**** BEGIN LOGGING AT Tue Jan 28 02:59:57 2020 Jan 28 03:24:34 New news from stackoverflow: How to store pyc files with setuptools3 in bitbake recipe in the same way setuptools does? Jan 28 06:37:33 Is it really necessary to install a more recent host compiler in order to build openjre-8 in meta-java? This is how I made it work, but now all the native tools will be compiled with the new compiler as well. Fully untested, I guess and things might break. Jan 28 06:37:41 RP, I know you're a busy person, but if you have any tips for debugging why my "bitbake openjre-8" keeps rebuilding, still no new sigdata is generated so that I can run bitbake-diffsigs, I'm all ears Jan 28 06:40:41 @kroon: Do you see this behavior on every branch or only master? I currently build openjre-8 on the zeus branch and it does not rebuild. Jan 28 06:41:33 RobertBerger, I'm on master branches of everything, except for meta-java where I'm on the 242 update branch Jan 28 06:45:40 @kroon: I guess the C14 host compiler dependency still exists there ;) Jan 28 06:46:08 RobertBerger, yes I'd think so Jan 28 06:46:54 @kroon: just switched to remotes/origin/g0hl1n/wip-jdk8u242ga and will try to see how it behaves with zeus branches Jan 28 06:47:29 @kroon - not well: openjdk-7_99b00-2.6.5.bb:7: Could not inherit file classes/features_check.bbclass Jan 28 06:48:04 RobertBerger, yeah I think you'll need master oe-core if you want to test Jan 28 06:54:21 kroon: you dont see any sigdata ? Jan 28 06:55:46 perhaps try bitbake-diffsigs -t openjre-8 do_configure Jan 28 07:08:32 khem, it tells me it can only find one matching sigdata Jan 28 07:09:03 khem, so I run "bitbake openjre-8" again Jan 28 07:09:21 which task keeps running again and again Jan 28 07:09:29 run it on that Jan 28 07:13:27 khem, 1. bitbake -c cleansstate openjre-8 --- 2. bitbake openjre-8 --- 3. bitbake openjre-8 # again, it rebuilds, fetch, unpack, etc.. Jan 28 07:13:45 khem, looking in stamps/, I only have one *.sigdata.* for the tasks Jan 28 07:14:07 So i cant run diffsigs Jan 28 07:14:46 run it on first task perhaps fetch Jan 28 07:15:04 but it seems quiiite wierd Jan 28 08:18:17 Hi, i'm trying to build my sdk but i get: package perl-module-dynaloader-5.24.4-r0.armv7ahf-neon requires perl-module-config, but none of the providers can be installed Jan 28 08:19:26 one solution is not installing target-sdk-provides-dummy but i think this is not the option. the other option is not installing mysql but i need mysql in the os Jan 28 08:23:17 runqemu tries to use wrong DEPLOY_DIR_IMAGE, how do I force it to use correct one? Jan 28 08:52:57 kroon: stupid question, but the sigdata does not change between rebuilds? (e.g. is it always the same file with same timestamps?) Jan 28 08:57:29 kroon: I had issues with mismatch during recipe parsing, maybe there's something going on there as well? Try to uncomment the two lines: https://git.yoctoproject.org/cgit.cgi/poky/tree/bitbake/lib/bb/siggen.py#n181 This should create a few more sigdata IIRC and maybe that'll be helpful? Jan 28 09:39:44 qschulz, thanks for the hints, i'll try checking those things Jan 28 09:46:54 Hi, is there a way to synchronize flash partitions (linux dts <-> uboot dts)? Jan 28 09:49:22 GeneralStupid: IIRC if mtdparts is set, it's possible to make U-Boot fixup the linux dts (maybe a KConfig option). You can also pass mtdparts to the kernel command line. Jan 28 09:49:33 GeneralStupid: I'd advice you to go to #u-boot and ask there :) Jan 28 09:50:00 Hello :) I have a custom dts file in conf/machine/MACHINE_NAME/custom-2.0.dts, KERNEL_DEVICETREE = "custom-2.0.dtb" seems loaded but my target is still undetected in the arch/arm/boot/dts/Makefile, I have to edit Makefile from bbappend ? o.0 Jan 28 09:50:26 qschulz, disabling rm_work solves it :-/ Jan 28 09:52:01 PinkSnake: the dts should be in a bbappend for the linux kernel recipe. Yes, you need to update the makefile as well. Jan 28 09:52:14 qschulz: