**** BEGIN LOGGING AT Tue Feb 16 03:00:33 2021 Feb 16 07:33:45 yo dudX Feb 16 07:34:19 Can I update my yocto dunfell version to the latest one? If yes, how? Feb 16 07:44:58 Just replacing the current poky directory with the latest one will work? Feb 16 07:49:40 vijay: usually one has git cloned poky, and just does a pull then. Feb 16 07:51:12 LetoThe2nd Got it, thank you. Feb 16 08:38:31 Good Morning! Is there any mechanism in KAS to archive/snapshot all used layers? Similar to how I set BB_GENERATE_MIRROR_TARBALLS to have a snapshot of all the sources used for a build? Feb 16 08:38:31 I guess `kas for-all-repos [...]` can be utilized to archieve this, but It seems like this is a common use case and there might be a ready-to-use solution out there, that I've just not found yet. Feb 16 08:42:26 gm, on zeus, bitnbake is building an out of tree module, but using local host headers. How can i force to use proper kernel headers version ? Feb 16 09:06:58 angelo__: by fixing the include paths? make sure variables passed by Yocto to the oot driver's Makefile are not overridden in the latter. Feb 16 09:17:13 qschulz, thanks Feb 16 09:19:41 dwagenk: snapshotting is usually done on build level, rather than setup level as far as i can tell. Feb 16 09:29:05 My fokus is on having everything available to reproduce a certain Image years later. Having the DL_DIR archived and my own custom layer only get's me partway if an external layer is not available online anymore. So that's my reasoning for trying to snapshot all layers (and document what kind of host system or container was used for the build). Could you elaborate on the 'snapshotting at the build level' a little more? Thx! Feb 16 09:35:39 dwagenk: the rationale is completely clear to me, no worries. but i'm not certain that the approach is really the best way. generally, my elaboration is: look at the sections on license compliance in the manual, they give a lot of information on archiving the sources and metadata. Feb 16 09:45:21 RP: you mean a map for SSTATE_MIRRORS or changing the UNINATIVE_URL ? Feb 16 09:53:23 dwagenk: just mirror all layers in your internal git repo? Feb 16 09:55:24 LetoThe2nd: The section providing-compilation-scripts-and-source-code-modifications is pretty close to what I'm looking for, but tailored towards the "provide sources+meta for external use", whilst I don't see it fit for "backup everything needed to reproduce the build or restart development for internal use". Since both are steps that should be done at every release I guessed it might be more tightly integrated. Feb 16 09:56:26 s/repo/server/. basically one additional repo per layer you're using Feb 16 09:58:49 dwagenk: well i personally would got with qschulz' suggestion anyways - mirror the stuff you want, and the kas config keeps the revision hashes to select. this makes it way easier to actually work with stuff later, instead of just having a humonguous tarball that you can bdil but not really use to maintain things. because rebuilding a couple of years later is usually done for a reason (fixes, modifications) and not Feb 16 09:58:49 just for fun (hey, lets build this to see if we can build it!) Feb 16 09:59:10 s/bdil/build/ Feb 16 09:59:40 qschulz: yes, That's one of the common approaches I've seen. Definitely better than just archiving the flat file structure. Feb 16 09:59:59 dl9pf: map for SSTATE_MIRRORS, changing the URL should be straightforward Feb 16 10:00:11 dwagenk then why did you ask about saving the flat file structure? Feb 16 10:05:11 It's a common task, so having to redo the (minimal) tooling around it every time feels a little redundant. Thus that's something I felt like it could be more tightly integrated to the layer-management. E.g. adding a global or per-repo configuration option in KAS for "mirror-to" that ensures the used refspec is present on the internal mirror and pushes or warns if it is not. Feb 16 10:05:53 Thanks btw. for the detailed conversation here! Feb 16 10:10:29 dwagenk: humm... send patches to the kas folks? ;) sounds like a neat idea to me. Feb 16 10:17:20 Sorry, was not completely clear there (archiving/snapshotting), actually both can be useful, depending on the use case IMHO, but the flat-file snapshot is documented already, right. Feb 16 10:18:23 > humm... send patches to the kas folks? ;) sounds like a neat idea to me. Feb 16 10:18:23 It's on my list now. In general polishing the pre-release workflow and taking a look into what fits in where and could be contributed back. Feb 16 10:32:57 dwagenk: \o/ Feb 16 11:59:39 * LetoThe2nd needs moar coffee, has budgetary presentation for new build server in a few! Feb 16 12:23:06 paulg: https://github.com/systemd/systemd/issues/16896 Feb 16 12:25:32 * RP suspects this is like pouring petrol on a fire :/ Feb 16 12:28:51 RP: kab00m! Feb 16 12:28:56 RP: well, he clearly does not want it :) but "update your kernel to not see this message" is mmm... Feb 16 12:29:13 like people update their kernel, it's one of the last things that get updated Feb 16 12:29:32 (the last being probably anything prior to the kernel in the boot process) Feb 16 13:57:25 qschulz: I update my bootloader and kernel all the time :) Feb 16 14:04:11 marex: by people I meant companies :) Feb 16 14:05:04 qschulz: I can name a few :) Feb 16 14:05:48 qschulz: btw the trick is to have AB update option all the way to the BootROM Feb 16 14:18:58 linux-yocto-dev module build seems to fail with gcc-10.x Feb 16 14:19:29 tlwoerner: with what error ? Feb 16 14:21:41 marex: | /z/build-master/meta-rockchip/build/tmp-glibc/work-shared/roc-rk3328-cc/kernel-source/crypto/aegis128-neon-inner.c:150:3: error: incompatible types when initializing type 'unsigned char' using type 'uint8x16_t' Feb 16 14:21:44 etc… Feb 16 