**** BEGIN LOGGING AT Thu Oct 19 03:00:02 2017 Oct 19 06:27:17 Is it possible to create an image with grub installed in it? So that i can just dd it on a stick and it will have grub on it? Oct 19 07:09:27 Is this channel dead? :< Oct 19 07:10:54 possible: yes, of course. is there a wic script already prepared for this: not sure, might be. Oct 19 07:11:23 channel dead: certainly not. but YP has crunch time right now, and its too early for the USians to be awake Oct 19 07:12:25 Well true, i'm in the "wrong" timezone for questions :P Oct 19 07:12:56 have you already looked at the wks scripts that come along with poky? Oct 19 07:13:07 what do you mean by "wic script"? Oct 19 07:14:53 http://www.yoctoproject.org/docs/2.3/dev-manual/dev-manual.html#create-wic-images-wic Oct 19 07:15:50 Interesting. I did not hear about this yet but i am a novice in bitbake... Oct 19 07:16:11 Just to understand: I need a wks file and then wic can create my image? Oct 19 07:16:25 its rather new, right. but basically the way for you to go. yes that *should Oct 19 07:16:39 *should* be possible from my understanding Oct 19 07:16:46 Awesome! Thanks a lot! http://www.yoctoproject.org/docs/2.3/dev-manual/dev-manual.html#using-a-provided-kickstart-file I think this is already exactly what i need Oct 19 07:17:07 if there is a matching kickstart file already around, all is well Oct 19 07:18:43 One last question: Do i place the wks file in the build/conf/ folder nex to the local.conf? Oct 19 07:20:35 no idea, i know it exists but i don't use it :-P Oct 19 07:20:58 Haha ok :P I guess i'll just figure it out Oct 19 07:22:53 people usually put it in wic folder Oct 19 07:53:34 wic tells me Error: Couldn't find HDDDIR, exiting. My receipt only generates ext4 image. How do i enable the hddimage? Oct 19 08:04:44 styler2go: which wks do you use ? Oct 19 08:08:30 I tried mkefidisk Oct 19 08:10:18 Maybe i understand it wrong. I am using wic create mkefidisk -e Oct 19 08:15:26 https://bugzilla.yoctoproject.org/show_bug.cgi?id=10501#c7 Here he fives two possible fixes, i would like to use the first one, enable hddimg. Does anyone know how i enable hddimg again? Oct 19 08:15:27 Bug 10501: normal, Medium, 2.3 M2, eduard.bartosh, RESOLVED WORKSFORME, do_image_wic fails with "Error: Couldn't find HDDDIR, exiting" Oct 19 09:04:16 hey, I'm trying to package a svn repository that has externals pointing to various `trunk/`s. So bitbake's `svn up` won't see a change. Thus bitbake will reuse the cached source. Oct 19 09:04:47 is there a way to force bitbake to *not* use a cache for a specific package/recipe? Oct 19 09:05:07 T_UNIX: look at what externalsrc does, i'm sure it cn be beaten into shape Oct 19 09:05:19 (at the cost of reproductibility, of course.) Oct 19 09:07:54 LetoThe2nd: thanks. Sounds good. So it won't try to unpack from a previous fetch? Oct 19 09:08:28 T_UNIX: externalsrc basically is a "cache nothing, always rebuild" as far as i get it Oct 19 09:12:19 LetoThe2nd: okay. Cool, thanks. I was just wondering whether the embedding process would be like: "well, you didn't change the recipe, nor the version. So I'll just handle the unpacked cache." Oct 19 09:19:27 Do you know if there will be an linux-fslc-imx_4.9 ? Oct 19 09:19:46 Will wic rebuild the hddimg if there are changes? Oct 19 09:46:24 styler2go: if you put IMAGE_FSTYPES = "wic" WKS_FILE = "mkefidisk.wks" it should do it every time i think Oct 19 09:58:25 nayfe, that's awesome, thank you Oct 19 10:00:05 styler2go: np, if i can help ;) Oct 19 10:01:24 My goal is to create a minimal image completely without busybox, i am using