Thanks, i tried that mtdparts parameter but it did not work for me. Maybe i need to reconfigure the kernel Jan 28 09:52:49 GeneralStupid: I don't remember but if there's a partition layout in the kernel dts, maybe mtdparts won't do shit /me shrugs. Go to #u-boot, they'll know Jan 28 09:53:16 qschulz Ok thx, sould be not difficult to add task based on machine name do to the job no ? Do you know it's not done inside Yocot itself ? Jan 28 09:53:20 kroon: and you don't have two sigdata for rm_work? Jan 28 09:53:47 PinkSnake: why and what do you want to make it part of task? Jan 28 09:54:13 kroon: try to find which task executed when rm_work is enabled fucks up your build Jan 28 09:54:49 kroon: "build" with "-g" and see the difference for you recipe in task-depends.dot Jan 28 09:55:32 qschulz Don't know but looks like a basic step to make a custom configuration and edit Makefile in a bbappend looks strange. Thank you for your time :) Jan 28 09:58:37 PinkSnake: suggestions and their implementations are welcomed :) Jan 28 10:00:07 PinkSnake: you need to put your DTS in a directory available to your recipe, so you'll need a bbappend just for that anyway. if you want to make the compilation of those DTS auto-magical, then do some tricks in tasks. Jan 28 10:01:23 PinkSnake: technically, dtb-y += mydts.dtb appended at the end of https://elixir.bootlin.com/linux/latest/source/arch/arm/boot/dts/Makefile should do the trick. You do that for every dts passed in the SRC_URI and that might do the trick Jan 28 10:01:28 qschulz Yes I got my dtb correctly, but i'm just looking for a proper way to do that, and to be honest i have made custom tasks to do that ;) I will take a look to edit Makefile with an other task :) Jan 28 10:01:32 (put the dts in the correct place :) ) Jan 28 10:03:28 qschulz: but they are also very slow :) So i will try it out first ;) Jan 28 10:24:37 PinkSnake: if you want to avoid bbappending to the kernel to push in you dts, you can always build it in a separate recipe using the devicetree.bbclass from oe-core Jan 28 10:25:59 nrossi As i have already bbappend to tweak the kernel i'm okay with bbappending the kernel, but thank you for the advise :) Jan 28 10:31:28 nrossi: didn;t know about this one :) thx Jan 28 10:38:32 qschulz nrossi Maybe you know which Yocto variable I could you use to add the dtb instruction in the right "section" inside the Makefile (pattern: dtb-$(CONFIG_ARCH_${SOC_FAMILLY}) += \) ? Jan 28 10:39:28 PinkSnake: I'm not sure you should care actually. dtb-y and that's it. Jan 28 10:39:43 Hoo nice ok thx :) Jan 28 10:39:48 PinkSnake: if you want to know, just look for dtbs which are using the same dtsi and put it in the same section Jan 28 10:42:55 qschulz Currently I know the section but a want to put all the process inside a linux-add-custom-dts.inc or something like that. Thank you guys, that seems to work like a charm with basic append dtb-y to the Makefile :) Jan 28 10:43:56 PinkSnake: honestly, it's just a way to compile less dtbs. Considering how small the DTBs are in size... meh Jan 28 11:25:50 toaster stopped at parsing recipes, 98% is done. How can I see what is wrong? Jan 28 11:44:27 how should I interpret this error "runqemu - ERROR - build/deploy/images/qemuarm64 not a directory valid DEPLOY_DIR_IMAGE" Jan 28 11:46:07 from the weird wording im not sure what is not valid, but anyways, my real DEPLOY_DIR_IMAGE is different, how I make runqemu use the correct one? Jan 28 12:05:55 If I want to share .dts file for a custom board between u-boot and linux kernel, conf/machine folder is the better place to put it ? Jan 28 12:15:38 Hi. So, my problem; I have a build server that multiple people use to build Yocto images, sometimes the same one, more often not. Jan 28 12:15:46 This takes up a lot of disk space. Jan 28 12:16:08 Is there a way to reduce the disk space consumption besides a shared downloads folder? Jan 28 12:16:59 INHERIT += "rm_work" Jan 