14:22:22 tlwoerner: what kernel version is that ? Feb 16 14:22:34 5.11-rc Feb 16 14:22:43 tlwoerner: I've been building 5.10 and 5.4 (latest stable) with debian 5.10 on both arm32 and arm64 just fine Feb 16 14:22:53 hmmmm, next too btw Feb 16 14:23:02 yes, linux-yocto builds fine for me Feb 16 14:23:30 tlwoerner: the above seems part of kernel crypto (i.e. something which depends on arm64 neon?) Feb 16 14:23:55 marex: the rk3328 is 4xA53 with crypto Feb 16 14:24:16 i pulled the defconfig from what armbian is using Feb 16 14:25:06 it builds fine if i use (for example) the linaro 8 toolchain Feb 16 14:25:27 gcc 10 is pickier on types i guess Feb 16 14:26:04 tlwoerner: did you try compiling it out of OE with gcc 10 ? Feb 16 14:26:22 marex: good idea, i hadn't thought of that Feb 16 14:27:27 I can try it on some arm64 right now Feb 16 14:30:37 tlwoerner: seems to build fine for latest next Feb 16 14:31:07 RP, isnt that the same link I put in the patch? Feb 16 14:32:56 marex: i'm guessing my non standard defconfig might be the cause (?) Feb 16 14:33:27 Hi everyone, since switching to gatesgarth I get these sigaborts as described here (https://wiki.yoctoproject.org/wiki/Pseudo_Abort). The inode mismatches seem to come from using a docker container (crops/poky) for my builds. Did anyone here have this problem before and perhaps found a fix? Feb 16 14:33:31 tlwoerner: well, does it compile out of OE ? :) Feb 16 14:33:36 with the same defconfig that is Feb 16 14:34:11 paulg: yes, its just updated a bit. I've tweaked the patch as requested in review and sent it upstream Feb 16 14:34:34 oh - ok -- I didn't read to the bottom - let me do that. Feb 16 14:35:45 you are a braver man than me - I thought about sharing it there, so other poor sods could make use of it, even if Captain Systemd didn't agree. Feb 16 14:36:02 In the end I decided I didn't need the drama of as you say - more gas on the fire. Feb 16 14:36:45 paulg: I think we do need to push the YP perspective a little now and again Feb 16 14:37:37 Yep - can't argue that. I'm happy to hide behind the hedge and eat popcorn and watch. ;-) Feb 16 14:38:32 * paulg thinks RP just did a "Here, hold my beer..." moment. :-P Feb 16 14:38:56 paulg: Just thought you might want to see where it goes :) Feb 16 14:42:37 totally! Especially since my fingerprints aren't anywhere to be seen. Feb 16 14:43:42 RP: Fixed the acpica repro issue.... was a little more involved than I expected (had to make another bison fix) Feb 16 14:44:29 * JPEW is so cold.... I should move my servers below my desk and use them as a space heater Feb 16 14:45:03 Except I shut them off to be nice to the electric grid in hopes I can avoid a rolling blackout Feb 16 14:45:04 * tlwoerner it always in shorts Feb 16 14:45:26 :s/it/is Feb 16 14:47:43 the neighbours think it's hilarious that i go outside in shorts all winter long to throw wood in the boiler, even at -20°C Feb 16 14:48:10 tlwoerner: how do you like the wood boiler? Feb 16 14:48:23 wmat: *love* it! Feb 16 14:48:40 tlwoerner: bought a camp in the north and am considering one Feb 16 14:48:57 wood is virtually free, and i get free domestic hot water from it too Feb 16 14:49:16 I thought you were in the city for some reason Feb 16 14:49:42 the pumps take next to no electricity, so even if we get a 3-day power outtage in the middle of winter, we're still at 26°C inside with hot showers :-D Feb 16 14:49:54 (on our backup generator) Feb 16 14:50:30 wmat: we live on a 30-acre horse farm just outside Tottenham Feb 16 14:51:12 wmat: i moved out of the city almost 17 years ago and wouldn't want to have to go back ;-) Feb 16 14:51:23 oops 14 Feb 16 14:53:09 wmat: your usage of the word "camp" implies you're not from the GTA originally, i've noticed people who grew up north of Barrie tend to use that word instead of "cottage" Feb 16 14:54:29 ? Feb 16 14:57:41 tlwoerner: cottage just sounds so pretentious ;) Feb 16 14:57:57 lol Feb 16 14:57:57 but yes, I grew up in northern ontario Feb 16 14:58:42 wmat: cool. my wife grew up in the Sudbury area Feb 16 14:58:58 tlwoerner: I was west of there in Elliot Lake Feb 16 14:59:08 if you were in Minnesota, it's a Cabin.. :) Feb 16 14:59:28 could be a completely modern house, but if it's in the north of the state, it's a Cabin.. :) Feb 16 14:59:36 heh Feb 16 14:59:54 cabin sounds nice. "camp" is that place your parents sent you as a kid during the summer ;-) Feb 16 14:59:59 and yes, cottage would be pretentious to say around here.. Feb 16 15:00:02 tlwoerner yup Feb 16 15:00:23 regional colloquialisms are interesting Feb 16 15:00:27 yup Feb 16 15:01:05 JPEW: cool, thanks. I worked out a few repro issues and sent them so hoping we can improve the stats a bit! Feb 16 15:01:07 to me a camp is something you visit once every couple years and is most likely run down, like maybe a hunting camp Feb 16 15:01:09 I'm guessing it has to do with what people originally did in the area.. Northern MN is called "Cabin country" here.. cause the rich would buy land up in the woods (that was more or less un-inhabitable), but a get away for the weekend, but never live there.. Feb 16 15:01:15 JPEW: power issues sound awful :( Feb 16 15:01:22 and most were just that, log cabins originally Feb 16 15:01:37 but cottage is the place you visit most weekends and keep up Feb 16 15:01:42 tlwoerner ya, that is the image I have.. but originally it probably did come from 'hunting camp' Feb 16 15:02:36 if you have cable tv, or one of the ott streaming services, there's a channel called "cottage life", and a magazine of the same name too Feb 16 15:02:49 to me cottage is a second home, that is used 'regularly'.. i.e. "ohh so you can afford TWO expensive homes" Feb 16 15:03:06 cottage is also (to me) a very east coast term Feb 