**** BEGIN LOGGING AT Wed Aug 03 02:59:58 2016 Aug 03 07:06:51 hi, how to I load the layers I have entered in sources/layers.txt? Aug 03 07:11:12 or maybe there is a newer mechanism to download layers? Aug 03 07:27:15 morning Aug 03 07:46:32 hi, when someone is upgrading init-ifupdown with opkg, can one be sure that by the time opkg returns, the network interface is up again after getting brought down by the installation process? Aug 03 08:11:14 mornin' Aug 03 08:26:49 is there a way to stop the recipetool from trying to scan the license file? Aug 03 09:01:30 https://nopaste.xyz/?552e2bd468e51930#bcq9pvUUM8sA5rQdhVGxiWZv0lY9Bp5+jzITxcscVks= <- is that behavior intended? Aug 03 09:09:56 according to hexdump -n 1 the first byte is either 20 (for 5 of the licenese files or 54 for the commercial license) Aug 03 09:40:45 anyone got an example of a NOMMU image configuration? Aug 03 09:43:10 IIRC linux can't work without a MMU... Aug 03 09:44:49 boucman_work, but it can.... http://nommu.org/ Aug 03 09:53:41 boucman_work, I would like to know why you feel a kernel requires an mmu Aug 03 09:55:18 I don't "feel" that... I just remembered that a requirement to port linux to a new HW was that the HW needed to have a mmu Aug 03 09:55:29 I didn't know about an external patch to remove that requirement Aug 03 09:56:02 I have no idea if nommu is supported by yocto, or how much work it would be to support it... Aug 03 09:58:18 there was a fork, and it was merged into mainline linux, https://en.wikipedia.org/wiki/%CE%9CClinux Aug 03 10:00:46 though it looks non-trivial to link user-space applications for non-mmu kernels Aug 03 10:01:49 jkroon_: good luck Aug 03 10:02:33 "it is just software" :-) Aug 03 10:03:10 lpapp, wish redengin good luck, I'm not the one goin' down that road .. :-) although it would be interesting Aug 03 10:03:58 yes, it is challenging, I guess. Aug 03 10:04:26 * jubr has had a running distro on Blackfin, nommu, /w uclinux + uclibc Aug 03 10:04:57 * jubr also remembers the porting hell and data-aborts in a nightmarisch way :) Aug 03 10:05:46 jkroon_, yeah, it takes a specialization of the toolchain for placement independence Aug 03 10:05:56 that was back in 2008 tho... Aug 03 10:06:25 incidentally, that was about the time when I played around with uclinux on Blackfin. Aug 03 10:06:32 bah, what happened to linux on the toaster... maybe I'm just old Aug 03 10:10:17 jubr, was there a common pattern of the code causing aborts? Aug 03 10:17:40 Hi there! I've got a question : is Uboot included in the .jffs file that is inside my images folder ? My goal is to flash a new Uboot to a board where I added SDRAM - I used a different receipe following what the provider of the board said... Aug 03 10:25:48 redengin: I think it was related to unaligned mem accesses & uclinux not being able to trap those with workaround code. Too long ago :) Aug 03 11:40:56 Hi all. How can i specify some extra flags being supplied to CMake for a cmake based project in my recipe? Aug 03 11:41:34 I've found PACKAGECONFIG_CONFARGS and EXTRA_OECMAKE in the cmake.bbclass but i'm not sure which one is preferably used Aug 03 11:53:21 PACKAGECONFIG_CONFARGS is for options provided by various PACKAGECONFIG options Aug 03 12:20:21 is it possible to set a preferred version of a package for a particular machine? Aug 03 12:39:27 Doing a quick reboot of the AB cluster to update nfs mount units. Aug 03 13:00:59 Cwiiis: grep -r --color PREFERRED ../sources/*/conf/machine/ Aug 03 13:01:47 Cwiiis: grep -r --color PREFERRED ../sources/poky/meta/conf/machine/ (the other one only matched on Freescale's release) Aug 03 13:03:06 jubr: that's not exactly what I asked - are you saying you can only do it by having a machine-specific include? Aug 03 13:06:10 Hi there, I am having some problems splitting the debug binaries into the correct package. Aug 03 13:07:07 I modify my recipe to add a -ptest package. That works fine, but the test debug binaries are added to