**** BEGIN LOGGING AT Fri Apr 24 02:59:59 2015 Apr 24 06:08:06 Hello everyone! Apr 24 06:08:53 Im looking for some help regarding device trees. Basically I want to know what peripherals are connected to the board and where are they connected - i.e. on what pins Apr 24 06:09:10 My system is running a yocto build on a cyclone5 dev board Apr 24 06:09:37 you're back! Apr 24 06:09:52 yes! Apr 24 06:09:55 haha :) Apr 24 06:10:05 hopefully a little more educated now Apr 24 06:10:10 do you still need help on your image? Apr 24 06:10:36 I got it up and running and I might need to slightly customise it at some point Apr 24 06:10:43 cool Apr 24 08:12:45 gstreamer1.0-plugins-good-alaw-1.2.3 was registered as shlib provider for libgstalaw.so, changing it to gst-plugins-good-alaw-0.10.31 because it was built later Apr 24 08:12:52 how can I determine what is causing this? Apr 24 08:13:03 I don't want gst-plugins0.10... Apr 24 08:15:11 Hi! I'm trying to build a bootable hddimg with syslinux. The created image is 32MB, but it has partition offsets & size are incredibly huge (between 200 and 900 GB) Apr 24 08:15:43 and the image is not bootable with qemu nor on the target system Apr 24 08:23:01 good morning Apr 24 08:23:09 think I found it Apr 24 08:23:11 paulg: so did you have any further ideas yesterday? Apr 24 08:23:53 CromFr, which image are you baking? Apr 24 08:25:55 redengin: core-image-minimal Apr 24 08:26:37 where do you see these partitions? Apr 24 08:27:27 redengin: i'm using fdisk: fdisk -l core-image-minimal-m-ab3000.hddimg Apr 24 08:29:08 thens its set to take everything Apr 24 08:30:36 redengin: I just tried cfdisk which crash with a scary segfault ^^ Apr 24 08:31:00 have you tried dd? Apr 24 08:33:00 redengin: I tried to flash the hddimg on the target using dd, but partition tables are not better and it still don't boot :/ Apr 24 08:33:28 what platform? Apr 24 08:34:26 Hello, does anybody know why I have to use PREFERRED_PROVIDER_virtual/jpeg to set the jpeg provider and why PREFERRED_PROVIDER_jpeg doesn't work? Apr 24 08:35:49 redengin: nothing exotic, it's an intel atom machine with EFI/Bios boot Apr 24 08:46:10 how can I make sure only gstreamer1.0 is build in stead of gstreamer0.10? Apr 24 08:46:13 now I have both... Apr 24 08:47:40 wv: I think you'll need to find what is using gstreamer 0.10 and disable / fix it Apr 24 08:48:17 morning all Apr 24 08:49:11 bluelightning, how can I do that best? Find out which is using gstreamer-0.10 Apr 24 08:55:33 wv: one way would be bitbake -g whatever_you_were_building then grep -- '-> \"gstreamer' *.dot | grep -v gstreamer1 Apr 24 09:55:55 where can i find a list of modules / drivers available in kernel or yocto ? Apr 24 10:01:46 miandonmenmian_: make menuconfig in your favorite kernel tree :) Apr 24 10:04:08 <_4urele_> LetoThe2nd miandonmenmian_, I think it is easier to do "bitbake linux-yocto -c menuconfig" Apr 24 10:08:21 _4urele_: depends on your starting point, but agreed for some of those ;) Apr 24 10:08:45 paulg: sorry, got the connection lost Apr 24 10:24:43 oh, there is a diffconfig command for bitbake, interesting. Apr 24 10:28:56 anyone can shed some light on "the algorithm used by yocto for creating/merging configs into a single config " would be really helpful Apr 24 10:59:38 bluelightning: i am facing now the problems you mentioned yesterday: error: file /lib/libc-2.18.so conflicts between attempted installs of my-package-1.0-r0.armv7a_vfpv3 and libc6-2.18-r0.armv7a_vfpv3 -- i am using TCMODE="my-external-tc" so all needed dependencies are installed by my-external-tc and conflicting with the one from my-package (which are the same, my-external-tc). is there a way to force to use the ones from either my-package or from my-extern Apr 24 11:00:42 i set: PREFERRED_PROVIDER_virtual/libc = "my-external-tc" Apr 24 11:10:19 hello Apr 24 11:10:47 I have this image recipe with some added packages (qt4-x11-free, gstreamer1.0, ...) Apr 24 11:11:02 but when I build it via bitbake -c populate_sk thisimage Apr 24 11:11:11 I don't get the qmake