16 15:04:04 cottage to me is a giant sack of extra maintenance. Feb 16 15:04:28 paulg: haha, true. not to mention the traffic up and back Feb 16 15:04:28 lol, they all are.. :) Feb 16 15:05:06 …as the entire metropolitan populate tries to get up and back before monday Feb 16 15:05:14 population Feb 16 15:05:15 tlwoerner: I cut through the back roads and pop out on HWY400 north of Barrie, it works out perfect Feb 16 15:05:37 ya, part of the reason I've never bothered getting a 'cabin'. I'd love to have land and woods up north -- but the commute makes me not want to do that Feb 16 15:06:07 (the part of Northern MN I'm talking about is basically all lakes and woods.. population density is VERY low most of the year..) Feb 16 15:06:09 I count on that to keep the masses down south Feb 16 15:06:36 and a LOT of the cabins are three season, but need to be winterized.. Feb 16 15:06:51 wmat: my family had a cottage just south of Parry Sound for ~20 years when i was younger. when my dad retired they bought one of the business on the lake and ran a cottages/campground for a couple years Feb 16 15:07:10 tlwoerner: nice Feb 16 15:07:35 i like to believe my farm is the best of both worlds ;-) Feb 16 15:07:37 even that area is too crowded for me Feb 16 15:07:50 i could commute to Toronto if i had to, but i'm not in the city Feb 16 15:07:53 tlwoerner same.. I have my 20 acres.. and endless maintenance Feb 16 15:08:01 i'm just not on a lake Feb 16 15:08:04 and as of 2 weeks ago.. fiber to the home Feb 16 15:08:10 bonus Feb 16 15:08:17 I have to keep reminding myself, lake in MN = mosqitos.. Feb 16 15:08:26 yo folks, are you planning your holidays or what? Feb 16 15:08:28 fray: yea, i saw that! me too. i got connected Feb 1st Feb 16 15:08:42 fray: sounds like Scotland (midges) :) Feb 16 15:08:53 RP, ya I've heard htat Feb 16 15:09:04 RP: do midges bite? Feb 16 15:09:08 good trout bait Feb 16 15:09:09 fray: just got ~15k€ budget for 2xepyc 7702 build machine approved. yay me! Feb 16 15:09:34 LetoThe2nd: nice! Feb 16 15:09:54 LetoThe2nd: can we all share it? ;) Feb 16 15:09:59 LetoThe2nd NICE.. Feb 16 15:10:15 tlwoerner: er, yes. Horribly. We get them in the forests even in northern England where I am Feb 16 15:10:15 I'm just waiting to be able to afford a large threadripper for myself Feb 16 15:10:24 fray: i was going to poke fun at you because you went from … 30(?)Mbps to 1000? heck, i would have killed for 30Mbps Feb 16 15:10:37 RP: ouch, not fun Feb 16 15:11:03 I was "this close" to ordering starlink when they finally tuned on Fiber Feb 16 15:11:17 tlwoerner: they're the main reason I've not been to Scotland as much as I could have when younger, my parents hate them Feb 16 15:11:20 fray: plus 6TB on pcie nvme Feb 16 15:11:21 30/3 was bad.. 1000/1000 is NO PROBLEM (other hten cost) Feb 16 15:11:22 * LetoThe2nd celebrates. Feb 16 15:11:23 * paulg wonders if midges are like blackflies or sandflies. Feb 16 15:11:28 tlwoerner: having experienced them, I can understand it Feb 16 15:11:37 mosquitoes love me. once my wife and i were walking in the woods and i was complaining about all the mosquitoes. she turned to say "stop complaining" and noticed the *cloud* i had around me (and the none that were around her) lol Feb 16 15:11:52 RP does DEET or similar prevent swarming//bites? Feb 16 15:11:56 RP: midges sound horrible. Much like our Black Flies in Canada but for a lot longer period of time. Feb 16 15:12:03 LetoThe2nd: yay, crazy build systems are cool Feb 16 15:12:28 RP: s/crazy/properly powered/g Feb 16 15:12:41 fray: kind of, not entirely Feb 16 15:12:46 fray: for the last 7 years i've had 3Mbps down and only .7 up, getting fibre was unbelievable Feb 16 15:13:09 wmat: they're there April to October, yes :/ Feb 16 15:13:13 fray: i went with the smallest package: 100Mbps, which i can't even imagining wouldn't be enough (haha) Feb 16 15:13:18 tlwoerner sounds like my neighbor.. was getting just under 30 down, and .5 up Feb 16 15:13:31 (he just built the house in August) Feb 16 15:13:50 So he called the ISP and got a radio link setup.. 40/20 is what he gets over that.. 40-60ms latency Feb 16 15:13:56 1000/1000 fiber sure is nice. I can transfer target images from my build server to home faster than I can copy them to the target :) Feb 16 15:14:01 he asked about latency on the fiber.. I said "4-6" Feb 16 15:14:02 lol Feb 16 15:14:50 to be fair 100/20 was the minimum I wanted.. ($100 a month) 100/50 was $150, 100/100 was $200, and 1G/1G was $300 Feb 16 15:14:54 fray: i was getting an order of magnitude less, *3* down, not 30 (lol) Feb 16 15:15:25 my parents (in the city) were getting similar.. 3 down/.5 up.. $100 a month.. Feb 16 15:15:33 finally moved htem to cable internet.. 200/10 now Feb 16 15:15:54 my plan though for the 1G/1G is to close my local office and move the computers here, but I need to finish building the office in the basement.. Feb 16 15:16:00 i never though in my lifetime that i'd see fibre to the farm Feb 16 15:16:02 200/10 was what I had with Spectrum before I moved into Google Fiber territory. Feb 16 15:16:21 (I just moved the second laundry area from where the office is, to under my stairs over the weekend.. more work then I was originally expecting for sure) Feb 16 15:16:33 georgem_home: you make it sound like you moved for the Internet (lol) Feb 16 15:16:39 and first time I've done drain/waste/vent changes of this magnitude Feb 16 15:16:58 tlwoerner: it was certainly a factor other things being equal :) Feb 16 15:16:59 georgem_home lol my parents have Spectrum.. Feb 16 15:17:29 fray: yeah. I guessed by the 200/10 speed Feb 16 15:17:32 I figure the fiber to my house ("free" other then service cost) probably added $10k or more to the value of my house Feb 16 15:17:44 fray: i