**** BEGIN LOGGING AT Mon Nov 30 03:00:00 2020 Nov 30 07:46:31 good morning Nov 30 07:48:14 yo dudX & GM mckoan Nov 30 07:56:17 hi guys, what are the differences between these three? https://git.yoctoproject.org/cgit/cgit.cgi/poky/tree/oe-init-build-env https://github.com/Freescale/fsl-community-bsp-base/blob/master/setup-environment and https://github.com/toradex/meta-fsl-bsp-release/blob/toradex_morty-4.9.51-8qm_beta2-bring_up/imx/tools/fsl-setup-release.sh ? Nov 30 07:56:23 why so many different scripts for the same setup? Nov 30 09:19:25 * qschulz waves at people :) Nov 30 09:24:18 * LetoThe2nd rewaves to qschulz with an anchor: https://youtu.be/th4Czv1j3F8 Nov 30 09:43:25 Can I submit patches to the mailinglist as attachment? Fighting with email clients... Nov 30 09:48:03 ernstp: no Nov 30 09:48:15 we need them in the mail to be able to review them inline Nov 30 09:48:23 in the mail = as body of the mail Nov 30 09:49:01 though some people have contrib branches in poky-contrib so maybe there's something you can do here, it's documented somewhere in the manual IIRC Nov 30 09:49:22 ernstp: why not just git-send-email? no need for fighting. Nov 30 09:49:56 LetoThe2nd: git-send-email is not very nice with either Office365 or Gmail :-) Nov 30 09:50:29 a lot of people use it with gmail, not that hard Nov 30 09:51:55 mcfrisk: you have to temporarily enable the less secure authentication method... Nov 30 09:53:31 or set up a dedicated account. a very common practise. Nov 30 09:53:48 (jump through the hoops once, be done) Nov 30 09:55:09 (a GitLab/Gerrit like system would be neat :-) ) Nov 30 09:56:10 she said jehova! Nov 30 10:06:12 git-send-email and gmail is fine, just set an application password Nov 30 10:06:18 * rburton has been using it for years Nov 30 10:07:08 https://support.google.com/accounts/answer/185833?hl=en Nov 30 10:07:16 if you're not using two step you can just auth like normal Nov 30 10:07:41 ernstp: ^^^ Nov 30 10:09:39 rburton: thanks! Nov 30 10:22:48 ernstp: https://www.openembedded.org/wiki/How_to_submit_a_patch_to_OpenEmbedded Nov 30 10:28:45 Hello everyone, I have another one to discuss :) https://stackoverflow.com/questions/65071824/how-to-set-the-wifi-registry-in-recipe Nov 30 10:29:09 Do you know a more elegant way of doing that other than through a systemd? Nov 30 10:36:26 'morning there ! Nov 30 10:38:53 howdy PaowZ Nov 30 10:46:41 you're fine over there ?? ready to tackle December and its load of upcoming surprises ? Nov 30 10:47:54 will the jester turn into a pixie or cute goblin ? :) Nov 30 10:48:40 no real surprises in sight here. besides a disturbing lack of glühwein. Nov 30 10:52:56 https://gist.github.com/fenrig/56eaf220191356819a8f33f77f1ea67b Nov 30 10:53:18 in my bbapend I'm creating a dir and putting a symlink, but in my resulting image its not included (put some comments) Nov 30 10:53:42 ..reminds me my mother last time I visited her, she wanted to please me with some Christmas Beer.. the cheap sticker glued onto the bottle shall had warned me.. and my tong just confirmed.. xD Nov 30 10:54:17 there are good christmas drinks around here. i can confirm. Nov 30 10:58:53 you guys have nothing to learn from, in terms of beer, no doubt about that.. Nov 30 11:06:00 would anyone know why a ${libdir}/engines/*.so is not included in the resulting image rootfs? Nov 30 11:06:13 I tried checking logs but found nothing of interest there Nov 30 11:09:19 fenrig: oe-pkgdata-util find-path /usr/lib/engines/*.so Nov 30 11:09:25 in which package is your symlink? Nov 30 11:09:31 Probably because you're not packaging it, or sothing your .so gets packaged in the -dev package because it isn't version.ed Nov 30 11:09:44 libssle-client Nov 30 11:09:48 its an openssl engine Nov 30 11:09:54 oh wow, that line got messed up. Nov 30 11:10:04 fenrig: have you checked with the line