**** BEGIN LOGGING AT Mon May 03 02:59:56 2021 May 03 07:42:07 yo dudX May 03 07:48:02 good morning May 03 08:50:21 Hi. Does anyone know about Bitbake command for getting all packages with their version? May 03 08:52:53 NVM I think I've found it May 03 10:33:49 Hi... I am looking for advice on how to properly use https://github.com/grz0zrg/fbg "library" in my custom meta-layer. May 03 10:33:50 Basically this projects provides couple of .c & .h files I need during compilation of my custom demo program. May 03 10:34:38 I would create fbg_git.bb recipe... and fbg_demo_1.0.0.bb May 03 10:36:00 Basically fbg_demo would need some of the source files from fgb_git.bb during build time... May 03 10:36:06 any examples and ideas? May 03 12:02:59 hello, how can I easily make npm available in my custom recipe step? May 03 12:03:29 I do not want to use npm.bbclass because it sets nodejs as RDEPENDS May 03 12:04:09 but I need the command "npm run build" to build a custom Vue.js project May 03 12:08:21 eduardas: take npm.bbclass as a starting point and beat it into shape. May 03 12:13:00 eduardas: DEPENDS = "nodejs-native" May 03 12:13:37 but carefully read the npm.bbclass anyway, because npm does not cross-compile by default May 03 12:13:56 works for some usecases (using npm to build static website (JS+HTML) May 03 12:14:36 depends on your definition of works, because it is very easy to accidentially bypass do_fetc May 03 12:14:39 do_fetch May 03 12:17:02 qschulz: this is a static website, so I believe it should work for my case May 03 12:17:31 qschulz: so thanks for the advice May 03 12:18:06 that's what we did end up doing recently so I guess it's good enough :) May 03 12:18:40 like i said - it depends on your requirements. May 03 12:18:41 LetoThe2nd: yes, having a modified npm.bbclass was something I was considering, but I am not familiar with the whole npm ecosystem. So I really want a simpler solution, to be honest. May 03 12:20:10 I really expected this to be more obvious in Yocto nowadays, to be honest May 03 12:20:25 eduardas: thats all legit. what i jsut want to make properly understood is that many, many js based build processes do random server fetches. so unless yo take great care to avoid/cache those, you will probably end up in a situation where you can reproduce the system, but not the js/html application in a couple of months time. May 03 12:20:39 i.e. I kinda expected this usecase to be already covered by npm.bbclass May 03 12:21:03 nope, npm.bbclass targets another usecase. May 03 12:22:44 doesn't npm have methods to define fixed versions of dependencies? May 03 12:22:52 LetoThe2nd: yup, probably will not produce reproducible builds May 03 12:23:46 eduardas: AIUI, npm.bbclass is meant to put a specific npm package onto the target. May 03 12:24:52 eduardas: not exactly for packaging your own application. May 03 12:27:49 LetoThe2nd: thanks for pointing that out. Lots of room for improvement, I see... May 03 12:28:16 LetoThe2nd: just as a sanity check... as far as I understand npm.bbclass uses the host's version of npm? May 03 12:28:23 improvement is complicated and painful when it comes to packaging JS May 03 12:28:45 i.e. bitbake does not build an independent version of npm for itself? May 03 12:29:18 it does build the version it can find in the layers. what else shouldiit build? May 03 12:29:53 I could not actually find it in the nodejs work directory May 03 12:30:06 npm itself I mean May 03 12:30:42 if you have DEPENDS += "nodejs" it ends up in the sysroot-path of the depending recipe May 03 12:36:22 LetoThe2nd: should I define a static website package as allarch? May 03 13:10:58 quit May 03 13:51:00 eduardas: i would say so, yes. May 03 14:48:57 khem: Ya, we use meta-mingw May 03 14:49:05 khem: Unfortunately May 03 14:51:40 JPEW: I've broken it with master as the pkgconfig fix only works with gcc 11 :/ May 