just call someone. it stops being worth my time very quickly :-) Feb 16 15:17:59 local phone company installed 48 fibers down the street, and 4 fibers to each home.. (there are only 12 homes in about 2 miles..) Feb 16 15:18:34 I found out, apparently wood shipments from Canada are stopped or way down.. _4_ pressure treated 2x4s.. $45 Feb 16 15:18:36 OUCH Feb 16 15:18:52 we're out of lumber up here as it is Feb 16 15:18:52 yikes Feb 16 15:18:55 (I had to resupport the stairs, the supports were hanging about 1 inch ABOVE the concrete floor.. Feb 16 15:19:20 had rotted out over the last 130 or so years.. they were nailed to the block walls.. but that isn't really 'right' Feb 16 15:19:27 it was the second thing we ran out of (after toiler paper, another wood product) Feb 16 15:20:15 I'm strongly considering building my basement office space with metal studs due to cost.. my only worry, I've never used them before and I wasn't planning to 'learn a new tech' along the way Feb 16 15:20:30 (basement office == server lab and partial maker lab) Feb 16 15:20:38 it's as if they know we'll all be staying home for the foreseeable future… (cue dramatic music, lol) Feb 16 15:20:51 :) Feb 16 15:21:03 fray, get yer own damn trees and leave us alone. ;-) Feb 16 15:21:29 will trade trees for vaccines. Feb 16 15:21:34 :) Feb 16 15:21:44 oh geez, don't get started on *that* one Feb 16 15:21:58 otherwise, we're keeping the door locked for even longer :D Feb 16 15:22:08 * zeddii runs Feb 16 15:22:15 lar thudering j* murphy... Feb 16 15:22:29 :D Feb 16 15:22:51 * tlwoerner didn't want to spell it all out Feb 16 15:28:20 Quick question; if I want to do post-processing removal of a file in the rootfs, isn't it just to do: Feb 16 15:28:34 remove-command() { Feb 16 15:29:06 rm -r ${IMAGE_ROOTFS}/path/to/thefile.txt Feb 16 15:29:09 } Feb 16 15:29:26 ROOTFS_POSTPROCESS_COMMAND += "remove-command;" Feb 16 15:29:34 in image recipie? Feb 16 15:37:45 not sure functions can have dashes in their name Feb 16 15:41:47 let's try snake_case then :) Feb 16 15:44:59 Hmm, might've actually been it Feb 16 15:47:52 Looks like it's building atleast, now I just gotta confirm it works Feb 16 15:51:23 Aww... dammit Feb 16 15:53:21 why doesn anybody put files in images that samebody does not want to be there? Feb 16 15:57:17 LetoThe2nd: More like, we have an image, and a recipe that creates said files, but I want to disable autostart for one file in particular Feb 16 15:57:51 wertigon: append said image to disable said autostart :) Feb 16 15:58:28 LetoThe2nd: you can do that with systemd? Feb 16 15:58:45 I have a systemd service I want to NOT start Feb 16 15:58:49 meh, "append said recipe". thats what i meant. sure, why should the systemd recipe be special? Feb 16 15:59:25 LetoThe2nd: I have two images though Feb 16 15:59:35 One regular one, where I want the service to autostart Feb 16 15:59:38 One where I don't Feb 16 15:59:47 For testing purposes Feb 16 16:00:14 But I still want to be able to call systemctl start service Feb 16 16:02:58 hm Feb 16 16:03:13 I thought "Great I'll just remove /etc/systemd/system/multi-user.target.wants/my-service.service" Feb 16 16:03:21 But that file is still around after I do that Feb 16 16:05:48 well then sprinkle bb notes (orwhatsitcalled) to find out why it doesn't work Feb 16 16:06:45 I'm suspecting systemd wants tree generator is called again as a post-post-processing Feb 16 16:07:50 Testing removing the WantedBy from the service itself then :) Feb 16 16:07:58 If I'm right, it will remove that Feb 16 16:08:20 Otherwise I'll remove the service completely - not optimal, but gets the job done atleast Feb 16 16:10:55 Thanks for the help :) Feb 16 16:24:48 hello #yocto! I am running petalinux, but I think the question stands for plain yocto as well. I want my image recipe to also create FIT images containing a kernel, dtb, and rootfs. Since the kernel and dtb binaries are produced by virtual/kernel (linux-xlnx), I need to wait for those recipes to deploy in order to start building the fit image. I have a task e.g. do_generate_fitimage, and do_generate_fitimage[depends] += "virtual/kernel:d Feb 16 16:24:48 o_deploy", and IMAGE_POSTPROCESS_COMMAND += " do_generate_fitimage; ". But the image recipe does not honor this dependency and runs do_generate_fitimage too early. What am I doing wrong? Feb 16 16:26:35 marble_visions: how do you know it runs too early? Feb 16 16:27:27 qschulz: since files in ${DEPLOY_DIR_IMAGE} are not populated when the task runs Feb 16 16:29:05 if I first run bitbake -c linux-xlnx, and then bitbake -c my-image, all is good. if I run only bitbake -c my-image (which should figure out the kernel dependency) it says files are missing. Feb 16 16:29:22 which files are missing? Feb 16 16:29:40 and do you have the kernel set to point to linux-xlnx in your configuration? Feb 16 16:30:40 PREFERRED_PROVIDER_virtual/kernel ??= "linux-xlnx" Feb 16 16:31:08 you can do bitbake my-image -e and then inspect the variables for PREFERRED_PROVIDER_virtual/kernel to check Feb 16 16:31:15 fray: I will doublecheck, but I think petalinux should have taken care of this. thanks for the suggestion Feb 16 16:31:46 are you using petalinux's "eSDK" and/or direct interfaces? petalinux ? If so, yes it should be doing all of that Feb 16 16:32:10 I am using petalinux 's Feb 16 16:32:37 so in reality I am entering verbatim petalinux-build -c my-image Feb 16 16:32:53 and petalinux-build -c linux-xlnx Feb 16 16:33:11 Ok, then I'd suggest you check Xilinx's forums.. I don't think many of us here have even used petalinux-build.. (I've used it a few times, but I'm not experienced enough to help diagnose this.. sorry) Feb 16 16:34:06 petalinux-build is a wrapper on the regular bitbake behaviors. (Some things