28 12:17:08 also a shared DL_DIR and SSTATE_CACHE Jan 28 12:17:41 Ok, sounds easy enough :) Jan 28 12:18:17 Thanks Jan 28 12:18:34 These are set in local.conf right? Jan 28 12:19:33 or distro or site or whatever, but yes they're global Jan 28 12:20:03 Ok, good :) Jan 28 12:20:24 Then I'll just make sure to put them in a good palce Jan 28 12:20:25 *place Jan 28 12:20:53 Big thanks for such a quick answer ^^ Jan 28 12:38:59 (on the other hand, unless you're in some form obliged to keep old builds, suffient disk space is cheap) Jan 28 12:54:39 Anyone has experience in using some on-demand VPS service for yocto builds in CI? It's a small scale project using a cloud based CI-system, where it doesn't make sense to have a dedicated VPS or server for doing builds. Jan 28 12:58:34 how can I see why toaster stopped parsing recipes? Jan 28 13:10:51 Hi, I'm looking for a hint how to integrate udev rules from lvm2 or libdevmapper into image? Jan 28 13:11:35 I have all required tools eg cryptsetup but it stuck because it waits for some event from udev Jan 28 13:12:14 I found that there need to be /lib/udev/rules.d/*dm-nofity.rules Jan 28 13:12:34 I see these files on build directory but they are not available on target Jan 28 13:13:09 does anyone know what package, recipe or simething else need to be added to copy these rules into image Jan 28 13:13:12 ? Jan 28 13:20:35 erbo: the problem is you want to share the sstate-cache Jan 28 13:20:52 erbo: don't share that and you're rebuilding the compiler all the time Jan 28 13:21:38 but sharing a gigabyte of data to on-demand servers might be a problem: travis has a cache button but it just puts it into S3 which then takes forever to download... Jan 28 13:23:08 rburton: yeah I definately want to keep sstate between builds, since 95+% of changes just involves building an app + regenerating image. Jan 28 13:23:36 so if your server has persistant storage, you're sorted Jan 28 13:29:50 I'm working with a BSP that is throwing "satisfy_dependencies_for" errors during do_rootfs only when I add EXTRA_IMAGE_FEATURES = "tools-sdk" to local.conf. What's the best way to work around this to get a functional image? Jan 28 13:30:17 Can I exclude the problematic recipes from dev-pkgs somehow? Jan 28 13:30:34 rburton: amazon ebs storage seems fairly afforable. 10$/month for 100GB "General Purpose SSH". Should be plenty for sstate-cache and dl_dir I think. Jan 28 13:30:46 give it a go Jan 28 13:30:47 and report back :) Jan 28 13:32:28 yeah it would be an intersting experiment to set it up as a trial. if it happens I'll summarize my findinds and share them Jan 28 13:53:19 erbo: I'm currently working on a setup using Backblaze B2 ($5/TB/month storage) for storing sstate cache, downloads and release artifacts. Builds are done in throw away containers Jan 28 13:53:38 I'll probably be writing that up as a blog post after I get back from FOSDEM & travel Jan 28 13:55:21 paulbarker: nice! what are you using to trigger the builds? Jan 28 13:55:59 is it normal that the kernel gets build with a different toolchain compared to the rest? i've got in build/tmp/work: all-oe-linux armv7ahf-neon-oe-linux-gnueabi overo-oe-linux-gnueabi x86_64-linux (not suprised about the last one, but shouldn't it be only one for the target?) Jan 28 13:56:36 erbo: GitLab CI. The runner is on a physical server and uses Docker to execute jobs. In principal the same idea can be used regardless of CI implementation Jan 28 13:56:46 smurray: Debian's reproducibility work assumes you have the "same environment": https://wiki.debian.org/ReproducibleBuilds/Howto#Introduction Jan 28 13:57:16 The trick is to use Cloudflare with Backblaze B2 so that download bandwidth isn't charged Jan 28 13:57:32 cunning Jan 28 13:59:51 paulbarker: that sounds really interesting, looking forward to blog post. Where do you blog btw? Jan 28 14:00:14 erbo: Nowhere right now, I'll figure that out later Jan 28 14:00:39 :) Jan 