**** BEGIN LOGGING AT Wed Dec 11 02:59:59 2013 Dec 11 07:42:01 hello. Dec 11 07:42:42 is there a way to compile qt console application, and not import whole qt4x11 ? Dec 11 07:45:48 ak77: have you looked into the qt4e examples? Dec 11 07:49:36 LetoThe2nd: where would that be? (looking in /yocto/poky/meta/recipes-qt/) Dec 11 07:50:19 ak77: ./meta/recipes-qt/images/qt4e-demo-image.bb Dec 11 07:50:32 and whatever it pulls in, of course. Dec 11 08:19:27 Can somebody please tell me, what's wrong with this line: SRC_URI = "git@192.168.4.35:kernel;protocol=git"? Dec 11 08:19:46 I get an error: MalformedUrl: The URL: 'git@192.168.4.35:kernel;protocol=git' is invalid and cannot be interpreted Dec 11 08:20:18 I've tried "protocol=ssh", but it's still the same Dec 11 08:25:22 kbart: I think you should have a / instead of : after the ip/host, try git@192.168.4.35/kernel;protocol=git" Dec 11 08:25:52 and protocol ssh if that's what you want Dec 11 08:33:14 erbo, still the same. But if I do "git clone git@192.168.4.35:kernel" manually, it works perfectly.. Dec 11 08:34:10 kbart: oh, you miss the git:// in the beginning Dec 11 08:34:43 SRC_URI = "git://git@....." Dec 11 08:37:44 Ah, that was it! Thank you erbo. Dec 11 09:34:48 morning all Dec 11 10:19:02 Hi, does anyone use icecc with yocto ? Dec 11 10:19:35 oswin: I think JaMa has been or at least knows people who have Dec 11 10:20:50 actually I'd be interested to know what the performance benefits of using it are (i.e. numbers) especially with the new blacklisting Dec 11 10:23:06 bluelightning, I've heard that is wasn't that advantageous, but in which case and with how many computer, I don't know... Dec 11 10:24:16 oswin: I'd heard the same in the past, in that the overhead of sending out the compile jobs meant that there wasn't much of an improvement; however nobody has come up with any numbers recently Dec 11 10:24:30 I have to assume that people who are using it have measured the improvement :) Dec 11 10:24:51 I hope so ^^ Dec 11 10:28:12 hi all i have been using poky with meta-atmel and when i try to load multi gadget,it shows the following error http://pastebin.com/ktV8LjUT can you tell me what is that issues Dec 11 10:37:11 good morning Dec 11 10:39:50 oswin: the last time i tried it with about 40 cores the benefit was perhaps getting 1/3 - 1/4 less buildtime Dec 11 10:40:11 oswin: but that was almost a month ago Dec 11 10:44:40 gagi, did you put the scheduler on the fastest computer ? Dec 11 10:45:28 oswin: yes, we are using icecc for almost all of our non-yocto builds in our office and it runs on a 8-core xeon server Dec 11 10:46:32 gagi, and what's the usual benefit ? 1/2 ? Dec 11 10:46:47 oswin: i think the problem is/was that many packages don't accept the parallel build Dec 11 10:47:14 oswin: it depends on how many machines you have, i would say > 1/2 Dec 11 10:47:16 gagi, (i mean on non yocto) Dec 11 10:47:54 gagi, ok good, thanks! Dec 11 11:54:24 oswin: bluelightning: yes I don't use it personally, because I work from home, but people in office are using that Dec 11 11:55:10 oswin: bluelightning: with a lot of computers connected to icecc and very big builds it gets really useful, building qt with -j 150 is a lot faster :) Dec 11 12:06:13 JaMa: right, with monsters like qt and webkit there ought to be some pretty big gains Dec 11 12:07:07 JaMa: btw, among other things Scott has just documented icecc.bbclass in the manual; perhaps I should get him to add a hint or two like that as well Dec 11 12:07:30 http://www.yoctoproject.org/docs/1.6/ref-manual/ref-manual.html#ref-classes-icecc Dec 11 12:13:31 I had working recipe with some compilation and installation. Then I added a single file I want to get deployed to the device (http://pastebin.com/xtKvvs8k) for this extra file it claims it was isntalled but not shipped, any clues? Dec 11 12:14:45 bluelightning: I have few comments, is there open bug for this or can I say them here and you'll proxy them? Dec 11 12:15:32 JaMa: I can proxy them, but it may be better for you to email to scott directly about it (cc me if you like) Dec 11 12:15:47 ack Dec 11 12:16:12 JaMa: I think he prefers srifenbark at gmail.com FYI Dec 11 12:17:32 ak77: basically you need to add something appropriate