are just 'pass through', in that it sources the eSDK environment for you, and runs a bitbake command.. but other things actually do a variety of commands.. Feb 16 16:34:22 will have a look in the xlnx forums again as well. I was wondering if maybe this was bitbake-rooted Feb 16 16:34:29 the petalinux-build -c linux-xlnx _SHOULD_ just be sourcing the environment, and then running bitbake linux-xlnx ... but I've not verified that Feb 16 16:35:31 unfortunately I don't have any petalinux installs handy, or I could figure out exactly what to source so you could run the bitbake command directly for verification Feb 16 16:36:16 (The Yocto Project eSDK documentation migth tell you..) Basically petalinux is an eSDK that has a wrapper. It can either be used via the wrapper, or you can go into the eSDK and follow the YP steps to work on it that way.. I just don't remember the steps.. sorry) Feb 16 16:36:47 (full disclosure, I work for Xilinx in their Yocto Project group -- but I don't use petalinux very often) Feb 16 16:37:23 fray: nice Feb 16 16:37:57 i have the lines written down somewhere Feb 16 16:38:09 so that I can drop to a yocto env and run bitbake directly Feb 16 16:39:26 ya, if you do that and 'bitbake my-image' shows the same behavior.. use the bitbake my-image -e to verify the virtual I mentioned.. should point to linux-xlxn.. assuming ti doesn, then I'm not sure what is wrong.. Feb 16 16:39:30 which files does it say are missing? Feb 16 16:40:04 there are some bootbin related files that don't always get regenerated if your tmp directory is reset.. (known issue) Feb 16 16:40:25 DTBs I am installing from a linux_xlnx_deploy_append, and the kernel fitimage files Feb 16 16:41:06 it's possible it's related to this, but I'm not sure. I don't do a lot with the actual deploy/fitimage files.. (I mostly do builds, compilers and userspace integration) Feb 16 16:41:47 okay, the things you mentions we're great to get me going again. thanks! :) Feb 16 16:43:00 btw, is your team, or Xilinx in general... do you know if there is effort into making a setup that supports multiple vivado exports into one petalinux project? Feb 16 16:43:21 At present, I'm not aware of anything like that. Feb 16 16:43:45 The current workflow uses XSCT (which is a command line version of part of vivado) to process things, so it SHOULD be possible.. Feb 16 16:44:30 so petalinux includes XSCT, reads the vivado database export, and runs from there? Feb 16 16:44:48 we're working to eliminate the need for xsct and allow vivado to directly export everything that is required.. then YP (and probably petalinux) workflows can just use an input file per configuration.. Feb 16 16:45:30 marble_visions more or less yes.. the HDF/XSA (or whatever it's called) export from Vivado is interpreted by XSCT to give the build the settings it needs for whatever step... Feb 16 16:46:10 but it would be an intermediary conf file, right? vivado won't export machine meta- layers? Feb 16 16:46:24 (I don't know when we're going to have this new workflow, but we're actively working on it) Feb 16 16:46:40 yes.. vivado itself doesn't know anything aobut Yocto Project and related.. Feb 16 16:48:09 good to know! Feb 16 16:48:16 Basically today (AFAIK) it's Vivado -> vivado hardwre model -- then XSCT reads this model and does whatever it needs.. Feb 16 16:48:42 in the future we hope it'll be Vivdo -> DTB + hardware model ... then the DTB can be directly processed and a YP config generated Feb 16 16:49:20 so it's still separate, but it won't require XSCT.. Feb 16 16:51:07 definitely a step in the right direction.. is the hardware model used in Yocto, or it gets passed to other departments? Feb 16 16:51:36 maybe you'd need it for uboot and fsbl configuration? Feb 16 17:00:51 afternoon, building from inside a docker image I have the following error: Please use a locale setting which supports UTF-8 (such as LANG=en_US.UTF-8). Feb 16 17:08:17 fray: qschulz: yep, can confirm that running the bitbake enviroment, the issue still exists, and PREFERRED_PROVIDER_virtual/kernel="linux-xlnx" Feb 16 17:08:33 I should create a minimal reproducible example.. Feb 16 17:08:58 the only other thing to try.. bitbake -c clean linux-xlnx -- then re-run the my-image.. if the problem goes away. Thens omething being done in the linux-xlnx is not being cached properly Feb 16 17:18:05 fray: hmm.. does it matter that do_generate_fitimage (and it's [depends]+=...) lives inside a bbclas that my-image inherits? can that get in the way of dependecy resolution? Feb 16 17:20:53 it might, I'm not really familiar with that part Feb 16 17:27:30 okay, I will add everything to the my-image recipe Feb 16 17:45:01 mmmmmm Feb 16 17:45:14 does IMAGE_POSTPROCESS_COMMAND functions actually have depends support? Feb 16 17:45:18 do* Feb 16 17:46:01 not directly Feb 16 17:46:03 good question.. I didn't think it wouldn't have. would and alternative be to use "add task before .. after .." ? Feb 16 17:46:15 marble_visions: yup Feb 16 17:46:26 if you add IMAGE_POSTPROCESS_COMMANd, you need to also specify the dependencies manually.. Feb 16 17:46:27 or, put the dependency on the task taking care of postprocess_command Feb 16 17:46:35 (I'm forgetting how, but I've had to do it in the past) Feb 16 17:47:22 wooow, okay, let me finish this build and will switch to check if it succeeds Feb 16 17:47:58 do_image_complete is the task taking care of said variable Feb 16 17:48:27 so like do_image_compiled[depends] += ":do_deploy" ? Feb 16 17:48:56 or just DEPENDS_append_pn-my-image = " recipe1" Feb 16 17:49:11 (depends on if it's something being run, or data being needed from a do_deploy) Feb 16 17:50:05 very quickly read the code and it does not seem to fulfill [depends] varflags Feb 16 17:50:34 fray: DEPENDS