**** BEGIN LOGGING AT Fri Jan 16 03:00:15 2015 Jan 16 06:46:33 LetoThe2nd: i see. so you're saying that booting a modern stripped down linux kernel that's good enough to reach a shell, is difficult in 4 or 8MB of RAM? Jan 16 06:47:18 is that even with compressed RAM and swap? Jan 16 07:37:10 Is there a way to configure a channel for smart when generating the rootfs? For ipk I used FEED_DEPLOYDIR_BASE_URI but I can't seem to find this for rpm Jan 16 07:57:51 dtm: i'm not saying super complicated, but it will certainly require thoughtful selection of kernel features, and won't offer much more than a busybox Jan 16 08:02:44 dtm: from a first and very rough estimate, for example my not-too-bloated kernel on a sama5 eats up roughly 4m Jan 16 08:02:52 err 10MByte Jan 16 08:03:41 so you'll certainly hit areas where you have to take out substantial parts when you want to run the whole system in 4M Jan 16 08:59:01 hello. i apologise for my lamenting; i did manage to build a working kernel and image after all Jan 16 09:00:43 now i wonder what's the best way to build additional packages, e.g. dropbear Jan 16 09:01:44 wiki FAQ suggests using IMAGE_INSTALL_append Jan 16 09:03:51 i'm afraid using that and running `bitbake core-image-minimal` will still do a lot more work than just build that single package that i want to add Jan 16 09:04:06 well then how about bitbake dropbear? Jan 16 09:06:39 is there a way to get valid targets/recipes for bitbake? Jan 16 09:07:45 you can just browse your layers and see what .bb files you find :) Jan 16 09:07:59 so basically find . -iname '*.bb'? Jan 16 09:08:12 basically. Jan 16 09:10:44 dRbiG: bitbake-layers show-recipes Jan 16 09:11:14 (it'll give you basically the same information as a find *.bb though) Jan 16 09:12:09 LetoThe2nd: ok. Jan 16 09:12:50 mago_: that works fine, thank you Jan 16 09:13:18 yeah, bitbake dropbear ought to work Jan 16 09:13:50 but first, let's see if my today's kernel will have working ethernet Jan 16 09:14:01 just doing bitbake dropbear won't include dropbear in your rootfs image though, it'll only build the dropbear binary package Jan 16 09:14:01 LetoThe2nd: what do you mean about busybox? that's just an embedded app suite on top of linux Jan 16 09:14:24 mago_: that's exactly what i'm looking for Jan 16 09:14:28 LetoThe2nd: so i mean yeah once you got linux ported, then i assume you'd be able to compile busybox and such Jan 16 09:14:34 i don't want to rebuild the rootfs everytime i want to add a package Jan 16 09:14:57 dtm: i know what busybox is, thank you. Jan 16 09:15:23 LetoThe2nd: i know you do. so i was confused as to why you're answering about it, to a question about the kernel ^_^ Jan 16 09:15:32 dRbiG: ok; if you want to install binary packages on your device, you will need a package manager on the rootfs. So you probably want IMAGE_FEATURES += "package-management" Jan 16 09:15:42 dtm: i mean that such a constrained system will not be able to run a reasably sized, comfortable userspace. you'll be stuck with busybox, and probably even static linking. Jan 16 09:15:51 LetoThe2nd: oh yes definitely Jan 16 09:16:12 LetoThe2nd: reasonably sized is not an objective Jan 16 09:17:04 LetoThe2nd: i was wondering if someone could slim the kernel as you say, with ram and swap compression (if we end up with swap), and then maybe SDL or some embedded opengl framebuffer Jan 16 09:17:38 dtm: sdl or ogl certainly, no way. Jan 16 09:17:42 okay. Jan 16 09:17:53 Is there a way to configure a channel for smart when generating the rootfs? For ipk I used FEED_DEPLOYDIR_BASE_URI but I can't seem to find this for rpm Jan 16 09:17:55 mago_: yes, added that early on Jan 16 09:18:00 LetoThe2nd: i couldn't find any info about any embedded opengl. there used to be an embedded mesa. Jan 16 09:18:06 but it was deprecated i guess Jan 16 09:18:09 dtm: and remember that compression and swap are also costly in terms of ram, before they kick in. Jan 16 09:18:20 LetoThe2nd: ok. i haven't ever used them. Jan 16 09:18:29 hey bluelightning Jan 16 09:18:36 morning mago_, all Jan 16 09:18:42 for a kernel that boots in 4mb incl userland, you'll probably have to get rid of swap, modules, and about all filesystems. Jan 16 09:19:15 .. or go back to, maybe 2.6.18 