03 14:52:17 RP: Ah... hmm May 03 14:53:59 JPEW: I'm hoping to merge gcc 11 soon which would solve the problem. There are other fixes for it in master-next from khem too May 03 14:54:20 RP: Ya. Thanks for pushing those, I was going to this morning but saw they were already there :) May 03 14:54:44 RP: The glib <-> pkgconfig dependency loop is a bit annoying May 03 14:54:45 RP: break stuff! break stuff! :) May 03 14:58:52 LetoThe2nd: I have managed to fail to replace the battery in my Garmin and cracked the screen in the process. I also tried to service a seatpost and that isn't working any more. Doing really well at it today :/ May 03 15:03:05 RP: awesome! perfect way to start out a week, if you ask me. May 03 15:08:14 LetoThe2nd: I'm sulking as I have a better track record than this. I'm missing some obvious step with the seatpost. The Garmin is now working ok having seen computer power, wouldn't power up out on the bike. May 03 15:10:16 RP: hey, my day started out by without prior planning having to roll out MS teams to all of our accounting/sourcing. May 03 15:20:18 Somehow I just removed zeddii's kernel patches from my mailbox without trace too :/ May 03 15:20:29 LetoThe2nd: ouch :/ May 03 15:24:25 khem: I merged gcc11, thanks! May 03 15:24:35 JPEW: ok to merge master-next on meta-mingw? May 03 15:24:53 RP: Yep May 03 15:25:16 khem: Thanks for the MinGW 7 -> 8 upgrade. It was on my TODO list May 03 15:25:31 JPEW: done. May 03 15:25:42 * JPEW wishes AUH worked for nativesdk May 03 15:29:48 JPEW: we don't have that many nativesdk only recipes do we? May 03 15:30:01 JPEW: thanks for the logging fix btw. We need to rip apart the logging stuff :/ May 03 15:30:39 RP: Ya, we don't have much nativesdk only, which is why the AUH (devtool really) probably isn't very high priority :) May 03 15:30:52 Yes, logging needs another good look May 03 15:35:50 RP, subliminal work-avoidance? Nuke all mails from zeddii ? May 03 15:40:13 * RP locks uninative 3.1 in. Still scares me that I'm fixing elf sections given how little I know about them May 03 15:40:25 I was recently irritated by the handling of defconfig in linux-yocto style kernels. May 03 15:40:25 The task scripts (run.do_kernel_configme) and a note in meta-freescale then pointed me to the KCONFIG_MODE variable. This selects, whether a provided out of tree defconfig is applied with `allnoconfig` or `alldefconfig` flag. `allnoconfig` is the default, while the `alldefconfig` handling would be closer to what I expect from dealing with the kernel in other contexts. May 03 15:40:25 Short: I think this should be documented and wanted to know if kernel-dev/common:changing the configuration (https://www.yoctoproject.org/docs/latest/mega-manual/mega-manual.html#changing-the-configuration) would be the right spot. I'd prepare a patch for it. May 03 15:46:37 dwagenk: a patch would be welcome May 03 15:46:40 michaelo: ^^^ May 03 15:59:36 dwagenk, RP: and I will be happy to review it! May 03 15:59:52 Good to know about "alldefconfig", I didn't know this one May 03 16:04:27 dwagenk: and yes, kernel-dev/common looks like a good location May 03 16:06:42 OK, I'll be sending a patch May 03 16:23:33 anyone seeing "qemu: uncaught target signal 4 (Illegal instruction) - core dumped" from update_font_cache postinst after rebuilding with gcc-11? May 03 16:25:04 Looking at the core-image-tiny-initramfs recipe and I am wondering why it is using the same COMPATIBLE_HOST as initramfs-live-install May 03 16:25:29 For example building for arm works fine... May 03 16:49:32 LetoThe2nd: seatpost now works again :) May 03 17:58:06 Is anyone using Linux and reading signal of a rotary encoder? Kernel somehow supports it, and on RaspberryPi there is a device tree overlay, but I'm looking for advice on how to read it. May 03 17:59:12 manuel1985: it probably sends KEY_EVENTs? so probably more or less what