**** BEGIN LOGGING AT Mon Jun 13 02:59:58 2016 Jun 13 06:18:38 Hi guys, Im new to yocto. Can someone help me understand how is yocto different from ubuntu in terms of terminal operations? Jun 13 06:19:04 I want to use yocto to work on some FFT calc and also OpenCV. Jun 13 06:19:57 Will the distro provide me resources to work on audio processing and OpenCV without any hassle? Jun 13 06:49:55 sonicabt: i'd guess that a lot of people are using poky-generated images to run opencv or such, but there's no guarantee for it being "hassle-free" Jun 13 06:51:25 sonicabt: first of all you need to get the nomenclature straight: yocto is an umbrella project for a lot of stuff, you probably want to use poky (which is the yocto projects reference distribution). and poky itself is based on the openembedded/bitbake technology. Jun 13 06:54:11 sonicabt: the essential point to get your head around is that all OE-based stuff is more geared towards cross-development and reproductibility, whereas desktop-style distributions encourage interactive use and and in-target development Jun 13 06:54:31 Understood Jun 13 07:08:51 So here's the thing, I wanted to connect a camera and microphone to intel edison(or any other processor which could do image and audio processing simultaneously), read data convert it into text, store it in database and use BLE (bluetooth low energy) to send the data to my smartphone. Jun 13 07:09:51 ok, and what is the question now? Jun 13 07:09:54 :-) Jun 13 07:10:01 i read about intel edison and they say it uses yocto. I dont even know how i can hook it all up. But i really want to be able to. Now that i understand the difference between OE stuff and a liux distro. Jun 13 07:10:19 well you're doing it backwards. Jun 13 07:10:31 oh. ok. Jun 13 07:10:55 the first thing would be to develop your application and make it work on your development host Jun 13 07:11:29 im listening. Jun 13 07:11:42 once that works, you can evaluate its performance and memory needs, then decide on the needed hardware platform, and then (as the last thing) integrate it into the packaging process of whatever distribution you want to use Jun 13 07:12:52 wonderful. Jun 13 07:13:26 ty LetoThe2nd Jun 13 07:13:31 YW Jun 13 07:16:03 could you enumerate on development host? Jun 13 07:21:15 LetoThe2nd: Could you help me understand what you meant by development host in the context of what i want to build? Jun 13 07:21:44 sonicabt: well the personal computer that you are sitting in front of. that you want to use for typing in source code. Jun 13 07:22:03 empirical Jun 13 07:22:29 imperial? :-) Jun 13 07:22:43 hey! this is not yocto specific, but maybe you know the solution: When I boot up the target machine, the serial input does not reach the prompt Jun 13 07:23:06 though the machine output is transfered in the opposite direction Jun 13 07:23:20 T_UNIX: hey! this is not yocto specific, but: find out where it stops Jun 13 07:24:16 ah, by, "reach the prompt" I meant that the prompt is displayed, but the keyboard input isn't arriving Jun 13 07:24:53 everything seems to boot nicely except for that I don't have keyboard input :-/ Jun 13 07:25:05 T_UNIX: ah, then if it has never worked before, check 1) the wiring 2) the echo settings of your terminal emulator Jun 13 07:25:51 everything worked with a different kernel+init system :-/ Jun 13 07:25:57 hm Jun 13 07:28:03 so you can interact with the bootloader prompt and such, just the serial terminal in linux userland doesn't get the input? Jun 13 07:28:35 LetoThe2nd: exactly :-/ Jun 13 07:29:35 T_UNIX: for alck of better suggestions, i'd advise to set up an ssh login, use that and then try to find out which processes are really running, and which is listening on the port Jun 13 07:29:59 I was wondering whether it - maybe - was switching input to 'regular' (e.g. USB) vs. serial Jun 13 07:30:11 LetoThe2nd: okay, got ssh access working Jun 13 07:47:01 LetoThe2nd: so even when the login prompt is shown, I can see lines I drop to it via ssh -> 'echo foo >> /dev/ttymxc0' Jun 13 07:48:23 it seems like just