**** BEGIN LOGGING AT Thu Mar 20 02:59:59 2014 Mar 20 07:14:06 koen: do you know anything about which Graphics SDK can be used with which kernels? I'm trying linux-yocto 3.10 and neither 4.05.00.03 nor 4.08.00.01 compiles. Mar 20 07:16:43 haven't tried sgx in a long, long time Mar 20 07:39:12 koen: I suspected so. Any idea who is a good person to ask? Mar 20 07:44:15 prabu, if he's still with TI Mar 20 07:44:20 if not, try Chase Mar 20 07:48:29 thanks Mar 20 07:49:13 morning Mar 20 08:12:58 what is the typical use-case for the bitbake client/server mechanism? Mar 20 08:49:28 good morning Mar 20 09:43:06 morning all Mar 20 09:49:44 hi bluelightning Mar 20 09:55:06 how to inherit fontcache (in my recipe) only if DISTRO_FEATURES has x11? Mar 20 09:55:46 hi all. I am running a quite ancient version of linux kernel in my system that I build with openembedded. currently I am having problems with detecting 4K monitor with xrandr and wondering if it could be related to my linux kernel version? Mar 20 09:55:52 kernel version is 2.6 Mar 20 09:57:01 This is my current recipe: http://pastebin.com/hX99vH2K Mar 20 09:57:14 but I'm not entirely happy with the if x11 stuff Mar 20 09:58:47 anyone who can comment on my question? Mar 20 09:59:33 cumba: it doesn't really sound like an OE question; it would be entirely dependent on the video drivers you are using Mar 20 10:00:55 bluelightning, thanks for your reply. I know that xrandr is related to the graphic driver but the problem is the graphics driver is the latest version. And if I use ubuntu with the same graphics driver version it works fine and it detects 4K monitor. Mar 20 10:01:06 so I was wondering what it could be rather than graphics driver Mar 20 10:01:29 cumba: you could be right, but I couldn't say for sure Mar 20 10:01:38 ok, thanks Mar 20 10:02:16 hi bluelightning, some time ago we discussed about making my own custom DISTRO_FEATURES. now I would need to do something similar but with COMBINED_FEATURES. e.g. i need to 'extend' COMBINED_FEATURES, should i be worry about anything, or is it okay to do so? Mar 20 10:02:19 matt10: looking at how fontcache.bbclass works, you could do something like FONT_PACKAGES = "${@base_contains('DISTRO_FEATURES', 'x11', '${PN}', '', d)}" Mar 20 10:03:10 ndec_: AFAIK COMBINED_FEATURES is only ever checked for specific features, so if you add something new I can't think of a way that would cause problems Mar 20 10:03:33 matt10: (that's after the inherit) Mar 20 10:03:43 hi matt10 Mar 20 10:03:45 er Mar 20 10:03:48 hi mckoan Mar 20 10:04:06 (overzealous tab-completion there, sorry) Mar 20 10:04:20 bluelightning: ok. do you know why COMBINED_FEATURES is done 'manually', instead of just cmobining automatically all features which are both in DISTRO_FEATURES and MACHINE_FEATURES? Mar 20 10:04:51 ndec_: it was a long time ago but I assume the thinking was that only certain features make sense to be controlled in both places Mar 20 10:05:30 right, but that wouldn't hurt to compute the 'joined' list automatically, no? Mar 20 10:09:01 bluelightning: hi, I posted a (v2) patch for fontconfig Mar 20 10:10:03 bluelightning: my problem that my recipe has two "sides": one for console only font and another for x11 (and in that case I'd like to inherit fontcache to have postinst stuff) Mar 20 10:11:47 Hi Mar 20 10:12:28 matt10: setting FONT_PACKAGES = "" (as the line I wrote above will effectively do when x11 is not in DISTRO_FEATURES) will disable the functionality of the fontcache class Mar 20 10:12:31 hi pt Mar 20 10:12:46 I have source code that has .mk files, I want to build that in oe-core Mar 20 10:13:22 top dir has Makefile in which i'm calling the main.mk file Mar 20 10:15:15 will this .mk file be executed after that Mar 20 10:15:17 ? Mar 20 10:17:08 er, well if you're calling it from the Makefile, I would have to assume so Mar 20 10:17:51 by default oe_runmake just calls make (with some arguments), and of course that expects to find the normal Makefile, but you have one of those Mar 20 10:18:34 if you need to do something custom when compiling you can always define your own do_compile function (or extend the existing one with do_compile_prepend / _append) Mar 20 10:20:56 yes i already i have the Make file in the top dir which oe_runmake calls, next in that Makefile i'm calling i'm calling the main.mk file Mar 20 10:21:44 so my question is will these .mk files run in the oe environment Mar 20 10:21:55 re Mar 20 10:24:50 pt: as I said, if you're running them from the makefile, I can't see why not Mar 20 10:27:28 ok, so if i use $CC there it will take for whichever toolchain i have set? Mar 20 10:28:34 you should use the CC variable yes Mar 20 10:32:07 ok thanks