to FILES so that the additional file gets picked up Dec 11 12:18:39 ak77: and put into the appropriate package (probably the main package, so I'm guessing you want FILES_${PN} += "${base_libdir}/firmware/" ) Dec 11 12:24:23 bluelightning: ups already sent to intel.com, but will bounce Dec 11 12:24:26 bluelightning: that did the trick.. but what added compiled file implicitly to FILES ? Dec 11 12:25:27 ak77: there's a default value for FILES for a number of standard packages in PACKAGES; these are set in meta/conf/bitbake.conf Dec 11 12:26:41 bluelightning: i see. thank you. Dec 11 12:28:29 bluelightning: i have another problem with that recipe, if you can take a look... do_rootfs fails due to bluez4 (>= 4.101) dependancy (i use bluez5), where does that dependancy come from? Dec 11 12:58:30 :/ Dec 11 13:11:43 Hi everyone. I really can't figure out how to use PREMIRRORS to circumvent our firewall. The problem is git. PREMIRRORS_prepend = "git://.*/.* http://git.yoctoproject.org/git/ \n" does not work. I would like to just set protocol=http for all the git:// repos. Is this not possible? Dec 11 13:28:27 my recipe compiles .c files and has -lbluetooth, where does information which package provides this library is hold? generated? Dec 11 13:30:37 it keeps pulling bluez4, altough i have bluez5, which provides same library (libbluetooth.so.3) Dec 11 13:33:24 ak77: well, when your application is being packaged the system will just look at what libraries it links to and then what package(s) those libraries are provided by, and set the dependencies accordingly Dec 11 13:34:17 ak77: if you have bluez4 in your sysroot (because you built it at some point before), or it's linking to bluez4 from the host, you can get problems like the one you are having Dec 11 13:57:22 hi there Dec 11 13:57:33 trying to send a patch to openembedded-core Dec 11 13:57:53 using git send-email Dec 11 13:58:20 However i'm being given a response 'the mailinglist requires people to subscribe before posting' :| Dec 11 13:58:24 I'm subscribed Dec 11 13:58:37 Xz: your mails are not coming from the email that is subscribed Dec 11 13:58:58 (my .gitconfig sets my from address) Dec 11 14:00:58 rburton: I have exactly same address in git Dec 11 14:03:24 Xz: use git format-patch and send the email with your email client to check if that works Dec 11 14:07:04 Hello. Dec 11 14:08:13 With yocto 1.4 there seem to be http mirrors out of the box for everything but yocto 1.5 doesn't work behind the firewall, nothing but trouble with some packages... Dec 11 14:10:08 Is there a way to build only the rootfs and not the rootfs + kernel? Dec 11 14:12:54 there must be somethig wrong with my git configration Dec 11 14:16:04 bluelightning: i'm not linking to bluez4 from the host, where is this sysroot, and how can I rebuild it completly Dec 11 14:17:27 I have user.email and sendemail.from set to my email which is subscribed (tried subscribing again - message back, that I'm already subscribed) Dec 11 14:17:35 still cannot send email from git Dec 11 14:17:46 I must be missing some other settings maybe? Dec 11 14:18:30 git version 1.7.12.2 Dec 11 14:59:39 ak77: you shouldn't need to rebuild it completely; if bluez4 has been built cleaning it out would just be a matter of bitbake -c clean bluez4 Dec 11 14:59:54 ak77: the sysroot is at tmp/sysroots// Dec 11 15:18:45 Hi, quick question. I'm trying to compile fsl-image-gui for freescale imx53qsb, but it fails with "ERROR: Task 437 (/media/yocto/yocto/poky-dora-10.0.0/fsl-community-bsp/sources/poky/meta/recipes-qt/qt4/qt4-x11-free_4.8.5.bb, do_compile) failed with exit code '1'" Any tips on where to go from here? Dec 11 15:21:27 Leo___: can you pastebin the full error? Dec 11 15:21:46 blue lightning, yea of course, one second Dec 11 15:22:46 bluelightning: thank you Dec 11 15:23:55 http://pastebin.com/ff4fDeHe Dec 11 15:24:15 hope this is what you need, just started out playing with all this today Dec 11 15:27:13 Leo___: that is an odd error, I haven't seen that before Dec 11 15:28:00 bluelightning: Ah, just my luck haha. should i file a bug report somewhere? Dec 11 15:28:29 Leo___: well, I'd suggest maybe asking about this on the meta-freescale mailing list Dec 11 15:28:46 the first error is