the direction dev machine -> target machine is 'broken' O.o Jun 13 07:50:24 the cursor stops blinking during typing. Though I guess that's just gnome-terminal's way of indicating "you're typing, I'm getting the input" Jun 13 08:02:43 hm Jun 13 08:02:46 sorry, no idea there Jun 13 09:54:59 Hi Jun 13 11:26:31 Seems related: https://www.muffwiggler.com/forum/viewtopic.php?t=162268&highlight= Jun 13 12:03:20 anybody familiar with meta-linaro toolchain ( https://git.linaro.org/openembedded/meta-linaro.git ) Jun 13 12:06:59 bluelightning: can you suggest somebody for toolchain specific question - I am struggling with meta-linaro toolchain :( Jun 13 12:13:43 darkhorse__: ndec perhaps? Jun 13 12:16:22 bluelightning: thanks. I have been using yocto for some time now. Haven't been able to get my head around toolchain recipes.. Is there a good reference that explains how the hell things actually work! Jun 13 12:17:34 darkhorse__: not really, there's a brief overview of some of the recipes involved in the toolchain build process in the manual but it doesn't describe everything Jun 13 12:18:09 bluelightning: :-( so it's all in a few peoples' heads then.. Jun 13 12:18:40 pretty much... most people don't need to dig into it though Jun 13 12:19:20 I don't think anyone would object to documenting it in more detail, it's just not something we've put time into up to this point Jun 13 13:30:46 ndec: hi, I've been referred to you for meta-linaro layer issues.. Jun 13 13:32:07 is there anything I can add to local conf to say, "don't use sstate for this recipe?" Jun 13 13:44:10 CTtpollard: what do you actually want to do? Jun 13 13:44:26 to force a recipe to rebuild i tend to use -C unpack Jun 13 13:44:47 or is it so massive that its a hindrance to put in sstate? Jun 13 13:45:10 rburton: I'm having a problem with a recipe across multiple build pipelines sharing sstate, I want to tell bitbake to not look in sstate for that specific package Jun 13 13:57:00 sorry rburton having issues on my end Jun 13 14:02:41 rburton: did you get my comment "I'm having a problem with a recipe across multiple build pipelines sharing sstate" ? Jun 13 14:06:22 yeah, not sure if theres a good way to opt out of sstate per recipe. RP? Jun 13 14:15:29 CTtpollard, rburton: you can deltask the particular sstate task Jun 13 14:15:52 RP: any example of how I can do this? :) Jun 13 14:19:12 CTtpollard: do_shared_workdir_setscene in kernel.bbclass is one way to do it I guess Jun 13 14:38:39 darkhorse_: you could also try #linaro or #96boards Jun 13 14:46:39 Hmm, -C seems to cause a lot of unhappiness for externalsrc in this setup. devtool modify -x a recipe, then clean and -C fetch or -C configure, and it chokes trying to unpack. will have to see if hte behavior is consistent or specific to this recipe / build environment Jun 13 16:11:01 RP: question, who generally does the recipe upgrades, is it the maintainer of each recipe? and are they tracked? I don't see much in the 2.2 schedule about any upgrades, so i'm guessing they happen outside of bug tracking. have someone at work wondering what version a certain recipe will be, which of course is hard to predict since we don't know what upstream is doing, but .. Jun 13 16:44:30 kergoth: http://recipes.yoctoproject.org/rrs/recipes/2.2/M2/ is how it gets tracked. No point in individual bugs for it Jun 13 16:44:54 kergoth: there is quite some automated tooling but you've just made me realise that perhaps it isn't as well known as I thought Jun 13 16:46:26 i've seen the available upgrade mails on the yocto list, and i've seen the updating of the config files in the metadata about the tracking, but i imagine not everyone has seen those things, so knows even less than i do (which is admittedly little). i was more curious about the process with regard to the release, obviously an unanticipated random recipe upgrade in the middle of the release cycle could affect the schedule Jun 13 16:46:44 hadn't seen that web site, though, quite cool Jun 13 16:52:26 kergoth: there are certain core pieces we like to plan like