOffred123: simply having the file present would not break the build so I still maintain something is running Mar 30 12:57:55 I have removed .lock and .all sock files. Because I have not found that .sock file in netstat I think that no process was attached to it Mar 30 13:31:25 hello, I am having a weird issue with Linux 5.5.12 when i.MX6UL SDMA firmware load fails during boot even though the file seems to be in the right location Mar 30 13:31:56 I have Yocto usrmerge enabled, so I wonder if default kernel loading follows symlinks Mar 30 13:32:03 heya. Is there a good way to add CXX compile flags to my SDK via the SDK's image recipe? Mar 30 13:32:51 imx-sdma 20ec000.sdma: Direct firmware load for imx/sdma/sdma-imx6q.bin failed with error -2 Mar 30 13:33:09 file exists as /usr/lib/firmware/imx/sdma/sdma-imx6q.bin Mar 30 13:33:43 lib is symlinked to usr Mar 30 13:36:29 lib -> usr/lib Mar 30 13:42:29 mckoan: i think i found it Mar 30 13:42:42 sven^: CXX flags should be passed when you have your SDK .sh sourced Mar 30 13:43:04 eduardas: I'm not sure it actually follows link? you can test for yourself I guess Mar 30 13:43:18 mckoan: try adding KCONFIG_MODE="--alldefconfig" right below the require of linux-yocto.inc. Mar 30 13:56:56 New news from stackoverflow: Output of scripts during bitbake missing Mar 30 13:59:39 qschulz: I think I had this discussion several times already. We are at the point where we are pretty sure we want to enforce it on an SDK basis. Right now we have like 5-10 medium sized projects all developed for the same platform and it's a ton of work to add new warning classes for all targets in all projects. So as a team of developers we decided it would be best to set them in the SDK we all use. Mar 30 14:00:39 also, from a platform maintainer's perspective this will help to ensure a certain level of code quality on all included projects, which is kind of an important point Mar 30 14:05:03 Hi all. I've got a weird problem where the symlink for the image is not being created, but the image is there and the log claims the python method create_symlinks was successful. Mar 30 14:05:13 Anything obvious spring to mind? Mar 30 14:05:21 sven^: ah my bad, misread. Thought you were asking how to pass CXX flags. I do not know if you can pass CXX flags that is different from the one used to compile all SW. But I have no experience in SDK so I'm gonna stop misguiding :) Mar 30 14:05:48 JoeR: actual error? logs? expectations? etc :) Mar 30 14:05:54 lol Mar 30 14:06:18 So it's a class I modified an age ago to make an sdcard image with a pair of partitions to ping pong between Mar 30 14:06:38 This was wayyyy back in dizzy days. So now I'm moving to thud and this little gremlin has appeared. Mar 30 14:07:11 DEBUG: Shell function do_image_ext4 finishedDEBUG: Executing python function create_symlinks Mar 30 14:07:12 NOTE: Creating symlink: /home/joer/distros/thud/fsl-community-bsp/ph069-qx/tmp/work/ph069_qx-poky-linux-gnue Mar 30 14:07:12 abi/core-image-qx-full/1.0-r0/deploy-core-image-qx-full-image-complete/core-image-qx-full-ph069-qx.ext4 -> c Mar 30 14:07:12 ore-image-qx-full-ph069-qx-20200330110946.rootfs.ext4 Mar 30 14:07:12 DEBUG: Python function create_symlinks finished Mar 30 14:07:48 So it looks like it happily did something, and yet the actual sym link is never created. Mar 30 14:08:17 qschulz: I actually want (I think) to modify the values exported by the SDK .sh, like CXXFLAGS. I saw BUILDSDK_CXXFLAGS, but the description sounds wrong to me Mar 30 14:08:49 sven^: that was my understanding yes. I don't know much about SDK unfortunately /me shrugs Mar 30 14:09:15 ok.. I'll keep waiting, maybe someone knows. Thanks :) Mar 30 14:10:16 Hmm no. Maybe scratch that. Maybe I'm being a fool.... Mar 30 14:16:00 qschulz: I think I might know what the problem is Mar 30 14:16:04 https://git.phytec.de/linux-mainline/commit/arch/arm/configs/imx_v6_v7_defconfig?h=v4.14.93-phy&id=885c386014a0c555d416558918c4bb010da745d4 Mar 30 14:16:18 this is not set in mainline defconfig for some reason Mar 30 14:19:02 don't mainline defconfigs assume using the standard linux-firmware package with its default