**** BEGIN LOGGING AT Tue Feb 04 03:00:52 2020 Feb 04 03:55:53 New news from stackoverflow: Include precompiled zImage in yocto project Feb 04 08:02:19 hooray, yet another downvote-battle on SO. Feb 04 08:20:41 and, is it just me or is linked ins newsfeed completely unusable? Feb 04 08:35:24 LetoThe2nd: we need to fix the bot Feb 04 08:35:34 LetoThe2nd: well, tbf you're not really answering the question :) Feb 04 08:35:36 LetoThe2nd: sorry, I think I'm confused Feb 04 08:37:09 LetoThe2nd: I promise I didn't downvote :) Feb 04 09:03:44 qschulz: oh i'm pretty sure that the poster of the first answer downvoted me. his answer consisted of a link to the 1.2 quickstart, the usual image of the openembedded data flow and a suggestion to do IMAGE_INSTALL_append Feb 04 09:04:30 qschulz: so i commented him saying "sorry, this is just nowhere near the question, and the IMAGE_INSTALL is technically wrong in that context. please don't take it personally." Feb 04 09:04:46 obviously another one who can't take critique Feb 04 09:05:29 qschulz: feel free to add/improve/... :) Feb 04 09:05:42 RP: confused? Feb 04 09:07:05 The real problem with SO is to me that there "I just want to" questions galore, and if you try to help you're being punished for it. Feb 04 09:07:43 LetoThe2nd: AB issues. They explain with so little context that you can only answer wrongly most of the time :/ Feb 04 09:08:05 * LetoThe2nd will probably just letSO be then. Feb 04 09:08:26 and file it under "failed community service attempts" Feb 04 09:08:32 Instead of fixing the actual bug. But that's actually the issues with most of programmers (I still have those moments and I still work with others doing the same) Feb 04 09:09:01 LetoThe2nd: Well... Don't take downvotes personnally. You did your best in helping people, if they don't want to hear abuot it. Meh, whatever Feb 04 09:09:20 qschulz: sure, but on irc you can have a discussion. and on a mailing list, you can't just let your own rubbish vanish, making all others look stupid. Feb 04 09:09:49 qschulz: i don't take the downvote personally, i take personally the system basically punishes those who want to help. Feb 04 09:10:09 LetoThe2nd: reminds me of: https://twitter.com/srostedt/status/1132009918444318721?s=20 Feb 04 09:10:55 qschulz: exactly. Feb 04 09:11:01 that fits so much. Feb 04 09:11:21 LetoThe2nd: yeah, I've a hard time with SO. Sometimes answers are completely not related to the questions. Like "how do I do this in vim" and an answer is "one would do it this way in emacs"... Feb 04 09:11:48 i don't ask there anyways. Feb 04 09:11:57 LetoThe2nd: I haven't understood your "and on mailing list" Feb 04 09:12:15 LetoThe2nd: I asked one question on it years ago. Upvoted a lot, never got an answer :D Feb 04 09:12:38 i've worked long and hard (believe it or not) to know folks i can ask *personally*. Feb 04 09:13:18 qschulz: on a mailing list, if you give a totally unrelated answer, you can't hide it afterwards. its in the archive, you can't just delete it like your crap output on SO. Feb 04 09:13:53 LetoThe2nd: aaaaaah I see. Yup, agreed. Feb 04 09:15:10 :) Feb 04 09:16:19 LetoThe2nd: people don't want to be held responsible anyway... And it's hard for people to take the risk of looking stupid. Ego thing. I don't like it when I'm being told but at least I learn :) Feb 04 09:16:46 yup. Feb 04 09:19:23 probably another reason why i am having such a hard time to recruit other streamers. Feb 04 09:25:31 LetoThe2nd: I would never have presented anything at a conference if I hadn't been forced by my previous employer. And... I actually like it (not the moment before, nor the first 2min, sheer panic). It's very nice. I learned so much preparing the thing. So I guess you've to get into people's home, put them in front of a camera and force them to stream to have streamers :) Feb 04 09:25:40 Hello. If /etc/network/interfaces is missing. Does it depend on menuconfig settings? If it does, which one? Feb 04 09:25:42 -to stream Feb 04 09:25:59 nacknick: /etc/network/interfaces is a debianoid thing Feb 04 09:26:26 And if I want to set static IP to my NXP's board, how