evtest is doing but in yuor own code May 03 18:01:36 qschulz: The thing is in my whole linux career I never came across events. May 03 18:02:19 Does it fire a callback of my process? May 03 18:04:30 I want to have a process run every 10ms and read the decoded value of the rotary encoder. Can I have that process have an own cpu core to himself? May 03 18:05:42 I thought of the process running every 10ms be a gstreamer element. May 03 18:17:42 manuel1985: IIRC the input subsystem sends events/signals whatever it's called to userspace to notify something changed May 03 18:18:02 obviously, most of the time, it's not polling, otherwise responsiveness would be horrendous May 03 18:20:37 well I was wrong :) May 03 18:21:09 it seems you just continuously read /dev/input/eventX and I assume this is blocking until there's something that is sent there May 03 18:21:59 I would assume that it's a character device, meaning if you don't read when it's sending something, it's lost (think UART) May 03 18:22:27 but I might be wrong, maybe the rotatory device isn't exposed as an input device, in which case, well something else May 03 18:51:11 hello all. writing a recipy for OpenTTD, but i have no idea how to compile the host tools needed for that as its on recipy and make it availible May 03 18:51:11 its currently a cmake build one with a flag to make tools only and one to make the dedicated server May 03 18:51:12 Any point to where to start to read about doing any of this would be much appriciated or an example. May 03 18:51:13 The tools made will be needed to compile Openttd binary May 03 18:51:53 (ofc i want any similar situation:)  not exact soluiton :)) May 03 18:56:39 Mrsan: may I ask what kind of embedded device you want yocto on, with OpenTTD :P. May 03 18:56:55 malinus_ im making a image for Rpi4 and for x64 May 03 18:57:17 primarly, but im making a official meta for it May 03 18:57:18 Mrsan: why are you using YOCTO. I'm just corious, I don't have anyhting to help/add :) May 03 18:57:39 malinus_ why not:)  (i dont want the overhead of the other oses :)) May 03 18:58:14 Fair enough, it just sounded exotic enough that I figured there might be some real-life application for it :P May 03 18:58:24 (real-life, meaning product or something) May 03 18:59:00 nah no product:)  im doing this as a training exersice mostly May 03 18:59:30 malinus_: Learn By Doing: ) May 03 18:59:51 sure thing, sounds fun :) May 03 19:01:29 so what im looking for really is where ucompile custom tools to use duiring compile of another recipy and how to get them connected :) May 03 19:07:48 Mrsan: You may need to write a -native recipe for the host tools May 03 19:08:11 JPEW any examples where to start? May 03 19:08:38 Mrsan: I had to do something similar for zdoom: https://github.com/JPEWdev/yocto-doom-demo/blob/main/meta-doom/recipes-games/zdoom/zdoom.inc May 03 19:08:50 Hopefully you could do it a *little* less grossly May 03 19:08:54 :) May 03 19:09:25 JPEW atleast its a start:)  Thank you:) May 03 19:10:46 Mrsan: np. Like I said, it's a gross example because it uses a shared work dir (I don't recally why.... maybe there were relocation issues?). Anyway, I think you'd be better trying a dedicated -native recipe that the target recipe can DEPEND on May 03 19:11:35 but now i have an idea in anycase :) May 03 19:27:14 hey folks. is there any sort of callback I can hook into from a kernel module to know that a shutdown/halt/reboot has been issued? customer char device driver is keeping a watchdog alive on embedded board, needs to be self-aware to actually permit shutdown to occur. May 03 19:27:43 alternatively, from user space, any way I can hook into shutdown and `rmmod` the module? would have the same effect. May 03 19:47:02 michaelo: RP: sent the documentation patch for KCONFIG_MODE. First try didn't go through because I wasn't subscribed to the docs