meta/recipes-extended/packagegroups/packagegroup-core-full-cmdline.bb for that Oct 19 10:03:37 minimal without busybox sounds kinda contradictional Oct 19 10:03:51 Well, as minimal as it can get without busybox Oct 19 10:03:52 unless you're not replacing busybox with toybox or such Oct 19 10:04:11 I am not dependend on disk space, only in functionality "minimal" Oct 19 10:05:21 then all depends on your defintion of functionality. mind, there is actually no need to have an init system, libraries,.. you don't even need a shell Oct 19 10:45:15 LetoThe2nd, isn't "sh" the minimum dependency? Oct 19 10:45:56 no, why? Oct 19 10:47:04 I am trying to create a system which just executes one sh script Oct 19 10:47:35 well then your usecase probably depends on sh. but technically, you can totally have a linux system without any kind of sh Oct 19 10:48:59 That's interesting. You could run a kernel without any kind of "program" in userspace? Oct 19 10:49:26 that too, but that would require patching the kernel. Oct 19 10:49:40 Oh i see Oct 19 10:49:47 but even without that, have a single statically linked binary, and init=$YOURBINARY Oct 19 10:49:50 done. Oct 19 10:52:10 Nice Oct 19 10:52:35 But then i need my own wic script do i? Or is there a VAR für the wic GRUB kernel command? Oct 19 10:52:54 given the assumption that your binary has no external dependencies whatsoever, no device nodes, no nothing Oct 19 10:53:20 in such a case, one probably would not use OE at all Oct 19 10:55:08 what is the total usecase of this "product" anyways, if one might ask? Oct 19 10:55:37 just "run one script" sounds like the script will have a lot of dependencies anyways. Oct 19 10:56:17 This very minimal linux was just of theoretical knowledge if it WOULD be possible. Oct 19 10:56:27 Yes, my script has a lot of dependencies Oct 19 11:23:44 I am now trying to run my image but it's behavior is very weird, any idea what could make it that weird? https://p.styler2go.de/9458113 (I am just trying to enter ls+enter) Oct 19 11:24:58 is this some serial console? Oct 19 11:25:24 This is a record of virtualbox window Oct 19 11:25:35 Same behavior when i run it on a device though Oct 19 11:27:00 Might be a wrong inittab? Oct 19 11:27:10 no idea, ad hoc. it seems to toggle between sh and bash, for reason i don't have a clue Oct 19 11:40:08 nrossi: I don't exactly understand "HAS_PLATFORM_INIT" .. it has some names, such as "zynq_microzed_config". And the only corresponding value I can find is the UBOOT_MACHINE in the machine config (/conf/machine/<...>.conf) .. but I don't see how it corresponds to recipes-bsb/platform-init/* Oct 19 11:42:24 JoiF: HAS_PLATFORM_INIT is for u-boot, it defines that a specific config has platform init files in the u-boot source tree Oct 19 11:43:41 JoiF: if your using platform-init recipe that provides virtual/platform-init, you don't want to have your config in HAS_PLATFORM_INIT. If it is though (e.g. your using a existing u-boot config but want to provide a custom init) you can use FORCE_PLATFORM_INIT = "1" and it will be ignored Oct 19 11:43:58 nrossi: Ahh. e.g. ./tmp/work/mud_zynq7-poky-linux-gnueabi/u-boot/1_2017.01-r0/git/board/xilinx/zynq/zynq-microzed/ps7_init_gpl.c Oct 19 11:44:04 JoiF: yep Oct 19 11:44:34 nrossi: So, since I have my own, I should actually NOT put my board-thing to HAS_PLATFORM_INIT? Oct 19 11:44:42 JoiF: Correct Oct 19 11:45:38 JoiF: HAS_PLATFORM_INIT is just a limitation of not being able to know the content of the source tree at parse time Oct 19 11:47:51 But isn't there anything is should do to tell it that it has to pull