**** BEGIN LOGGING AT Fri Feb 09 03:00:01 2018 Feb 09 08:02:40 Hi Feb 09 08:02:44 is there any option to let a site.conf flow into an OE build automatically from the build host? Feb 09 08:02:58 I would have expected that pre-setting BBPATH to a hosts conf file location would solve that, but it seems to be overwritten unconditionally and there is only a 'bypass' for the TOPDIR lookup. Feb 09 14:40:04 ejoerns: that is a harder problem to solve than you'd think, I think there is an open bug about it Feb 09 15:07:51 In the past I've handled that via adjusting the templateconf, i.e. inject ~/.oe into the BBPATH via the template bblayers.conf. But there's a big prblem with loading files external to your build directory — it's easy to forget they're being used later on, and then you wonder where the hell something is coming from Feb 09 15:16:53 RP: too bad, but I already assumed handling to be semi-trivial when scrolling through the code :( Feb 09 15:17:26 kergoth, yes, thus I really wanted to use it only for site-specific settings! ;) Feb 09 15:25:49 ejoerns: I would quite like to make it work, I just ran into roadblocks last time I tried Feb 09 15:26:50 I'd like to see a note printed about site.conf being loaded, as a special case, either always or when it's loaded from a location outside of $TOPDIR and $COREBASE/.., if we go that direction Feb 09 15:27:19 kergoth: yes, as you say its all too easy to do something interesting Feb 09 15:27:51 i had it loading it by default at mentor, and multiple engineers ended up forgetting about it at some point :) Feb 09 15:28:00 minor thing, but worth keeping in mind Feb 09 15:28:24 kergoth: I can see it happening all too easily... Feb 09 15:41:12 kergoth +1 Feb 09 15:42:01 I really need site.conf or something similar to make my workflow sane. I change local.conf to exactly the same foo too often and I hate repetition Feb 09 15:42:35 but that foo is my own configuration, not the team's, so local.conf.sample doesn't scale Feb 09 15:44:04 I actually end up in the reverse of forgetting where settings came from, I assume they were set and they weren't because I forgot to setup local.conf Feb 09 15:44:13 * moto-timo facepalm Feb 09 15:44:22 ha Feb 09 15:44:26 oops Feb 09 15:44:54 I'll admit I am not spending enough time to sharpen my tools Feb 09 15:57:46 moto-timo: site.conf in a meta-moto-timo +++ Feb 09 15:58:07 meta-ross has a site.conf which sets a few useful global assignments Feb 09 15:58:26 hello Feb 09 15:58:29 rburton: of course! sigh. blind spot Feb 09 15:58:54 not a bad idea, need to remember to add it to your bblayers every time of course, unless you add it to the template, but not a bad idea Feb 09 15:59:11 but running bblayers is easy and I do it anyway Feb 09 15:59:23 I mean bitbake-layers add-layer of course Feb 09 16:00:15 i am trying to run bitbake in an alpine container, but it fails on the 8th task "OSError: [Errno 32] Broken pipe" could this be cgroup limit somewhere ? or somethign else ? Feb 09 16:00:39 in worker.serve() Feb 09 16:13:00 found the problem... Feb 09 18:38:40 wakko: +1 for alpine container, we have never yet build OE on non-glibc systems Feb 09 18:39:50 wakko: could you send a comment to github.com/crops/yocto-dockerfiles ? or create an issue asking for alpine based container Feb 09 18:40:07 wakko: we have been planning for that as officially supported but just didn't do it yet Feb 09 18:40:35 it might be interesting, especially with pseudo Feb 09 18:41:09 rewitt has said multiple times that he thinks alpine is what the containers should be based on for BKM Feb 09 18:41:48 pseudo is sticky bit, few years ago I use to taunt Electric Cloud on these kind of tools now we have one Feb 09 18:41:59 it might also be interesting to do it with AUH and run all the default machines, including qemux86_musl which is the biggest miss for most devs Feb 09 18:42:17 rewitt might say anything Feb 09 18:42:22 lol Feb 09 18:42:50 khem: have you had a chance to look at the state of the bitbake world failures on qemux86? most of those are musl problems Feb 09 18:43:13 khem: and I just don't have time to look at them all. I look at the one's I know about or care about Feb 09 18:43:21 