mailinglist. May 03 21:28:29 If two layers have a recipe of the same name, how do you determine which layer the recipe should come from? I'm getting different behavior from regular build and an eSDK build.. May 03 21:35:58 fray: obviously if the versions differ, the usual version preferences would apply if set.. if not, or if the version is hte same, it'd be layer priority. May 03 21:36:23 fray: recipe version order will be determined by layer priority. but parsing is determined by BBLAYERS order May 03 21:39:32 so the BBFILES order has the recipe I expected first. but it's not being picked up. May 03 21:39:33 so, if you have namespace collisions (same version recipe, bbclass, .conf or .inc), check your BBLAYERS order - I believe the first one in the list will win May 03 21:39:47 that's what I thought.. and I thought it did it fia BBFILES May 03 21:40:40 BBFILES="/recipes-*/*/*.bb poky/meta/recipes-*/*/*.bb ...." May 03 21:40:55 regular build it picked it from my layer.. but eSDK it picked it from poky layer May 03 21:41:00 poky _repository_ May 03 21:41:41 and unfortunately the name is the same.. cause I'm trying to work around a stupid error in gatesgarth w/o patching the poky repository.. cause last time we did thi to work around a problem, we had users bitch furiously that "we broke things" May 03 21:41:57 well, yeah BBLAYERS is easier for understand if all layers append, but yeah, if some layers prepend, it's more complicated and need to check BBFILES May 03 21:41:58 "what do you mean it works with Zeus, when you have to patch Zeus".. May 03 21:42:18 in this case, we have one prepend.. to hopefully put it at the beginning (which it does) May 03 21:42:37 contents of the BBFILES looks the same to me in both cases May 03 21:43:24 kergoth: isn't the layer priority first? I'm pretty use you can downgrade a recipe to lower version by adding it to a layer with higher priority (without setting P_V) May 03 21:43:33 s/use/sure/g May 03 21:44:14 both layers currently have the same prioity, 5 May 03 21:44:26 JaMa: yes, that's correct. higher-priority layer with older recipe will still win May 03 21:45:03 fray: that sucks, that why we override all priorities in our generated bblayers.conf May 03 21:45:35 yeah, though only in the absence of an explicit prreference May 03 21:46:25 i dislike that behavior, as well as the mismatch in overriding between bblayers order and layer priority.. the meta-mentor setup scripts reorder bblayers by layer priority, added a bitbake-layers command sort-layers May 03 21:46:27 fray: if layer priority is the same, recipe version will play next. if recipe versions are the same and recipe filenames are the same, BBFILES order is next May 03 21:47:31 kergoth: sometimes it's confusing, but sometimes it could be your last resort to fix things unobtrusively... May 03 21:48:22 eh, sometimes, usually i prefer just adding explicit preferences or bbappending a DEFAULT_PREFERENCE instead, though. or override a layer's BBFILE_PRIORITY in bblayers.conf.. May 03 21:49:42 I just adjusted the priorites and I'm trying again.. I just find it odd that regular and eSDK get two different results May 03 21:51:10 kergoth: heh, the worst I had to do is override 3rd-party layer's LAYERSERIES_COMPAT, which was too conservative... May 03 21:51:22 if hte two layers have the same priority and the version isn't specified, the selection could be non-deterministic... i can't recall if bbfiles order is consistent right now May 03 21:51:35 ahh yes, i've done that too, to hack around old nxp layers :) May 03 21:52:24 Ya, that's why I checked the BBFILES order, it's the same in both cases, so it's something else causing different behavior May 03 21:52:45 and maybe it's just purely non-deterministic in this case, but if that happens.. it might be something to look at (in the future) May 03 22:18:41 RP: seeing this http://sprunge.us/UfuZ37 