or such. Jan 16 09:19:32 LetoThe2nd: i would guess that you'd have to basically have a single-process and kernel image Jan 16 09:19:40 bluelightning, is there a way to configure a channel for smart when generating the rootfs? For ipk I used FEED_DEPLOYDIR_BASE_URI but I can't seem to find this for rpm. I saw your message in the mail archive but couldn't find any new things on it https://www.mail-archive.com/yocto@yoctoproject.org/msg14641.html Jan 16 09:19:41 yes. Jan 16 09:19:43 not unlike a normal cartridge game Jan 16 09:20:34 LetoThe2nd: so you couldn't get SDL or opengl in 8MB RAM? is there an embedded opengl? Jan 16 09:20:59 dtm: think for a second how much only the framebuffers would need. Jan 16 09:21:27 LetoThe2nd: ok Jan 16 09:21:37 dtm: 4mb kernel+userlang almost certainly means "no display, or some super tiny monochrome hack at maximum" Jan 16 09:22:07 i believe linux 2.0 and 2.2 were used for many embedded devices even when 2.4 was current, in 1999 Jan 16 09:22:33 i remember that someone had *some* really stripped version running in 1MB of RAM. Jan 16 09:22:47 dtm: feel free to subtract 1999 from 2015 to find out how many years it has been since that Jan 16 09:23:14 i do feel so very fancy free, yes. Jan 16 09:23:15 dtm: and of course - go back to ancient software, with ancient feature sets, then you get ancient resource footprints. Jan 16 09:23:30 yeah. Jan 16 09:23:33 RagBal: there is PACKAGE_FEED_URIS... somehow we have managed not to document that, but it's basically just a space-separated list of URIs for the repos Jan 16 09:24:55 bluelightning, will look into that, thanks! Jan 16 09:25:48 LetoThe2nd: so do you know offhand what RAM minimum would be considered normal for embedding the smallest opengl or other 3D library implementation, even an old one, that people may still use today? Jan 16 09:26:34 seeing that he typical opengl library size is between 2MB and 4MB, I4m not sure how you want to use opengl with only 4MB of RAM Jan 16 09:26:54 dtm: i don't have numbers, but i guess something along the lines 64mb ram, 32mb rootfs should be enough to get opengl ready. Jan 16 09:26:54 abelloni: yeah i'm saying that there's a 4MB and an 8MB profile Jan 16 09:26:59 seems i got the ethernet working; now i need dhcp and dropbear Jan 16 09:27:08 LetoThe2nd: okay Jan 16 09:27:15 i wonder if i will get systemd service files within the packages Jan 16 09:27:18 dtm: neither 4mb nor 8mb will suffice. like i said. just do the math on the framebuffers. Jan 16 09:27:26 LetoThe2nd: ok. Jan 16 09:27:51 LetoThe2nd: well i do thank you for the insight on a quasi-silly question <3 Jan 16 09:28:23 it's a historically significant point of general curiosity Jan 16 09:29:11 in part, we were wondering if it was actually feasible for the people who did claim to have gotten linux and x11 working, back in that time Jan 16 09:29:32 mysteriously unreleased! :-D Jan 16 09:30:01 they didn't claim any 3D software. Jan 16 09:30:52 so i know that an old linux and X11 is possible in 4MB, because i did that on a 386, and it sucked. Jan 16 09:31:16 5MB with swap, anyway. Jan 16 09:47:17 so you say a modern kernel+userspace can fit in 4MB. what kernel version? and apparently ram compression is out of the question. mesa-sdl 3.4 from 2007 takes about 2MB RAM according to http://wiki.tcl.tk/1131 and might require an older kernel. so that's ~6MB. a single max res framebuffer (640x480x24bit) is 2457600 bytes so I guess we'd do low res 320x240 at 614400 bytes. So assuming you ported all this to the hardware, that leaves room for a pretty big app Jan 16 09:47:17 and some other libs. then you factor in how wrong all assumptions are, and it's "maybe"? Jan 16 09:51:20 dtm: read up here (and on dependent buzzwords) http://www.cnx-software.com/2012/02/08/pengutronix-uclinux-3-2-for-energy-micro-efm32-cortex-m3-geckos/ to get an impression of what can be don on such low end platforms with significant effort Jan 16 09:51:44 or http://www.cnx-software.com/2011/12/30/linux-for-cortex-m3-m4-microcontrollers/ Jan 16 09:52:00 LetoThe2nd: oh, buzzwords! thanks! Jan 16 09:52:44 i think i remember reading about uclinux in about 1997, on the 68000 Jan 16 10:02:21 LetoThe2nd: it's very kind of you to look up URLs. that first