paths? Mar 30 14:27:03 New news from stackoverflow: Bitbake - use different toolchain in a recipe Mar 30 14:55:06 guys, if I'm using branch zeus of a certain layer, but I need a recipe that is only available in the master branch, what do you recon is the best way to use it? I want to avoid "checkout master -- recipe" Mar 30 14:55:54 cjdc3: how about creating your own layer and using cp? add some praying to a deity of your choice that backporting is simple. Mar 30 14:56:19 isnt' that even worse Mar 30 14:56:20 ? Mar 30 14:56:30 I become the owner and maintainer of that recipe Mar 30 14:56:48 I won't be getting future updates automatically Mar 30 14:57:09 New news from stackoverflow: Gstreamer plugins are not getting compiled on yocto 2.7 Mar 30 14:57:40 cjdc3: you become the owner for your state and therefor have a strong incentive to not stick to old versions :) Mar 30 15:00:00 cjdc3: or approach the layer maintainer to do the backports for you, as well as providing the support. might require inserting coinse. Mar 30 15:00:02 cjdc3: if you don't want to maintain recipes, upgrade to master :) Mar 30 15:00:53 qschulz y that's what I'm doing now...but it's quite some hefty work due to all dependencies Mar 30 15:02:32 cjdc3: so you want a magical solution that provides everything you need, in specific revisions you need, on a release you want to define? Mar 30 15:02:44 cjdc3: sounds like work. either do it yourself, or pay somebody. Mar 30 15:03:18 LetoThe2nd doesn't need to be magical, just smart Mar 30 15:03:47 cjdc3: souds like a smart way to spend money. Mar 30 15:04:39 upgrading everything to master is a smart option...just brainstorming on whether better options exist Mar 30 15:12:41 cjdc3: next release out in a month or so (nothing written in stone as always), will be LTS, so it's a pretty damn good reason to upgrade IMHO Mar 30 15:13:41 qschulz y thanks for the info. in that case y, might as well upgrade Mar 30 15:27:16 New news from stackoverflow: phytec phyBOARD iMX-6 performed poorly when running qt5 opengles application from flash instead of sd card (fps halved) Mar 30 15:50:07 "ERROR: linux-qoriq-4.1-r0 do_fetch: Fetcher failure for URL: 'git://git.freescale.com/ppc/sdk/linux.git;nobranch=1'. Unable to fetch URL from any source." I dont find it on git.yoctoproject or git://github.com/NXP/ , any suggestions ? Mar 30 15:51:38 eliazzz: screaming at NXP sounds legit. Mar 30 15:52:31 any NXP people here then :D ? Mar 30 15:57:51 eliazzz: this is yocto channel, try https://community.nxp.com Mar 30 15:57:58 eliazzz: https://github.com/Freescale/meta-freescale/commit/8605319d17b08da79534681dc85b0a23e2e87c73 Mar 30 15:58:04 use code aurora Mar 30 15:59:19 will do, thx Mar 30 15:59:43 qschulz: that's ppc stuff Mar 30 16:02:44 eliazzz: http://git.yoctoproject.org/cgit/cgit.cgi/meta-freescale/tree/recipes-kernel/linux/linux-qoriq_4.19.bb?h=master Mar 30 16:03:33 and still not 4.1 Mar 30 16:03:45 LetoThe2nd: indeed :-D Mar 30 16:04:01 eliazzz: so it probably boils down to scream at yourself for using outdated stuff AND at nxp for not keeping outdated stuff online :) Mar 30 16:04:49 ah :) how did that happen...thanks! that looks excellent Mar 30 16:06:14 mckoan: I meant that they changed all repos from freescale.com to codeaurora and just gave one exmaple :) Mar 30 16:08:05 I was using coriq linux recipe from git://git.yoctoproject.org/meta-fsl-ppc Mar 30 16:08:45 qschulz: ;-) Mar 30 16:08:53 (https://layers.openembedded.org/layerindex/branch/master/layer/meta-fsl-ppc/) Mar 30 16:11:11 eliazzz: "For QorIQ targets, please use meta-freescale layer for pyro and newer releases. " Mar 30 16:26:05 I have a strange-looking shared-state reuse failure. I've rsync'd my sstate from CI server as usual, and "bitbake -n " starts to show it wants to rebuild tons of stuff. Taking the kernel as example, comparing one by one the sigdata extracted from both envs using "-S none", bitbake-diffsigs on do_package.sigdata files show nothing (only on do_build.sigdata does rm_work_all introduce some "noise"). Yet bitbake fetch/.../compile/...:package my kernel. Mar 30 16:26:28 What am I missing here ? Mar 30 16:26:42 (on sumo) Mar 30 16:30:04 Hello, I am trying to get s splash screen working on a raspberrypi3 using meta-raspberrypi layer, but no success :( It should be enabled by default according to some bugreports I found, but I can still see only the console log and at the end, just before the loging propt this message: umount: can't unmount /mnt/.psplash: no such... Mar 30 16:30:54 Tried setting up SPLASH variable and also IMAGE_FEATURES += splash, but nothing worked :( Mar 30 16:31:13 I am on poky zeus, but even tried master Mar 30 17:27:44 New news from stackoverflow: libi2c-dev, libserial-dev how to add to "Yocto"? Mar 30 17:31:11 I was tasked to udpate a sumo-based yocto project, but I'm already stuck building the existing one: when I `bitbake qt5-image -c populate_sdk_ext` and install the resulting SDK, `devtool --help` fails with an error that `/opt/poky/2.5.1/sysroots/x86_64-pokysdk-linux/lib/ld-linux-x86-64.so.2` is not installed. Mar 30 17:32:00 Does that sound familiar? Do I need to install the standard SDK before installling the ext-sdk? and if so: why is the location `/opt/poky` baked into the ext-sdk's binaries? Mar 30 18:27:56 New news from stackoverflow: Yocto: update target with new PHP module Mar 30 18:28:13 problame: yes you need to install the sdk to use it. when you do, you have the opportunity to select an alternate location Mar 30 18:28:35 that's the whole point of using a sdk, anyway :) Mar 30 18:30:08 yann: well that's definetly confusing, as I didn't find any hint about having to install _both_ in the docs. It seemed to me that the ext SDK just bundles more stuff for development. Mar 30 18:30:33 yann: anyways, it should be possible to tell the ext-sdk that the standard SDK is installed in a non-default location, right? Mar 30 18:30:44 oh sorry I mis-read you Mar 30 18:31:14 so no you don't need to install the standard sdk first Mar 30 18:32:00 when you installed the ext-sdk, it asked you about a location, right ? you selected a place different from /opt/poky ? Mar 30 18:32:23 yann: well the binaries installed by the ext-sdk need a linker to be installed at /opt/poky/2.5.1/sysroots/x86_64-pokysdk-linux/lib/ld-linux-x86-64.so.2, otherwise you get `no such file or directory` when trying to start `devtool` Mar 30 18:32:44 the ext-sdk defaults to `~/poky_sdk`, not `/opt/poky` Mar 30 18:32:52 the standard-sdk defaults to `/opt/poky` Mar 30 18:33:04 (and that matches with what I read in the docs) Mar 30 18:33:37 remember: the project is still on sumo (2.5), not the current version of yocto. but honestly: I doubt that these paths changed in the meantime. Mar 30 18:34:44 that may be a bug somewhere. I confess I have not started to use the est-sdk yet (although it's on my todo list). I already started to wonder why those default locations are different, though. Maybe you can try to create a symlink in /opt/poky to get out of this situation until someone more knowledgeable about this can answer. Mar 30 19:07:01 yann: yeah, created that symlink, the python interpeter launched, ran into the next error right away: something about locals Mar 30 19:07:04 *locales Mar 30 19:07:29 perhaps it doesn't help that - because it's 2.5 - i need to run the whole setup in an ubuntu:16.04 docker container Mar 30 19:44:16 RP: sorry for the delay in the install-buildtools second round. I was sick over the weekend :( Mar 30 19:44:38 * moto-timo is tired of sinus infections Mar 30 19:54:04 moto-timo: sorry to hear that, but thanks. I actually did other things over the weekend so its not a problem Mar 30 19:54:19 moto-timo: I did just mention one small possible tweak, not sure if we want to do that or not though Mar 30 19:54:35 moto-timo: just glad to have the patches really! Mar 30 20:04:25 RP: are you thinking the check_gcc_version should be using host_gcc_version ? or do you want host_gcc_version to be updated/fail if < 5.0? Mar 30 20:05:38 RP: or possibly both ? Mar 30 20:09:45 moto-timo: change host_gcc_version to a new function which returns a version and then have check_gcc_version use the new function Mar 30 20:09:58 RP: roger Mar 30 20:10:04 moto-timo: just wondering Mar 30 20:10:15 moto-timo: My mind was jogged by a recent patch to fix it Mar 30 20:11:12 RP: serendipity. I wasn't thinking of that function Mar 30 21:18:16 JPEW: if you have a moment to ponder something, https://autobuilder.yoctoproject.org/typhoon/#/builders/52/builds/1728 Mar 30 21:18:39 JPEW: if not I'll ponder tomorrow Mar 30 22:02:56 Hi there. I cannot build gstreamer1.0-1.16, using zeus. It seems that a recipe for gstreamer1.0-plugins-base (bbappend) has been added, but is just refering to mainline gstreamer, so imx-specific stuff is not included. Specificly, I'm lacking gstimxcommon.h, which is in the Freescale gstreamer repo, but not in mainline. So basicly, gstreamer1.0 for IMX is not buildable in zeus. Anyone know anything? :-) Mar 30 22:14:32 kspr: meta-freescale has some complicated workarounds in it, it's easy to run afoul of them. What are you trying to accomplish, actually? Based on some recent experiments I've done here, the -imx recipes for gstreamer in meta-freescale build fine if you're using the default configuration with vivante Mar 30 22:16:49 smurray, thanks, right now I'm just trying to build an image (I'm upgrading from an old Yocto..). I'm using meta-freescale, and building the default version of gstreamer, which is 1.16. It fails when building imx-gst1.0-plugin, as it cannot find imxgstcommon.h Mar 30 22:19:04 kspr: when you say "building the default version of gstreamer", do you mean you've set PREFERRED_VERSION for the gstreamer recipes to pick up the non -imx versions? Mar 30 22:19:32 smurray, nope, PREFERRED_VERSION for gstreamer is not set. Mar 30 22:20:49 smurray, maybe I need to set it to 1.16.imx? Mar 30 22:22:46 kspr: ah, my bad, I'm looking at master, not zeus branch. meta-freescale zeus branch has 1.14-imx. If you need imx-gst1.0-plugin, you'll likely need to set PREFERRED_VERSION_gstreamer1.0 = "1.14.imx" (and gstreamer1.0-plugins-base, gstreamer1.0-plugins-good, etc. that you need) Mar 30 22:23:30 kspr: the alternative would be switching to master or backporting the 1.16.imx recipes it has Mar 30 22:23:42 kspr: or dropping imx-gst1.0-plugin if you don't need it Mar 30 22:26:10 smurray, I tried master, it failed the same way. It also seemed that gstreamer1.0-plugins-base actually has 1.16, in the MM_04_05_01+ branches. The issue seems to be that the bbappend for 1.16 in meta-freescale is not refering to that repo, but to the mainline which is not compatible. So for me, it seems that the recipe for gstreamer1.0-plugins-base 1.16 in meta-freescale is just not really finished. But, it's a good idea trying 1.14.imx, I would guess that Mar 30 22:26:10 it is more stable :-) Mar 30 22:27:30 kspr: the bbappend in meta-freescale, i.e: recipes-multimedia/gstreamer/gstreamer1.0-plugins-base_1.16.%.bbappend ? Mar 30 22:27:47 smurray, yes Mar 30 22:27:55 kspr: it only tweaks PACKAGECONFIG_GL, not sure why that would be a big problem? Mar 30 22:29:04 smurray, it is a problem because that the bbappend inherits from poky/meta/recipes-multimedia/gstreamer/gstreamer1.0-plugins-base_1.16.bb, which is fetching source from the mainline gstreamer repo. And that repo does not have the imx-specific files (eg. imxgstcommon.h). Mar 30 22:30:06 kspr: sure, but you probably shouldn't expect imx-gst-plugin to build Mar 30 22:30:26 smurray, why not? :-) Mar 30 22:30:57 kspr: it's in the name, "imx", the error you're seeing clearly shows it needs NXP's hacked up version Mar 30 22:31:38 smurray, so what you say is that the issue is that there's no gstreamer1.0-plugins-base_1.16.imx.bb? Mar 30 22:32:24 smurray, I mean, gst-imx-plugins 1.16 has been included in zeus, and is dependend on the plugins-base package.. Mar 30 22:32:27 kspr: as I said above, it's present on meta-freescale master branch Mar 30 22:32:40 smurray, yes, I know :-) Mar 30 22:32:52 kspr: dude, it's not my layer, complain to NXP Mar 30 22:33:14 smurray, sorry, I'm not complaing at all. Just wanna make sure that I understand :-) Mar 30 22:34:00 smurray, but I'll try 1.14.imx and see if that works. Thanks a lot for your help! :-) Mar 30 22:34:12 kspr: np, good luck Mar 30 22:34:18 Thanks **** ENDING LOGGING AT Tue Mar 31 02:59:57 2020