definitely specific to meta-fsl-arm Dec 11 15:29:01 bluelightning: ok, will do. thank you Dec 11 15:35:02 Denwid: I tried using email as well - the same problem Dec 11 15:54:18 Hello I would like to understand a little something before digging in the creation of one recipe. My custom software needs the library libpq (from postgresql 9.2 which is not yet available in yocto/open embedded) and libpqxx (which is also not available). To compile my software in bitbake system do I need to add recipe for libpqxx and postgresql or as soon as those library are available on my Dec 11 15:54:18 local system it will compile correctly ? Dec 11 15:55:06 weebet: you need recipe for your dependencies. Dec 11 15:55:27 the build won't use any 'local' libraries, without doing very dirty hacks... Dec 11 15:55:48 ndec : thank you . this is bad news, I tried to create recipe for postgresql but I was not able to be successfull to create the recipe. Dec 11 15:56:15 weebet: there is a postgresql 9.2 recipe btw: http://layers.openembedded.org/layerindex/branch/master/recipes/?q=postgres Dec 11 15:56:31 i was about to say the same thing ;-) Dec 11 15:56:56 I am surprised, the last time I have looked, it was for 8.4 or something like this Dec 11 15:57:50 looking at the git history it was upgraded in August of this year Dec 11 15:58:31 Great :-) Thank you Dec 11 16:00:33 I would have another question, when I log onto poky, it will only take care of the 8 first character. is there any plan to solve this issue ? I have seen that there were a lot of patch to apply to solve this, but never found any howto to correct this. Dec 11 16:05:13 I noticed that module_strip.bbclass is empty, is there some other way to make sure kernel modules are stripped? Dec 11 16:06:12 19 degrees here. ;-) Dec 11 16:06:44 erbo: iirc its automatically handled by package.bbclass now, nothing you need to do Dec 11 16:06:47 afaik anyway Dec 11 16:06:59 * Crofton|work curses kit kat and bluetooth headset Dec 11 16:07:45 ah turning it on is a big help Dec 11 16:09:54 erbo: coincidentally I'm about to send a patch to drop that empty class Dec 11 16:12:14 rp: i believe i just found another bug, thanks to the verify commit on branch change. git ls-remote lists both refs/tags/ and remotes/head/foo/bar/. do you have any ideas on how to solve that? Dec 11 16:13:09 kergoth: I'll have to look through package.bbclass, because my modules are not stripped :) Dec 11 16:13:14 i.e., when what i was looking for was the sha1 of refs/tags/, instead i get back the sha1 of some random branch that happens to end with the same string as the tag i was looking for. Dec 11 16:14:31 zibri: hmm, that could be quite nasty :/ Dec 11 16:18:00 weebet: it seems you just need to set ENCRYPT_METHOD in /etc/login.defs Dec 11 16:18:10 I wonder if there's a good reason why we don't do that by default... Dec 11 16:19:03 Thank you bluelightning Dec 11 16:21:33 zibri: git ls-remote refs/tags/ Dec 11 16:21:39 afaik that should work fine, no? Dec 11 16:21:43 http://stackoverflow.com/search?q=yocto Dec 11 16:22:00 http://stackoverflow.com/search?q=openembedded Dec 11 16:22:03 oh, i'm guessing in this case it's from srcrev, so we don't know what it's referring to? Dec 11 16:22:09 kergoth: yes, indeed. but this is in the git fetcher. i'm not sure that change is fine there Dec 11 16:22:39 but maybe try that first, if nothing comes out, try refs/head/. Dec 11 16:23:24 refs/heads/ even Dec 11 16:23:37 hmm yes, thats's an idea Dec 11 16:23:54 not the prettiest, i'll admit :/ Dec 11 16:25:10 not many other options if the user gives us something ambiguous, other than to just fail :) Dec 11 16:26:23 Is anyone else having problems with m4-native package? It always fails at configure phase for me. Dec 11 16:26:41 weebet: actually that only applies if you have "shadow" in your image; if you're using busybox's version of login/passwd then it should be supporting >8 char passwords out of the box Dec 11 16:26:43 vicenteolivert: pastebin the problem please Dec 11 16:26:43 configure: error: C compiler cannot create executables Dec 11 16:26:54 vicenteolivert: pastebin the config.log please :) Dec 11 16:27:03 Ok. Dec 11 16:30:58 kergoth, rp: thanks for your input! i'll try to prepare a patch and send