May 03 22:18:52 this is on aarch64 build machine May 03 22:19:00 I wonder if uninative has something to do ? May 03 22:56:18 khem: what makes you think that is uninative? May 03 22:56:30 khem: aarch64 has been on in testing on the AB... May 03 22:56:50 khem: the busybox reproducible failure is still there, thought it had been fixed :/ May 03 23:07:39 RP: the busybox one yeah it will take some iterations, I think the one I sent makes the beginning, May 03 23:08:15 applets.h is a generated file and the tools used in generation are not building consistent output May 03 23:08:18 khem: looking at the diffoscope output, I'm betting its a sort being run under different locales May 03 23:08:42 khem: ( and A-Z sorting differently May 03 23:08:47 its just re-ordering the lines which effectively May 03 23:09:35 yeah, thats where first I wanted to sort the cfgs and then next patch is to sort .config itself before feeding it to make oldconfig May 03 23:10:12 problem is that you need two systems to see the difference so not easy to debug locally May 03 23:10:45 on single system the header file always comes out one way May 03 23:11:25 so for now, I would say hold the long options patch May 03 23:11:39 the other one can go in. I think it has nothing to do with gcc11 May 03 23:12:03 gcc11 just makes it more evident because of dwarf-5 debug info defaults May 03 23:13:19 I think another fix I wanted to try was to call make gen_build_files explicitly in busybox do configure May 03 23:13:26 so we can control when applets.h is generated May 03 23:16:15 khem: I think I've seen this before, its the wildcarding in the gen_build_files.sh script May 03 23:16:43 khem: the wildcards expand differently depending on the locale setting, e.g. non-utf8 vs utf8 May 03 23:17:26 I see May 03 23:17:34 so what was the way to fix it ? May 03 23:17:36 sort ? May 03 23:26:43 khem: trying to remember :/ May 03 23:28:11 something like this https://git.openembedded.org/openembedded-core/commit/?id=5541ba76ccc0c416f315bc0dc14a20a33059bd5f May 03 23:28:51 khem: that is a makefile fix, this is a shell script May 03 23:29:30 khem: more like http://git.yoctoproject.org/cgit.cgi/poky/commit/?id=2d2f017240cd2f0ee02635cf0a7d2a612b946ac1 May 03 23:31:47 RP: interesting I wonder is the issue at hand due to dash Vs bash. May 03 23:32:29 and why dont we set locale globally for builds once for all May 03 23:33:04 khem: we do, which made the rsync issue all the more puzzling May 03 23:33:19 hmm May 03 23:35:11 khem: I can't reproduce the behaviour offhand May 03 23:35:15 perhaps its using some host tools which is obstinate :) May 03 23:36:31 khem: it will be some bash/dash host tools version + default locale problem May 03 23:37:59 We should just containerize the build perhaps :) May 03 23:39:54 khem: better to fix these bits of non-determinism ;-) May 03 23:45:49 RP: should something like this help perhaps http://sprunge.us/180nel May 03 23:47:03 RP: we allowed developers to have different distros and it quickly becomes unaffordable support burden, and this year infra team has started deploying containers and the support ticket graph is visibiliy very low May 03 23:53:55 khem: something like that might work May 03 23:54:10 khem: I understand but we are not doing so badly with the wider project May 03 23:54:26 I am trying locally so see if it generates different .h with bash/dash lets see May 03 23:54:30 if I can reproduce May 03 23:55:10 khem: I need to sleep. Good Luck. If you don't get anywhere I can have a look tomorrow May 03 23:56:25 thanks RP gn :) May 04 00:01:36 RP: looks like bash/dash issue May 04 00:17:02 RP: patch is on ml May 04 00:17:52 I tested it with local /bin/bash and /bin/dash and output it same before it was not. It was much like what we see in diffoscope so I think this should fix the issue **** ENDING LOGGING AT Tue May 04 02:59:57 2021