should I do it? Feb 04 09:26:39 qschulz: i've actually thrown chocolate at the audience in my presentations many times. Feb 04 09:26:53 nacknick: depends on the network manager you want to use Feb 04 09:26:53 nacknick: if you want interfaces etc, install ifupdown Feb 04 09:27:31 (with ifupdown counting as a network manager for the sake of this conversation, rburton++) Feb 04 09:27:59 LetoThe2nd: I don't care which one. Just want to set static IP simply and permanently. Feb 04 09:28:06 rburton: Thanks. I try Feb 04 09:28:10 I'll* Feb 04 09:28:10 cool kids use connman or network-manager and have mDNS do its magi Feb 04 09:28:13 magic Feb 04 09:28:23 all my images i can ssh into using the name bob.local Feb 04 09:28:31 * LetoThe2nd is not cool. uses networkd Feb 04 09:28:56 networkd works well for the use cases it works well in Feb 04 09:28:56 * qschulz uses ip commands and udhcpc Feb 04 09:29:16 RP: if my usecases don't fit, i make them do so :P Feb 04 09:29:19 rburton: ^^^^ Feb 04 09:29:22 (oops) Feb 04 09:29:30 ha Feb 04 09:30:26 rburton: Do you use local DNS? Feb 04 09:30:37 nacknick: no, mDNS is awesome Feb 04 09:32:24 rburton: Does it require install mdns on the endpoints? Or some other settings in the network Feb 04 09:33:07 I mean, if I want to use it in my LAN for my boards, should I rebuild all their images with mdns? Feb 04 09:34:23 nacknick: i put avahi into all my images for convenience. its certainly nice being able to just let DHCP do its thing and mDNS do the resolution. Feb 04 09:35:02 you need to have the daemon running on the machines that you want to be able to find, and the machines that do the searching need the mdns resolver. Feb 04 09:35:19 all linux distros have both parts Feb 04 09:35:33 So just add avahi package to my boards? Feb 04 09:35:38 but it is a bit more high tech than just a static ip in interfaces, so your choice Feb 04 09:35:46 avahi-daemon Feb 04 09:36:05 https://layers.openembedded.org/layerindex/branch/master/recipes/?q=avahi Feb 04 09:37:19 rburton: actually an interesting approach. need to keep this on my look-at-list. Feb 04 09:37:39 LetoThe2nd: surprised systemd doesn't do the job of avahi yet Feb 04 09:37:47 considering lennart wrote avahi Feb 04 09:37:54 rburton: hehe Feb 04 09:38:38 rburton: i'm highly pragmatic when it comes to systemd. it fits my bill better than any other solution available at the moment, so i use it. Feb 04 09:39:44 LetoThe2nd: i stand corrected https://wiki.archlinux.org/index.php/Systemd-resolved#mDNS Feb 04 09:40:03 just need to use the resolved Feb 04 09:40:07 hrhrhr Feb 04 09:40:27 thats what i meant. if i now decide to use it, probably just need to flip some config bit somewhere. Feb 04 09:40:34 -> fits use case. Feb 04 09:40:56 LetoThe2nd: if you test it, report back :) Feb 04 09:41:47 rburton: shall do so. probably 2021 or 2022 depending on my workload. Feb 04 09:41:56 that soon? Feb 04 09:42:32 yup. im totally underworked. Feb 04 09:43:27 LetoThe2nd: do you have a bugzilla account? :) Feb 04 09:43:41 and believe it or not, i'm jsut right at this moment writing some parts of code that do feed networkd. Feb 04 09:43:44 RP: ayup. Feb 04 09:56:51 rburton: I added `avahi` to `IMAGE_INSTALL_apped` and got the following error when building *core-image-minimal-1.0-r0 do_rootfs: Could not invoke dnf.* - Do you have any idea why? Feb 04 09:57:01 New news from stackoverflow: Yocto: adding "tree" to image, where are output binaries Feb 04 10:00:08 RP: want me to file a bug that people are generally stupid? Feb 04 10:00:53 LetoThe2nd: I'll pass on that thanks ;-) Feb 04 10:01:06 RP: hum why the question then? Feb 04 10:02:26 ah. now i got it. Feb 04 10:02:29 LetoThe2nd: you were saying you were underworked ;-) Feb 04 10:02:40 LetoThe2nd: we could fix this :) Feb 04 10:03:15 RP: the problem is that this would not increase my workload, only my frustration. because i would have to pass on each and every bug for my incompetence. Feb 04 10:04:02 LetoThe2nd: that is simply not true ;-) Feb 04 10:04:49 hrhrhr Feb 04 10:04:53 i wish it was so. Feb 04 10:05:50 anyways. objections if i make the next livecoding