is a bad idea for something you need after deploy task has run Feb 16 17:50:54 because there is no guarantee deploy will run before your DEPENDS is resolved Feb 16 17:51:03 yes.. DEPENDS is needed if you are running a command provided by a native.. deploy is needed if you are using a deployed item Feb 16 17:51:04 and also, it populates the sysroot of the recipe for no reason Feb 16 17:51:35 ya, the sysroot is why you'd want it in depends Feb 16 17:51:42 fray: my bad, read too quickly. absolutely, it should be only native recipe Feb 16 17:52:21 recipes* in DEPENDS of an image recipe... and... there might even be better variables (I think there are some specific ones when they are related to a certain IMAGE_FSTYPE for exampole Feb 16 17:52:36 fray | so like do_image_compiled[depends] += ":do_deploy" ? Feb 16 17:52:39 yes ^ Feb 16 17:52:49 (without the typo :) ) Feb 16 17:53:04 ya, that's what I've done for deployed bits in the past Feb 16 17:53:26 i.e. I need a firmware that is in do_deploy, and my processing code moves or configures it in some way Feb 16 17:54:05 fray: mmmmmm Feb 16 17:54:23 for machine specific firmware... it should be in EXTRA_IMAGEDEPENDS Feb 16 17:54:34 I think? Feb 16 17:54:42 Too late, my brain is fried :) Feb 16 17:54:53 good luck and have a nice day/evening Feb 16 17:55:00 qschulz: wait! Feb 16 17:55:12 you are being owed one beer :) Feb 16 17:55:21 fray: you too Feb 16 17:55:29 if you guys drink, that is Feb 16 17:55:29 ha.. no problem Feb 16 17:56:31 I think deps are now being picked up with the addtask mechanism.. waiting for the build to finish to confirm Feb 16 18:04:25 the only time I've ever had water where I'm building the office is when the lock wall (where the laundry is) completely filled with water.. once I drilled those drain holes, no more water on the other side of the wall.. Feb 16 18:04:28 oops Feb 16 18:05:37 https://www.irccloud.com/pastebin/LdTVUGSR/ Feb 16 18:05:46 RP: ^^^ core-image-minimal Feb 16 18:06:26 on dunfell Feb 16 18:06:36 ooooooh Feb 16 18:08:28 dl9pf: nice. If you send a patch enabling rpm I can enable on the autobuilder Feb 16 18:08:43 dl9pf: we can test there and confirm with world Feb 16 18:08:49 (for master) Feb 16 18:08:51 ok Feb 16 18:09:52 dl9pf: I'm surprised, I though it was not enabled for a reason! Feb 16 18:10:22 i just changed package_classes to rpm and images to core-image-minimal to start Feb 16 18:11:32 fray: qschulz: got it to work! woohoo! Feb 16 18:11:52 dl9pf: we already know the package contents is ok so if the rpm side works, its worth wider testing with the full config on the autobuilder Feb 16 18:11:54 so I'll keep that in mind, rule of thumb, use addtask Feb 16 18:12:05 roger Feb 16 18:30:00 sent Feb 16 19:08:04 jonmason: https://autobuilder.yoctoproject.org/typhoon/#/builders/113/builds/728 - breaks in the release build! :/ Feb 16 19:08:10 (its gatesgarth) Feb 16 19:10:40 RP: on it. Is there a bug I need to worry about? Feb 16 19:11:42 jonmason: swat isn't up to it yet Feb 16 19:12:04 jonmason: this was the 3.2.2 build but I'm assuming it would be a meta-arm fix and the main release is ok? Feb 16 19:12:36 yes, I'm assuming there is a bug with something not applying cleanly Feb 16 19:12:46 trying to replicate locally to confirm Feb 16 19:16:11 actually, i think it is running master meta-arm on gatesgarth, which we don't test. Probably better to drop that from compat Feb 16 19:42:50 JPEW: I rebuilt the image (meta-python-ptest-image) with IMAGE_FSTYPES wic.qcow2 and tried to launch it with VirtManager. It hung on /dev/sda1 systemd task Feb 16 19:51:13 JPEW: and the prior attempt I was using virtio bus for the disk with core-image-minimal from dl.yp.org 3.2.1 Feb 16 19:51:18 JPEW: which worked Feb 16 19:52:04 JPEW: https://gist.github.com/moto-timo/27e39256ce3aad4da7b8ac0e0273d4fc Feb 16 19:52:30 JPEW: I've been playing around with disk size and memory (and most recently bus) Feb 16 19:57:17 JPEW: with 'runqemu kvm nographic slirp' the image boots fine, so it's something about the VM environment I think Feb 16 19:58:27 moto-timo: Ya, sounds like it Feb 16 19:59:02 I forget why you're wanting to run in KubeVirt instead of qemu+kvm? Feb 16 20:03:24 Ultimately, kubernetes native/cloud native in a real cloud. Feb 16 20:04:15 And to be honest, I haven't completely grasped your workflow. Feb 16 20:10:39 Is there a way to get the command line runqemu would run without actually running it? Feb 16 20:17:54 moto-timo: I think so? But honestly I'm not even using runqemu, I just let labgrid run the system QEMU: https://github.com/JPEWdev/yocto-doom-demo/blob/main/ci/qemux86-64/env.yaml Feb 16 20:18:16 I think runqemu would work just a well though Feb 16 20:28:22 rburton: do the outreachy projects have a scheduled end date? Feb 16 20:45:02 tlwoerner: March, although extensions are possible Feb 16 20:45:08 hi all -- I still cannot build systemd with PACKAGECONFIG_append = " importd" Feb 16 20:46:33 moto-timo: there you are, i just looked for your nic and didn't find it :-) Feb 16 20:46:37 systemd compilation fails with "Run-time dependency {libapparmor,glib-2.0,gobject-2.0,gio-2.0} found: NO (tried pkgconfig and cmake)" Feb 16 20:47:02 moto-timo: if we do have a YP event, it would be nice if one or more of the outreachy participants could present their work Feb 16 20:47:46 vdl: are you setting PACKAGECONFIG append in your local.conf? You need to specify the package name (pn) which is being appended: PACKAGECONFIG_append_pn-qemu = " " Feb 16 20:48:20 vdl: note the syntax (the underscores are important, the space after the first " is important) Feb 16 20:49:15 moto-timo: oops sorry bad copying, I'm using PACKAGECONFIG_append_pn-systemd = " curl xz zlib bzip2 gcrypt importd" from my distro conf. Feb 16 20:50:29 