Apr 24 11:11:19 nor gstreamer1.0 packages Apr 24 11:11:56 bitbake meta-toolchain-qt5 Apr 24 11:12:06 yes, but I need qt4 Apr 24 11:12:13 yes.. sry Apr 24 11:12:19 will it then also include the gstreamer1.0 packages? Apr 24 11:12:46 to answer my own question: no it doesn't, that is why I tried this populate_sk in the first place :) Apr 24 11:13:29 populate_sdk should include gstreamer1.0 Apr 24 11:13:57 but it doesn't Apr 24 11:14:02 neither the qmake is not in place Apr 24 11:14:09 for my host machine Apr 24 11:14:20 where with meta-toolchain-qt4 it is Apr 24 11:15:47 ls -l Apr 24 11:15:52 *sorry Apr 24 11:40:48 how does meta-toolchain-qt decides which packages are needed? Apr 24 11:44:27 wv: have you tried to extend your SDK image like this: https://github.com/balister/meta-sdr/blob/master/recipes-images/images/native-sdk.inc Apr 24 11:46:09 StMartin81 Apr 24 11:46:11 no... Apr 24 11:46:20 so I just create a native-sdk.inc as is? Apr 24 11:46:33 how does it get included somewhere? Apr 24 11:48:45 I'm not sure if I have understood this mechanism, but at least for cmake I think it works like this: cmake_2.8.12.2.bb has a BBCLASSEXTEND = "nativesdk" statement Apr 24 11:49:02 ah, no, I see that you require it in your images Apr 24 11:49:14 yes Apr 24 11:50:59 you would need a "require" statement in you thisimage.bb Apr 24 12:07:13 bluelightning: are you aware of any linux changes between dylan and daisy that is significant? Apr 24 12:19:25 is there any syntax to have bitbake foo package-index run package-index only once foo completed? (other than bitbake foo && bitbake package-index which would do parsing twice) Apr 24 12:31:33 does Yocto provide a chroot environment for testing some command line tool without the target being at hand? I know that it can generate qemu images, but I prefer chroot console for a console util over a very slow qemu gui. Apr 24 12:49:10 is there a way not to build/install a toolchain? Apr 24 12:50:25 How can I add the gstreamer1.0 libraries to my sdk? Apr 24 12:52:39 lpapp: you could mount the ext3 image and chroot there? Apr 24 12:57:26 gst-meta-video makes me use gstreamer0.10 Apr 24 12:57:32 but I want gstreamer1.0 in stead Apr 24 12:57:33 howto? Apr 24 13:00:07 lpapp: I don't personally keep track of individual kernel changes I'm afraid Apr 24 13:01:26 lpapp: no I don't believe we have such a chroot tool Apr 24 13:02:52 wv: you need to replace that with gstreamer1.0-meta-video Apr 24 13:03:34 <_4urele_> does someone knows if there is a layer where I could find python-beaker? Apr 24 13:08:12 bluelightning: i am facing now the problems you mentioned yesterday: error: file /lib/libc-2.18.so conflicts between attempted installs of my-package-1.0-r0.armv7a_vfpv3 and libc6-2.18-r0.armv7a_vfpv3 -- i am using TCMODE="my-external-tc" so all needed dependencies are installed by my-external-tc and conflicting with the one from my-package (which Apr 24 13:08:17 are the same my-external-tc). is there a way to force to use the ones from either my-package or from my-external-tc? Apr 24 13:09:16 ericbutters: there are probably a load of PREFERRED_PROVIDER values you'll need to set Apr 24 13:10:26 bluelightning: i set: PREFERRED_PROVIDER_virtual/libc = "my-external-tc" -- but i still get error for libc Apr 24 13:12:32 ericbutters: I'm assuming that is not sufficient if the other one is being built Apr 24 13:12:48 but then, you wouldn't have the headers without building it... Apr 24 13:14:42 bluelightning: is there something like PROVIDES -= or PROVIDES_remove? Apr 24 13:14:53 that wouldn't solve it Apr 24 13:22:53 ericbutters: I really feel like you're going about this the wrong way and causing yourself a load of unnecessary pain as a result Apr 24 13:23:45 if the original image has a few special things in it that you want to reuse as binaries, sure - extract those individually, package them up and install them via recipes Apr 24 13:27:02 bluelightning: yes, you are right. Apr 24 13:43:30 bluelightning, yes, but I'm a bit puzzled about how everything comes together in