my own platform-init into the build? Or is that done by default? Oct 19 11:48:27 When I say "pull my own", I mean use the one I'm supplying in my recipes-bsp/platform-init/... Oct 19 11:49:12 JoiF: the platform-init recipes PROVIDES virtual/platform-init, u-boot depends on that if you set SPL_BINARY = "spl/boot.bin" and its zynq(mp) Oct 19 11:49:59 (actually any SPL_BINARY value triggers the depends) Oct 19 11:50:58 Ok. It should work, then. Oct 19 11:51:08 (since that's what I'm doing.) Oct 19 11:51:52 Btw, Anyone here who'll be in Prag on Monday? :p Oct 19 11:52:38 and on sunday, and on tuesday, and on wednesday, and on thursday... Oct 19 11:53:01 Yay :D Oct 19 11:53:32 \m/ Oct 19 11:53:52 I'm actually going tomorrow .. I'll be staying until Sunday, 29th Oct 19 11:53:53 if you see a weirdo with beard wearing a heavy metal shirt, thats me. Oct 19 11:54:16 you're at OEDEM then too? Oct 19 11:54:39 Haha. You mean.. like everyone at the conference?? ;) Isn't that the stereotype? ;) Oct 19 11:55:04 actually, not. seems like elce is really getting close to the classic industry. Oct 19 11:55:15 No, just ELCE. But my friend lives in Prag and I'm staying with him .. and seeing more of.. Prag. ;) Oct 19 11:56:23 a shame Oct 19 11:57:00 oedem usually has a pretty good discussion culture Oct 19 11:58:19 09:00 on a Sunday?? Are you guys insane?? ;) Oct 19 11:58:40 nobody said we have to show up sober. Oct 19 11:58:46 Haha Oct 19 11:59:06 and insane... yep, probably. Oct 19 12:01:40 It looks interesting. But I don't know if I should commit to showing up there in a full capacity. This isn't a thing were you just drop in and sit quietly at the back and listen in, right? Oct 19 12:04:24 Naahh.. I should probably stay away from it. I think this is all in quite capable hands ;) Oct 19 12:04:34 usually its not for the passive kind, yes. Oct 19 12:05:07 and its rather full AFAICS anyways. but i thought i'd ask as i have totally no clue if i already know you or not. Oct 19 12:05:07 My only "problem" with OE is that my company's mail-spam-filter blocks every other message that comes from the mailing lists. And that's not OE's fault. ;) Oct 19 12:35:00 Haha, I understand :) Oct 19 12:58:04 Is there some "easy" way to add files to my final image? Oct 19 12:58:38 styler2go: the kinda coprrect way is http://www.yoctoproject.org/docs/current/dev-manual/dev-manual.html#packaging-externally-produced-binaries Oct 19 13:02:47 styler2go: write recipe to package files, put package in image Oct 19 13:03:02 hmmm ok Oct 19 13:03:39 e.g. RTFM behind given link :) Oct 19 13:03:40 and then i do_install { install -m 0755 } Oct 19 13:06:10 yes, i will. thank you :) Oct 19 13:41:25 Is there a way to exclude a layer or a bbappend depending on which machine I'm building for? Oct 19 13:42:31 tasslehoff: you can certainly limit the effects of the bbappend to specific machines, with the postfix identifiers (orwhattheyrecalled) Oct 19 13:44:04 LetoThe2nd: I have to build for a variscite board and for sabre. The bluez5_%.bbappend in meta-variscite-fslc demands a variscite-bt.conf no matter what I build. Oct 19 13:44:22 I want to see if I can work around it in my own layer :) Oct 19 13:57:33 variscite has a u-boot-fw-utils.bb, and there is also a u-boot-fw-utils_2017.01.bb. even if I set a preferred version for my machine, the variscite one is chosen. Oct 19 14:15:17 Hm. By some magic the version actually becomes 1:v2017.03+gitAUTOINC+b3e7f7518b Oct 19 14:15:29 Probably when including uboot-config Oct 19 14:31:21 tasslehoff: maybe a layer priority problem ? Oct 19 14:33:25 Sometimes bitbake