vdl: so your error is telling you that you are missing RDEPENDS... use bitbake -e systemd and grep for ^RDEPENDS Feb 16 20:50:56 vdl: otherwise, there might be another error that one of those dependencies cannot build for some reason. Feb 16 20:50:56 * vdl is typing... Feb 16 20:51:53 vdl: but since you are giving us the compile log for systemd, I suspect you need to add DEPENDS/RDEPENDS Feb 16 20:53:38 moto-timo: there you go: http://ix.io/2PFe Feb 16 20:54:21 my browser blocks that url Feb 16 20:55:46 moto-timo: RDEPENDS_systemd=" kmod dbus util-linux-mount util-linux-umount udev (= 1:244.5-r0) util-linux-agetty util-linux-fsck systemd-serialgetty volatile-binds ldconfig libnss-myhostname update-rc.d systemd-vconsole-setup" Feb 16 20:59:11 moto-timo: I've added all of these packages to IMAGE_INSTALL_append but I still get the same failure Feb 16 21:01:40 vdl: you need it at build time, I suspect, so you need to bbappend systemd recipe with DEPENDS Feb 16 21:04:00 moto-timo: all these systemd "Run-time dependencies" are glib I think. Is there such package in yocto? Feb 16 21:04:25 I can try to monkey patch poky's systemd recipe Feb 16 21:06:41 I don't see what provides glib and libapparmor Feb 16 21:10:33 vdl: please use layers.openembedded.org and search recipes... help yourself :) Feb 16 21:12:18 vdl: also, oe-pkgdata-util is handy for finding out what files are in what package http://git.yoctoproject.org/cgit/cgit.cgi/poky/tree/scripts/oe-pkgdata-util Feb 16 21:22:15 moto-timo: I didn't know layers.openembedded.org could search recipes, thank you. glib-2.0 found in openembedded-core ;-) Feb 16 21:23:22 vdl: happy to point you at one of my daily tools :) Feb 16 21:23:56 vdl: I also sometimes just sit with my iPad and roam cgit repos... Feb 16 21:24:23 but that's a slow meandering path vs. a real search Feb 16 21:28:42 moto-timo: iPad Oriented Development, eh? :D Feb 16 21:30:26 moto-timo: stupid question, "openembedded-core" is poky/meta, right? Feb 16 21:36:04 vdl: correct Feb 16 21:36:33 vdl: combo-layer stitches it into poky and makes it look like poky/meta (and poky/scripts) Feb 16 21:36:52 vdl: that's also how bitbake gets into poky/ Feb 16 21:44:09 vdl: the key point is that poky is just a convenience repo that is bitbake + openembedded-core + meta-yocto all glued together Feb 16 21:47:19 rburton: simplifies what I say yet again :) Feb 16 21:47:51 * moto-timo blames 5 hours of sleep Feb 16 21:52:46 rburton: that's very confusing in fact ^^. I'd expect just a meta-poky distro layer and that's it. I'd make sense for such glued-together repo to be named "yocto" if such thing should exist Feb 16 21:52:57 it definitely should not be called yocto Feb 16 21:53:02 thats one of the things it should not be called Feb 16 21:53:07 because its an example Feb 16 21:53:22 the poky distro is just one way of glueing the bits together into something that works Feb 16 21:53:24 the reference distro is poky, not yocto Feb 16 21:54:18 yep, poky's a distro, hence a standalone meta-poky layer would make sense Feb 16 21:54:41 http://git.yoctoproject.org/cgit/cgit.cgi/meta-yocto/ <-- that? Feb 16 21:54:49 bring your own bitbake and oe-core, and it works Feb 16 21:54:55 as that's all poky is: those glued together Feb 16 21:55:07 (there were reasons it wasn't make meta-poky and i can't recall them) Feb 16 21:56:23 vdl: yocto is more than just a buildsystem. bitbake+oe-core+meta-yocto are the baseline buildsystem for yocto, but that's one project under its umbrella. admittedly the main one, but yocto wouldn't make sense as a name Feb 16 21:57:54 http://git.yoctoproject.org/cgit/cgit.cgi/poky/tree/meta-poky/conf/distro/poky.conf Feb 16 22:01:04 so you could get rid of the "poky" repository for production projects, right? Feb 16 22:01:46 as he said, poky repo is just bitbake+oe-core+meta-poky, whether you want it as as ingel repo or individual components is really up to you and how you want to structure your project Feb 16 22:02:20 that's just this "poky" repository that I would've named "yocto", since it contains meta-poky and meta-yocto-bsp, and is the glued-together repo, but nevermind Feb 16 22:02:30 it's not yocto. Feb 16 22:02:39 that's not what it is, so it doesn't make sense to call it that Feb 16 22:02:51 it's an easy way to get and build poky, the reference distribution Feb 16 22:04:49 its like saying they should have called the web server 'apache' it's not apache. the super project is called apache, the web server is apache httpd Feb 16 22:07:30 I see Feb 16 22:08:05 Yocto is the means by which multiple companies, many of which are competitors, are able to collaborate on a common embedded linux baseline. the build tooling for it is a big part of that, but not the only part. Feb 16 22:12:57 bitbake has its own repo, right? Feb 16 22:22:50 I might get rid of the poky repository then, not a fan of the glued-together approach (even though I understand the purpose) Feb 16 22:33:03 vdl, yes, bitbake is it's own repo Feb 16 22:33:19 https://git.openembedded.org/ Feb 16 22:33:25 so is core Feb 16 22:52:58 oi Feb 16 22:54:32 vdl: have you made your own manifest repo yet? Feb 16 22:54:44 nerdboy: I'm using kas Feb 16 22:55:28 sorry, i need more context for that... Feb 16 22:56:04 no idea what kas is Feb 16 22:58:11 moto-timo: it seems like adding meta-openembedded/meta-perl and meta-security allows to build systemd with importd, but it brings in a required "apparmor" distro feature. Doesn't this sound too much just for building importd? Feb 16 23:03:10 i meant making something like this: https://github.com/VCTLabs/vct-boundary-bsp-platform Feb 16 23:03:59 you can make/test custom distro stuff, oe-core vs poky, etc Feb 16 23:04:34 I do not like "repo". kas is another tool from siemens, better