it tomorrow Dec 11 16:32:13 rburton: http://www.pastebin.ca/2496272 Dec 11 16:33:58 rburton: Ok, it's failing because the CFLAGS I have passed. But those CFLAGS should be valid... Dec 11 16:34:07 bluelightning : how to know if I use shadow / busybox ? also, I am not using the latest repo how to checkout only the postgresql recipe ? I tried to git clone git://git.openembedded.org/meta-openembedded/meta-oe/recipes-support/postgresql but I have a remote error: acces denied or repository not exported /meta-openembedded/meta-oe/recipes-support/postgresql Dec 11 16:34:12 Folks, can someone with a redhat bugzilla account see if they can access: https://bugzilla.redhat.com/show_bug.cgi?id=813070, we are seeing an userspace nfs readdir issue that is referenced in here possibly. Dec 11 16:34:13 Bug 813070: was not found. Dec 11 16:35:12 rburton: it looks like it was compiling for i686 instead of mips. Dec 11 16:35:18 vicenteolivert: you did ask for m4-native Dec 11 16:35:26 or at least, it's building m4-native Dec 11 16:35:50 so where you specified those cflags, you need to use them for just target builds and not native Dec 11 16:36:04 weebet: git can't be used to fetch individual files like that, but you can fetch / check out the latest branch, copy the files out (or clone the repository again in a separate directory) Dec 11 16:36:27 rburton: I have specified those cflags in my local.conf Dec 11 16:36:36 rburton: is not correct? Dec 11 16:36:36 weebet: re shadow / busybox, you can do: ls -l `which passwd` Dec 11 16:37:21 vicenteolivert: extend TARGET_CFLAGS so you're not changing native CFLAGS Dec 11 16:37:42 Ah... Dec 11 16:37:55 bluelightning : no shadow in (... PATH ...) so I uses busybox Dec 11 16:38:33 and when I do : passwd it tells me that the password must have minimum of 5, maximum of 8 Dec 11 16:39:29 I work with gumstix layer under danny. I guess I work with an old repo Dec 11 16:39:36 rburton: TARGET_CFLAGS_append = " -mmicromips -mno-jals" , is that the right way to do it? Dec 11 16:40:01 vicenteolivert: that should work. yeah. Dec 11 16:40:13 rburton: ok, thanks. I'll let you know if it works. Dec 11 16:40:43 I NEED STICKERS! Dec 11 16:41:08 rburton: quick question. Is it possible to build only the rootfs image and not the kernel image? Dec 11 16:41:26 weebet: if it used shadow, it would definitely say /usr/bin/passwd -> /bin/busybox (or busybox.suid) Dec 11 16:41:39 weebet: er, I mean if it used busybox Dec 11 16:44:19 "/usr/bin/passwd -> /bin/tinylogin" Dec 11 16:44:34 Or better..., is there a way to simply use my own ".config" file to build the kernel? Dec 11 16:45:25 weebet: ah of course, tinylogin Dec 11 16:46:04 vicenteolivert: you can build your own kernel if you don't like the yocto kernel config thing Dec 11 16:46:52 rburton: ok, but how can I avoid building the yocto kernel? Dec 11 16:47:21 vicenteolivert: i recommend getting yocto to build a kernel using your config Dec 11 16:47:38 weebet: in danny we used tinylogin , before we used busybox's integrated version of the same Dec 11 16:47:57 weebet: but it seems that behaved differently w.r.t encryption methods Dec 11 16:48:58 weebet: to be honest I'm not sure how to fix that; you might try googling "tinylogin 8 character" or something like that Dec 11 16:49:24 Thank you bluelightning. I guess I will work to get the latest Dec 11 16:49:30 repo Dec 11 17:12:39 Jefro, ! Dec 11 17:12:44 OH NOES! Dec 11 17:14:46 rburton: despite of using TARGET_CFLAGS_append, eglibc-initial fails anyway: http://pastebin.ca/2496286 Dec 11 17:15:25 "mipsel-poky-linux-gcc: error: unrecognized command line option '-mmicromips'" Dec 11 17:15:36 are you sure you're using gcc options that the gcc you're building knows about? Dec 11 17:15:56 hmm Dec 11 17:17:07 gah Dec 11 17:18:36 rburton: -mmicromips is a valig CFLAGS for a mips gcc. Dec 11 17:20:42 vicenteolivert: well, clearly not for the gcc that it's running. Dec 11 17:21:30 rburton: I don't have any idea about how to fix that :S Dec 11 17:23:58 vicenteolivert: http://gcc.gnu.org/onlinedocs/gcc-4.8.2/gcc/MIPS-Options.html#MIPS-Options says otherwise Dec 11 17:25:19 vicenteolivert: i'd guess those variables are mentor/cs specific Dec 11 17:25:30 (kergoth might know) Dec 11 17:27:28 not