**** BEGIN LOGGING AT Wed Dec 30 02:59:59 2015 Dec 30 05:26:11 hi i have created a yocto layer but getting an error :ERROR: BBFILE_PATTERN_new-layer not defined Dec 30 12:47:11 exit Dec 30 13:43:47 Hi Dec 30 13:45:21 can somebody tell me how to rebuild a package? "bitbake -c clean mplayer2; bitbake -c build mplayer2" => nothing is done: "Attempted 1691 tasks of which 1690 didn't need to be rerun and all succeeded" Dec 30 13:46:32 chep: try "bitbake -c cleansstate mplayer2" Dec 30 13:47:17 then "bitbake mplayer2" Dec 30 13:48:05 mario-goulart: thanks a lot, it works Dec 30 13:48:48 You're welcome Dec 30 23:51:53 I've noticed some applications I'm compiling with bitbake don't have the relevant shared code library lookups in them compared to a regular Makefile build. Anyone seen this? Dec 30 23:56:03 freach42: its possible that the application's makefile relied on overlinking behavior of ld Dec 30 23:56:09 and got by thus far Dec 30 23:56:17 but now defaults have hange Dec 30 23:56:21 changed in linker Dec 30 23:56:36 so you need to explicitly call out for libraries you need at runtime Dec 30 23:56:39 on linker cmdline Dec 30 23:56:43 for the app Dec 30 23:56:51 indirect linking may not work Dec 30 23:57:07 which package is causing it ? Dec 30 23:58:27 this is a set of shared libraries I've created and an app that links to them Dec 30 23:58:38 OK Dec 30 23:58:39 trying to use bb for this environment Dec 30 23:58:52 then show me how you are linking the app Dec 30 23:59:02 if I do an objdump of (regular build, bb build) Dec 30 23:59:48 the "regular" build shows my libraries as NEEDED dynamic relocs Dec 30 23:59:56 you should still use makefiles or cmake Dec 31 00:00:03 to build your component Dec 31 00:00:12 the bb build only shows libc.so.6 Dec 31 00:00:28 I'm using Makefiles Dec 31 00:00:33 OK you need to show the code Dec 31 00:00:51 I know.... :) Hold on I need to sanitize it.... Dec 31 00:01:53 while I'm doing that, how do I get fully verbose compile/link command lines from bb? Dec 31 00:02:17 bb just calls make. the output of the commands run inthe task is in the task log Dec 31 00:03:14 Bitbake "exe" Dec 31 00:03:14 Dynamic Section: Dec 31 00:03:14 NEEDED libc.so.6 Dec 31 00:03:14 INIT 0x0000157c Dec 31 00:03:14 FINI 0x00003a2c Dec 31 00:03:14 INIT_ARRAY 0x00016000 Dec 31 00:03:16 Makefile "exe" Dec 31 00:03:18 Dynamic Section: Dec 31 00:03:23 NEEDED libMYLIB_tx.so Dec 31 00:03:24 NEEDED libMYLIB_rx.so Dec 31 00:03:27 NEEDED librt.so.1 Dec 31 00:03:28 NEEDED libpthread.so.0 Dec 31 00:03:30 NEEDED libgcc_s.so.1 Dec 31 00:03:33 NEEDED libc.so.6 Dec 31 00:03:35 NEEDED ld-linux.so.3 Dec 31 00:03:52 is the task log a file? Dec 31 00:04:10 don't spam the channel Dec 31 00:04:26 post log.do_compile Dec 31 00:04:35 to some pastebin place Dec 31 00:04:37 sorry, thought you wanted it pasted Dec 31 00:04:43 will do next time Dec 31 00:04:59 the logs are in build tree Dec 31 00:05:31 cd tmp*/work/*/work///temp Dec 31 00:06:16 ok, so if it's just calling make with the new Makefiles another dev added I just need to make sure -his- make is getting the proper $(LD) -lMYLIB_tx -lMYLIB_rx Dec 31 00:07:16 (I'm about two hours into bitbake) Dec 31 00:08:03 no worries everything starts with first step Dec 31 00:08:14 or first stumble Dec 31 00:08:32 so did you write a recipe for your packages ? Dec 31 00:08:44 another dev did that Dec 31 00:08:59 OK then paste your makefiles and recipes somewhere Dec 31 00:09:31 I had written Makefiles, he added new recipes. The project isn't open-source (yet) can't disclose Dec 31 00:10:02 but the /temp/ directory was helpful, I see all of the run. files Dec 31 00:11:05 I would recommend a lot of reading start with getting started to Developer manual Dec 31 00:11:30 agreed, I was hoping to