session a little less YOEP-centric, but more like "why you should use a proper build system for you c/c++ stuff, and here's an example"? Feb 04 10:06:56 i feel like this is a topic many newcomers struglle with, given the amount of c-file-manually-compile-recipes. Feb 04 10:12:25 nacknick: as i said, install avahi-daemon Feb 04 10:13:59 nacknick: or if you use systemd already then just tell it to do the mdns stuff, https://wiki.archlinux.org/index.php/Systemd-resolved#mDNS Feb 04 10:19:19 is it possible to have .ipk with opkg install interactive input? Feb 04 10:21:13 doubt it very much Feb 04 10:21:40 at least, you can't need it because you can't do interactive in a rootfs session Feb 04 10:23:14 so you might be able to check if there's input during the postinst Feb 04 10:39:37 rburton: thx, so the only solution would be to use .deb Feb 04 10:40:27 mckoan: whats the bigger picture of the use case? Feb 04 10:51:35 LetoThe2nd: I know it's weird but I want to install a binary package which has a config file (one of many possible). I need to define which one at install/reconfig time Feb 04 10:52:24 mckoan: and it is absolutely required that all the variants are installed under the exact same name? Feb 04 10:55:41 mckoan: line of thought is, you could have this thing provide a virtual/something. for your x configuration options, just create x packages something-1, soemthing-2,... which conflict and each provide virtual/something Feb 04 10:55:53 so you can use the package name as selector. Feb 04 10:56:24 nobody says you cant have a number of packages providing almost identical payloads. :) Feb 04 10:57:56 LetoThe2nd: I see, thanks Feb 04 10:58:05 mckoan: just an idea. Feb 04 10:58:16 LetoThe2nd: ;-) Feb 04 10:58:27 mckoan: neat side effect, this works interactively and non-interactively. Feb 04 11:16:59 where can I discuss device trees on freenode? :P Feb 04 11:40:38 I'm puzzled by a kernel compilation issue (rockchip 4.4 kernel, using warrior toolchain): the kernel builds fine from bitbake, but using a yocto sdk built from the same source, the build fails when attempting to link arch/arm64/kernel/vdso/vdso.so.dbg, with ".../aarch64-shadow-linux/8.3.0/real-ld: fatal error: -shared and -static are incompatible" Feb 04 11:40:59 that does not seem to inspire google much Feb 04 13:57:49 New news from stackoverflow: bitbake populate_sdk for image fails with error Feb 04 14:03:10 hm, that happens only when using gold - the same call with ld.bfd shows no problem... Feb 04 14:11:46 damned, kernel-arch.bbclass is too clever :) Feb 04 14:27:56 New news from stackoverflow: Yocto: adding "tree" to image, where are output binaries? Feb 04 14:28:47 RP: ipk vs deb should be a subTest(), I'm not sure why it wouldn't get both Feb 04 14:31:54 JPEW: once self.fail() is called, game over Feb 04 14:32:07 JPEW: tracked down the ipk problems btw Feb 04 14:32:33 RP: Oh right I've been carrying around an incomplete patch to fix subTest :( Feb 04 14:33:36 Does anyone have a build of tcp-wrappers around? If so could they see what permissions ${S}/*.c have? Are they r--r--r-- or rw-r--r--? Feb 04 14:34:04 I think quilt is changing permissions in some cases to apply patches but not sure i can prove it :/ Feb 04 14:37:05 RP: almost all are r--r--r-- except: Feb 04 14:37:31 ./weak_symbols.c Feb 04 14:37:35 weak_symbols.c is probably the only one? Feb 04 14:37:39 correct Feb 04 14:38:11 JPEW: we end up with more files with rw in some builds. Just don't know how. I'm suspicious that this recipe has a lot of patches Feb 04 14:38:30 RP: Ya, that would make sense Feb 04 14:39:26 what year was YP started? ( <- gsoc application) 2010? 