28 14:00:57 paulbarker: cool looking forward to the writeup Jan 28 14:01:28 paulbarker: does backblaze let you expire objects based on last access etc automatically Jan 28 14:01:53 rburton: No, sadly not. You can expire from initial upload date if you want to Jan 28 14:01:57 shame Jan 28 14:02:25 I like to ensure rebuilds are possible without sstate so I plan to clean sstate once per month anyway, so it's no issue for me Jan 28 14:02:36 would be good if it self-pruned based on last access: prune everything not accessed for a month and then do weekly builds of everything you support Jan 28 14:02:45 fair enough Jan 28 14:56:51 crazy_imp: that's not the toolchain used to compiled. It's just to separate non-arch-specific (shell scripts), arch-specific for target, arch-specific for host, and machine specific (and some for multilib also but you have none at the moment) Jan 28 14:58:39 RP: re: the "virus in OE" mails, I think they were talking about actual processes and not modules or driver from the kernel. I guess they now have more init scripts or something that are run automatically at boot since they upgraded? Anyway Jan 28 15:01:42 Could we add custom dts files directly to kernel sources with simple 'cp' ? Because of (cp ${WORKDIR}/zynq-xxx-2.1.dts ${WORKDIR}/git/arch/${ARCH}/boot/dts/) doesn't work... but patch is okay Jan 28 15:03:17 qschulz: ah, so "arch-specific for target" explains why the kernel build dir is in there and "arch-specific for host" why the kernel headers go into it? Jan 28 15:12:39 qschulz: its hard to say exactly what they meant :/ Jan 28 15:13:27 Hi, i have two uboot defconfigs, i want to build both of them: UBOOT_CONFIG ??= "sd spi". But now it writes the SPI build into my wic file... How can i chose one Jan 28 15:17:21 JPEW: I've a series to remove makeinfo from hosttools Jan 28 15:17:37 never liked that being there in the first place Jan 28 15:25:28 GeneralStupid maybe UBOOT_CONFIG = "sd spi" Jan 28 15:27:25 RP: Cool! I'll look it over Jan 28 15:29:33 PinkSnake: youre right, it looks like the last one is taken for wic creation Jan 28 15:32:13 GeneralStupid more info -> https://elinux.org/Bitbake_Cheat_Sheet Jan 28 15:39:08 PinkSnake: thankyou Jan 28 15:57:32 JPEW: interesting, I wonder how reproducible the Debian packages end up in practice, i.e. can I easily set up a machine at home, build one and manage to get the same hash at the end Jan 28 15:57:49 smurray: I think thats the idea Jan 28 15:58:47 JPEW: heh, of course ;) It's enough of a time investment to try it that I've not bothered yet to satisfy my curiosity Jan 28 16:01:56 RP, maybe bitbake should warn or error when inheriting the same class multiple times Jan 28 16:04:26 crazy_imp: the kernel headers in x86_64-linux? that does not look right to me. Everything that is supposed to be run on the host (native recipes) will be compiled in that directory Jan 28 16:04:59 no, headers are in overo-oe-linux-gnueabi Jan 28 16:05:24 RP: yeah I don't know, everything is so weird. Same for the one with the NAND issue, it's digressing in many different ways. Good luck with those :/ Jan 28 16:05:33 crazy_imp: yes, that is arch-specific for the target Jan 28 16:06:42 crazy_imp: read it too quickly... Jan 28 16:09:18 crazy_imp: the important point is that it shoulnd't matter to you where the recipes is being built (outside of the obvious target vs host) Jan 28 16:12:30 qschulz: just feared that there's something wrong and it used different toolchains to build stuff Jan 28 16:14:31 crazy_imp: nope, you can check that by reading the ${WORKDIR}/temp/(run|log).do_compile Jan 28 16:17:11 YPTM: Minutes at https://docs.google.com/document/d/1ly8nyhO14kDNnFcW2QskANXW3ZT7QwKC5wWVDg9dDH4 Jan 28 16:42:34 dreyna: I can't edit, but the question about videos should say "OE Workshop" not "YP Summit" Jan 28 16:43:34 jonmason - fixed, thanks Jan 28 16:43:58 tlwoerner: I'll post links to social media (and probably mailing