yocto Apr 24 13:44:08 I already found out that most of gst-0.10 was generated because of gst-va-intel Apr 24 13:44:21 but now it still depends of qt4 Apr 24 13:44:34 qt4 requires gst-plugins-base-0.10 Apr 24 13:44:43 however, that's not the biggest issue Apr 24 13:45:09 biggest issue is that I don't seem to manage how to include the gstreamer1.0 things in my -c populate-sdk Apr 24 13:45:20 well, with qt4 I believe you are stuck unless you disable building phonon or the phonon gstreamer backend entirely Apr 24 13:45:36 yes, but actually, I don't care about the gstreamer0.10 Apr 24 13:45:42 if it's there, it's there Apr 24 13:45:52 if the libraries are in the image, they should be in the SDK Apr 24 13:46:01 but I'd like to have the gstreamer1.0 pkgconfig included as well in my sdk Apr 24 13:46:09 well, that's the thing, they aren't Apr 24 13:46:17 they are in the image Apr 24 13:46:23 I can use them on my device Apr 24 13:46:39 but I want to build an application in qtcreator and deploy to my device Apr 24 13:46:46 for that I need this toolchain Apr 24 13:46:48 packages appear in the SDK because they are either -dev / -dbg packages or dependencies thereof Apr 24 13:47:22 so you mean that I would have to put gstreamer-meta-base-dev Apr 24 13:47:24 or something? Apr 24 13:47:33 I am guessing that because of some unusual dependency relationship that the package which contains the pkgconfig file somehow is not pulled in; you'd have to do some digging to find out why that wasn't the case Apr 24 13:48:26 if you must you can force something to appear in the SDK using TOOLCHAIN_TARGET_TASK_append = " packagename" Apr 24 13:48:35 (don't forget the leading space in the value) Apr 24 13:49:16 but for that you will need to know which package actually contains the pkg-config file Apr 24 13:50:00 logically it ought to be one of the -dev packages, but then if that were true it ought to be in the SDK already if it's a counterpart of a non-dev package that's in the image Apr 24 13:50:17 this sounds like a bug to me Apr 24 13:55:26 bluelightning: ok, thanks! Apr 24 13:59:16 hi folks, I would like to export ssl-staticdev and libcurl-staticdev to my SDK, but Apr 24 13:59:21 SDKIMAGE_FEATURES = "dev-pkgs dbg-pkgs staticdev-pkgs" doesn't work Apr 24 13:59:51 I had to put RDEPENDS_${PN} = "libcurl-staticdev openssl-staticdev" Apr 24 14:00:03 the question is, why SDKIMAGE doesn't work? Apr 24 14:00:08 bluelightning, but they are not included as dev in my image? Apr 24 14:00:26 second question, why RDEPENDS? they are DEPENDS, being statically linked means they aren't in the target Apr 24 14:00:47 I just did a IMAGE_INSTALL += "\ gstreamer-1.0-meta-base etc" Apr 24 14:00:53 never dev Apr 24 14:02:55 hmm, I guess no recipe atm uses wic for creating an image? Apr 24 14:03:20 and what's the difference between IMAGE_INSTALL and IMAGE_FEATURES? Apr 24 14:03:33 packages vs features Apr 24 14:04:44 ? Apr 24 14:04:59 and if I say I want gstreamer1.0 in my image Apr 24 14:05:10 is that a package or a feature? Apr 24 14:05:23 sounds like a package Apr 24 14:05:42 ah, I see Apr 24 14:05:49 like a feature is: I want psplash Apr 24 14:06:00 psplash itself is not a package Apr 24 14:06:06 you can see core-image's class on how features are matched to package groups Apr 24 14:06:23 (I'm just trying to figure out why gstreamer1.0 is not added to my sdk :) Apr 24 14:06:26 wv: as to details about specific contents of our gstreamer packages I don't know, sorry Apr 24 14:06:42 as it is added to image_install Apr 24 14:07:11 wv: logically if you have the gstreamer libraries in your image then building an SDK from that image should give you the necessary development files - if not, that's a bug Apr 24 14:07:37 ok Apr 24 14:07:42 LocutusOfBorg1: how are you building the SDK? Apr 24 14:07:47 Well, maybe I trye with this force thing Apr 24 14:08:16 wv: if you wouldn't mind it would be great if you could file a bug in our bugzilla Apr 24 14:09:09 just to be sure: the sdk is build using bitbake -c populate_sdk myimage Apr 24 14:09:10 