the recipe -dbg package, which is not what I wanted Aug 03 13:07:24 so I added a line like this: PACKAGES += "${PN}-ptest-dbg" Aug 03 13:08:32 now the test debug binaries end up on the -ptest-debg, but also the files that earlier went into ${PN}-dbg now end up in ${PN}-ptest-dbg as well Aug 03 13:08:53 now ${PN}-dbg only contains the src files Aug 03 13:09:33 I tried using FILES_${PN}-dbg += and FILES_${PN}-ptest-dbg += in my recipe but that didn't help Aug 03 13:38:16 (in case it's handy, PREFERRED_VERSION_package-name_machine is what I wanted) Aug 03 13:56:38 melonipoika: packages listed earlier in PACKAGES pick up files before later ones. if you add ${PN}-ptest-dbg to the beginning of PACKAGES, it will pick up all the files that match FILES_${PN}-ptest-dbg before any other packages are considered. i've never dealt with ptest packages myself, so not sure what the recommended practice is. maybe the tests are supposed to be stored in a separate path that does no Aug 03 13:56:44 t conflict. Aug 03 13:57:48 you could either set PACKAGES explicitly to an order that makes sense, or rearrange it with something like the following: Aug 03 13:58:01 PACKAGES_remove = "${PN}-dbg" Aug 03 13:58:26 PACKAGES_prepend = "${PN}-dbg ${PN}-ptest-dbg" Aug 03 13:58:38 urr... *PACKAGES_prepend = "${PN}-dbg ${PN}-ptest-dbg " Aug 03 13:58:57 but might still be problems. hopefully you understand the issue now at least. Aug 03 13:59:30 the glossary entry for PACKAGES doesn't seem to mention that earlier packages pick up files first. pretty bad omission. Aug 03 14:03:40 you'll find the default PACKAGES and FILES_* values in meta/conf/bitbake.conf btw Aug 03 14:08:18 Cwiiis: from Freescale: ../sources/meta-fsl-arm/conf/machine/include/imx-base.inc:PREFERRED_VERSION_directfb_mx6 = "1.7.4" Aug 03 14:09:01 _mx6 is the soc-part of a _${MACHINE} override var - does that answer your question? Aug 03 14:09:48 such vars can be set anywhere in your .conf Aug 03 14:11:00 jubr: yup, thanks Aug 03 14:11:26 Hi there! I've got a question : is Uboot included in the .jffs file that is inside my images folder ? Aug 03 14:12:28 Kakounet: Generally, uboot is not included in the rootfs. It doesn't really work that way. Aug 03 14:13:14 You have to create some kind of partition layout to put onto the raw flash. uboot usually goes at a specific address depending on your SOC. Aug 03 14:13:25 I'm a bit lost... I just want to flash uboot, but the doc I have is not really clear Aug 03 14:14:08 Ok, I saw that on the doc, it has to go on a specific address, i'll dig in that direction, thanks ! Aug 03 14:14:15 u-boot is the first thing your SOC will run. The SOC doesn't have hardware to traverse a JFFS2 file system. ;) Most of the time there is a hard-coded flash address you place u-boot at. Aug 03 14:16:08 Kakounet: https://github.com/openbmc/openbmc/blob/a244dcdcd45c0b55023095fb6cc3e5dcc0e090b6/meta-openbmc-bsp/meta-aspeed/meta-ast2400/conf/machine/include/ast2400.inc Aug 03 14:16:09 https://github.com/openbmc/openbmc/blob/2fe86d90044af218ced8f42fdded6b136f1046d2/meta-phosphor/classes/kernel-cuimage.bbclass Aug 03 14:16:15 Looks at UBOOT_LOADADDRESS Aug 03 14:16:38 This is the original Yocto bbclass that is related too: https://github.com/openbmc/openbmc/blob/2fe86d90044af218ced8f42fdded6b136f1046d2/yocto-poky/meta/classes/kernel-uimage.bbclass Aug 03 14:16:46 Thxs ! Aug 03 14:17:05 I am reading that (for my board) http://architechboards-hachiko-tiny.readthedocs.io/en/latest/board.html#flashing-nor Aug 03 14:17:18 0x18000000-0x18080000 spibsc0_loader (offset: 0x00000000) Aug 03 14:17:34 sounds like a correct address ? Aug 03 14:17:44 spibsc0_loader: contains u-boot (u-boot.bin) Aug 03 14:18:30 _loader would be u-boot, _bootenv would likely be the u-boot environment variables. Aug 03 14:18:35 It's just that it's written : Warning The operation is prone to failure, use it at your own risk. :( Aug 03 14:18:46 * Kakounet is scarred Aug 03 14:18:48 For instance, we