lists) about the OE Workshop videos as soon as we figure out where it'll be hosted, etc Jan 28 16:49:46 RP: I did a last AB run last night, all green (for the things I can fix) for v5.4. I can send the series, but not bump the defaults. Or will that be a distraction with the reproducibility issues ? Jan 28 16:51:49 zeddii: no, feel free to send Jan 28 16:51:58 zeddii: I saw a multilib failure? Jan 28 16:52:32 zeddii: "INIT: PANIC: segmentation violation! sleeping for 30 seconds." Jan 28 16:52:37 zeddii: that sounds kernel? :/ Jan 28 16:53:10 not sure. I've never been able to see that one here either. so there's that and the 32bit musl one. Jan 28 16:53:28 zeddii: right, that segfault on musl doesn't sound good either :( Jan 28 16:53:50 is that multilib one 32 bit as well, I'm still not very good and digging out the exact config. Jan 28 16:53:51 zeddii: we've never seen those outside the kernel change Jan 28 16:54:15 ahah. that one is mips64 for multilib. Jan 28 16:54:25 * zeddii still wants to kick that out the door. Jan 28 16:54:27 zeddii: yes, mips64 Jan 28 16:54:45 zeddii: maybe victor could help? Jan 28 16:55:10 yah. I'll contact them. Jan 28 16:55:12 jonmason: awesome. last year Crofton|road set up an "events.openembedded.org", maybe we could give that a poke? Jan 28 16:58:05 * tlwoerner wonders which TI SoC denix was referring to that needs a 32-bit SPL but a 64-bit (full) U-Boot? Jan 28 17:00:45 tlwoerner: TI announced a new K3 family few years ago and there are multiple SoCs in that family already - e.g. AM65x, J7, etc Jan 28 17:22:52 tgamblin: that professor finally got back to me, he says he hasn't taught YP/OE specifically in any of his classes :-( Jan 28 17:23:26 tgamblin: i know he has taught (and probably put together) a "software build and release" course, and i was under the impression YP/OE was part of that course in some way Jan 28 17:23:31 https://www.senecacollege.ca/ssos/findwithoutsemester/sbr600/sict Jan 28 17:26:49 tlwoerner: The outline seems like a good place to start, though. I wonder if I can get some interest at Carleton U Jan 28 17:27:30 tgamblin: carletonU has *the best* motto of any U in the world! Jan 28 17:27:56 Carleton: where the "k" stands for "quality" (it works better heard than read) Jan 28 17:28:14 lol Jan 28 17:28:37 lol Jan 28 17:29:11 I can only really speak to graduate studies in the systems/computer engineering department Jan 28 17:29:36 Its reputation is therefore mostly unknown to me :) Jan 28 17:31:04 tgamblin: i went to RyersonU myself, but this professor, Chris Tyler, does a really good job teaching Open Source things at Seneca Jan 28 17:34:05 back in the days before RedHat made ARM an "official" port, Chris' department was doing the Fedora ARM builds: https://wiki.cdot.senecacollege.ca/wiki/Fedora_ARM_Secondary_Architecture Jan 28 17:36:00 tlwoerner: sounds like he is pretty hands-on with the subject matter, then. My immediate concerns with taking the idea of a Yocto-centric course to the local schools is that a professor with no background/interest might end up trying to teach it Jan 28 17:36:53 tgamblin: then *you* should teach it! :-D Jan 28 17:36:54 s/concerns/concern/ Jan 28 17:37:23 If only! Jan 28 17:39:14 pfft. too many of my university profs had "no background/interest" in several of the courses i took! Jan 28 17:39:52 my 2nd year "data structures in C" course was taught by someone who didn't even know C (never mind data structures in C)! Jan 28 17:49:13 tlwoerner: Yeah, that happens unfortunately. I can remember a few instances of that. Ironically (maybe), I've seen some computer science courses be more practical than those listed as software/systems/computer engineering... Jan 28 17:57:17 New news from stackoverflow: Is the bitbake immediate assignment behavior expected? Jan 28 18:30:01 RP, is that really expected bitbake behaviour, described in the SO question above ? Jan 28 18:31:08 a little confusing for someone used