what is your view of world ? poky or enlarged view Feb 09 18:43:37 in this case literally what JaMa/armpit are running Feb 09 18:44:01 ok, yes I look at those often and fix as I find that 25th hr Feb 09 18:44:31 understood. just wanted to make sure you were aware there are musl problems being uncovered Feb 09 18:44:39 the one's left are actually not musl's problems, and they need some ninja level understanding of those codebases Feb 09 18:44:47 we need some more people looking at that Feb 09 18:44:55 some of these codebases are older than me Feb 09 18:45:12 oh. I didn't catch the ninja problems... did not look close enough Feb 09 18:45:16 yeah. ancient Feb 09 18:45:32 yes, I have a musl world build that I run for rpi3 whenever machine is bored Feb 09 18:46:01 I just want to spread the work around to more devs. AUH running qemux86_musl helps Feb 09 18:46:04 oe-core world builds are clean for musl Feb 09 18:46:09 right Feb 09 18:46:13 and for all arches Feb 09 18:46:16 right Feb 09 18:46:32 on AB we only test x86 for musl but thats ok Feb 09 18:46:58 baby steps :) Feb 09 18:47:36 khem: as the topic is libcs, out of curiosity, did you ever try bionic and OE together? Feb 09 18:47:43 this baby is almost 2 years old now Feb 09 18:48:15 and yet some are not aware it was born Feb 09 18:48:17 lol Feb 09 18:48:18 tbr: whats benefit of bionic Feb 09 18:48:45 I think I should market more code less Feb 09 18:48:47 khem: if one would want to recycle some binary only stuff that came with a device Feb 09 18:49:14 most prominently graphics drivers Feb 09 18:49:20 The Khem musl and clang Road Show, coming soon to a town near you Feb 09 18:49:30 tbr: bionic is too much androidware so if we were to build android'ish distro it might be interesting Feb 09 18:50:07 tbr: there are now shims available for musl to run glibc based precompiled binaries which is sort of cool Feb 09 18:50:42 moto-timo: yeah, I should get my lederhosen and wide brimmed hat Feb 09 18:51:06 khem: yeah, I agree. It's just something that has been going around in my mind. E.g. the Hikey960 has zero proper graphics support for glibc, only the android builds. Feb 09 18:51:11 khem: and grow a manly beard Feb 09 18:51:25 long hipster beard Feb 09 18:51:28 of course there is libhybris, but that's kinda meh Feb 09 18:51:56 * moto-timo thinks musl and clang should be cartoon characters Feb 09 18:52:26 one of these days someone must design an opensource GPU and put things to rest Feb 09 18:53:03 * moto-timo hopes the RISC-V people will do that Feb 09 18:53:15 and i want a pony Feb 09 18:55:05 moto-timo: yes, days of opensource are beginning in h/w realm so interesting things will happen Feb 09 19:32:37 khem: My question to you is with musl being available, is there much point in using (or extending bionic to be POSIX) anymore? Feb 09 19:32:47 Other than just "'cuz"? Feb 09 19:33:24 bionic seems to solve the reason it was created: to make the Android system C code work. Feb 09 19:33:41 musl has a grander scheme. To be an alternative to glibc. Feb 09 19:35:13 behanw: yes, if you have andoid things that you want to work on non android systems Feb 09 19:35:20 fairly narrow area Feb 09 19:35:28 Fair enough. Feb 09 19:35:37 But indeed a niche answer. ;) Feb 09 19:35:49 khem: musl has impressed the hell out of me. Feb 09 19:36:00 but I would rather then implement a wrapper on top of musl ldso Feb 09 19:36:21 yes musl is KISS Feb 09 19:36:50 no meltdown or spectre if we follow KISS :) Feb 09 19:38:36 see https://github.com/adelielinux/gcompat Feb 09 19:38:54 for bionic we need to replace 'g' with 'b' Feb 09 19:39:00 how hard would that be :) Feb 09 19:46:07 so you are interested in alpine container as a host for building oe distros ? Feb 09 19:46:27 i though it was a stupid idea as oe can/could produce its own host Feb 09 19:47:13 then i might look further into my problem (still not fixed) the ld from musl is giving me a segfault atm Feb 09 19:49:03 python[28055]: segfault at 564300000000 ip 00007f745847e6c7 sp 00007ffe24a49038 error 4 in ld-musl-x86_64.so.1[7f745842e000+89000] Feb 09 19:53:59 (1/3) Upgrading musl (1.1.18-r2 -> 1.1.18-r3) **** ENDING LOGGING AT Sat Feb 10 03:00:01 2018