is really frustrating :D Oct 19 14:33:41 s/bitbake/software Oct 19 14:33:45 Error: exec_cmd: cp /mnt/upboard/build-recovery/tmp/work/up_board-poky-linux/recovery-image/1.0-r0/recovery-image-1.0/hddimg/EFI/BOOT/* /mnt/upboard/build-recovery/tmp/deploy/images/up-board/recovery-image-up-board-20171019142129/build/hdd/boot/EFI/BOOT returned '1' instead of 0 Oct 19 14:33:50 Any idea what could cause this? Oct 19 14:34:01 (error message while creating wic) Oct 19 14:34:18 cp set errorcode 1 Oct 19 14:34:30 of course cp doesn't say what the error codes are, so that's no good Oct 19 14:34:32 ERROR: Function failed: do_image_wic Oct 19 14:35:05 styler2go: well its just a cp command. look at both parts of it and try to undertand what might have failed. Oct 19 14:35:18 path problem i think, as there is a build folder in image deploy folder Oct 19 14:35:26 -> is the source there, -> is the destination path there. Oct 19 14:35:32 oh well.. now it worked after just retrying it for the third time Oct 19 14:37:27 Also, this was an automatic process by wic so i don't know how i could check why it didn't have the file/folder/whatever it tried to copy Oct 19 14:47:34 nrossi: Not getting those "unable to halt processor" errors anymore, but I get nothing on my serial-console :/ Oct 19 14:52:30 JoiF: if you have jtag, check where the PC is, and see if its hung in spl Oct 19 14:56:30 nrossi: After loading u-boot.elf? Oct 19 14:57:17 JoiF: do you get output from SPL? or is it U-Boot that is not doing anything? Oct 19 14:57:17 After I load/download u-boot-spl.bin (which I dd from boot.bin) there's nothing being printed Oct 19 14:57:31 No, no output at all Oct 19 14:58:07 JoiF: then that is where the failure is Oct 19 15:00:08 pc is at: 00007ddc, 00007de8, 00004a54, 00007de4 .. those are the addresses I see if I continually do "rrd pc" .. it's jumping between something around those parts. But at this point, I haven't loaded u-boot.elf yet. Oct 19 15:00:43 If I load u-boot.elf it stops at 00000040 Oct 19 15:01:20 JoiF: thats a exception vector, u-boot-spl has hung by the looks of those addresses Oct 19 15:05:01 JoiF: looking at a u-boot-spl binary here, 0x7xxx might be serial functions? Oct 19 15:06:16 Got something now Oct 19 15:06:57 Now it hangs after displaying "Trying to boot from RAM" Oct 19 15:09:05 JoiF: thats normal for a jtag boot, since there is nothing to boot from. You should be able to load u-boot.elf once you've seen SPL get that far Oct 19 15:10:52 Yes. But still after loading u-boot.elf nothing happens. Oct 19 15:10:57 I suspect my memory config Oct 19 15:15:07 Hmm, nope Oct 19 15:15:16 (as far as I can tell) Oct 19 15:20:34 Trying to boot from RAMNo valid device tree binary found - please append one to U-Boot binary, use u-boot-dtb.bin or define CONFIG_OF_EMBED. For sandbox, use -d initcall sequence 04061b90 failed at call 04046a08 (err=-1)### ERROR ### Please RESET the board ### Oct 19 15:22:19 JoiF: hmmm, is this u-boot.elf? and you have OF_EMBED=y? Oct 19 15:23:09 I do now. And now u-boot finally works. ;) Oct 19 15:23:31 JoiF: :) Oct 19 15:23:35 *high-five* Oct 19 16:32:40 hi ppl, I have a question regarding building python recipes, so when I run `bitbake python-pyopenssl` I get an error like ERROR: Nothing RPROVIDES 'cpp' (but /builds/release/meta-openembedded/meta-python/recipes-devtools/python/python-pycparser_2.17.bb RDEPENDS on or otherwise requires it) Missing or unbuildable dependency chain was: ['python-pyopenssl', 'python-cryptography', 'python-cffi', 'python-pycparser', 'cpp']. Oct 19 16:32:58 It seems that I need to add cpp somewhere