avoid that.... :P Dec 31 00:12:22 kergoth: have you worked with extendible SDK Dec 31 00:12:40 kergoth: My devs are not happy with build times Dec 31 00:12:49 log.do_compile has most of the info I need for now. what's funny is my exes built for a different platform (but same CPU) run on this box Dec 31 00:12:59 build times are bad Dec 31 00:13:07 _really_bad. Dec 31 00:13:26 they are not bad if you set up right Dec 31 00:13:32 are you using sstate ? Dec 31 00:13:34 existing build system is very parallel, 16-30 simultaneous GCC Dec 31 00:13:49 depending on build machine. Dec 31 00:14:05 It looks like bb is going to take at least 4x as log Dec 31 00:14:07 OE has dual parallism Dec 31 00:14:07 long Dec 31 00:14:24 it runs several tasks and each task can have its own -j Dec 31 00:14:25 I'd love to exploit that, or this dev is going to get shivved. Dec 31 00:14:50 my usecase is something uncommon Dec 31 00:15:01 so dont get the notion that build times are bad Dec 31 00:15:15 we benchmark our build all the time. Dec 31 00:15:32 build times are subjective Dec 31 00:15:33 down to 19 minutes for kernel/libs/app Dec 31 00:15:51 but poor incremental build implementation Dec 31 00:16:00 e.g. with yocto/oe we brought down our effective build time to less than 20 mins Dec 31 00:16:08 from over 2.5 hrs Dec 31 00:16:23 I'd love to know your secret Dec 31 00:16:25 yes incremental build is where yocto will help Dec 31 00:16:42 my developers are spoilt Dec 31 00:16:48 as I said this guys gonna get it if he quadruples the build times. Dec 31 00:17:08 We need it where I am because we cross-compile for multiple platforms. Dec 31 00:17:21 same as me. Dec 31 00:17:28 I cross compile 14 machines Dec 31 00:17:37 we do over 2500 builds each day Dec 31 00:17:38 about the same here Dec 31 00:17:41 yikes Dec 31 00:17:58 so, parallelism in "global" build is done how? Dec 31 00:18:08 we use precommit verification Dec 31 00:18:18 and Continuous integration Dec 31 00:18:42 so our turnaround time for a bad commit is important Dec 31 00:18:50 so most devs are just pulling version N or N+1 of libraries? Dec 31 00:18:50 so far its about 30 min to 1hr Dec 31 00:18:53 compiled Dec 31 00:19:03 yes Dec 31 00:19:09 thats the bad part Dec 31 00:19:30 yep, we end up having to tinker with the kernel a lot to work around silicon bugs Dec 31 00:19:32 but we use BB_SRCREV_POLICY = "cache" Dec 31 00:19:49 so it does not fetch new changes everytime underneath a build Dec 31 00:19:53 where is that stored, or is in an env variable? Dec 31 00:19:59 local.conf Dec 31 00:20:05 we use AUTOREVs in recipes Dec 31 00:20:16 so that infact will be quite disruptive Dec 31 00:20:34 since every commit will start affecting all devs Dec 31 00:20:53 and most of the time they dont want to change things like this behind the scenes Dec 31 00:20:59 what about explicit parallelism to make? Dec 31 00:21:09 although I would prefer them to act as integrator :) Dec 31 00:21:30 parallism is calculated at start Dec 31 00:21:36 based on build machine config Dec 31 00:21:52 you can also lock it to something static if you like Dec 31 00:23:11 we just do chrt -b 0 nice ionice make -j10 -l10 Dec 31 00:24:16 BB_NUMBER_THREADS Dec 31 00:24:22 and PARALLEL_MAKE Dec 31 00:24:37 so I echo BB_NUMBER_THREADS=10 >> local.conf ? Dec 31 00:26:14 which release are u using Dec 31 00:27:00 1.28.0 Dec 31 00:27:08 yocto 2.0 ? Dec 31 00:27:31 that was from bitbake --version Dec 31 00:27:40 I don't believe yocto is installed Dec 31 00:28:13 BitBake Build Tool Core version 1.28.0 Dec 31 00:30:35 regarding my concern about the final exe that's build against the libraries... The shared libraries look to have the right dependencies amongst them, just not the final exe **** ENDING LOGGING AT Thu Dec 31 02:59:58 2015