toolchain/kernel, the rest happens if/as/when until we freeze. Depends a lot on the upstream and what testing says about the update Jun 13 16:53:36 * kergoth nods Jun 13 16:53:37 okay, thanks Jun 13 16:54:28 kergoth: I think it might be worth summarising what tools we have and what we do on the mailing list. There are others who actively work that process and I'll ask them to write something up Jun 13 16:54:49 that'd be helpful, thanks. i'm sure others would find it interesting/useful as well Jun 13 17:41:31 the issue with the package maintainers is the definition may not be clear on what is expected. Jun 13 17:41:57 or what exception is of them Jun 13 17:54:34 Hello Jun 13 18:01:29 Hello Jun 13 18:05:26 hi all, I am trying to use C header from one package in another, but it fails to find it with "No such file or directory" Jun 13 18:05:50 I added the package providing header to DEPENDS in recipe Jun 13 18:07:00 How exactly do you make the header available to the other recipe? Jun 13 18:07:57 Linteriste: in the recipe providing header I install it under /usr/include Jun 13 18:08:48 but now I am thinking I do not need it in image Jun 13 18:09:24 how can I make it available to the recipe? Jun 13 18:09:57 For which purpose do you need the header? To compile? Jun 13 18:10:01 maciejjo: did you see if the header is in the sysroot path? Jun 13 18:10:52 when I extract rootfs I can see the header under /usr/include Jun 13 18:11:21 you wont use the rootfs header Jun 13 18:11:49 if you want to compile the package before install it, you will need the headers on the sysroot path Jun 13 18:12:21 and how to add files to sysroot? Jun 13 18:12:41 it is automatic when you put the package as a depends Jun 13 18:13:03 so you can look at tmp/sysroots/ Jun 13 18:13:07 ok Jun 13 18:13:12 and see if it is there Jun 13 18:14:45 it is there Jun 13 18:14:56 sysroots/beaglebone/usr/include/prussdrv.h Jun 13 18:15:05 ok, so it can be an error on the makefile Jun 13 18:15:21 try to compile the package with devshell Jun 13 18:15:35 and see if everything is ok Jun 13 18:15:38 ok Jun 13 18:17:58 same result Jun 13 18:18:51 did see the CFLAGS if the sysroot path is in there Jun 13 18:18:57 like --sysroot=.... Jun 13 18:19:30 on the cflags Jun 13 18:19:39 no, I do not have this Jun 13 18:19:40 you will need to put -I= Jun 13 18:19:52 --sysroot= isn't in cflags. it's in CC Jun 13 18:20:00 thanks Jun 13 18:20:29 i confused with -I Jun 13 18:20:51 the -I have to be changed to -I= Jun 13 18:21:07 hi.. lets say i want to use poky to just copy files to workdir and create rpm from that. no fetching/patching/configure/compiling for cross platform etc. just simple task of creating an package from an location.. so i do not want bitbake to create a toolchain etc.. how to do this? Jun 13 18:21:13 so the --sysroot option will search on the right path Jun 13 18:26:08 btooth: you can use the INHIBIT_DEFAULT_DEPS and some other variables that exclude the cross-compiling stuffs Jun 13 18:26:25 look at the allarch class into the poky Jun 13 18:27:16 or you can disable the default tasks like do_compile[noexec] = "1" Jun 13 18:27:35 and the recipe will just copy the files and generate the rpm Jun 13 18:31:09 igor2: so what are you saing is I need to add -I=/usr/include to gcc options in my Makefile? Jun 13 18:33:24 maybe, it depends how is your makefile Jun 13 18:33:40 but it can work, you have to try and test Jun 13 18:34:01 with devshell you can change the makefile and try everything Jun 13 18:34:17 until you discover what is wrong on your makefile Jun 13 18:34:27 ok, I will try to figure it out, thanks for help Jun 13 18:34:38 you're welcome Jun 13 18:36:22 igor2: what kind of machine to use for this? i bitbake to take seconds for this from scratch, not building anything, maybe pseudo/fakeroot but thats all.. Jun 13 18:37:48 i mean, download poky and call bitbake from scratch to generate rpm from a folder.. Jun 13 18:39:07 i didnt understand what you want exactly Jun 13 18:39:55 if you want a recipe that only copy files and generate a package Jun 13 18:40:31 