2011? https://en.wikipedia.org/wiki/Yocto_Project Feb 04 14:39:40 tlwoerner: 2010 Feb 04 14:39:47 thanks :-) Feb 04 14:40:19 0.9 was released Oct 2010 iirc Feb 04 14:40:46 I wonder if patch behaves differently Feb 04 15:42:57 halstead: can you please add me to the gsoc mailing list? will the yocto gsoc mailing list show up on https://www.yoctoproject.org/community/mailing-lists/ Feb 04 15:43:19 RP: I have send pulls which are also in kraj/pu on poky contrib, one of them is musl fix so you want to pick it up Feb 04 15:44:28 RP: btw. that bison-native race is impacting quite a bit of builds Feb 04 15:44:48 one way I found was to depend on external libunistrings but that does not seem optimum Feb 04 15:50:26 khem: should we revert bison for now? Feb 04 15:50:36 khem: are those upgrades in master-next? Feb 04 15:50:49 RP: I will look into fixing it today Feb 04 15:51:12 no I sent pulls earlier last week and git-send emails Feb 04 15:51:30 but all are in kraj/pu branch on contrib too Feb 04 15:52:24 khem: ok, I'm really struggling to do all the review and fix these reproducibility issues and keep track of everything else :( Feb 04 15:53:03 RP: I have changed my workflow a bit, where kraj/pu will be more collection of other patches too so you have some help Feb 04 15:53:45 RP: I could throw it as AB perhaps once in a day or 2 or so Feb 04 15:54:18 khem: I've queued musl and netbase in next locally Feb 04 15:54:51 RP: OK the webkit will take long to recompile but perhaps take it in for nightly Feb 04 15:55:18 YPTM: Armin is on Feb 04 15:58:13 New news from stackoverflow: boost and btrfs-tools failed with Yocto build Feb 04 16:00:37 YPTM: Joshua Watt here Feb 04 16:00:39 YPTM: Randy is in da house. Feb 04 16:00:55 JPEW20: is this the new improved Joshua 2.0? Feb 04 16:01:05 YPTM Michael joined. Feb 04 16:01:36 tlwoerner: lol, no. I just signed in on my laptop from the web interface. Can't install hexchat easily on Windows 10 :( Feb 04 16:01:43 lol Feb 04 16:02:10 tlwoerner, We haven't made a GSoC list. Is there a discussion about it? Feb 04 16:02:11 if anyone wants to join the monthly technical call: https://www.yoctoproject.org/public-virtual-meetings/ Feb 04 16:02:26 halstead: i just added myself to the list! Feb 04 16:02:36 https://lists.yoctoproject.org/g/gsoc/ Feb 04 16:03:00 vmeson: I was just about to ask for the link Feb 04 16:03:13 aehs29_: :) Feb 04 16:04:03 tlwoerner, Thanks. I found it. Feb 04 16:07:34 I'm trying to build 'qt5-opengles2-test' from meta-qt5 and I'm getting: "error: 'glEnableVertexAttribArray' was not declared in this scope", while trying to build. Anyone encountered this before? Feb 04 16:07:34 " Feb 04 16:13:23 can I tell devtool to ignore certain tags? Feb 04 16:14:09 someone, at some point, tagged something with a numeric value which is grokked as the newest revision, even though it is quite old. Feb 04 16:15:28 tlwoerner, Added to https://www.yoctoproject.org/community/mailing-lists/ Feb 04 16:15:38 halstead: thanks! :_) Feb 04 16:15:40 :-) Feb 04 16:15:52 * tlwoerner nose is out of join! Feb 04 16:15:57 s/join/joint Feb 04 16:16:30 fullstop: I fail to understand your question Feb 04 16:17:16 qschulz: https://pastebin.com/raw/fx1emFjY Feb 04 16:17:49 There is a tag which is something like dlog553 Feb 04 16:17:58 the latest version is actually v1.29.35 Feb 04 16:18:13 but devtool update wants to move it to 553 because 553 > 1.29.35 Feb 04 16:18:56 I can go and delete the tags in git but that seems kind of harsh Feb 04 16:21:44 fullstop: 553.1.29.35 is your only hope Feb 04 16:21:56 fullstop: (i can't answer your question :) ) Feb 04 16:22:00 :-) Feb 04 16:23:46 fullstop: devtool has a way to say "use this branch" for the "devtool add" command, but i don't think there's an "ignore this tag/branch" option for "devtool upgrade" Feb 04 16:23:59 sounds like a good gsoc project.... ;-) Feb 04 16:24:24 magic is great as long as it works. Feb 04 16:24:39 and, yes, it sounds like a good gsoc project Feb 04 16:24:42 https://wiki.yoctoproject.org/wiki/Stable_Release_and_LTS Feb 04 16:31:24 fyi: https://wiki.yoctoproject.org/wiki/Yocto_3.1_Schedule Feb 04 16:34:13 gotta drop, have another call Feb 04 16:34:15 thanks guys Feb 04 16:38:14 yptm: https://wiki.yoctoproject.org/wiki/GSoC Feb 04 16:40:30 YPTM: minutes here: https://docs.google.com/document/d/1ly8nyhO14kDNnFcW2QskANXW3ZT7QwKC5wWVDg9dDH4 Feb 04 16:44:30 armpit, Shall I go ahead and add the docs list and repo as a patchwork project? Feb 04 16:46:39 one more devtool question.. devtool upgrade recipe-name fails for a recipe which has files in a machine specific directory