correct? Apr 24 14:09:17 JEEBsv: no, it's external, although there has been some minor discussion about a special image class which would allow you to have it called as part of a build Apr 24 14:09:24 wv: yes Apr 24 14:09:36 ok Apr 24 14:10:24 I'm just looking at how to move from hddimg to something closer to the directdisk thing. wic would seem to be the latest part of infra for that kind of stuff Apr 24 14:10:25 and then the result is in tmp/deploy/sdk/poky-eglibc-i686-myimage-corei7-toolchain-1.6.2.sh Apr 24 14:10:25 ? Apr 24 14:11:00 and preferably when doing bitbake image-name instead of having to run wic manually afterwards Apr 24 14:11:24 an image class would sound good Apr 24 14:11:42 wv: yes Apr 24 14:24:31 bluelightning, yes, bitbake my_target_image -c populate_sdk Apr 24 14:24:47 BTW using yocto/dora Apr 24 14:24:50 not dizzy Apr 24 14:25:09 (because I need an older gcc in order to avoid kernel segfault on imx28) Apr 24 14:45:44 paulg: ping Apr 24 14:49:37 phonon is build as 3th party app from qt Apr 24 14:49:44 is there a way I can disable it? Apr 24 14:54:54 wv: I believe QT_PHONON = "" should do it Apr 24 14:55:22 (based on "git grep phonon") Apr 24 14:57:46 nice, I didn't know that command :) Apr 24 15:05:54 So, this ncurses files ownership problem I ran into yesterday only occurs if you combine package_ipk with meta-sourcery. Either package_rpm with meta-sourcery or package_ipk without it end up with correct ownership. What the heck? Apr 24 15:08:36 bluelightning: how is devtool going? Apr 24 15:09:02 lpapp: discovered some issues that we're in the process of fixing, but reasonably well Apr 24 15:45:17 JaMa: a review of the latest patchset to oe-devel, with proper (I hope) layer.conf, pretty please? Apr 24 16:13:51 Hello. I want to bring "perf" into my image and I'm using IMAGE_INSTALL_append = " perf" to do that. bitbake core-image-minimal fails with a QA Issue: ERROR: QA Issue: perf: The compile log indicates that host include and/or library paths were used. Any idea what I'm doing wrong? Apr 24 16:17:20 nicolae_: what's the exact erorr line? I did not spot a fix in: git log meta/recipes-kernel/perf/ -- are you on master or a released branch and what is your host distro? Apr 24 16:18:36 I'm on master, distro Fedora 21. Looking through the log, I'm seeing this: cc1: warning: include location "/usr/include/slang" is unsafe for cross-compilation [-Wpoison-system-directories]. After googling, this pops up http://patchwork.openembedded.org/patch/35257/ . Checking it now. Apr 24 16:21:01 nicolae_: that's commit 4e0daf05ed04e9fb4343032c6290a379f53b89f3 fyi, anyway keep digging. Apr 24 16:24:37 grepping reveals that perf needs slang for perf-tui. I have only enabled perf-scripting in PERF_FEATURES_ENABLE, since I don't know what tui is. Apr 24 16:25:22 tui = text user interface (probably ncurses), as in gdb-tui Apr 24 16:25:42 err gdbtui Apr 24 16:27:58 nicolae_: hint from mty linux.git tree: grep -ir tui tools/perf/config/ --> tools/perf/config/Makefile: msg := $(warning slang not found, disables TUI support. Please install slang-devel or libslang-dev); Apr 24 16:28:40 so is there a missing dependency (you could disable the tui support if you just need to plough through to another goal). Apr 24 16:30:15 It uses the wrong path, "CFLAGS += -I/usr/include/slang" Apr 24 16:41:33 nicolae_: do you know how to fix that? patch coming from you RSN? :) Apr 24 16:48:19 Is it possible to verify if a flag is enabled on PKGCONFIG (EXTRA_OECONF) ? If it is enabled I want to take an action Apr 24 16:52:11 vmeson: I haven't investigated yet. http://patchwork.openembedded.org/patch/35257/ seems to address this issue. Apr 24 16:57:10 no perf QA warnings here. Apr 24 16:59:23 I'm using linux 3.18 Apr 24 17:00:42 heh. that's the nature of perf. if it is a specific kernel, then you do need to take care of any patches to it. or make sure one of the sed commands fixes your source appropriately. Apr 24 17:01:11 so yo are in the right place looking into those other perf fixes Apr 24 17:06:33 Is