**** BEGIN LOGGING AT Mon Aug 06 03:00:00 2018 Aug 06 10:09:57 rburton: we need to build 2.5.1, question is whether to pull in the binutils changes... Aug 06 11:15:57 rburton: glibc doesn't look so happy :( Aug 06 11:43:47 erm no Aug 06 11:44:09 | checking version of make... 3.82, bad Aug 06 11:44:12 what! Aug 06 11:45:33 huh have they bumped the hard requirement to 4.x? Aug 06 11:45:47 GNU make 4.0 or later is now required to build glibc. Aug 06 11:45:48 yes! Aug 06 11:46:53 well thats going to bail on a few of the builders then Aug 06 12:16:41 rburton: that is going to make things interesting :( Aug 06 12:39:43 evil plan: can we detect the host make version at parse time and add a make-native DEPENDS if required Aug 06 12:40:14 (now i want to hack on versioned depends that also work for host dependencies again) Aug 06 13:02:15 rburton: which mut? I don't see bash and libjpeg-turbo in http://git.yoctoproject.org/cgit/cgit.cgi/poky-contrib/log/?h=ross/mut Aug 06 13:02:22 JaMa: haven't pushed yet Aug 06 13:02:26 ok Aug 06 13:02:36 but sure I'm fine with split Aug 06 13:49:04 rburton: that would wreak havok with sstate and determinism :/ Aug 06 13:49:13 wimp Aug 06 13:49:27 hi, i'm looking for a term to describe the "recipes-x" level of a layer, which seems akin to gentoo's package categories Aug 06 13:49:32 rburton: didn't say it wasn't possible ;-) Aug 06 13:49:58 opello: the recipes directories are just to not have 100 subdirectories in a single directory Aug 06 13:50:16 they're *not* package categories which are the SECTION field in the recipe Aug 06 13:50:38 rburton: given the build time of make-native we could probably just add it as a dependency Aug 06 13:50:48 yeah Aug 06 13:50:54 rburton: yeah, i understand the intent, i didn't see a canonical way to talk about those buckets Aug 06 13:51:01 rss means the sysroot race issues we once had aren't an issue Aug 06 13:51:51 opello, just refer to them as directories that attempt to organize the content Aug 06 13:51:59 that is the extent of it.. they have no other system meaning Aug 06 13:52:26 ok, i was hoping the people that use this stuff more might have a standard way of referring to these intermediate organizational directories :) Aug 06 13:52:43 that is all they are directories.. seriously.. Aug 06 13:53:12 yeah, i do understand that Aug 06 13:53:24 if you look in conf/layer.conf you will see a line similar to: BBFILES += "${LAYERDIR}/recipes-*/*/*.bb" Aug 06 13:53:31 the pathing is explicitly stated in the layer.conf Aug 06 13:53:35 you can see that any search pattern will find them Aug 06 13:53:56 I've seen small layers (in the past) just do: ${LAYERDIR}/recipes/*.bb (while don't recommend that), it is valid Aug 06 13:54:25 right, preference is to follow existing behavior to avoid 'surprises' when people go between things Aug 06 13:54:48 yes, witht eh technical requirement that location to bb files has to be specified in the BBFILES Aug 06 13:55:12 i really wanted to call them 'recipe category directories' but that felt wrong per SECTION being a construct :) Aug 06 13:55:50 but thanks, you've both helped! appreciate it! Aug 06 14:14:32 RP, morning.. are the builders loaded or is there an issue ? Aug 06 14:15:21 * armpit sumo-nmut should have been done by now Aug 06 14:18:53 armpit: isn't it mostly done? Aug 06 14:19:14 armpit: nightly-x86 was running albeit slowly last time I looked Aug 06 14:22:27 k, was expecting it done by now.. Aug 06 14:37:44 * armpit sent patches to list for completeness Aug 06 14:42:06 New news from stackoverflow: Adding new spi module on IMX8 QM using Yocto Aug 06 15:10:33 does the "image" class provide the rootfs? Aug 06 15:11:42 yates: yes Aug 06 15:12:56 bluelightning: thanks. i should have asked "only the rootfs", and i think the answer to that is "no." it's the entire "image", including linux kernel, u-boot, rootfs, etc. right? Aug 06 15:14:42 yates: correct yes - though most of image.bbclass is to do with the rootfs, beyond that it's mostly in image_fstypes Aug 06 15:14:44 i am trying to generate a small rootfs suitable for using as the cpio.gz in an initramfs Aug 06 15:15:28 yates: you may wish to look at meta/recipes-core/images/core-image-minimal-initramfs.bb Aug 06 15:17:08 bluelightning: https://paste.fedoraproject.org/paste/VuYxMw0cuIdlvkZjv5TPyA Aug 06 15:17:23 is that only for non-arm cores? Aug 06 15:18:05 yates: I meant have a look at it for reference Aug 06 15:19:49 armpit: merge the binutils fix and build 2.5.1? Aug 06 15:20:42 yep. Aug 06 15:21:43 just updated stable/sumo-next Aug 06 15:22:25 * armpit now to fix rocko-next weird build failure Aug 06 15:24:08 