sure offhand, would have to go pester the folks in that department Dec 11 17:27:45 vicenteolivert: so i'd guess either 1) patch your gcc 2) stop using those options 3) use mentor's toolchain layer Dec 11 17:27:57 Uhm... Dec 11 17:28:44 I'll try removing those options first..., but I guess the board will not boot then. Dec 11 17:38:09 * pidge is running the poky-contrib cleanup right now. any deleted branches will be in poky-contrib-archive Dec 11 17:40:28 nice Dec 11 17:54:16 okay, i see no openssh-server in oe-classic DISTRO_FEATURES Dec 11 17:54:36 what's the proper way to punt dropbear in favor of openssh? Dec 11 17:54:45 moin, btw... Dec 11 17:57:29 mr_science: OE-Classic questions should really be asked in #oe Dec 11 17:57:46 mr_science: good morning to you also :) Dec 11 17:57:56 stupid default tab... Dec 11 17:57:58 thanks Dec 11 17:58:29 ... and the new fetch code explodes if you try to use a tag in SRCREV Dec 11 17:58:35 this stupid branch check is really starting to get on my nerves Dec 11 18:07:04 anybody use smart package manager? Dec 11 18:11:51 i used smart to install a package. subsequent installs error out because e.g. libc6 "is already installed" Dec 11 18:16:53 j8: that is very strange Dec 11 18:17:02 j8: is that something that would be easy for me to reproduce here? Dec 11 18:24:51 bluelightning: i am on dylan. i installed base-files to test out a fix. that succeeded, so i tried installing packagegroup-core-boot which yielded the error Dec 11 18:25:33 j8: hmm, packagegroup-core-boot itself should already be installed, unless you're building a highly customised image (or you're upgrading...) Dec 11 18:29:01 bluelightning: yea, i'm doing this because i thought the image i got didn't have any smart package data Dec 11 18:31:04 j8: was package-management in IMAGE_FEATURES when this image was built? Dec 11 18:31:30 <_tid> hello :) Dec 11 18:31:44 <_tid> i was wondering if there was any way to include kernel modules into the ubi image ? Dec 11 18:32:35 ok, poky-contrib got a first pass cleaning. more cleaning in a bit. Dec 11 18:39:43 bluelightning: i've got package-management in a IMAGE_FEATURES_append, so now i'm wondering if i'm doing it right Dec 11 18:40:06 j8: with a leading space in the value? Dec 11 18:43:10 j8: turning this around, if you do "smart query --installed" does it actually show the full list of packages that should be installed on the system or just the few that were installed when you installed base-files? Dec 11 18:53:37 bluelightning: can't say right now because i am hacking things right now. should i be seeing a /var/lib/smart directory in the built image? Dec 11 18:54:59 j8: you should yes Dec 11 19:06:51 bluelightning: strange, i'm not getting anything in /var/lib/smart. should i be able to do IMAGE_FEATURES = "x11" and then IMAGE_FEATURES_append = " package-management" Dec 11 19:07:25 j8: you should yes Dec 11 19:08:21 j8: I'd try bitbake -e your-image | grep ^IMAGE_FEATURES= Dec 11 19:16:27 bluelightning: IMAGE_FEATURES does have package-management Dec 11 19:18:22 j8: does your image recipe inherit image or core-image? Dec 11 19:18:56 bluelightning: core-image Dec 11 19:19:18 j8: there isn't a line adding "remove_packaging_data_files" to ROOTFS_POSTPROCESS_COMMAND is there? Dec 11 19:20:46 bluelightning: that line ended up in my *image*.bb - is that eliminating the smart metadata? Dec 11 19:21:20 j8: yes, it will be Dec 11 19:21:21 j8: you need to remove that Dec 11 19:22:16 mystery solved Dec 11 23:11:59 I'd be interested in learning more about Yocto, but I get put off by the naming that is used. What is Yocto, Poky, sato, etc, etc.. Why are these names being chosen, and what relevance do they have to what they identify Dec 11 23:12:27 Why is the QUick Start Guide such a jumbled mess of information Dec 11 23:13:14 Should I really feel confident that anyone other than script kiddies are involved with this project ? Dec 11 23:13:17 yocto is an umbrella project which includes many things, of which poky is one. poky is its buildsystem Dec 11 23:13:28 did you not see all the major corporations which are members of yocto? Dec 11 23:13:32 go look at the website Dec 11 23:13:53 what do you mean build system ? You mean a replacement for unix