store our MAC address into the _bootenv. Aug 03 14:18:59 Kakounet: i'd just like to point out that prior to tinkering with that rom, i'd make ABSOLUTELY sure you have some other means of booting the device, from sd card or something similar Aug 03 14:19:08 Kakounet: You have a way to flash the board even if it is bricked? Maybe read the whole flash off and save a copy. Aug 03 14:19:28 if you have a jtag probe or such, yes. Aug 03 14:19:44 I 2nd that. Even if you flash u-boot correctly it is possible your u-boot you compiled doesn't have all the drivers you need for this board. Aug 03 14:20:22 SOC vendors are notorious for not getting their u-boot hacks upstream... Aug 03 14:21:19 My team had to basically re-write all of the drivers for u-boot and the kernel for the AST2400 / AST2500 chips because the vendor supplied changes were for a 5 year old version of u-boot and Linux. Aug 03 14:21:21 I made all that with a virtual machine given by the provider of my board, so I think it's ok but yes, I will try other means of flashing uboot Aug 03 14:22:17 its just essential that you have some means of flashing u-boot without relying on a running and functional uboot itself. Aug 03 14:48:25 ubootception! :) Aug 03 14:50:44 * jubr remembers seeing a redboot that could start the n+1 version of itself from mem & then flash the running version so you limit the chance of bricking the device Aug 03 14:53:03 we currently work with i.MX6: they have a nice USB serial DL mode: talks directly to the ROM. Saved our hides a few times over :) Aug 03 14:56:06 jubr: I'm also working on an i.MX6 with Buildroot : for the newbie I am it's much more easier :) Aug 03 14:57:45 Buildroot? *shivers* Aug 03 14:58:05 You realize you are in the #yocto channel? :) Aug 03 15:00:00 jubr: most of us use both depending on customer needs... Aug 03 15:03:32 boucman_work: We are the customer in that respect - thanking the lord every day :) Aug 03 15:04:05 hehe Aug 03 15:45:17 anyone else have been having troubles with kernel panics on raspberry builds? Aug 03 16:00:58 if ${sbindir} is /usr/sbin, what variable (usually) represents /sbin ? Aug 03 16:01:43 ah, base_sbindir :) Aug 03 16:36:53 anyone fancy #mozgnt / #mozbeer / #mozwine in a little bit? Aug 03 16:58:13 I'm trying to understand what gets put into a regular package and what gets put into a -dev package. Are there any pointers on that? Aug 03 16:59:13 We are trying to reuse the autoconf-archive recipe and for some reason the /usr/share/aclocal contents are getting put into the nativesdk-autoconf-archive-dev and not nativesdk-autoconf-archive. The dots aren't connecting in my head. Aug 03 17:16:27 Hello Aug 03 17:16:34 I'm having a bit of an issue starting toaster locally. Aug 03 17:17:06 "source toaster start" fails with "command not found: python3 /home/jedix/installs/yp/openembedded-core/bitbake/bin/../lib/toaster/manage.py" Aug 03 17:17:20 however, running th command by hand works Aug 03 17:17:31 python3 /home/jedix/installs/yp/openembedded-core/bitbake/bin/../lib/toaster/manage.py checksocket "localhost:8000" Aug 03 17:49:08 hiya. I've just set up a layer for etnaviv that we juuuust got working with mesa and X on imx6 Aug 03 17:49:48 We have been consumers rather than participants in yocto up to this point Aug 03 17:50:13 Anything I should do to advertize the existence of this etnaviv layer? Aug 03 18:29:13 ah Aug 03 18:29:20 toaster didn't work because I use zsh Aug 03 18:47:46 stwcx: the value of FILES_${PN}-dev determines what gets put into the -dev package. you'll find the default value in meta/conf/bitbake.conf. Aug 03 18:49:14 for each package in PACKAGES, the patterns in FILES_ determine what files are put into that package Aug 03 18:50:14 packages that appear earlier in PACKAGES "win" when it comes to picking up files in cases where many patterns match the same file Aug 03 18:51:03 (which i want to add to the PACKAGES glossary entry, but i just submitted a bunch of stuff and don't want