**** BEGIN LOGGING AT Fri Mar 11 02:59:58 2016 Mar 11 07:50:58 I have KERNEL_IMAGETYPE set to zImage. Can I make it generate vmlinux as well? Mar 11 11:57:10 anyone remember what creates the file /etc/version in rootfs? Mar 11 11:58:41 rburton, hmm, who makes /etc/timestamp Mar 11 11:58:48 and there is /etc/version Mar 11 11:59:03 and do they appear in images from Angstrom Mar 11 12:11:50 Crofton|work: meta/classes/rootfs-postcommands.bbclass Mar 11 12:12:54 that is timestamp Mar 11 12:13:16 interesting we have two files with about the same thing in them :) Mar 11 12:16:25 IIRC /etc/timestamp isn't really about the contents, it's about the date on the file Mar 11 12:19:40 http://stackoverflow.com/questions/35927774/how-to-get-rootfs-build-timestamp-at-runtime Mar 11 13:05:30 Has anyone ever seen wic fail with an error like "exec_cmd: install -m 0644 -D /etc /etc returned '1' instead of 0"?? Mar 11 13:55:48 Crofton|work: if you're still looking /etc/version gets written in lib/oe/rootfs.py Mar 11 13:56:15 ah Mar 11 13:56:19 mainly curious Mar 11 14:16:55 Alright, so I've figured it out that my issue was with how I'm using IMAGE_BOOT_FILES but the curiosity remains for me because the documentation seems to indicate it's possible to copy whole directories of files to the boot partition as such: IMAGE_BOOT_FILES = "mydir/*;destdir/". However that format yields the error: Malformed boot file entry: ;destdir/. What is the appropriate syntax for that variable if the docs are indeed wron Mar 11 14:22:34 Never mind...looks like my version of oe-core isn't new enough to get that feature. Dangit. Mar 11 14:46:38 Is there a way to call a bb.utils function without having to assign it to a variable? I'm using "${@bb.data.append(,,d)}" on a single line of a task, but it's throwing an error because the function returns None, and then the recipe evaluates "None", and exits with the error it has no idea what that means. Mar 11 14:47:15 use an anonymous python function, that's what they're for, to run python code at the end of parsing Mar 11 14:47:44 kergoth: How would I do that in a regular task (sorry, I'm very new at this) Mar 11 14:48:26 you don't Mar 11 14:48:49 an anonymous python function is a function that runs at the end of parsing, not at task execution time Mar 11 14:49:16 and of course, a ${@} is run when the task is expanded, not when it's run Mar 11 14:52:02 so then define in the recipe: python () { d.appendVar(,) }? Mar 11 14:55:14 (I apologize, I'm having trouble reading between the lines of what you're saying; the documentation seems to be pretty sparse on how to use these things, and usually recipes seem to be using the ${@} syntax to set recipe variables Mar 11 14:58:08 there are hundreds of examples all over the metadata of use of anonymous python to alter the metadata Mar 11 15:00:06 https://www.yoctoproject.org/docs/1.6/bitbake-user-manual/bitbake-user-manual.html#anonymous-python-functions may be of use Mar 11 15:04:41 Yes, I came across that on my own. I suppose I'm more curious at this point why that's the preferred usage in this case, but ${@} is provided when you want to fetch the return value. I was off looking for a way to execute the function without regard to the return value Mar 11 15:04:59 Or at the very least, have the task execution ignore the return value...which I suppose I could just by assigning the output to a dummy variable. Mar 11 15:05:05 But that seems like a silly thing to have to do. Mar 11 15:05:16 (just to ignore the return value) Mar 11 15:05:20 ${@} is only used for inline python, and it will *only* be run if the variable it's in gets expanded Mar 11 15:05:34 if you put it in a dummy variable, that will do precisely nothing Mar 11 15:05:41 unless the varaibel is actually referenced, and therefore expanded Mar 11 15:06:02 you're trying to use it to do something it's not designed for. don't do that Mar 11 15:06:54 it's used to add logic to variable expansion, that's all. if you want to do something else, you're using the wrong mechanism Mar 11 15:07:35 I'm not sure that's a fare statement if the function is, like appendVar, meant to update a variable and I want my task to update a variable, to say that I'm doing it wrong by assigning the output to a dummy variable within the task. Mar 11 15:07:50 When the task is expanded, regardless of the dummy variable, shouldn't the whole task be expanded? Mar 11 15:07:59 tasks cannot alter metadata in a way that will affect anything outside that task Mar 11 15:08:05 tasks get isolated copies of the metadata Mar 11 15:08:18 Or better, wouldn't I get around that by dummyvar := ${@...