but I am not sure how to do this Oct 19 16:35:55 stavro: I think gcc should provide cpp. What happens when you do 'bitbake cpp'? Should show gcc RPROVIDES cpp. Oct 19 16:37:26 laplante: I get ERROR: Nothing PROVIDES 'cpp'. Close matches: mcpp Oct 19 16:38:08 Hmm seems like your toolchain isn't set up properly then? Oct 19 16:40:21 stavro: What happens when you do 'bitbake gcc'? Oct 19 16:42:33 laplante: I get ERROR: Nothing PROVIDES 'gcc' ERROR: gcc was skipped: it has an incompatible license: GPL-3.0-with-GCC-exception & GPLv3. Hmm that might be the issue... Oct 19 16:49:16 Yeah, you'll have to come to terms with that particular instance of GPL-3.0 or use clang instead. Oct 19 16:53:56 rburton, or RP do either of you going to run builds today/tonight ? Oct 19 16:54:51 armpit: I'm not Oct 19 16:55:01 k Oct 19 17:13:26 rburton: youre the maintainer for initramfs-live-boot, any reason why it RDEPENDS on udev instead of ${VIRTUAL-RUNTIME_dev_manager}? Oct 19 17:15:11 rburton: my point is what if I dont want udev but busybox-mdev instead Oct 19 17:30:19 Can anyone provide some tips on tracking down the source of "do_rootfs: libevdev-dev not found in the feeds"-style errors during image generation? I have confirmed that the RPMs exist (in my case under deploy/rpm/cortexa9hf_neon), and are not empty (as verified with packages_split). At a loss on how to proceed... Oct 19 17:31:38 aehs29: ha, wasn't aware i was ;) the live-boot code uses udevadm etc, so it will expect those commands to exist Oct 19 17:33:01 rburton: I use it for poky-tiny, and eudev is bringing about 600 extra KB which is about 45% of the whole userspace anyway Oct 19 17:33:51 rburton: So I think I might have to create an init-live-tiny recipe Oct 19 17:37:13 rburton: wait but systemd-udev does provide udevadm as well, why wouldnt it be possible to have initramfs-live-boot + systemd? Oct 19 17:43:40 aehs29: systemd rprovides udev (you can't build eudev and systemd at the same time) Oct 19 17:45:14 rburton: I see your point, any other suggestions besides creating an init-live for tiny? Oct 19 17:46:07 aehs29: port init-live away from using udev? presumably not possible. do you really want *live*? they're nonsense. just boot properly. Oct 19 17:46:38 hands up who actually has write-once media to use live images on Oct 19 17:46:51 rburton: its not entierly live, its the tiny-initramfs image, it does work just fine using the same recipe with busybox-mdev instead Oct 19 17:47:23 rburton: since it actually drops to shell before anything else happens Oct 19 17:47:45 yeah. split the recipes up? Oct 19 17:50:02 rburton: yeah I dont see any other way thats less intrusive Oct 19 18:06:50 Regarding my do_rootfs failures, I ran with verbose logging and see a bunch of warnings like this (one for all, one for cortexa9hf_neon, and one for my machine arch): "NOTE: [my image name] do_rootfs: Previous repo file missing: [yocto project path]/build/tmp/work/[arch]/[my image name]/1.0-r0/rpms/cortexa9hf_neon/repodata/primary.xml.gz" Oct 19 18:07:15 Is it possible my image recipe is missing a dependency to build the package database(s) or something? Oct 19 18:20:45 Might have a lead: looks like my image recipe overwrote IMAGES_FEATURES which contained (probably) important stuff like ${ROOTFS_PKGMANAGE_BOOTSTRAP}, package-group-coreboot, ${COMMON_INSTALL}, etc... Also was using += instead of _append, contrary to what docs say. Oct 19 19:18:21 laplante: as it is a -dev package there should be a FILES_${PN}-dev adding some files, do you add ${PN}-dev to the image ? Oct 19 19:22:34 nefe: It's the built-in