I gave you? Nov 30 11:10:04 how can I move it to not get packaged in dev Nov 30 11:10:12 Probably because you're not packaging it, or your .so gets packaged in the -dev package because it isn't versioned. Nov 30 11:10:13 no not yet :) Nov 30 11:10:21 Check the docs for the SOLIBSDEV variable Nov 30 11:11:01 ERROR: Unable to find any package producing path /usr/lib/engines/libssle-client.so Nov 30 11:11:13 same with * btw Nov 30 11:12:06 fenrig: are you even sure your bbappend is applied? Nov 30 11:12:19 bitbake-layers show-appends Nov 30 11:12:22 yeah I can see it reflected in the image dir in the recipe Nov 30 11:12:46 oe-pkgdata-util list-pkg-files -p [recipe] Nov 30 11:12:59 great to see what went where, maybe another package took the file first? Nov 30 11:14:40 i can see the bbappend :D Nov 30 11:14:45 but I put a typo :D Nov 30 11:14:48 its engines Nov 30 11:15:12 ah no i put it right :o Nov 30 11:15:41 the bbappend is matched so its applied Nov 30 11:15:55 maybe I have to run bitbake again? Nov 30 11:17:49 check the do_install log just in case lnr is failing but not reporting an error Nov 30 11:18:52 COMPONENT is set correctly, right Nov 30 11:22:35 yeah Nov 30 11:23:45 i you look at my gist, then you will see that the lnr was done but just that when putting in the rootfs that engines dir and the subfile are not being "imported Nov 30 11:23:56 "imported/installed" Nov 30 11:27:01 fenrig: do you happen to have a INSANE_SKIP somehwere in your recipe that would silence a file not being installed in any apckage? Nov 30 11:27:36 fenrig: have you cleaned your tmpdir before running oe-pkgdata-util? Nov 30 11:27:45 you need the workdirs/tmpdir for it to be successful IIRC Nov 30 11:29:24 AHA now I have it, it's indeed in a dev file Nov 30 11:29:31 libssle-client-dev: /usr/lib/engines/libssle-client.so Nov 30 11:29:59 putting FILES_SOLIBSDEV to "" Nov 30 11:30:29 is not enough I guess Nov 30 11:31:36 aha it seems it puts unversioned symbolic links in PN-dev still Nov 30 11:31:53 qschulz:wrong, oe-pkgdata doesn't care about work dir at all, that's the point Nov 30 11:32:14 it works on the pkgdata, which exists as long as the package was built in some way (from sstate counts) Nov 30 11:32:20 rburton: for some reason, when i clean my workdir it does not work anymore (on thud though so don't know) Nov 30 11:37:07 qschulz: depends how you clean. the data is in tmp/pkgdata. Nov 30 11:37:22 so if you delete tmp then you'll need to bitbake the right recipes to extract the pkgdata from sstate Nov 30 11:44:43 does anyone know how to exclude symlinks from being put in the dev package? Nov 30 11:47:53 ok I can override FILES_${PN}-dev but I wonder if thats a good idea Nov 30 11:51:46 Setting SOLIBSDEV = "" usually does that. Nov 30 11:52:10 omg :D ok thx I was playing with solibs Nov 30 11:52:50 However, you may have to explicitly add ${libdir}/engines to FILES_${PN}. I'm not sure whether it's matched by the default. Nov 30 11:53:11 will do :D Nov 30 11:53:21 I probably need that insane skip on dev-so Nov 30 11:54:11 Nope, leave that off. insane skip is almost never needed. Nov 30 11:59:27 SOLIBSDEV doesnt seem to do it :/ Nov 30 12:06:31 manually putting FILES_${PN}-dev does the trick :o and insane_skip on dev-so offcourse Nov 30 12:06:38 otherwise I would run in qa issue Nov 30 12:34:59 or you know... just version it :) Nov 30 12:47:30 qschulz: böring! Nov 30 13:59:18 is there a simple way to build a single package using a different gcc version? the only way I can think of right now is to use a multiconfig build but this seems pretty overkill to me. Nov 30 14:02:31 JaBen: I think meta-clang may show how to switch compilers Nov 30 14:03:43 @RP: thanks, will take a look Nov 30 14:10:26 in a scenario where you manage your kernel configuration through devtool menuconfig, what is the best practice to