with "IsADirectoryError: [Errno 21] Is a directory:", where the directory printed is files/machine-name/. Feb 04 16:47:52 https://pastebin.com/raw/Wnhj4wbq Feb 04 16:48:43 I have file://init/*.conf in my SRC_URI and different machines pull in different files. Feb 04 16:54:24 this is now an after lunch problem. :-) Feb 04 16:58:57 RP: gentle reminder about testing again with logrotate patch v3 Feb 04 17:35:37 tgamblin: thanks, added locally Feb 04 18:28:45 New news from stackoverflow: Yocto package generation without modifying distributions on targets (ARM & x86) Feb 04 19:49:08 hello. I m trying to boot my yocto hddimg generated from yocto build on virt-manager. It is core-image-minimal image. I m not able to find any configuration in virt-manager, which will boot this image. Is there any doc that I can use to boot the hddimg Feb 04 19:57:10 basically I want to boot my hddimg (or bzImage and ext4 rootfs) on qemu Feb 04 20:00:58 centos, its always the blummin centos workers :( Feb 04 21:05:38 * zeddii surived fosdem Feb 04 21:05:48 khem: around? Feb 04 21:05:52 now some Zzzzzzzz’s before I can do anything useful. Feb 04 21:23:29 can i override a variable for only one function? such as EXTRA_OEMAKE ? Feb 04 21:25:49 mischief1: What do you mean 'only one function' ? Feb 04 21:26:16 mischief1: you can override everything, I think... Feb 04 21:27:15 roussinm: i only want to change a variable for the duration of one function/task. Feb 04 21:30:15 mischief1: task overrides do exist Feb 04 21:30:37 EXTRA_OEMAKE_task-XXX = "y" iirc Feb 04 21:32:12 RP: So I could something like : EXTRA_OEMAKE_compile_append ? Feb 04 21:32:33 _append_task-compile Feb 04 21:32:49 oh I see! Feb 04 21:33:15 hi all o/ Feb 04 21:33:37 I've been having a bit of trouble getting a kernel module to build out-of-tree... was wondering if the symptoms sound familiar to anyone... Feb 04 21:34:04 I've tweaked the hello-mod example to demonstrate the issue - http://git.yoctoproject.org/cgit/cgit.cgi/poky/tree/meta-skeleton/recipes-kernel/hello-mod/files Feb 04 21:34:05 i think i have to override the whole task anyway Feb 04 21:34:21 I'm seeing: hello: Unknown symbol __class_create (err 0) Feb 04 21:34:26 and other symbols Feb 04 21:34:28 i'm trying to fix our recipe for backports which has a hardcoded invocation to /sbin/depmod.. so stupid Feb 04 21:39:19 more info: https://gist.github.com/attie/e47055eaca1f28074e2858e9fe8225ab Feb 04 21:39:29 if anyone else has the ability to give it a go, that would be much appreciated Feb 04 21:39:38 or any thoughts on what's going on here... Feb 04 22:10:45 hello Feb 04 22:12:29 given that I have root access to a device, that I can run it in fastboot mode and flash an image, that I have the image (one *.ubi, one boot.img and one recovery.img), can I somehow reverse engineer my system and create a yocto distro that would work on it? Feb 04 22:14:02 Nothing illegal planned: I would like a yocto image for my device, the manufacturer is fine with me playing around with it, but doesn't want to go through the trouble of giving me information about it. In other words, I have an NDA and I can do whatever I want to make a prototype, but it doesn't have to cost them anything xD Feb 04 22:15:39 I am pretty new to Yocto, so I don't really realize what is needed to start from scratch. I would guess that I should try to get the kernel config from my device, so that I can compile it from Yocto, but then would that and the fact that it is an armv7 be enough for me to create a minimal image? Feb 04 22:24:10 JPEW: http://git.yoctoproject.org/cgit.cgi/poky-contrib/commit/?h=rpurdie/t222&id=d5be254b04ff3708981b4b6467d29223f13f4efa Feb 04 22:24:19 * RP can't decide how horrible that is Feb 04 22:26:05 palate: it's defenitley a very good start Feb 04 22:26:28 if you're able to rebuld the kernel and maybe reuse the uboot it should be booting already Feb 04 22:26:40 from there it should be a piece of cake ;) Feb 04 22:27:24 did you find the config information on the original root filesystem? Feb 04 22:27:38 and maybe start with the same kernel version Feb 04 22:28:03 it might be necessary to