the documentation maintainer to feel like he's being spammed. :P) Aug 03 18:52:31 stwcx: you can also check out the value of FILES_${PN}-dev with 'bitbake -e ' Aug 03 18:53:04 ${PN}-dev appears before ${PN} in PACKAGES, so it will "win" if the files could be picked up by either one of those Aug 03 18:54:17 Ulfalizer: Thanks. Let me read and absorb what you wrote. Aug 03 18:59:09 Ulfalizer: I see 'FILES_${PN}-dev' contains ${datadir}/aclocal . So that's where it is coming from, thanks. Aug 03 18:59:55 no problem. i don't even know what aclocal is for to be honest. Aug 03 19:00:13 The autoconf-archive package is a collection of aclocal macros. Does it make sense to add 'nativesdk-autoconf-archive-dev' to the nativesdk package group, or trying to make the non-dev package have the aclocal stuff for this particular package? Aug 03 19:00:27 ah, some autotools thingy Aug 03 19:00:35 They are a collection of additional commonly used autotools macros. Aug 03 19:00:40 Yep. Aug 03 19:01:15 why can't they be in the -dev package? it seems they'd make most sense there. Aug 03 19:01:28 since those macros would only be needed for development Aug 03 19:01:49 you could still make sure to install the -dev package if you think the users would want those macros installed by default Aug 03 19:02:09 They can be and by default they are. That works great for 'DEPENDS_package = "autoconf-archive-native"'. Aug 03 19:02:48 What gave us trouble initially was that we added RDEPENDS = "nativesdk-autoconf-archive" to nativesdk-packagegroup-sdk-host.bbappend. Aug 03 19:02:52 Which ends up doing nothing useful. Aug 03 19:03:31 RDEPENDS = "nativesdk-autoconf-archive-dev" does work though, but it seems odd to have to add a dev package to the RDEPENDS of the SDK. Aug 03 19:05:42 stwcx: another option would be to do RDEPENDS_${PN}-dev = "nativesdk-autoconf-archive-dev" and adding my-packagegroup-dev to IMAGE_INSTALL (or whatever the right name for the nativsdk portion of the SDK is again) Aug 03 19:06:08 (the reason i disappeared for a while is because i can't remember the name of that variable. there's one for the host portion of the SDK, and one for the target portion. :P) Aug 03 19:07:07 TOOLCHAIN_HOST_TASK it is :) Aug 03 19:08:36 RDEPENDS_${PN}-dev += "nativesdk-autoconf-archive-dev" even. i just made the same mistake i added a warning for in http://www.yoctoproject.org/docs/2.2/ref-manual/ref-manual.html#var-RDEPENDS ... Aug 03 19:09:38 Do the dev packages get automatically installed in the SDK? Aug 03 19:10:36 nope, hence why you'd need to list e.g. "nativesdk-autoconf-archive-dev" in TOOLCHAIN_HOST_TASK if you want to install it Aug 03 19:11:29 Ulfalizer! Aug 03 19:11:32 i assumed you had some packagegroup that listed stuff you wanted to install. in that case that packagegroup could generate two packages "my-packagegroup" and "my-packagegroup-dev", and you could list those. Aug 03 19:11:38 mario-goulart: o.O Aug 03 19:11:53 mario-goulart: i should be drinking beer and stuff :P Aug 03 19:12:06 Yeah! :-) Aug 03 19:17:03 stwcx: btw, packagegroups can be overkill sometimes, imo. nothing wrong with just adding stuff directly to TOOLCHAIN_HOST/TARGET_TASK. it's still possible to split that up over multiple bbappends, or to add comments for groups of packages by doing TOOLCHAIN_FOO_TASK += "...." and putting a comment above that. Aug 03 19:17:45 packagroups is for when it's handy for the package manager on the target to treat those packages as a unit, but that might not always be needed Aug 03 19:18:21 that might be a personal opinion though :P Aug 03 19:19:22 * Ulfalizer has seen packagegroups nested within packagegroups nested within packagegroups :| Aug 03 19:35:07 Ulfalizer: My biggest concern with the TOOLCHAIN variable was I saw some classes that did ?