suited IMHO. Feb 16 23:05:54 it's actually not so bad unless it's android... Feb 16 23:06:27 I won't argue on that :) Feb 16 23:12:06 Isn't it supposed to be a `dhcp` package in `meta/recipes-connectivity`? I don't find it in gatesgarth, but when I google for it I seem to find something on some repos 🤔 (disclaimer: I'm new to Yocto) Feb 16 23:12:57 (I'm trying to do this in my image recipe: `IMAGE_INSTALL_append = " dhcp"`) Feb 16 23:13:32 (I would like to have a dhcp server running, so that I can connect over USB, get an IP and SSH into my board) Feb 16 23:21:13 vdl: kas over repo any day. Feb 16 23:21:14 kas doesn't phone home every time you run it and it actually gives you ERROR messages if it can't clone. Feb 16 23:21:14 repo: crickets Feb 16 23:22:09 disclaimer: i made that thing ^^ at least 5 yrs ago.. Feb 16 23:22:45 vdl: I haven't knowingly built importd, so I have no idea. But I've had to bring in selinux just to build yadda yadda and set 'pam' distro feature to build foo. It happens. Feb 16 23:23:22 lot's of people like repo and use it well. It just isn't for me. Feb 16 23:23:31 ^Lots Feb 16 23:23:40 grammar has gone to the winds Feb 16 23:24:04 moto-timo: ;) btw importd still fails even though Run-time dependencies apparmor and glib stuffs are found... Feb 16 23:51:50 dl9pf: https://autobuilder.yoctoproject.org/typhoon/#/builders/118/builds/9 - it says all the rpms are different on that fedora worker! :/ Feb 16 23:52:20 dl9pf: looks like the problem may be host specific, will wait and see how the other builds look Feb 16 23:55:30 ok need to see it locally, checking master next Feb 16 23:56:21 can I limit the images and targets w/o hacking reproducible.py ? Feb 16 23:57:12 dl9pf: hacking that is easiest. FWIW when it says packages will be copied to /srv/autobuilder/autobuilder.yoctoproject.org/pub/repro-fail/oe-reproducible-20210216-3b7npcqi Feb 16 23:57:46 dl9pf: it means here: https://autobuilder.yocto.io/pub/repro-fail/oe-reproducible-20210216-3b7npcqi/ Feb 16 23:58:20 dl9pf: could pick a couple of the rpms out and run diffoscope locally Feb 16 23:59:09 ack Feb 17 00:04:10 rpm metadata ... some hashes and the BUILDTIME ... all in the rpm header Feb 17 00:24:04 SIGMD5, SHA1HEADER, SHA256HEADER, BUILDTIME ... everything else is fine Feb 17 00:25:07 ya, it's important when checking RPMs on a rebuild to only diff the actual contents that are not actually package specific.. Feb 17 00:25:22 for security purposes packages do get information when they were built to avoid 'look-alikes' Feb 17 00:25:42 it might be possible set those values -- but if you are actually building them then you wanted a new one for some reason. Feb 17 00:26:13 From a 'diff' perspective, there is an rpm-diff somewhere.. which will avoid the signatures and buildtime.. (or at least used to) so you can look for equivalency Feb 17 00:27:41 uff ... https://github.com/rpm-software-management/rpm/issues/1467 Feb 17 00:31:11 I've done things with RPM if FIPS mode before.. it's not as hard as people think.. Feb 17 00:31:29 FIPS doesn't say you can have MD5 or SHA1.. it says you can't use them in security critical components. Feb 17 00:32:11 So for RPM, you would need to document where they are used, and that they are NOT security critical. And then provide an implementation of the functions that can ONLY be used in those particular places and not "general usafge" Feb 17 00:32:53 This is no different then other protocols that require md5 hashes or sha1 hashes.. again not for security but for 'magic numbers', as security should be handled with other methods Feb 17 00:33:32 i mean in 'this will stick around for quite some' Feb 17 00:33:51 it will that is what happens when you have an installed base.. Feb 17 00:34:02 the work to dump MD5SUM and SHA1 _started_ almost 10 years ago Feb 17 00:34:14 so basically our diff needs to adapt and skip these Feb 17 00:34:16 Red Hat though rejected many of the changes for various reasons (good and bad) Feb 17 00:34:34 yes, there is already an 'rpm-diff' that will skip those headers (or there was last time I used it) Feb 17 00:34:53 https://github.com/kholia/ReproducibleBuilds/blob/master/rpm-compare Feb 17 00:35:00 or rpm-diff Feb 17 00:35:05 dl9pf: We really do want deterministic rpms. Can we not set BUILDTIME to SOURCE_DATE_EPOCH? Feb 17 00:35:39 that is what I'd recommend.. but I do question why we are rebuilding RPMs (if they're already cached).. Feb 17 00:36:09 if the input to the RPM (binaries and metadata) are different.. then we know the RPM will be different.. If they are the same, then we know the RPM output is the same (except for those fiends) Feb 17 00:36:13 'er.. fields Feb 17 00:37:01 RPM itself should never be generating specific output, only packaging. Considering under the hood it's a compresses cpio with additional metadata (the header) Feb 17 00:38:11 fray: the input is the same. We just need to find a way to be able to set those fields the same too and then this should work Feb 17 00:38:30 dl9pf: keep in mind we can patch rpm if needed Feb 17 00:38:47 the hashes are just that, hashes. So if something changes in the header they'll change.. we can't "set" them.. or validation will fail.. Feb 17 00:38:47 prefer not to obviously Feb 17 00:38:58 the date is the only piece we really can set.. Feb 17 00:39:11 fray: so it might just be BUILDTIME and if we have to patch it to set that... Feb 17 00:39:17 yes Feb 17 00:40:09 * RP -> Zzzz Feb 17 00:40:48 night Feb 17 00:42:34 ok, let's ponder tomorrow about BUILDTIME Feb 17 00:45:55 (for FIPS, integrity validation can still used unapproved hashes, but must be backed up by an approved mechanism. This is where package signatures come in as the 'approved' backup) **** ENDING LOGGING AT Wed Feb 17 02:59:56 2021