find a vendor specific kernel fork somewhere Feb 04 22:28:28 usually vendors provide their own layers containing BSP meta-data Feb 04 22:30:43 hello. I m trying to boot my yocto hddimg generated from yocto build on virt-manager. It is core-image-minimal image. I m not able to find any configuration in virt-manager, which will boot this image. Is there any doc that I can use to boot the hddimg Feb 04 22:30:52 basically I want to boot my hddimg (or bzImage and ext4 rootfs) on qemu Feb 04 22:32:57 creich: I'm not sure how to get that information :/. So the uboot would be this boot.img, right? I have no clue, but my guess would be that boot.img is the bootloader (possibly uboot), recovery.img is the one that gets me in fastboot mode, and the *.ubi image is the actual image. And you're saying that I could just try to create my own *.ubi without caring about the boot and recovery. Correct? Feb 04 22:34:07 creich: first I'd like to have a way to know if I'm right about the partitioning (i.e. "recovery.img" would bring the fastboot system). But I have no clue how to check that Feb 04 22:41:25 Well actually I know that with fastboot I flash a "system", a "boot" and a "recovery" partition. I'm not clear yet what the recovery would be Feb 04 22:46:05 Is setting TCLIBC="musl" in DISTRO.conf sufficient to enable musl on my distro? Feb 04 22:46:30 LetoThe2nd: good work on the live coding videos Feb 04 22:46:37 kriive: yes Feb 04 22:46:50 its as simple as that Feb 04 22:47:02 kriive: Where are those live coding videos? Feb 04 22:47:16 roussinm: https://www.youtube.com/watch?v=o-8g0TPVVGg Feb 04 22:47:52 You can find more on the Yocto Projects's youtube channel here: https://www.youtube.com/channel/UC2_mG3h-AxxK2oz08j2bz2g Feb 04 22:48:03 kriive: thanks! Feb 04 23:04:41 palate: first steps could be to boot the default system and tr to gather as much information as possible Feb 04 23:04:53 or at least mount the image if possible Feb 04 23:05:14 do you have access to the device running the default sw? Feb 04 23:05:17 in any way? Feb 04 23:05:35 creich: yes, through adb shell, I am root there Feb 04 23:05:54 ah ok, so it's an android running there? Feb 04 23:06:32 i don't know much about android configurations, but having a root shell is a very good starting point Feb 04 23:06:55 creich: nope, linux image. I don't know why they have fastboot and adb, nor where that comes from Feb 04 23:07:12 su you can get your kernel version and some extra info by 'uname -a' Feb 04 23:07:31 ok, maybe even better. but does not matter too much Feb 04 23:07:41 in any case it's a linux base :) Feb 04 23:09:31 creich: sorry to ask again, I was wondering if you saw my query about out-of-tree modules and unknown symbol errors...? Feb 04 23:10:13 attieg: might have seen, but forgotten about it. sorry ;) Feb 04 23:10:17 i can have a look Feb 04 23:10:18 ha, no problem Feb 04 23:10:33 would be much appreciated, thank you! Feb 04 23:11:02 happy to share the info again if you don't have it, it's not too far back ^^ Feb 04 23:11:10 fundamentally, this: https://gist.github.com/attie/e47055eaca1f28074e2858e9fe8225ab Feb 04 23:11:33 yep found it in the logs Feb 04 23:11:54 building in-tree is fine, zero problems... but I'd like to keep it out-of-tree if possible Feb 04 23:12:05 creich: right: Linux mpp 3.18.31-ga339981ec18d-... #1 SMP PREEMPT Wed Feb 13 09:57:43 CET 2019 armv7l GNU/Linux Feb 04 23:12:25 creich: so it's a pretty old kernel 3.18.31, right? Feb 04 23:12:52 palate: yeah... that's pretty old... the bit is a commit ID (some changes after the 3.18.31 tag) Feb 04 23:13:16 might be worth looking out for it, though I wouldn't be super hopeful about finding anything :-( Feb 04 23:14:23 attieg: hmm, I'll try Feb 04 23:14:28 palate: what's the SOC? a rokchip part? Feb 04 23:15:22 attieg: never tried out-of-tree build, but we can still try to find out ;) Feb 04 23:15:36 so you've made an in-tree build already that worked you say? Feb 04 23:15:38 attieg: the image says apq8009 in its name. From this page it seems like the only one listed is a snapdragon 212: https://en.wikipedia.org/wiki/Comparison_of_single-board_computers