make, or cmake ? Dec 11 23:14:28 no, it's higher level than that. it builds an entire linux distribution, all the components involved, bootloader, kernel, gui, everything, giving you a filesystem/image to put on your device Dec 11 23:14:37 it runs make under the hood to build individual components Dec 11 23:14:54 (among other tools) Dec 11 23:14:54 So, why is it called Poky ? Dec 11 23:15:35 yocto is a metric prefix, indicating its use for small systems Dec 11 23:15:41 no clue what poky originated from, personally Dec 11 23:15:43 don't know, don't care Dec 11 23:16:28 OK, so Poky is the build system, what's bitbake ? The executable that actually operates on "the build system" Dec 11 23:16:50 In which case, why isn't the build system called bitbake ? Dec 11 23:18:06 poky is yocto's buildsystem, it uses multiple open source components under the hood, of which bitbake is one Dec 11 23:18:21 those components existed before yocto or poky, and are used by other projects than just poky Dec 11 23:19:06 i'll admit it can be confusing, but the decisions weren't made randomly and without reason Dec 11 23:21:07 OK, so let's start over. Poky is the build system, but it uses a variety of tools to accomplish the build. So, let's clarify what we mean by a 'build' I regard a build as the process of a 'make' like system working through a set of makefiles and executing toolchain commands such as compile, archive, strip, link, etc. Dec 11 23:22:19 So, where does Poky and bitbake fit into that picture Dec 11 23:24:55 kergoth: poky is a sweet in japan Dec 11 23:24:56 i already explained that. it builds an entire linux system for an embedded system Dec 11 23:24:56 (usually transliterated to pokey, iirc) Dec 11 23:24:58 a linux distribution consists of a great deal of software, go read up on redhat and the like to get an idea of what all is involved Dec 11 23:25:04 but so does make, or cmake. Dec 11 23:25:07 it builds everything obeying their dependencies Dec 11 23:25:10 no, they don't Dec 11 23:25:28 well, in rare cases. buildroot is a make based system which is similar, but much less flexible Dec 11 23:25:38 make is generally used to build one piece of software Dec 11 23:25:47 not the hundreds of pieces of software involved in an entire linux distribution Dec 11 23:25:55 every distro has their own high level tools for this Dec 11 23:26:11 this is no different than those, other than goals and embedded focus Dec 11 23:27:52 so, Poky is the equivalent of something like 'standard' linux make menuconfig ? Dec 11 23:27:57 no Dec 11 23:28:04 make menuconfig is how you configure the build of a linux kernel Dec 11 23:28:11 that's one of many components in a linux distribution Dec 11 23:28:21 a kernel is just one small piece of a distro Dec 11 23:28:34 strictly speaking Poky is a reference distribution, showing how to assemble the pieces in Yocto (bitbake, oe-core, etc) into a distribution. Dec 11 23:28:46 well put Dec 11 23:28:46 its also the distribution that Yocto does QA on Dec 11 23:29:14 bitbake is the tool that decides what to build, and how. Dec 11 23:29:27 so bitbake is like cmake Dec 11 23:29:30 no Dec 11 23:29:34 oe-core is the "core" metadata, linux, glibc, gcc, etc. Dec 11 23:30:04 cmake is a buildsystem used to build a single piece of software, not unlike autoconf/automake/make/etc Dec 11 23:30:04 Smitty: its not a great comparison but bitbake/oe-core == make/makefiles Dec 11 23:30:13 its similar, but at a different scope Dec 11 23:30:30 what it does is similar, in that it runs tasks, obeying inter-task dependencies Dec 11 23:30:47 but tools like make/cmake are oriented toward building individual source files. make is file-oriented Dec 11 23:31:27 CMake, by reading CMakeLists.txt (and other cmake files) can build anything from a single object file to a library, a suite of libraries, or an entire distribution. It just depends upon how complicated you want it to be. How is that any different from bitbake Dec 11 23:31:42 Smitty: good luck building a distribution with cmake. Dec 11 23:32:00 Smitty: maybe you can - as kergoth said, the buildroot project builds distros using mostly Makefiles Dec 11 23:32:06 again, its a matter of scope and focus. yes, its theoretically possible to build a distro