you just need to create this recipe as I said before Jun 13 18:41:31 i fetch libs and binaries from a svn server. i use recipies just for file mapping to be installed by do_install. so there is no build envolved. but i want to use bitbake to parse my recipe and to call do_install and do_package_rpm. but all of that without creating sysroots for host or target. Jun 13 18:42:36 poky/yocto is not designed for that, i know.. but i want to use the classes poky ships and bitbake just for that purpose. Jun 13 18:43:51 I dont know if the bitbake will create the sysroot path and etc, but if you inherit allarch on your recipe, it will not need anything Jun 13 18:44:15 igor2: i will try Jun 13 18:47:30 igor2: using allarch does not help.. calling bitbake my-recipe will always first build tons of other tools Jun 13 18:47:50 all needed by MACHINE=qemu .. Jun 13 18:47:54 etc Jun 13 18:50:57 now I understand Jun 13 18:51:08 well, I dont know if it is possible Jun 13 18:53:57 anyone here? some bitbake, or poky experts? i know, i just could use ipkg/rpm, but i have all the file mapping ready in bb files and bitbake and poky ships with all needed classes etc to do this. I just need to tell bitbake that nothing has to be build for a target machine.. just the tools needed for do_install and do_package_ipk Jun 13 19:08:54 btooth: maybe the -b option can do what you want Jun 13 19:09:06 bitbake -b recipe Jun 13 19:09:18 it will only build the tasks of your recipe Jun 13 19:59:21 Hmm, meta-intel is unconditionally modifying linux-firmware. If it's a generic fix, it belongs in oe-core, if not, it should be conditional on the MACHINE/INTEL_COMMON_PACKAGE_ARCH Jun 13 20:14:15 kergoth: bug please! Jun 13 20:14:21 * rburton -> sofa Jun 13 20:53:45 kergoth: did you have a look at the runqueue patch btw? Jun 13 22:20:52 Hmm, would be nice if there was a good way to hint to bitbake that certain tasks need to run in isolation. I keep having to manually scale down BB_NUMBER_THREADS/PARALLEL_MAKE to compile qtwebkit, webkitgtk, etc due to OOM kills, then scale it back up after building them Jun 13 22:22:27 ideally bitbake would know which tasks parallelize on their own and then take as many tokens from a pool as the task will start processes Jun 13 22:23:00 i.e. a make -j8 would take up 8 executors, and on a 16 core machine bitbake would only start 8 other tasks (or 1 other make -j8 compilation) Jun 13 22:23:25 that would solve these problems, wouldn't it? Jun 13 22:24:19 richard had a WIP setup with bitbake operating as the make jobserver Jun 13 22:24:22 i wonder whatever happened to that Jun 13 22:24:45 that's not entirely sufficient, though, since make isn't hte only process which consumes resources Jun 13 22:24:48 it'd be a start, though Jun 13 22:26:24 yeah, I was wondering how hard it would be to emulate a make jobserver, and whether that's a stable API Jun 13 22:28:55 But unfortunately I don't have the time to implement any of this, so we'll just turn down the number of threads and ignore the problem for now :/ Jun 13 22:30:32 there it is, "[OE-core] [POC PATCH] Add shared make jobserver support" from last feb. "I'm sending this out so at least the code is available to people. Its not ready for merging in its current form but might be the basis for someone else to finish this up" Jun 13 22:31:01 Hm, I might put it in my local copy and see what happens Jun 13 22:32:42 that's what i was thinking too Jun 13 22:32:44 i'm not sure it needs a make intercept, could always create the fifo in an event handler instead of at first call of make and set up the makeflags there Jun 13 22:32:45 hmm Jun 13 22:34:45 hmm Jun 13 22:34:46 the make intercept could also use os.execv() instead of subprocess Jun 13 22:34:52 looks pretty simple other than that, though. Jun 13 22:35:35 oh, i see, it passes the actual file descriptors on the commandline, so the child needs to inherit those, the intercept makes more sense now Jun 13 22:36:35 In any case, that's a topic for tomorrow. Night. Jun 13 22:36:45 night **** ENDING LOGGING AT Tue Jun 14 02:59:59 2016