Feb 04 23:16:14 creich: correct - if I place my files (or the modified hello files) into the tree, and build them there, it works... Feb 04 23:16:15 have you checked dmesg as stated within the error message you got during modprobe? Feb 04 23:16:45 the snippet in the gist is all the output I get (that's a serial console, and running dmesg gives no additional info) Feb 04 23:17:18 attieg: so if that was https://www.intrinsyc.com/computing-platforms/open-q-212a-system-on-module-som/, it says that it has "OE Yocto support" Feb 04 23:17:20 I did notice, that from modinfo, there are a couple of differences - for example my *big* driver has dependancy on the CAN module, which isn't listed when building out-of-tree Feb 04 23:17:59 ... but the dependancy is listed when building in-tree Feb 04 23:18:16 attieg: https://stackoverflow.com/questions/29578931/unknown-symbol-class-create-err-0 Feb 04 23:18:22 have you seen this? Feb 04 23:18:35 creich: yeah, license is all good Feb 04 23:19:26 i also checked the license in the recipe / *.bb just incase that was being considered (ew)... but i've copied the hello-mod as exactly as I can Feb 04 23:19:41 (hence the tweaks to a simple hello-mod example... even that doesn't work out) Feb 04 23:19:48 hmm, ok. just a lucky shoot ^^ Feb 04 23:19:56 ok Feb 04 23:20:09 minimal examples are always good :) Feb 04 23:20:28 palate: ah, neat! is that just using the same SOC? or is that actually your target? Feb 04 23:20:59 creich: yeah :) if nothing else, it allows someone else with a handy yocto & board to give it a go too... Feb 04 23:21:29 I'm very suspicious of my environment at the moment - the mfg bsp, etc... Feb 04 23:21:59 as i said, i never tried out-of-tree build. but i could give it a try Feb 04 23:22:10 attieg: no clue: I have the device and I am trying to figure out what it is to make a yocto image for it Feb 04 23:22:28 let me check, but i should have a complete workspace with a minimal running bsp and shell ready to insert some stuff Feb 04 23:22:31 if you have the time and stuff to hand, that would be very kind... but no pressure Feb 04 23:22:40 RP: w.r.t. bison issue, do you know if AB slaves have libunistring installed on host ? Feb 04 23:23:04 RP: I can reliably reproduce it more on systems without libunistring Feb 04 23:24:09 palate: is the thing in the photo on that page you linked the thing on your desk? Feb 04 23:24:56 khem: how do I test? Feb 04 23:25:26 attieg: well, I would need to open the device. It's some kind of gamepad. Should I try to open it? Feb 04 23:25:38 ssh into the node and check if /usr/include/uniwidth.h exists Feb 04 23:26:10 attieg: creich: btw I have a /proc/config.gz, so I guess I should be able to extract kernel config from that, shouldn't I? Feb 04 23:26:11 khem: pokybuild@ubuntu1804-ty:~$ ls /usr/include/uniwidth.h -la Feb 04 23:26:11 ls: cannot access '/usr/include/uniwidth.h': No such file or director Feb 04 23:26:33 palate: yep, that's a very good start Feb 04 23:26:49 if you could find the specific commit of the sources somewhere you're almost done ;) Feb 04 23:27:19 khem: a quick scan on debian and centos workers says no such header Feb 04 23:27:24 creich: ohoh, looking for that then Feb 04 23:27:34 attieg: ok, have to run a freh build which can take a few minutes, but i am preparing a ws :) Feb 04 23:27:48 RP:ok so it seems the builders dont have it ? Feb 04 23:28:07 s/freh/fresh/ Feb 04 23:28:23 RP: my failures are on ubuntu 18.04 and debian 10 Feb 04 23:28:47 which is all machines I have access to, it works ok on archlinux box Feb 04 23:28:54 but that does not count Feb 04 23:29:27 khem: looks like the workers do not Feb 04 23:29:49 creich: thanks! Feb 04 23:31:43 yw :) Feb 04 23:36:46 RP: it seems it never showed up on autobuilders ? Feb 04 23:36:56 khem: no Feb 04 23:37:06 khem: which is why I'm a bit puzzled by it Feb 04 23:37:21 yeah I see it it on different machines Feb 04 23:38:41 intermittent though Feb 04 23:49:24 creich: still cloning the linux kernel to try to find that commit hash... Feb 04 23:49:32 creich: is it important that I get exactly that commit? Feb 04 23:49:46 