finalize the .config modifications back to the recipe? Nov 30 14:10:46 I expected to have a sort of detool defconfig Nov 30 14:10:59 s/detool/devtool Nov 30 14:25:42 mckoan: bitbake -c savedefconfig virtual/kernel Nov 30 14:26:35 then you take the defconfig from the WORKDIR of the kernel and you can add it to your recipe Nov 30 14:27:37 qschulz: yes of course, I expected something à la devtool ;-) Nov 30 14:31:37 mckoan: not aware of it though I also don't use devtool that much. Something to add to devtool maybe :) ? Nov 30 15:24:05 just to clarify, there is no simple way to build and include multiple versions of a package, right? Nov 30 15:25:17 JaBen: you can with different recipes Nov 30 15:26:24 which are named differently of course Nov 30 15:28:15 @qschulz: but gcc_8.2.bb and gcc_9.3.bb would not work? or you it suffice to set the PN in one of them to e.g. gcc-8? Nov 30 15:29:59 JaBen: no it won't because you can only have one version of a recipe Nov 30 15:29:59 I am trying to use ADI's u-boot which doesn't build with gcc 9. so now I'm trying to somehow create two separate GCC build toolchains... fun stuff Nov 30 15:30:23 JaBen: try to find upstream patches for gcc9, shouldn't be too much of a hassle Nov 30 15:30:43 it's probably much easier than trying to get two GCCs Nov 30 15:30:51 (from my PoV obviously :) ) Nov 30 15:32:23 hm... Nov 30 15:38:26 Hello everyone, I have another one to discuss :) https://stackoverflow.com/questions/65071824/how-to-set-the-wifi-registry-in-recipe Nov 30 15:38:59 whoops... that was an accident Nov 30 15:57:43 rburton: hi, i'm looking at the autobuilder and only find this error in multilib, https://autobuilder.yoctoproject.org/typhoon/#/builders/44/builds/2789 Nov 30 16:03:02 thats not very useful is it Nov 30 16:07:18 rburton: it is a network error on mips32 failed to ping... may be some infra error Nov 30 16:18:21 alimon: I think there are other open bugs like that, its some kind of race :( Nov 30 16:18:29 alimon: in other cases if we rerun it just works :/ Nov 30 17:34:42 RP: ok, did you take care of master-next failures?, if not i can take a look to identify the patches that are failing, i just subscribed to yocto-builds and trying to identify new errors Nov 30 20:32:35 hi, i'm trying to override a function provided by a .bbclass in a single recipe, i redefined the function completely in a .bbappend, and it shows as i would like in `bitbake -e` but the base implementation is still being used ... is there some other way to change the function for a single recipe? (even inspect.getsource couldn't find the source, was my second pass attempt) Nov 30 20:59:32 alimon: Sorry, was in meetings. I did the minimum needed to keep the builds moving. I replied about the pseudo ones to peter Nov 30 21:00:07 alimon: I don't think there are any broken patches left in the queue right now, its only the intermittent failures that broke in some builds Nov 30 21:23:26 RP: ok good Nov 30 21:47:12 /usr/bin/python when meta-python2 is used is a symlink to python2 and I'm wondering if anyone can point me to the best way to change this to python3. "oe-pkgdata-util find-path /usr/bin/python" tells me this symlink comes from python-core, but I'm unsure where to go from here. Nov 30 21:48:01 (unfortunately still need python2, working on removing it, but want the default system python to be 3) Nov 30 22:04:08 Androo: you'll have to bbappend the recipes to do that Nov 30 22:04:18 just delete the symlink in python and create it in py3 Nov 30 22:13:33 rburton: I was thinking perhaps it was an install order issue? Like the python2 package is installed after python3, overwriting the symlink, and I could just change the ordering so that the reverse happens? Nov 30 22:33:41 Androo: I don't believe the python3 recipe installs the link at all **** ENDING LOGGING AT Tue Dec 01 03:00:01 2020