bluelightning: if we assume that recipe could be built for my machine, what files does it generate into tmp/deploy/images? Aug 06 15:24:24 is it JUST the rootfs image? Aug 06 15:24:31 yates: that's not defined by the image Aug 06 15:25:38 so i'm a little confused. i stated earlier: "i am trying to generate a small rootfs suitable for using as the cpio.gz in an initramfs" how does that recipe help? Aug 06 15:26:02 the recipe defines that rootfs, just not what happens to it afterwards Aug 06 15:26:14 i.e. it specifies the content Aug 06 15:26:58 if you were to specify "cpio.gz" in IMAGE_FSTYPES you'll get a cpio.gz file Aug 06 15:32:58 bluelightning: what did you mean by "that's not defined by the image". that statement confused me. Aug 06 15:33:14 yates: in general, any image will work for any distro and any machine, through use of variables like bluelightning just mentioned. recipe controls the list of packages installed, machine controls the filesystem/partitioning/etc which is specific to the hardware. Aug 06 15:35:45 an initramfs image is a rare exception since it uses a format that differs from other images. as bluelightning says, see IMAGE_FSTYPES Aug 06 15:35:46 * kergoth yawns Aug 06 15:39:12 when i bitbake my main image, i get not only the rootfs.ext4 file, but also the zImage and .dtb files. why wouldn't the core-image-minimal-initramfs.bb also do the same? Aug 06 15:40:10 RP, thanks Aug 06 15:40:12 and u-boot Aug 06 15:42:56 my intention/thought was to add a dependency to my normal image build which generated a special cpio.gz that would be booted at early initialization, then under certain conditions i would continue on to boot the "normal" rootfs from my normal image build Aug 06 15:44:24 yates: kernel and u-boot files come from those recipes Aug 06 15:44:30 the image recipe has nothing to do with putting those files there Aug 06 15:54:52 then the purpose of the image recipe/image class is to generate just the rootfs? Aug 06 15:56:48 part of the problem (for me) is that the term "image" seems to be ambiguous. sometimes it means the rootfs image, sometimes the entire distro. Aug 06 15:59:10 my vendor "distro" image recipe (which is based on fsl-image-gui.bb) has "inherit core-image distro_features_check". how does this result in a kernel zImage in deploy? Aug 06 15:59:12 image means 'rootfs image'. This can be extended [easily] to filesystem image or even disk image. Aug 06 15:59:50 kernel comes from do_deploy in the kernel receipes and is grouped under the images directory as part of that work.. Aug 06 16:00:16 I don't remember if the do_deploy settings for the kernel are in the kernel.bbclass or in the linux-yocto recipe(s)... but it's somewhere there.. Aug 06 16:00:29 that happens separetly from the rootfs images (users of the image.bbclass) Aug 06 16:00:46 [we had talked aobut this a bit last week, which is the only reason I have a clue where the kernel side comes from] Aug 06 16:01:17 thanks fray and others. if i don't respond for awhile it means i am attempting to digest what i'm reading.. Aug 06 16:01:56 in all cases, the content move from the workdir to the deploy dir using do_deploy... so you will likely never see a direct copy operation, if that is what you are looking for Aug 06 16:05:46 ok fray, this helps - thx Aug 06 16:07:20 yates: the image defines what packages are installed and the like, one of those usually results in a dependency on the kernel/bootloader (usually the machine adjusts IMAGE_DEPENDS or whatever to get the bootloader built), and then those recipes deploy their own files to the images dir Aug 06 16:07:36 as i said earlier, the image recipe has nothing to do with anything but building a rootfs, and it has dependencies on other stuff Aug 06 16:07:47 * kergoth gets more coffee Aug 06 16:11:20 (usually the transformation from rootfs -> filesystem -> disk happens through a series of post-processing commands, that happen to be executed by the image -- but the definitions for these operations are elsewhere) Aug 06 16:12:25 New news from stackoverflow: Programmatic tracking version of Yocto custom BSP image Aug 06 16:22:41 RP: firing glibc again with a make-native dependency Aug 06 16:23:57 rburton: ok, we really need -next again too with the perl change dropped Aug 06 16:24:08 rburton: its currently building sumo 2.5.1 Aug 06 18:04:56 ok, so i've tried to bundle my own cpio.gz initramfs image in with the kernel by a) setting INITRAMFS_IMAGE_BUNDE to "" in my build/conf/local.conf file, and b) added INITRAMFS_IMAGE = "swupdate-image" into my main distro image recipe. but it doesn't appear the image is getting bundled. i tested it