one is pretty good. 4MB RAM. but does uclinux always have no MMU support? it seems like you'd want MMU software support if you have MMU hardware, given how fragile things sound without it. Jan 16 10:03:39 from here http://www.uclinux.org/pub/uClinux/FAQ.shtml it sounds like there's no MMU or memory protection etc at all Jan 16 10:04:53 dtm: correct. Jan 16 10:05:02 one of the drawbacks if you want to go super tiny. Jan 16 10:06:04 LetoThe2nd: well ya better know what you're doing then! nothing new to a game console ;) Jan 16 10:09:04 LetoThe2nd: that uclinux faq i just pasted says it's linux 2.0.38. so it's immune to the ping crash, ok! :-D so the url you pasted says they have linux 3.2, so I guess they made a unique modernized fork of uclinux? which isn't part of the main project/ Jan 16 10:09:05 ? Jan 16 10:09:59 they have 755kB of RAM in use, by the linux 3.2 kernel and such, after bootup? on a system with 4MB total. wow. Jan 16 10:10:45 so i guess their version only runs on their hardware Jan 16 10:10:49 dtm: no idea what they have exactly, os its mostly out of my focus of interest. i know it exists, thats about all. Jan 16 10:10:55 ok Jan 16 10:10:58 but yes, this is *highly* hardware specific. Jan 16 10:12:10 okay. Jan 16 10:12:55 hence a lot more effort especially if you aren't targeting their ARM family Jan 16 10:14:47 a general question in comparing the stock linux source trees... is linux 3.2 a lot more modular and can thus be easily configured smaller than, say, the overall smaller 2.0 and 2.4? Jan 16 10:15:42 from my professional point of view (e.g., i do embedded systems for a living): this is only of academic interest, or if you are targeting a product with volumes in the 100k pcs/anno + X size. in any other case, the dev costs will easily outweigh the hardware cost savings. Jan 16 10:16:00 (and both are not in my focus.) Jan 16 10:17:09 yes academic interest Jan 16 10:21:17 I don't think you save anything on the hardware Jan 16 10:21:53 4MB of ram is probably more expensive than 128 now Jan 16 10:23:12 abelloni: ha! no kiddin? it's such an odd item huh? Jan 16 10:23:24 kinda like 18" dishwashers instead of 24"? lol Jan 16 10:29:45 I would say that it is a matter of volume Jan 16 10:31:27 exactly. Jan 16 10:32:11 we're having hell of a time usually getting rom/ram chips in the megabyte range for our older products here Jan 16 10:36:48 we have a customer designing a new board because DDR chip are getting too expensive Jan 16 10:36:53 versus DDR2 Jan 16 10:37:40 totally understandable Jan 16 10:38:21 LetoThe2nd: are you wanting to find such small ram chips because they take less power or what? Jan 16 10:38:48 i just bought an 18" dishwasher btw, originally priced about double the equivalent 24" dishwasher with identical componentry Jan 16 10:39:13 though with a lot of special Sears discounts to bring it back down ;) Jan 16 10:39:47 which they don't even sell in stores, because it's the same product but less popular one-off manufacturing Jan 16 10:43:14 dtm: no, we sometimes need those chips if we have to produce spare parts for hardware designed years, sometimes tens of years ago. Jan 16 10:44:54 LetoThe2nd: oh right. very good. Jan 16 10:45:03 so you can't just stockpile them Jan 16 10:47:04 you usually can't store smt components longer than a few years if you need reliable solder characteristics for manufacturing. Jan 16 10:57:14 LetoThe2nd: oh i didn't know that. i'd heard of hard drive lubricants gumming up on the shelf. interesting. Jan 16 11:01:30 LetoThe2nd: is MIPS still a hot thing for embedding like it was in the 90s with N64 and PlayStation and whatnot, or has that all gone to ARM? Jan 16 11:03:38 it depends on the product Jan 16 11:04:49 there's still a reasonable amount of mips out there Jan 16 11:05:14 my networked audio stuff is all mips, for exxample Jan 16 11:05:35 still trying to convince them to switch from buildroot to yocto too Jan 16 11:07:09 ok Jan 16 11:10:03 is R4300i (aka NEC VR4300, a variant of R4200) still a thing? does it get deployed in many products? is that relevant to current source trees? is support for that still buildable, say, in linux 3.2? Jan 16 11:11:02 i know some embedded stuff still gets modernized as the same product (miniaturized etc), or some gets totally obsolete Jan 16 11:11:08 many years later Jan 16 11:13:56 i'm totally not from mips land, so i can't say anything there. Jan 16 11:26:23 "There also some low cost Cortex-M boards such as the STM32F4-Discovery board available for around 15 USD, but there is unfortunately not enough RAM (192KB) to be able to run uClinux." <-- lol Jan 16 11:29:43 Hi, I am having problems while building an image sdk together with deb packaging system. Jan 16 11:30:26 I already had to add self.target_pm.install_complementary(self.d.getVar('SDKIMAGE_INSTALL_COMPLEMENTARY', True)) to DpkgSdk in poky/meta/lib/oe/sdk.py Jan 16 11:30:48 in order to get -dev packages installed automatically to it by using dev-pkgs feature Jan 16 11:31:26 After that, I can see in log.do_populate_sdk that it tries to install all the dev packages Jan 16 11:32:04 but when trying to install them, APT fails because it says some dependencie sfor some packages are broken because the deps are not going to be installed Jan 16 11:33:15 for instance, in IMAGE_INSTALL I have some gstreamer stuff. Gstreamer dev packages depend on pulseaudio-dev which in turn depends on avahi-dev, which in turn automatically depends on "avahi" package. Jan 16 11:34:30 but as avahi is not in my IMAGE_INSTALL var, apt doesn't want to install it to the sdk and then installation of the dev packages is not happening. Jan 16 11:36:24 I could fix it by running apt-get install -f after installing packages by modifying DpkgPM, but that's not the right way of doing it, as in fact it's hiding some important errors among packages being installed Jan 16 11:37:04 what do you think is the best approach to solve that? do you consider that a bug on bitbake? or on the recipes side? Jan 16 11:43:03 pespin: it would be a bug in the metadata, but I'm unsure of the exact correct fix Jan 16 11:43:56 the first thing to do would be to file a bug in bugzilla, if you wouldn't mind Jan 16 12:20:37 bluelightning: Hi, for the moment I created this bug + patch, which should solve the first part of the problem I faced -> https://bugzilla.yoctoproject.org/show_bug.cgi?id=7160 Jan 16 12:20:38 Bug 7160: normal, Undecided, ---, richard.purdie, NEW , SDKIMAGE_FEATURES not used by deb packaging system Jan 16 12:21:38 I will now fillup another bug regarding the whole context/problem Jan 16 12:22:55 good morning Jan 16 12:48:16 bluelightning: I also created the ther one regarding the dependencies -> https://bugzilla.yoctoproject.org/show_bug.cgi?id=7166 Jan 16 12:48:18 Bug 7166: normal, Undecided, ---, richard.purdie, NEW , Dependency problems with deb packaging system when installing dev packages on SDK generation Jan 16 13:01:12 why is linux-yocto using separate repos for each kernel version (3.14, 3.17 etc)? Jan 16 13:08:00 mago_: I'd assume it has something to do with the single meta branch, but zeddii might be able to explain further Jan 16 13:08:43 couldn't you have one meta-branch per version-branch too? Jan 16 13:09:00 Hi guys. Jan 16 13:09:12 hey Jan 16 13:10:25 Any idea how to get rid if the systemd error systemd-tmpfiles[108]: chmod(/var/tmp) failed: no such file or... Jan 16 13:11:31 systemd is trying to create these dirs from the tmp and var tmpfiles .conf files Jan 16 13:11:50 But these too are already created by base files as symlinks. Jan 16 13:11:54 two* Jan 16 13:12:41 Hence these errors: chmod(/var/tmp) failed, chmod(var/log) failed. Jan 16 13:13:01 Is this a known issue? Jan 16 13:16:22 Isn't smart capable of removing packages installed as dependency while removing something? Jan 16 13:16:35 I can't seem to find any parameter for that Jan 16 13:17:10 bluelightning: also, how does standard/base relate to the kernel stable branches? i noticed the 3.14 standard/base has quite a few commits that are not in gregk's stable repo on his 3.14.y branch. Do commits flow both directions here? standard/base stuff goes upstream and gregk commits keep getting merged into standard base? Jan 16 13:18:32 mago_: I'm not sure I'm afraid, another thing that zeddii / dvhart / tomz / nitink would be able to answer much better than I could Jan 16 13:18:55 ok, thanks Jan 16 13:19:26 one of them should be here shortly I would think Jan 16 13:39:57 * zeddii