}? Mar 11 15:08:34 then it will alter the metdadata for all tasks, in which case you should just be using anonymous python Mar 11 15:08:43 as it will not be specific toi the task in any way Mar 11 15:09:14 i don't see why you're jumping through hoops to accomplish something it's not intended for when we have actual mechanisms designed to programmatically alter the metadata Mar 11 15:11:04 I'm not sure I see what I'm doing as jumping through hoops so much as trying to find the expected way that was intended to be able to do what I'm trying to do...which is quite simple. I just need to copy a directory from my deploy image to the the boot partition. Mar 11 15:11:37 And I need to do it in a way that lets my recipes update IMAGE_BOOT_FILES rather than having to specify it in a layer or local.conf because by doing so, I blow away the MACHINE definition of that value Mar 11 15:12:50 I've been fighting with this mechanism for days by changing layer priorities, using weak-sets, etc. Nothing seems to help. My top-level software layer simply kills the BSP's weak-set (Xilinx) which is then appended by the vendor's BSP layer, so I lose out on the Xilinx BSP's files. Mar 11 15:13:39 And all of this effort is simply to get around the possibility of users needing to deploy this image as a read-only filesystem since some files do need to get edited after the bake simply because you can't predict the target network in our case. Mar 11 15:15:12 So to your comments about my not doing it the way the designers intended: that's all I'm trying to determine. How would the designers have intended this because so far, I haven't been able to figure it out. Mar 11 15:31:38 kergoth: Let's back up for a moment. Are you saying that it's not the intended design to use the bb.data.appendVar() method from a recipe? Mar 11 16:04:14 no, appendVar should be used, just not from ${@} Mar 11 16:10:49 Looks like khem sent some perl code to the list. :-) Mar 11 16:13:58 kergoth: alright, thanks Mar 11 16:28:02 anyone done a first boot recipe to enlarge an ext4 filesystm? Mar 11 16:48:30 That sounds like a clever way to give someone a copy of the internal filesystem based on a jumper setting... Mar 11 18:41:31 halstead, what happens when you click on latest release? Mar 11 18:41:32 https://www.yoctoproject.org/documentation Mar 11 18:42:33 Crofton|work, It shows the error message because Yocto 2.0.1 came out but the docs for it aren't up yet. Mar 11 18:42:41 Crofton|work, I'll have to get that fixed. Mar 11 18:42:45 thanks Mar 11 18:42:53 I am attempting to use th edocs Mar 11 18:43:07 because my question makes me look so stupid I can't ask it here :) Mar 11 18:44:46 You can find 2.0 under previous releases rwhile I fix this. Mar 11 18:45:34 I did Mar 11 18:57:53 where would I find an exmaple of how to amke a first boot script that actually runs on the device? Mar 11 19:14:08 hmm, how do I stop a build from fialing because I do not have a licence file Mar 11 19:14:53 you mean for LIC_FILES_CHKSUM? Mar 11 19:15:07 yes Mar 11 19:15:13 You can use CLOSED for LICENSE at least temporarily until you get proper license Mar 11 19:15:33 in this case I am trying to install a postinst script on the deivce Mar 11 19:15:39 and that is all the recipe does Mar 11 19:15:42 no sources etc Mar 11 19:15:50 then you can choose your license Mar 11 19:15:57 why not MIT like packagegroup recipes? Mar 11 19:16:11 you can copy&paset both LICENSE and LIC_FILES_CHKSUM from there Mar 11 19:16:33 ah, that is what I was trying to remember Mar 11 19:26:59 so whatpart creates the symlinks in my tmp-glibc/deploy/images/machine directory Mar 11 19:27:33 I'm trying to build a ext4 image core-image-minimal.ext4 and the timestamped ones exist, but not the core-image-minimal-machine.ext symlink Mar 11 19:27:39 fischerm, has lost the link to the image file Mar 11 19:27:54 for fitImage etc it works Mar 11 20:15:41 How do I get a postinst to run on the deivce itself? Mar 11 20:17:39 fail. exit non-zero Mar 11 20:17:53 the general idiom is to check if $D is set, and if so,w e know it's do_rootfs-time Mar 11 20:17:59 i.e. if [ -n "$D" ]; then exit 1; fi Mar 11 20:18:58 grr, I ahv ethat Mar 11 20:19:07 * Crofton|work debugs harder Mar 11 20:19:50 this recipe is empty except for postinst :) Mar 11 20:23:50 remember: $D, not ${D}, otherwise it'll never run Mar 11 20:28:43 and spell command you want to run porprerly Mar 11 20:29:42 * fischerm flips table ... stupid ext4 image Mar 11 20:32:27 rofl yeah I had the command wrong Mar 11 20:46:33 heh, oops Mar 11 21:34:03 I am having issues with commit 66205c6096ce9d8bc828bf9b61d927cb495f69b1 in the jethro branch. It is complaining about nothing providing 'virtual/nativesdk-nativesdk-libx11' which of course looks wrong. Mar 11 21:59:17 Crofton|work, New 2.0.1 docs are done. Thanks for the notice. Mar 11 21:59:32 np Mar 11 21:59:50 ah, nice, a non-broken docs page is a good thing Mar 11 22:30:38 Hello Mar 11 22:30:44 Is there any reason for systemd to require qemu? **** ENDING LOGGING AT Sat Mar 12 02:59:58 2016