bluelightning Mar 20 10:35:37 bluelightning: This is working fine, could you quickly review my recipe: http://pastebin.com/M5bchUH7 before I post it to meta-oe mailing list? Mar 20 10:39:50 can we source a script using bitbake file? Mar 20 10:53:34 i mean as source a shell script Mar 20 10:57:47 pt: within a shell function you can run just about any shell command yes Mar 20 10:58:13 pt: but remember that the default shell is /bin/sh, which may not necessarily be bash Mar 20 11:00:47 "source xxx" is a bashim but ". xxx" is posix, technically this should work Mar 20 11:10:40 but in bitbake it is not working Mar 20 11:15:43 could you pastebin the relevant portions of your recipe please? Mar 20 11:27:23 once i've customized an image to build using the "hob" ui, is there any way I can make it generate an image recipe so that I can store it for later use? and possibly invoke the build using the console UI ? Mar 20 11:31:43 mago_: I think there is an option to save the image recipe yes Mar 20 11:55:35 hi, i'm trying to explicitly add 'libegl' to an image (packagegroup), and since I have two providers for virtual/egl, I have set PREFERRED_PROVIDER_virtual/egl = "mali-bin". The preferred provider normally works fine, but bitbake picks mesa instead of mali-bin for satisfying the 'libegl'. I've tried adding PREFERRED_PROVIDER_libegl, but obviously this doesn't help. I can't seem to find a way to force OE to pick mali-bin instead of mesa Mar 20 11:55:56 any ideas? Mar 20 11:56:06 http://pastebin.com/tDnd7PGu Mar 20 11:57:09 The reason seems to be that OE adds virtual/libgl to the equation, which I don't understand or see where that is coming from Mar 20 11:59:01 (mali doesn't provide virtual/libgl, so I don't change the OE default PREFERRED_PROVIDER_virtual/libgl) Mar 20 12:20:55 simply: bitbake mali-bin is working ? Mar 20 12:44:30 matt10b: yes Mar 20 12:59:30 hi! Mar 20 12:59:33 I need help Mar 20 12:59:52 I just cannot figure out how to disable the thumb tune feature when building an image Mar 20 13:00:04 my tactic is Mar 20 13:00:51 DEFAULTTUNE = "cortexa8hf-neon" Mar 20 13:00:52 require conf/machine/include/ti816x.inc Mar 20 13:02:25 but somehow Mar 20 13:02:49 when I run bitbake it gives Mar 20 13:02:49 TUNE_FEATURES = "armv7a vfp thumb neon callconvention-hard" Mar 20 13:03:05 which translages into a gcc -marm -mthumb Mar 20 13:03:12 translates* Mar 20 13:03:19 normally DEFAULTTUNE is a DISTRO setting. are you sure it's not overriden in the DISTRO? Mar 20 13:03:34 I am using angstrom Mar 20 13:03:42 but Mar 20 13:03:47 I have my own machine Mar 20 13:03:54 not using the distro's default machines Mar 20 13:04:19 angstrom will set it indeed. Mar 20 13:04:27 do you know where? Mar 20 13:04:30 wait Mar 20 13:04:31 il grep Mar 20 13:04:34 see https://github.com/Angstrom-distribution/meta-angstrom/blob/master/conf/distro/include/arm-defaults.inc Mar 20 13:05:05 I have it Mar 20 13:05:06 I can see Mar 20 13:05:13 there is code that inserts the thumb Mar 20 13:05:29 what is the correct way to change that? Mar 20 13:06:09 wait Mar 20 13:06:17 features = d.getVar('TUNE_FEATURES', True).split() Mar 20 13:06:18 if 'armv7a' in features: Mar 20 13:06:28 there the TUNE_FEATURES are already set Mar 20 13:06:33 its just how it outputs them? Mar 20 13:07:12 ok this is the place Mar 20 13:07:14 thanks man Mar 20 13:07:19 I have something to play with Mar 20 13:07:43 you can certainly change 'tune' there. but be aware that you are changing a critical distro config, and you be in untested territory. Mar 20 13:07:43 they say I must change the value Mar 20 13:07:58 yea I need to give it a go Mar 20 13:08:15 is there a correct way of making that change Mar 20 13:08:17 ? Mar 20 13:08:22 like normally I would do a bbappend Mar 20 13:08:28 but that is an inc file? Mar 20 13:08:41 or must I branch Mar 20 13:09:23 at what point/phase in processing a recipe does Bitbake download stuff specified in SRC_URI? (if running standalone, as in no OE stuff at all) Mar 20 13:09:37 mago_: do_fetch Mar 20 13:09:38 pompomJuice: you have to patch, can't use bbappend for that. Mar 20 13:10:26 patch what? Mar 20 13:10:36 the arm-defaults file? Mar 20 13:10:57 because ./oebb update has to have the remote branch Mar 20 13:11:03 or... Mar 20 13:12:04 bluelightning, i see. my problem was that i had defined by own do_fetch. so you're not suppose to do that then, if you want to use the SRC_URI variable? Mar 20 13:12:04 new question Mar 20 13:12:46 mago_: correct Mar 20 13:12:52 how can I change the CC that is generated for a run.do_compile step? Mar 20 13:12:53 mago_: it's rare that people define their own do_fetch - what are you trying to do? Mar 20 13:13:38 it currently contains -marm -thumb -mfloat-abi Mar 20 13:13:52 I want it to be without the thumb for that package Mar 20 13:18:45 cant find include arm-defailts.inc anywhere Mar 20 13:20:47 pompomJuice: maybe you want ARM_INSTRUCTION_SET = "arm" in the recipe in question Mar 20 13:21:17 awesome! Mar 20 13:21:20 I will try that Mar 20 13:21:50 because I have this package from TI "multimedia-utils" that cannot do thumb and hard float together Mar 20 13:21:54 the rest works Mar 20 13:22:04 bluelightning, mostly trying to learn bitbake. i've created a small layer (not depending on OE) to experiment with dependencies, fetching of files and writing a few classes of my own. the specific case is a recipe which just downloads a textfile from the internets and reverses the contents, i later plan to try the bbappend mechanism to alter that behavior from another layer. that is why im fiddling with custom do_fetch etc Mar 20 13:22:33 mago_: ah ok, in that context, fair enough :) Mar 20 13:23:09 or rather, i was writing my own do_fetch with calls to bb.fetch2.Fetch() etc (as the manual suggests), but now i will just use SRC_URI directly, suites my case just fine Mar 20 13:23:52 mago_: right, the knowledge of SRC_URI is in the OE code that calls the fetcher, rather than the fetcher itself Mar 20 13:24:15 mago_ : I know how to do what you want to do Mar 20 13:24:44 you must write your own fetcher that supports txt files Mar 20 13:24:46 bluelightning, hm, so without OE, SRC_URI will not function? curious, because it seems to work just fine here, as long as i dont mess with do_fetch Mar 20 13:25:42 mago_: see python base_do_fetch() in meta/classes/base.bbclass - notice how it grabs the value of SRC_URI and passes it to the fetcher Mar 20 13:28:21 bluelightning, yes, but my tree does not contain that base.bbclass. it still gets downloaded, and I'm also getting warnings about not having a md5sum for my SRC_URI. so i guess bitbake on its own also acts on SRC_URI content Mar 20 13:37:37 mago_ Mar 20 13:37:46 you are on the right track Mar 20 13:37:57 but still on very basic levels Mar 20 13:38:02 do you know about yocto project? Mar 20 13:38:27 you can glean so much more information from the documentation there, than find it here Mar 20 13:38:47 I know what you are saying... its just very basic. You can get all those answeres for free in the docs Mar 20 13:39:26 OR differently put Mar 20 13:39:35 your questions have long complicated answers Mar 20 13:39:42 like your MD5 sum one Mar 20 13:39:51 bitbake looks at the URI you give it Mar 20 13:40:11 if it sees git://github.com/myproject.git Mar 20 13:40:29 it automagically uses git to fetch and builds it aswell using auto tools Mar 20 13:40:39 in your case Mar 20 13:40:42 pompomJuice, i'm well aware about checksumming, and what bitbake does with it. My question is only about what aspects of SRC_URI processing/support bitbake does on its own, and what parts of it is in OE-core layer Mar 20 13:41:13 aah Mar 20 13:41:15 I see Mar 20 13:41:16 ok Mar 20 13:41:34 then your best net is to go read the python inside sources/bitbake Mar 20 13:42:02 mago_: right, if you don't specify any URI in the constructor it does seem to grab SRC_URI itself Mar 20 13:42:03 oh Mar 20 13:54:22 bluelightning, but i'm not even instanciating bb.fetch2.Fetch() in my code? Mar 20 13:54:30 and nowhere in my base.bbclass either Mar 20 13:55:03 i'm only doing this: fetcher = bb.fetch2.Fetch(['http://ftp.sunet.se/pub/os/OS2/hobbes/00filesd.txt'], d) Mar 20 13:55:16 mago_: are you sure it's using the class you expect? bitbake has no intrisic idea of do_fetch Mar 20 13:55:30 *intrinsic Mar 20 13:56:55 bluelightning, well, do_fetch does not seem to exist. if i run "bitbake myrecipe -c fetch", it results in error: ERROR: Task do_fetch does not exist for target test Mar 20 13:57:10 so clearly, do_fetch is not defined and still something is downloading the contents of SRC_URI :) Mar 20 13:58:56 mago_: you must have a call somewhere then, bitbake won't do it by itself Mar 20 14:02:01 oh, wait. my bad; my recipe downloads two files, one using SRC_URI and one by manually calling Fetch(). the file names are very similar, and i was mistaken. it does not seem to touch the SRC_URI file Mar 20 22:35:05 so I'm trying to enable a kernel option so I'm patching defconfig, but it doesn't seem to be propagated to the .config in the build folder Mar 20 22:35:12 Is there a better way to do that? Mar 20 22:51:29 Don't forget: http://openembedded.org/wiki/OEDAM Mar 20 22:51:39 Get yer travel approved add your names! Mar 21 00:18:06 Crofton|work: i already approved myself Mar 21 00:18:32 had to argue with myself first, but i won... **** ENDING LOGGING AT Fri Mar 21 02:59:58 2014