to how gnu make handles immediate expansion Jan 28 18:34:43 kroon: it is how the system behaves, I'm not surprised by it Jan 28 18:34:59 kroon: not sure I'd say it was designed to do that but it is the way it works Jan 28 18:35:49 RP, ok Jan 28 18:36:55 kergoth: you'd probably roll eyes at that one :/ Jan 28 18:37:11 kergoth: keep wondering if we should "fix" this somehow Jan 28 18:37:14 RP, kergoth, maybe its worth mentioning something in the bitbake manual Jan 28 18:37:25 kroon: happy to have patches Jan 28 18:40:09 RP, I guess the difference compared to gnu make is that if FOO is undefined and the time of parsing the immediate expansion line, ${FOO} is preserved as it is instead of evaluating to nothing ? Jan 28 18:41:01 kroon: correct Jan 28 19:08:40 If I have a package which is a native library but also produces non-native artifacts, is it possible to have one part of it packaged as "allarch" ? Jan 28 19:14:51 fullstop: there is something called allarch.bbclass, don't know if that helps tho. Jan 28 19:17:41 RP, I sent a patch, i was a little lazy and didn't update the example code, but the example code is b0rked anyway so.. Jan 28 19:18:48 nowadays bitbake even throws an expansion error when a variable references itself Jan 28 19:19:15 regardless of wether its an immediate or lazy assignment Jan 28 19:20:01 kroon: immediate should work I'd have thought? :/ Jan 28 19:20:43 * RP probably needs to add this to the test suite Jan 28 19:21:33 RP, hmm. not according to my testing on master branches.. Jan 28 19:21:59 A := "A is ${A}" Jan 28 19:22:05 gives me an expansion error Jan 28 19:22:43 # expansion of A threw ExpansionError: Failure expanding variable A, expression was A is ${A} which triggered exception Exception: variable A references itself! Jan 28 19:30:40 How should I be fetching using go? There isn't a go fetcher, so should I just download in the do_configure section? Jan 28 19:32:51 Nathan22: use go-dep perhaps Jan 28 19:40:43 khem: What does dep do vs go get Jan 28 19:42:07 Bad question sorry. I came across this: https://github.com/golang/dep/blob/master/docs/FAQ.md#does-dep-replace-go-get which -- But is it best practice to use dep vs go get? Jan 28 19:43:27 After trying to simply go get, I'm now running into cert issues Jan 28 19:45:45 Nathan22: dep should perhaps resolve the deps automatically Jan 28 19:45:51 what kind of cert issues Jan 28 19:47:41 'unable to access ... error setting certificate verify locations' Jan 28 19:58:14 Go seems to be looking for an ssl cert in a directory that doesn't exist Jan 28 20:03:10 Nathan22: hmm ca-certificates ? Jan 28 20:03:25 Yeah Jan 28 20:03:47 Searching the environmental variables the path is referenced by SDKPATH and SDKPATHNATVIE Jan 28 20:35:38 roussinm: yes, i was hoping to mix allarch in Jan 28 20:45:05 fullstop: if you look inside insane.bbclass there is a specific check that if you have a recipe that inherits from allarch, you can't packaged architecture specific binaries... Jan 28 20:45:45 fullstop: package_qa_check_check function. Jan 28 20:46:28 so it is not possible Jan 28 20:47:57 It feels like it... maybe there is some trickery possible with overrides. Maybe a maintainer can help you here. Jan 28 21:16:13 fullstop: you mean you want a recipe that produces at least two packages, one e.g. for ARM and the other allarch? Jan 28 21:17:12 rburton: yes. In this case, a shared library is created and some Lua code is generated from one of the header files using the pre-processor. Jan 28 21:17:22 you can do that of course Jan 28 21:17:52 I'd like the have the Lua pieces be allarch and the shared library be.. whatever arch the recipe decides. Jan 28 21:18:12 i'd have to ask *why* Jan 28 21:18:27 debian does that to save gigabytes of storage on their archives Jan 28 21:18:53 but unless you're building for ten architectures and maintaining a feed, there's zero actual point Jan 28 21:19:04 In my case I need to be able