with such tools, and i already pointed out one example of this, but they don't tend to scale well, nor are they as flexible Dec 11 23:32:51 the reason we started openembedded and bitbake is that make was not doing the job Dec 11 23:33:01 Smitty: the difference to anyone who isn't a random hacker looking for a pet project is that bitbake exists, and lets-build-a-distro-using-cmake doesn't. Dec 11 23:33:55 anyway, it's just a tool, move on. Dec 11 23:33:57 next question? Dec 11 23:34:38 if it doesn't meet your needs, there are plenty of alternatives, but this meets our needs, and that of many other companies. yocto is a way for those companies to work together on a solid well tested core for their distributions, so they can focus on what really differentiates them Dec 11 23:34:55 OK, fine, I can understand that. I wouldn't want to try to build an entire distro with CMake, or make. So, Yocto provides the infrastructure and tools for building a distro. Poky is a reference distribution created by Yocto. Dec 11 23:35:09 yes Dec 11 23:36:10 there are plenty of others: angstrom is a popular example of a distro built using yocto pieces Dec 11 23:36:49 bitbake is a tool which can process 'recipes' and construct a set of commands to compile, link, etc Dec 11 23:37:34 build file systems, construct package feeds, etc Dec 11 23:37:36 yep. Dec 11 23:37:42 fine Dec 11 23:38:22 each recipe tells it how to build a particular piece of software. glibc, busybox, bash, whatever. they define dependencies between them, so it can build them in the right order Dec 11 23:38:24 * rburton -> bed Dec 11 23:38:31 it's also capable of cross-compilation, which is an important feature Dec 11 23:39:44 what's what's the sato part of core-image-sato trying to convey ? Dec 11 23:40:28 sato is a particular GUI environment. scroll down to SATO on http://www.pokylinux.org/about/ Dec 11 23:40:43 gnome mobile based gui for embedded devices Dec 11 23:40:55 one of many UIs which are used in embedded systems Dec 11 23:42:08 see also http://blogs.gnome.org/thos/2007/08/01/sato-has-landed/ Dec 11 23:42:19 (which was the original announcement, from ages ago) Dec 11 23:43:55 so, core-image-sato shoudl perhaps be called sato-core-image, since the image contains sato on top of core. Dec 11 23:44:18 Now, I get really confused by the quick start guide when I read this "By default, the target architecture for the build is qemux86, which produces an image that can be used in the QEMU emulator and is targeted at an Intel® 32-bit based architecture" Dec 11 23:44:55 it's a convention for all the core images to be named core-image- Dec 11 23:44:59 minimal, base, sato, etc Dec 11 23:45:24 as i mentioned, it can handle cross-compilation. that is, it targets a platform other than what you're running on Dec 11 23:45:33 whether that's a beagleboard, minnowboard, some custom dev board, or whatevewr Dec 11 23:45:48 the default config targets qemu, an emulation environment, so you can run it on your build machine under emulation Dec 11 23:46:19 Why would I have to specify qemux86, and not just x86. Surely the target (x86) can run in any virtual machine which can host an x86. Dec 11 23:48:41 there is a generic x86 target available also. the qemu one will build a kernel configured specifically for use in qemu, with all the drivers that it emulates, and also builds the tools necessary to run the output of hte build, including qemu itself Dec 11 23:48:59 and you don't have to specify it, qemux86 is default :) Dec 11 23:50:08 how does one target a virtual machine ? Surely you target a real architecture like x86, ARM, whatever Dec 11 23:50:28 no, you specify a specific "machine", not a generic architecture Dec 11 23:50:38 as i mentioned before, it builds the entire distro, including bootloader(s) and kernel Dec 11 23:50:41 whch aren't generic Dec 11 23:50:48 particularly in embedded Dec 11 23:52:07 are you sayng that the compiler is emitting different instructions for qemu hosted x86 than for true x86 ? Dec 11 23:53:00 that's not what i said, though it could be optimizing differently sure. the main thing i said was that the kernel was configured differently, among other things Dec 11 23:54:15 there are kernel compile options for running within a