there an easy fix to disable QA? Apr 24 17:12:05 nicolae_: modify ERROR_QA and WARN_QA, I guess. Apr 24 17:38:50 Hello. I is just my curiosity, how to know what packages is available ? I mean at all, in yocto or in some meta layer. For example in meta-qt5? Apr 24 17:39:08 packages, or recipes? Apr 24 17:39:16 packages) Apr 24 17:40:01 Because for example, there is qtquickcontrol recipe, but there is not packages with such name to include into rootfs Apr 24 17:45:35 ncgs: http://layers.openembedded.org/layerindex/branch/master/recipes/ Apr 24 17:48:16 yes, but there I can search only recipes. also there is "bitbake-layers show-recipes", but I am interesting packages. Apr 24 17:49:32 second example there is dhcp.bb recipe, but for appending to image , I should add dhcp-client and dhcp-server Apr 24 17:53:55 i think at the moment the best you could do would be to examine the recipes. my bb tool has a search-packages command if you want to look for something specific Apr 24 17:54:10 i expect the oe-pkgdata-tool could show some information after your build is done, as well Apr 24 17:54:17 oe-pkgdata-util, that is Apr 24 18:03:10 ok I'll try it, thank you Apr 24 18:07:14 I'm currently building with Daisy (v1.5) and need to bring in the athena libs, possibly from oe (libXaw). Is the athena libs in the later Yocto distros (e.b. 1.7 / 1.8)? Apr 24 18:14:06 1.5 is Dora Apr 24 18:14:09 https://wiki.yoctoproject.org/wiki/Releases Apr 24 18:14:17 Doh Apr 24 18:14:47 http://layers.openembedded.org/layerindex/recipe/950/ Apr 24 18:16:57 JaMa: yes, I've got the OE recipes now. Just before I go hacking the Yocto to get the libXaw recipe working (from OE), I was hoping I could find it in the later Yocto release. Apr 24 18:17:22 trying to save myself from upgrading Yocto at this late point in the game, if there is no benefit. Apr 24 18:35:37 kergoth Apr 24 18:35:41 kergoth: ping Apr 24 18:38:46 seebs: any thoughts on this? I'm seeing libncurses.so.5 owned by the build user inside the binary package, but only when using package_ipk, package_rpm is fine, and using package_ipk with an internal toolchain is also fine, it's only when using meta-sourcery with package_ipk. not sure where to look next, checked pseudo.log for ncurses, but all i see are path mismatches for pkgdata and /run/shm (typical), and some quasi-mismatches, none of which include Apr 24 18:38:46 the filename of the file with the wrong ownership Apr 24 18:38:52 T0mW: pong Apr 24 18:39:45 Usually when we had that with meta-sourcery, it was because the 32-bit objcopy had stripped a binary and done so in a way that involved recreating a file, but it wasn't running in the pseudo environment correctly. Apr 24 18:40:06 hmm, a bunch of these 'incoming path had trailing slash. Not unlinking' Apr 24 18:40:09 T0mW: do you understand the difference between meta-oe and oe-classic? Apr 24 18:40:19 "Just before I go hacking the Yocto to get the libXaw recipe working (from OE), I was hoping I could find it in the later Yocto release. Apr 24 18:40:27 " doesn't make any sense to me Apr 24 18:40:34 that was my first thought, but we are running NO32LIBS="0", and there aren't any library load errors coming from the commands that i can see Apr 24 18:40:41 so not sure how they'd be run in the wrong context Apr 24 18:45:47 * kergoth examines the ncurses WORKDIR under devshell with an eye toward ownership in both good and bad setups Apr 24 18:59:25 hmm,configure and compile don't run under pseudo, so a do_install should never use cp -p or -a, or it'd end up retaining that user ownership rather than the default root/root for new files created in that context, correct? Apr 24 19:01:33 I believe that's right. Apr 24 19:42:00 Interesting. in the bad case, ${D}${base_libdir}/libncurses.so.5.9 is owned by the build user, in the good case, ${D}${base_libdir}/libncurses.so.5.9 is owned by root. So something in the ncurses installation is behaving differently between the two cases Apr 24 19:42:01 huh Apr 24 19:42:02 * kergoth digs Apr 24 21:54:27 Would there be interest in