i.e. 3.18.31-ga339981ec18d-509563-ga9381f5e68f7 Feb 04 23:50:34 palate: most probably there are some specific changes necessary to get the hardware running (drivers) Feb 04 23:50:38 it really depends Feb 04 23:51:22 creich: i'm going to have to leave for today... i'll check in tomorrow - thanks for your help Feb 04 23:51:32 if you're sure about the hardware and can find some documentation and thus the drivers within the mainline kernel it should be ok to use a close version Feb 04 23:51:45 attieg: ok, just give me a ping Feb 04 23:52:04 i let the workspace run through, so we can go forward tomorrow :) Feb 04 23:52:36 will do, many thanks! Feb 04 23:52:40 creich: Like I could try to extract the kernel options from /proc/config.gz, take 3.18.something, create an *.ubi image and try to flash it as "system"? Feb 04 23:54:14 palate: sure, you can try that first Feb 04 23:54:29 maybe you don't even have to build an image yet Feb 04 23:55:01 creich: what do you mean? Feb 04 23:55:12 i don't know about fastboot, but in u-boot for example, you can interrupt the boot process and make it load some kernel and device tree from tftp Feb 04 23:55:41 with that you could at least check if you kernel will boot toll the point it should panic due to a missing rootfs Feb 04 23:56:11 but you could avoid the whole image building process which can take quite some time and bring it's own sources for mistakes Feb 04 23:56:39 so minimal viable should be a simple boot of the kernel itself i'd say Feb 04 23:57:29 right Feb 04 23:57:59 so I need to learn how to build the kernel only with yocto, and then how it works in fastboot. Feb 04 23:58:25 From some fastboot documentation, it seemed like the "boot" partition is flashed with boot.img and contains the bootloader and the kernel (if that could be) Feb 04 23:58:35 for that you can build the kernel without yocto if you like Feb 04 23:58:42 ah, ok. Feb 04 23:58:50 than you might need to build an image Feb 04 23:59:17 but as i said, i don*t know fastboot... check if there is any option to boot via network or smth easy Feb 04 23:59:35 creich: but yocto will give me exactly one image, right? And I can set it to be *.ubi. Or could I get a *.ubi with the system and a boot.img? Feb 04 23:59:37 or if you feel confident in building an image, you surely can do that too Feb 04 23:59:43 creich: right Feb 04 23:59:56 New news from stackoverflow: WebOS Build - Compilation faile - grub-efi_2.02 failed Feb 05 00:00:09 yocto *might* give you *some* image ;) Feb 05 00:00:16 :D Feb 05 00:00:31 if it is the right one will depend on your setting and the specific hardware Feb 05 00:00:42 that's why i'd like to avoid as much as possible Feb 05 00:01:13 if you could verify the exact board and maybe find some documentation, that could be very helpful i guess Feb 05 00:01:33 here is some very good course material you should have a look at Feb 05 00:01:52 i just used it myself recently to get some insights in the whole process Feb 05 00:02:06 https://bootlin.com/doc/training/embedded-linux/ Feb 05 00:02:40 you also could checkout the live coding videos from LetoThe2nd for further insight in the whole yocto stuff Feb 05 00:03:12 creich: awesome, thanks a lot! Feb 05 00:03:20 yw :) Feb 05 00:03:28 i'll leave for tonight as well :) Feb 05 00:03:37 good luck and maybe cya later Feb 05 00:05:16 good night :) Feb 05 00:08:07 RP: I think I see the issue LIBUNISTRING_UNIWIDTH_H is detected as uniwidth.h but actually rule says it should be lib/uniwidth.h Feb 05 00:11:56 khem: ah, that could explain it Feb 05 00:34:20 RP: there is 3.5.1 released last week I will give it a shot Feb 05 00:55:36 I've started a GSoC ideas page: https://wiki.yoctoproject.org/wiki/GSoC/Ideas Feb 05 00:56:48 tlwoerner, should that also be advertised on OE wiki ? Feb 05 00:57:33 * tlwoerner doesn't want to find himself in the middle of an "OE vs YP" discussion! Feb 05 00:58:36 i'm happy for any help OE would like to lend Feb 05 01:00:00 armpit: do you mean wiki.openembedded.net ? (lol) Feb 05 02:52:21 tlwoerner, yes wiki.openembedded only to spread the word **** ENDING LOGGING AT Wed Feb 05 03:00:08 2020