reads Jan 16 13:41:00 standard/base are common patches to the yocto kernel, these are not the same category as greg's stable patches, they can be features, or fixes that are specific to our use case. Everything merges into standard/base (LTSI, -stable, CVEs, etc) and then to all the BSP branches. Jan 16 13:43:17 hi, i'm creating a i386 bsp layer with yocto-bsp script and it always add a CONFIG_MATOM=y Jan 16 13:43:24 to kernel config Jan 16 13:44:10 when i boot up an amd geode, kernel complains about a missing instruction (movbe). Jan 16 13:44:30 *adds Jan 16 13:45:18 even if a choose i586 in the tuning options... Jan 16 13:46:36 Related to the problem I stated above: I pushed a patch to oe-core as I got no feedback. We can trace this topic on ml. Thanks anyway. Jan 16 14:39:48 is it possible with yocto to enter some env that will have the toolchain setup so that I can manually build software? Jan 16 14:45:11 bluelightning: hi, in one of my packages which inherits module.bbclass (for building out-of-tree modules) I assign to FILES_${PN} but it does come out be empty and hence not populating the main package. I haven't seen this behaviour before Jan 16 14:47:00 darkhorse: are you setting FILES_${PN} before or after the "inherit module" line? Jan 16 14:47:22 bluelightning: oh..before Jan 16 14:47:43 darkhorse: right, in this instance that won't work because module.bbclass sets it to "" explicitly Jan 16 14:47:50 move it to after, problem solved :) Jan 16 14:48:30 I can't tell yet but I do wonder if it's kept empty for a reason though Jan 16 14:50:39 ah, right... so the idea is that it automatically splits out all kernel module(s) produced by the recipe into kernel-module- packages just like the kernel itself, and then the main package is just an empty meta-package that pulls in all of those to help in the case where there is more than one Jan 16 14:51:38 so I guess it depends on what these extra files are for; if they should be associated with one of the kernel modules in particular the better thing to do would be to add those files to the individual module package instead Jan 16 14:53:14 bluelightning: i see.. so the module.bbclass is a bit special. i tried "bitbake -e recipe-name.bb to see who is setting it to empty but unfortantely could figure out Jan 16 14:53:47 bluelightning: moving it after 'inherit module' didn't change anything for me Jan 16 14:54:25 bluelightning: here's assignments that i can see when i dump environment Jan 16 14:56:16 bluelightning:after all different assignment that are happening to this variable, it kind of overrides it with the following message: rename (to) data.py:166 [expandKeys] Jan 16 14:58:12 darkhorse: somewhat confusingly you may find two entries for it as FILES_${PN} and FILES_, see if you can see both of those Jan 16 15:03:04 bluelightning: you are a star :-) Jan 16 15:03:47 bluelightning: i would have given up long ago, without your help Jan 16 15:04:03 I aim to please :) Jan 16 15:04:06 buy him a beer Jan 16 15:04:14 and come tofosdem and man the booth :) Jan 16 15:04:45 bluelightning: how can i buy him a beer? I am more than happy to Jan 16 15:05:14 bluelightning: sorry it was a qustion to Crofton Jan 16 15:05:43 well, no beers necessary unless you happen to be at one of the conferences I'm able to go to ;) Jan 16 15:13:33 hi all, bluelightning, I would like to build a image specific sdk ("bitbake my_image -c populate_sdk") containing ./sysroots/x86_64-pokysdk-linux/usr/bin/qmake. How can I add toolchain to my image? Jan 16 15:34:24 TuTizz: I do that by adding "inherit populate_sdk_qt5" to my image Jan 16 15:34:52 pespin, ok ty Jan 16 15:35:06 TuTizz: and this too I think -> TOOLCHAIN_HOST_TASK += " nativesdk-qtbase-tools-dev " Jan 16 15:47:45 whats the benefit of moving patches from your kernel recipes SRC_URI_append_mymachine into a mymachine.scc-file and a patch-directive? i don't really see the point, and doing so also makes bitbake unable to detect changes to patches Jan 16 16:00:21 zeddii: ^ Jan 16 16:00:30 bluelightning: getting very close to generating image identical to my existing build system output - init manager is one of the last remaining issues. how can i completely remove sysvinit support provided by yocto? I have already