and there appears to be no cpio.gz early init boot. Aug 06 18:05:35 and swupdate-image.bb generates the cpio.gz - this works - i can see the foo.cpio.gz file in my deploy/images/ Aug 06 18:07:56 "main distro image recipe"? Aug 06 18:08:20 yes, the reicpe which builds zImage + dtb + uboot + rootfs Aug 06 18:08:21 INITRAMFS_IMAGE is a variable used by the kernel, so it needs to be set globally, either in the local.conf or other appropriate global configuration file, not a recipe Aug 06 18:08:34 it is Aug 06 18:08:44 oh wait. Aug 06 18:08:44 then it's not set in your "main distro image recipe" Aug 06 18:08:46 no Aug 06 18:08:51 :) Aug 06 19:14:26 khem, ping Aug 06 19:22:50 khem, nm Aug 06 19:27:49 ok, it is still not working. Aug 06 19:28:44 CONFIG_INITRAMFS_SOURCE="" Aug 06 19:29:44 there is no file being specified to the kernel. isn't the oe initramfs mechanism supposed to automagically configure this? Aug 06 19:35:24 i have done the following: a) created a recipe for the cpio.gz inheriting from image which configures my desired packages and a few other things. that works, i get a ...cpio.gz in deploy Aug 06 19:36:11 b) added INITRAMFS_IMAGE = "swupdate-image" (my cpio.gz recipe above) and INITRAMFS_IMAGE_BUNDLE = "1" into my local.conf Aug 06 19:36:56 and i think that's all.. Aug 06 19:37:15 should there be a depends in the kernel recipe on swupdate-image? Aug 06 19:40:06 yates: those options are in local.conf or another globally available .conf location right? Aug 06 19:40:21 clsulliv: yes, in my build/conf/local.conf Aug 06 19:40:32 ok, cool Aug 06 19:41:04 yates: and nothing gets deployed to tmp/deploy/images/ that looks something like zImage-initramfs-blah? Aug 06 19:42:09 oh... Aug 06 19:42:12 yes Aug 06 19:42:33 you mean this is the image i have to put in my fat32 boot? :) :) Aug 06 19:42:36 yates: that file is your kernel + initramfs bundle Aug 06 19:42:41 doh! Aug 06 19:42:45 i hate it when i do that.. Aug 06 19:42:50 thank you! Aug 06 19:43:00 yates: np Aug 06 19:43:04 i was flashing the old one (had a script with that hardcoded) Aug 06 19:44:46 but the build is also generating (apparently) the original non-initramfs zImage - is that normal? Aug 06 19:44:59 yates: it is Aug 06 19:45:30 ok, very good. Aug 06 19:46:59 next problem: my fat32 partition is now too small... Aug 06 19:47:10 i got this.. Aug 06 19:49:48 can you create symlinks on a fat32? Aug 06 19:52:10 no Aug 06 19:52:20 piddle Aug 06 20:13:04 hi guys, someone have dnf problems when running dnf update command? mine is getting librepo.LibrepoException: (8, 'curl error: An unknown option was passed in to libcurl', 'An Curl handle error') Aug 06 20:13:17 someone have ever seen this before? Aug 06 21:01:17 Hi, I set up my toolchain and rootfs for the eclipse plugin and I'm testing a hello world program. I get an "configure: error: C compiler cannot create executables" error when i right click on my project and click reconfigure project. https://pastebin.com/81a0uVYm Aug 06 21:02:20 any help appreciated Aug 06 21:17:17 once you've booted an initramfs, is there a command to pivot to the external (e.g., in an ext4 partition) rootfs? Aug 06 21:17:45 (yay, i booted an initramfs! veryc cool!) Aug 06 21:17:54 \/ Aug 06 21:18:57 pivot_root per chance? yes! Aug 06 22:06:44 yates: I believe youre looking for switchroot Aug 06 22:15:13 armpit: yes Aug 06 22:16:23 khem, sorry. I ended up saying "nm". was going to bounce updating binutls in rocko to tip of 2.29.1 but I noticed "new features" got added Aug 06 22:16:38 so decided not to Aug 06 22:20:43 armpit: ok, I think it should be ok to move to 2.29.1 Aug 06 22:20:43 it is at 2.29.1... they have added commits after the release Aug 06 22:20:44 on the 2.29 branch Aug 06 22:28:07 khem, ok.. I will give test run on the update. I have 8 backports done against the current version. Aug 06 22:30:01 sounds good Aug 06 22:31:46 Hi, I set up my toolchain and rootfs for the eclipse plugin and I'm testing a hello world program. I get an "configure: error: C compiler cannot create executables" error when i right click on my project and click reconfigure project. Error: https://pastebin.com/81a0uVYm Config.log: https://pastebin.com/vvjJhFW3 any help apreciated Aug 06 22:34:17 kism: can you please file a bug on bugzilla.yoctoproject.org for component " Aug 06 22:34:36 kism: component "Eclipse Plugin" Aug 06 22:35:35 kism: please make sure you specify which OS (and version) you are using and which versions of YP and Eclipse Aug 06 22:48:46 * armpit sweat.. backports apply w/o issue on new binutls Aug 06 23:03:23 khem, thanks **** ENDING LOGGING AT Tue Aug 07 03:00:00 2018