virtual machine as apposed to native hardware ? Dec 11 23:54:42 not generally, no, but what drivers are enabled can be different, as qemu only emulates certain hardware Dec 11 23:55:48 OK, so really this qemux86 target is a subset of a generic x86 target I would imagine. Dec 11 23:56:42 don't really know, i've never tested the generic x86 stuff, it's a relatively recent addition. embedded tends to be quite specific, not building generic kernels that work on any system, though that's changing somewhat Dec 12 00:00:46 So, I'm quite familiar with cross compiling for embedded, and frankly, I don't see any difference between that and compiling native for non-embedded. I still use gnu cc, g++, ld, ar, etc, and of course have to specify the architecture, or specific CPU in some cases. although I'm not constructing an entire OS. Dec 12 00:03:13 So, Yocto is providing a framework to facilitate not only the cross-compiling, but also the construction of the file system, the boot image, etc. Is that right ? Dec 12 00:04:28 Hello everyone, has any of you figured out how to auto-connect a network interface on boot on Dora/Dylan? Dec 12 00:06:35 i'm assuming 'auto eth0' in interfaces isn't doing it :) Dec 12 00:07:13 kergoth: that only automatically brings up when I do ifup -a (vs ifup wlan0) Dec 12 00:07:32 'auto' is intended to bringup on boot also Dec 12 00:07:42 possible the hardware isn't ready at that point in teh boot process, though Dec 12 00:07:46 kergoth: reallly? hmm I see so that's where it's broken Dec 12 00:07:57 kergoth: that is highly possible Dec 12 00:08:08 My immediate question would be, Why would auto-connecting a network interface behave any different on Dora/Dylan than any other standard linux build ? Dec 12 00:08:09 you'd think it'd automatically come up when it becomes available though, via the udev bits Dec 12 00:08:15 kergoth: so I don't need a separte systemd service to do this? Dec 12 00:08:16 as hotplug used to Dec 12 00:08:33 ah, not sure exactly how ifupdown integrates with systemd, or if it does at all Dec 12 00:08:50 actually i think you can either use connman, or you could use systemd's built in interface stuff Dec 12 00:09:01 e.g. systemctl enable dhcpcd@eth0 or what have you Dec 12 00:09:10 kergoth: hmmm Dec 12 00:09:12 afraid i'm not sure off the top of my head Dec 12 00:09:36 Smitty: it isn't in general. but as with desktop distros, the variations in configuration between distros matter Dec 12 00:09:44 kergoth: so the 'auto' command only works when I have systemd correctly configured, right? Dec 12 00:10:10 i'd guess so, it wouldn't surprise me if the ifiupdown startup script just ran ifup -a Dec 12 00:10:13 :) Dec 12 00:10:36 Thanks kergoth, you have been helpful in clearing some of the mud from my eyes. Dec 12 00:10:39 ciao Dec 12 00:10:43 np Dec 12 00:10:43 kergoth: good stuff, thanks for the pointers. I am gonna go fix my systemd scripts now Dec 12 00:10:49 good luck :) Dec 12 00:10:57 kergoth: thanks =D Dec 12 00:11:24 of course wlan config gets complicated, e.g. wpa-supplicant configuration.. Dec 12 00:11:43 not that thats particular hard, i just meant another piece to check the systemd integration of Dec 12 00:11:52 kergoth: ya there are a number of components that i must align Dec 12 00:12:23 of course, if it uses ifupdown, can use the wpa-supplicant ifupdown integration and let ifup handle it, in theory.. Dec 12 00:12:27 * kergoth shrugs, as you say, lots of pieces Dec 12 00:13:32 kergoth: wait, of it's either systemd service + wpa_supplicat OR ifupdown + wpa_supplicant? Not systemd service + ifupdown + wpa_Supplicant? Dec 12 00:14:11 well, a systemd service could run ifup, and then you'd adjust interfaces appropriately, or it coul dhandle the wpa-supplicant/dhcp client bits itself presumably, never messed with that myself Dec 12 00:14:39 i'd guess it'd be less complicated to do the ifup route, but then you need tom ake sure it runs at the right time Dec 12 00:14:51 whereas systemd services which are bound to an interface will automatically run when teh interface shows up Dec 12 00:14:54 (afaik) Dec 12 00:14:58 e.g. the @ one i mentioned before Dec 12 00:15:04 kergoth: gotchat Dec 12 00:15:07 gotcha* **** ENDING LOGGING AT Thu Dec 12 02:59:59 2013