**** BEGIN LOGGING AT Thu Oct 01 02:59:59 2015 Oct 01 07:39:24 good morning Oct 01 07:53:14 morning mckoan, all Oct 01 10:21:53 belen, did you get toaster going with my image? Oct 01 10:23:22 Crofton|work: I did. I sent you an email with a link to the database and another one with instructions for replicating the feat ;) Oct 01 10:23:56 Crofton|work: I've passed the database file onto someone from WindRiver who will be setting up Toaster in the Yocto Project booth Oct 01 10:25:06 OK, so we are good for tool demo that closely matches hardware running in booth Oct 01 10:25:21 I need to go through your suggestions Oct 01 10:25:44 I cheat a little and do the real build against fido, but no one should notice :) Oct 01 10:26:11 Crofton|work: I think we are all set, yes. I built for the zedboard-zynq7, but I can try a different machine if you want me to Oct 01 10:26:21 that should be OK Oct 01 10:27:17 Mostly we are just showing how we can build complicated stuff with OpenEmbedded and have tools to analyze builds Oct 01 10:27:18 etc Oct 01 10:29:53 Crofton|work: well, all data is there. People should be able to look all the recipes and packages built, all dependencies between them, and the tasks that BitBake ran, all variable values, etc. It is a beast of an image, too! Oct 01 10:30:27 if it was easy we would use buildroot :) Oct 01 10:30:39 :D Oct 01 10:31:40 SO I just need to test my latest fido build :) and pack up the demo Oct 01 10:32:13 Being able to rebuild the demo a day before leaving and having it work is a YP benefit Oct 01 14:08:27 rburton1, I'm guessing you konw which file I can edit to change what matchbox runs when it starts? Oct 01 14:08:55 depends how you're starting matchbox really Oct 01 14:09:14 sato uses matchbox-session which is the shell scriipt that starts other things, and then exec's matchbox-wm Oct 01 14:15:46 IMAGE_FEATURES += x11-base :) Oct 01 14:16:18 I've done this once and forget Oct 01 14:17:25 yes I am trying to complete demo day before flying' Oct 01 14:25:15 Crofton|work: maybe have a look at mini-x-session Oct 01 14:29:27 bluelightning, found by grepping rootfs :) Oct 01 15:02:09 I beleive I set a new record, demo complete before I get check in email from airline Oct 01 15:28:15 Crofton|work: When the going gets tough, the tough get Yocto Oct 01 15:45:25 trying to figure out why postinst scripts are not doing what I expect. Here is the script: https://gist.github.com/cbrake/0ee2cfeb1b0255317ced Oct 01 15:45:41 the systemctl enable stuff seems to be working as expected, as the services are all enabled after first boot Oct 01 15:45:52 but none of the mkfifo of linking stuff is working Oct 01 15:47:01 (running Angstrom/Dizzy) Oct 01 16:13:49 Alright, has anyone used the "yocto build appliance" with kvm? And yes I know I'm marginally off-topic but oe-core can generate this crap too and I wonder if I'm not in some generic hell Oct 01 16:21:21 the BA is just a image that runs in a vm, so yes it should work in a qemu/kvm Oct 01 16:22:48 cbrake: well, the first thing to note is you probably want to move that into a conditional on D being set, since it'll run on the host at do_rootfs time too, cluttering up your /var/tmp, and if all that succeeds on the host, it won't run on target at first boot at all. Oct 01 16:23:00 cbrake: if you want to run that at do_rootfs time you need to operate against paths relative to D Oct 01 16:23:04 $D/var/cliff/.. Oct 01 16:23:20 postinsts are only run on target if they fail on the host Oct 01 16:24:19 given that mkfifo *can* run on the host, you probably just need to fix your paths. if anything cant' be run on the host, then you need to exit non-zero if D is set Oct 01 16:24:24 * kergoth yawns Oct 01 16:32:25 rburton1: Yes, it ought. Except I grabbed it and I'm getting a lot of "this isn't a valid disk" crap Oct 01 16:32:38 And I don't have super duper easy access to vbox/vmware atm Oct 01 16:32:56 I even converted from vmdk (which is supported) to raw, and still, nada Oct 01 16:33:24 The OE docs I found for making your own virtual images with OE assume you have the kernel around and can do the direct boot option Oct 01 16:33:30 I wonder if something wasn't packaged or what Oct 01 16:33:56 Sighing and checking freespace on a laptop with vbox installed tho Oct 01 16:34:29 huh, odd. you'd think the raw disk image would work.. worst case you could probably set up the raw