recipe (as of morty) in recipes-support. I don't see any FILES entries, but the packages-split directory shows files in it, so it must be autopackaged or something. The recipe does inherit from autotools. Oct 19 19:23:23 nefe: Also, I am seeing other packages besides -dev failing in the same way, e.g. shadow. In all cases I've checked so far, the packages exist (in tmp/deploy/rpm) and are not empty (according to packages-split) Oct 19 19:24:17 nefe: I appreciate your assistance, by the way :) Oct 19 19:47:56 laplante: maybe time to upgrade to rocko :) Oct 19 19:49:23 nefe: :) Yeah, I think I found the commit that fixes this: "package_manager.py: Generate correct RPM package names again", 01406f8bf69ade6370bcaebbfea4a8747758a152 Oct 19 19:49:45 nefe: "This restores the original functionality so that the Oct 19 19:49:46 installed-packages.txt files produced by inheriting buildhistory yet Oct 19 19:49:46 again contains a list of the names of the installed packages." Oct 19 19:57:58 laplante: et pas de soucis pour le "support" c'est tellement simple tout ça ;) Oct 19 20:03:50 nefe: Sorry to disappoint but I had to Google translate that. Despite my last name, I do not in fact speak French :/. And thanks again for the help! :) Oct 19 20:04:43 oops :) Oct 19 20:14:37 maybe one day :) Oct 19 20:25:44 rburton: halstead: patchtest is back Oct 19 22:26:14 garbados, Are you around for a question about http://git.yoctoproject.org/cgit/cgit.cgi/layerindex-web/tree/Dockerfile#n40 ? Oct 19 22:27:33 gabrbedd, Do 2 CMD instructions work in the same file now? I'm looking at https://docs.docker.com/engine/reference/builder/#cmd and it says only the last still. Normally I'd make a wrapper for the 2 commands. But I might be missing something. Oct 19 22:29:04 garbados, I'm looking at https://docs.docker.com/engine/reference/builder/#cmd and it says only the last will take effect.. Normally I'd make a wrapper for the 2 commands but I might be missing something. Oct 19 22:31:45 halstead: it looks to me as if they both should take effect - after all we have a CMD mkdir ... some lines above that Oct 19 22:32:17 * halstead checks again. Oct 19 22:32:27 halstead: if the gunicorn command blocks then that would prevent celery from starting though Oct 19 22:33:19 halstead: which it probably does :( Oct 19 22:37:03 looking at the docs I think this does need to be rewritten Oct 19 22:38:32 bluelightning, So in my production build I've added, "RUN mkdir /opt/workdir" because the earlier CMD is ignored. Oct 19 22:38:51 I'll go ahead and make the wrapper script. Oct 19 22:39:06 halstead: ok - if you could send me a patch that would be great Oct 19 22:39:08 thanks Oct 19 22:39:17 bluelightning, Will do. Oct 19 22:44:41 Can anyone please tell me if the external Linaro support in meta-linaro-toolchain is still supported? It seems to be missing some fundamental things (e.g. TCLIBC conf), and is broken in a variety of ways, at least for Linaro 7.1.1. I've fixed a lot of issues with it (and will submit patches), but still hitting issues. Oct 19 22:50:21 ndec: ^ Oct 19 23:05:55 Hello! Is there a way to check which recipe provides a particular package? Oct 19 23:06:04 Find and grep is not helping me Oct 19 23:06:31 bhargav: oe-pkgdata-util can help but only if the recipe providing the package you're looking for has already been built Oct 19 23:07:24 @bluelightning thanks, let me give that a spin Oct 19 23:32:38 Dear All, Oct 19 23:32:58 Is there any "autobuilded" toolchain sdk for armv4 ? Oct 19 23:33:11 just to download and install ? **** ENDING LOGGING AT Fri Oct 20 03:00:01 2017