tried setting VIRTUAL-RUNTIME_initscripts = "" in my distro.conf Jan 16 16:02:17 darkhorse: DISTRO_FEATURES_BACKFILL_CONSIDERED = "sysvinit" should do it, as mentioned in http://www.yoctoproject.org/docs/current/dev-manual/dev-manual.html#using-systemd-exclusively Jan 16 16:03:07 note that changing that will probably trigger a bunch of rebuilding I'm afraid, since lots of recipes look at DISTRO_FEATURES Jan 16 16:04:23 bluelightning: that's fine. i will try that. but let me mention that i actually don't want any init manager at all. so no sysvinit no systemd. i have my own 'pre-cooked' template files that i just want to copy over Jan 16 16:05:26 darkhorse: ok, that shouldn't be an issue - you'd do the above and also _not_ add "systemd" to DISTRO_FEATURES, and that should work Jan 16 16:06:20 building with no init system (well, more often with a custom flat init script) is definitely supported - poky-tiny does it for example Jan 16 16:06:30 Hi. Is there an easy way to get layer dependencies from http://layers.openembedded.org/? Jan 16 16:07:05 For example, I'd like to know what layers meta-browser depends on. Jan 16 16:07:58 mario-goulart: i think you can get the layer dependencies from the layers themself nowadays. Have a look in layer/conf/layer.conf and look for LAYERDEPENDS_.. lines Jan 16 16:07:59 mario-goulart: assuming you mean programmatically, yes, in fact Chong Lu is working on adding functionality to bitbake-layers to query that - see http://patchwork.openembedded.org/patch/86553/ Jan 16 16:08:35 mario-goulart: as you can see from my reply, it's WIP, but the patch may still be useful as an example of how to use the API Jan 16 16:08:38 mago_: right, but then I need a clone of each layer to get the full dependencies chain. Jan 16 16:08:59 mario-goulart: why do you want to extract dependencies for layers that you haven't cloned? Jan 16 16:09:00 yes and unfortunately most layers don't set LAYERDEPENDS Jan 16 16:09:27 mago_: to generate a manifest file (repo) for projects which use multiple layers. Jan 16 16:10:57 IIUC, I could get what I want by requesting http://layers.openembedded.org/layerindex/api/layerDependencies/ and http://layers.openembedded.org/layerindex/api/layerItems/ , but it seems that I can't do that atomically. Jan 16 16:12:01 Is it possible to have a recipe which has always to be run, i.e. bitbake see's fit to execute compile and install every time you run it. Jan 16 16:12:33 dguthrie: possibly depending on details - can you explain what you're trying to achieve? Jan 16 16:24:58 bluelightning: removing the sysvinit using DISTRO_FEATURES_BACKFILL_CONSIDERED = "sysvinit" didn't go very well. i get the following error during recipes parsing stage: Jan 16 16:25:43 bluelightning: ERROR: Please ensure that your setting of VIRTUAL-RUNTIME_init_manager (sysvinit) matches the entries enabled in DISTRO_FEATURES Jan 16 16:30:39 darkhorse: sounds like your init manager is still set to sysvinit Jan 16 16:36:19 rburton: yes..that's possibly because in my conf/bitbake.conf i see DISTRO_FEATURES_BACKFILL="pulseaudio sysvinit" but then i thought DISTRO_FEATURES_BACKFILL_CONSIDERED should remove it - no? Jan 16 16:39:20 no, that literally just messes with the distro features Jan 16 16:39:44 set that variable to systemd and you'll be sorted Jan 16 16:39:51 (pretty sure this is in the documentation :) Jan 16 16:42:16 rburton: thanks. i will try it out Jan 16 17:15:06 bluelighting: I am building from multiple build directories into a single deploy directory. The hardware I am using needs more images that just the kernel and root filesystem. The final recipe that combines these into a single image for the software upgrader can't tell if it's dependencies have changed, so I was the final software upgrade image creation recipe to execute it's tasks all the time Jan 16 17:30:54 zeddii_home: hi can you help me with some kernel initramfs issue? Jan 16 18:05:38 dguthrie: ah ok... you could mark the tasks that do that end processing as "nostamp", that would mean they execute every time they are called for rather than being stamped so they only run when the inputs change (which is the default) Jan 16 21:09:10 exit Jan 16 21:09:18 ¬¬ **** ENDING LOGGING AT Sat Jan 17 02:59:59 2015