disk on a loop device via losetup, mount the rootfs, and extract the kernel binary Oct 01 16:34:35 that's what i'd try Oct 01 16:35:10 Tartarus: if you're running on linux why do you want to run the BA? it's mainly for people running a windows host tbh. Oct 01 16:35:53 could be to repro what a customer is doing, or avoid host contamination.. though the former seems more likely, since a chroot or light container is quicker and easier isolation Oct 01 17:02:13 rburton1: It's because I'm in hell. I see over in #yocto a few weeks ago someone reported that on Ubuntu 14.04 building python for core2 (ie qemux86-64) they get a failure and illegal instruction. I'm reliably hitting that in both a 14.04 VM and a 14.04 chroot on my system. And I need to build on a nominally supported host for fido because I'm trying to Oct 01 17:02:13 use meta-rust and wow, uggg, there's some help needed there Oct 01 17:02:31 (rust-compiler builds a copy of llvm which in turn uses the host sytem, not Yocto gcc-native, for gcc) Oct 01 17:02:46 So I'm trying to find a new env I can do a qemux86-64 test on fido for meta-rust in Oct 01 17:03:01 FC21 installing in a VM now, was hoping the Y-B-A would be quicker to get going, but is not being Oct 01 17:03:25 I had it building on F22 on my laptop before the SSD in that gave up and died :( Oct 01 17:03:33 (RMA on that arrives today) Oct 01 17:36:58 argle-blargle, it does work in virtualbox at least Oct 01 17:37:03 So it's some sort of KVM related hell Oct 01 17:44:24 Yeah, "old" qemu stuff does not like that size vmdk, converting it to qcow2 with a more modern set of tools Oct 01 17:44:37 And when it turns out that it's too new a qcow2 I'll convert _that_ to raw Oct 01 17:44:49 Scream loudly, then get back to customer work Oct 01 17:59:51 double blargle, now I can boot it and have it hang Oct 01 18:00:11 Doesn't find a root Oct 01 18:00:17 Which is crazy since the image works in vbox Oct 01 18:06:50 efi vs non-efi boot, perhaps? *shrugs* Oct 01 20:25:47 kergoth: mkfifo can run on host, but can you package them in an ipk? Oct 01 20:26:10 so what is happening is its probably running on host, and not on target Oct 01 20:26:17 that makes a lot of sense Oct 01 20:27:24 yep. either exit 1 when D is set, or operate relative to $D. Oct 01 20:27:36 not sure on the fifo thing, ipk uses tar for the data, so if tar can contain it.. Oct 01 20:27:55 don't see any reason why not, it can hold devices and whatnot just fine, it just creates them when extracting Oct 01 20:28:12 I thought ipk used ar? Oct 01 20:31:11 an ipk is an ar archive containing control.tar.gz, data.tar.gz, and a version file Oct 01 20:31:35 control.tar.gz has the metadata, post/pre rm/inst scripts, etc, data.tar.gz is the actual files in the package Oct 01 20:31:41 same file format as .deb Oct 01 20:51:14 kergoth: good point, test fifos in tar ... Oct 01 20:53:32 seems I can put a fifo into a tar, but not extract it Oct 01 20:55:40 that seems odd, all it has to do is run mkfifo Oct 01 20:55:41 actually it works, must have done something wrong in my first test Oct 01 20:56:02 ok, so probably just need ${D} in my postinst, and everything will be happy Oct 01 20:56:12 to clarify, *not* ${D} Oct 01 20:56:21 ${D} will be expanded by bitbake, and it's always set to somehting Oct 01 20:56:29 $D is the shell environment variable, which will only be set in do_rootfs Oct 01 20:56:46 ahh, ok -- noted Oct 01 20:57:11 i think we actually have a qa test to fail the build if you use ${D} in a postinst nowadays Oct 01 21:15:06 * cbrake wonders if connman requires wireless extensions compatibility turned on ... Oct 01 21:15:28 i'd certainly hope not, given they're on hte way out Oct 01 21:22:12 appears like its working without -- phew Oct 01 21:26:24 * cbrake notices netctl in the archlinux wiki ... Oct 01 21:27:56 don't see any oe recipes yet for it Oct 01 21:40:14 aaand vanilla F21 cannot build m4-native and blows up Oct 01 21:44:14 * kergoth thinks Tartarus is having one of those days Oct 01 21:44:42 s/days/weeks/ Oct 01 21:44:59 Sunday the SSD in my few day old pixel 2 died Oct 01 21:45:05 Got the replacement one today Oct 01 21:45:16 Should have time to set it up again, more or less anyhow, before ELCE tho Oct 01 21:56:47 I think I'm going to confirm I can make python fail one last time, and then see about an older 1.8.x tag **** ENDING LOGGING AT Fri Oct 02 02:59:59 2015