a QA test for package_qa which identifies files whose uid/gid match that of the user running bitbake? Apr 24 22:05:30 I would Apr 24 22:26:25 kergoth: no one woudl say no to extra QA Apr 24 22:26:33 good point Apr 24 22:26:53 where it belongs in *_QA vars is another question, but adding the test wouldn't hurt Apr 24 22:27:15 at best it could be made optional if anyone complains about time Apr 24 22:27:42 * kergoth nods Apr 24 22:28:18 my only potential concern is if the user doing the build is e.g. the first non-system user, and someone creates a package that creates and uses such a user, but it seems unlikely, and i'm already excluding /home from the search Apr 24 22:29:11 in there a setting that say, in essence, regardless of the parallel/threads setting, don't ever do more than X do_fetch's at any one time? Apr 24 22:31:21 both richard and i have played with alternative schedulers to implement resource-driven constraints Apr 24 22:31:25 but i don't think anything is in there currently Apr 24 22:32:38 kergoth: ok, thanks Apr 24 22:33:43 is it possible to generate an sdk, then use it on a mac (osX) without the need for some sort of virtual machine? Apr 24 22:34:41 I don't think there's a darwin SDKMACHINE yet Apr 24 22:34:51 (SDKMACHINE is where the sdk will be installed/used) Apr 24 22:35:10 hmm Apr 24 22:35:12 kergoth: ok, thanks Apr 24 22:35:58 when i was at elc last year i noticed a lot of OE/Yocto people with macs Apr 24 22:36:05 i haven't used a mac since... 1998? Apr 24 22:36:15 i know some folks have worked on darwin support, but i *think* most of it was on being able to build on a mac, as opposed to targeting one Apr 24 22:36:18 not sure, though Apr 24 22:36:38 * kergoth does 100% of his work on a macbook air, but ssh's into VMs, either remote or local, generally Apr 24 22:36:51 ah Apr 24 22:37:16 some of the developers on my team use macs, natively, and i was wondering how to best support them Apr 24 22:37:25 maybe i'll look into doing something with vagrant Apr 24 23:15:18 Hmm, we should really pass the do_package_qa CachedPath instance into the qa test functions Apr 24 23:32:42 Anyone messed with gyp and the like in to to? Apr 24 23:33:32 Been messing with it and it's been a nightmare so far. Not due to yocto. Just wondering if anyone has any workarounds aside from patching with a new build system Apr 24 23:35:50 most of Mac-fans run Linux VMs on those powerful machines Apr 24 23:35:55 to do real work Apr 24 23:36:19 unless you are writing ios apps Apr 24 23:37:00 xulfer, yes I am going to mess with it and have in past Apr 24 23:38:16 kergoth: I tried to get it going on native darwin and Koen Apr 24 23:38:25 it went so far Apr 24 23:38:29 and we lost steam Apr 24 23:38:51 i tried many many years ago, kept running into portability issues with cp, sed, etc, mostly because i was too stubborn to add the gnubin path from brew's coreutils install to my PATH Apr 24 23:38:55 :) Apr 24 23:39:11 yeah I realized that I was installing all gnu tools Apr 24 23:39:38 I guess with toaster I can use mac as front end Apr 24 23:39:58 to an ugly but useful VM doing the dirty job Apr 24 23:40:01 at back Apr 24 23:40:19 and look cool may be Apr 24 23:40:34 * khem` uses Macs all the time Apr 24 23:43:03 I also run elementary/freya native on iMac Apr 24 23:43:30 time to sit in traffic Apr 24 23:44:32 toaster client sounds like a decent idea.. or even a bitbake UI as client to another machine if that works nowadays Apr 24 23:44:35 have fun :P Apr 24 23:48:28 Heh I just broke down and got a Linux box Apr 24 23:49:57 I'd like to put together a dedicated headless linux build machine at some point, ideally low power, quiet, and physically small, and set up a CI server that prepopulates my sstate cache nightly Apr 24 23:50:02 one of these days Apr 24 23:52:22 Check out soyoustart for that Apr 24 23:52:33 Cheap and fast. That's what I use for that Apr 24 23:53:10 I mean 40/mo for a ssd raid with 16GB is pretty good Apr 24 23:53:16 And a Xeon I believe **** ENDING LOGGING AT Sat Apr 25 02:59:59 2015