to distribute them separately. Jan 28 21:19:27 tgamblin: https://autobuilder.yoctoproject.org/typhoon/#/builders/101/builds/403 :/ Jan 28 21:19:28 anyway as allarch.bbclass says, PACKAGE_ARCH="any" is what sets all packages to be agnostic Jan 28 21:19:43 so PACKAGE_ARCH_${PN} might work Jan 28 21:19:50 RP: were you not riding? Jan 28 21:20:25 Let me try PACKAGE_ARCH_${PN}-subpkg="ANY" and see what happens. Jan 28 21:20:27 rburton: no, meetings finished too late :( Jan 28 21:22:14 RP: see my private chat on the other server then Jan 28 21:23:26 nope, I'll just live with allarch stuff being set to arm for now. Jan 28 21:23:34 for this particular recipe Jan 28 21:24:21 Thanks for the suggestion, though, rburton. Jan 28 21:28:11 fullstop: 'any' not ANY Jan 28 21:28:23 fullstop: if it doesn't work, file a bug Jan 28 21:29:03 Should it work with the "sub-package" like that? Jan 28 21:29:22 Also, I am on warrior if that matters. Jan 28 21:30:43 fullstop: maybe its a bug, maybe its a feature request. either way a bug is good Jan 28 21:35:06 RP: ParseLogsTest failing due to my changes? Or am I misreading it? Jan 28 21:37:04 tgamblin: failure with your changes in the build... Jan 28 21:37:23 tgamblin: not sure quite what I do with that! Jan 28 21:38:49 RP: Might need an extra step in the tests to change permissions on the log file. I'll focus on trying to figure out why those particular images were doing it prior to my v3, though Jan 28 21:42:33 tgamblin: ok, thanks Jan 28 21:43:48 tgamblin: hmm, second one: https://autobuilder.yoctoproject.org/typhoon/#/builders/109/builds/402 Jan 28 21:46:54 RP: alright, v3 is a scratch. Out of curiosity, how would I be able to look at the log files from the builders if I need to? Are they saved anywhere? Jan 28 21:47:45 RP: e.g. Log: /home/pokybuild/yocto-worker/qemux86-64-alt/build/build/tmp/work/qemux86_64-poky-linux/core-image-sato/1.0-r0/target_logs/dmesg_output.log Jan 28 21:50:57 tgamblin: they are preserved until the next build Jan 28 21:51:21 tgamblin: halstead can help you have a look at these failed builds Jan 28 21:51:35 RP: alright, thanks Jan 28 21:51:55 tgamblin: we need to stop another build working on this worker though so need to pause it asap Jan 28 21:56:23 RP: k. I can work without those logs for now Jan 28 21:57:06 Odd that I still can't recreate it locally. Maybe I should try that again, too Jan 28 21:57:52 RP, tgamblin I've paused ubuntu1804-ty-3 to ensure the qemux86-64-alt build isn't removed. Jan 28 22:46:46 Just curious -- I always understood Yocto/oe to be building a complete set of sandboxed native tools and runtimes for the host -- is this not true? I am able to build gn-native on Ubuntu 16.04 only if I install gcc-9 on the host machine otherwise it complains about missing c++17 STL libraries. Jan 28 22:47:23 See https://github.com/OSSystems/meta-browser/issues/353 Jan 28 22:49:27 jpsalm: native recipes are built using the host compiler Jan 28 22:50:15 but we do now support buildtools tarballs which ship a new gcc for that situation Jan 28 22:51:43 Interesting, thanks Jan 28 22:54:51 only in master Jan 28 22:56:30 yeah when i say "now" i mean "in the last two weeks" Jan 28 22:56:42 but that's the background and a possible solution Jan 29 00:13:24 * RP suspects he's broken meta-gplv2 again :( Jan 29 00:14:49 Uh oh Jan 29 00:20:23 RP: what are the actual builder machines? A cluster of core2duo boxes or something? Jan 29 00:22:31 ls Jan 29 00:53:27 tgamblin, I think there a few different flavors as we just refreshed some. MHalstead would know Jan 29 00:53:40 the specifics Jan 29 01:17:37 tgamblin, It's a blend of Xeon(R) CPU E5-2697 v3, E5-2670 v3, and one Huawei TaiShan 2280 Jan 29 01:18:00 halstead: armpit: ah, thanks Jan 29 01:34:25 halstead: any chance that builder is still accessible? Jan 29 02:59:43 tgamblin, It is still there. Just need to finish adding your ssh access. **** ENDING LOGGING AT Wed Jan 29 02:59:57 2020