= on them. I was worried about breaking something else. It looked like that package group was already predefined as a nice location to bbappend content into for our distro. Aug 03 19:40:55 stwcx: if you want to see how the TOOLCHAIN* variables currently get their values, you can run 'bitbake -e | less'. above the value for each variable, you'll see how it ended up with that value. might be helpful in figuring out what makes sense. Aug 03 19:41:41 can't see any classes setting TOOLCHAIN_(HOST/TARGET)_TASK directly in poky Aug 03 19:43:17 Yep. Aug 03 19:43:34 My general policy is to avoid setting anything that is ?= in poky unless I really know what I'm doing. ;) Aug 03 19:46:09 hrm... populate_sdk_base.bbclass does TOOLCHAIN_HOST_TASK ?= "..." do add some basic stuff it seems Aug 03 19:46:52 Yeah. Aug 03 19:47:11 So maybe TOOLCHAIN_HOST_TASK_append would allow it to still get those items? Aug 03 19:47:42 yep, that would work Aug 04 01:41:02 so I'm trying to figure out how to add udev, udev-extraconf and nodejs before running bitbake rpi-basic-image. Whats the best way to go about adding these packages to this image? Aug 04 01:54:41 r0r0: you could do IMAGE_INSTALL += "udev udev-extraconf" in rpi-basic-image.bb for example (or in one of the files it includes. i'm not familiar with that layer.) Aug 04 01:55:11 cool thanks I will give it a shot Aug 04 01:56:23 if you don't want to modify the layer, you could do it in a bbappend, or in a configuration file (make it IMAGE_INSTALL_append = " udev udev-extraconf" in that case) Aug 04 01:57:22 I tried adding: https://github.com/imyller/meta-nodejs and then setting EXTRALAYERS in my bblayers.conf but after that Im not sure what to do... Aug 04 01:58:05 Im also wondering if it might be easier to setup toaster and just customize stuff that way, haven't had a chance to play with that yet. Aug 04 01:58:25 i've actually never used toaster, so i'm not sure :) Aug 04 02:00:05 what's EXTRALAYERS? normally, you'd add the layer to BBLAYERS in ${BUILDDIR}/conf/bitbake.conf. Aug 04 02:00:15 after that, the recipes it provides will be available Aug 04 02:01:01 got me I was just trying to get node loaded based on their instructions on that persons github Aug 04 02:02:03 tried just pointing the nodes manifest in bblayers but when I ran bitbake again I didn't see any change in packages in the image that was built Aug 04 02:03:35 r0r0: to install node into the image, you'd have to add it to IMAGE_INSTALL Aug 04 02:04:02 if no recipe can be found for it (e.g. because they layer wasn't added properly), you'll get an error Aug 04 02:05:02 urr, should have been ${BUILDDIR}/conf/bblayers.conf above by the way :P Aug 04 02:05:31 Im gonna try it right now :-) Aug 04 02:06:38 so I should be able to just dump : IMAGE_INSTALL_append = "udev udev-extraconf" into my belayers.conf? Aug 04 02:06:42 or is the local.conf Aug 04 02:07:00 r0r0: local.conf, bblayers.conf is for layer-related configuration Aug 04 02:07:41 add the layer that provides node in bblayers.conf, and to IMAGE_INSTALL_append = " packages you want" in local.conf Aug 04 02:07:59 makes more sense now Aug 04 02:08:23 when I run bitbake again is should pickup on the changes I assume? Aug 04 02:08:28 you must add a space to the beginning in IMAGE_INSTALL_append = " foo bar" by the way. it's not added for you when using _append. Aug 04 02:08:28 don't have to for ex. make clean Aug 04 02:08:45 nope, changes are detected automatically Aug 04 02:09:10 I'm excited to show this off at work tmo, pretty cool stuff Aug 04 02:12:12 its building the image now, lets see if it works :-) Aug 04 02:14:30 nice Aug 04 02:14:33 worked perfect Aug 04 02:14:41 thank you Ulfalizer Aug 04 02:14:48 np :) Aug 04 02:23:51 so when you get a build error then remove the requested dependency from your local.conf file and get a Missing or unbuildable dependency what do you do? Aug 04 02:24:50 what's the error message? Aug 04 02:26:04 I think I got it figured out, I thought I had saved the file after removing what was causing that error but didn't... my bad. Aug 04 02:26:28 that time it built without error **** ENDING LOGGING AT Thu Aug 04 02:59:58 2016