**** BEGIN LOGGING AT Wed Apr 01 02:59:58 2015 Apr 01 08:06:28 morning all Apr 01 08:06:51 morning Apr 01 08:08:05 hi mag0 Apr 01 08:38:40 So frustrating... some bitbake kernel config process somehow refuses to insert CONFIG_NF_* into the kernel config Apr 01 08:39:35 I have tried linux/files/foo.cfg Apr 01 08:40:09 I have tried menuconfig -> safedefconfig -> linux/linux-fslc/machine/defconfig Apr 01 08:40:24 it throws those settings out Apr 01 08:40:42 but they are legal and all dependencies are met Apr 01 08:42:47 other settings that I introduce works fine though Apr 01 08:48:56 pompomJuice: have you tried checking the logs? Apr 01 08:49:19 I have Apr 01 08:49:23 I am not sure where to look Apr 01 08:49:27 when I grep defconfig Apr 01 08:49:32 or my foo.cfg Apr 01 08:49:36 all I see is it is copying them Apr 01 08:49:54 nothing about trying to combine defconfig with files/foo.cfg to create a new config Apr 01 08:50:18 I am not sure when or where that happens Apr 01 08:50:35 I always thought defconfig is copied to git/.config like the code suggests in kernel.bbclass Apr 01 08:50:54 all of a sudden I realise there is this elaborate system that builds a config Apr 01 08:50:59 and it is not working Apr 01 08:51:14 or it works for other settings Apr 01 08:52:40 pompomJuice: appending cfg fragments is only for linux-yocto kernels Apr 01 08:52:49 but if I try to add these settings... they get thrown out ... http://hastebin.com/moxapoxisi.ini Apr 01 08:53:01 ok... Apr 01 08:53:10 pompomJuice: what do you mean with 'combine' ? Apr 01 08:53:13 but those cfg fragments seem to work Apr 01 08:53:34 are the frags listed in SRC_URI? Apr 01 08:53:40 yes Apr 01 08:53:48 I then ditched the fragments system Apr 01 08:54:02 I did a manual bitbake -c menuconfig Apr 01 08:54:06 saved the config Apr 01 08:54:14 theb bitbake -c savedefconfig Apr 01 08:54:26 this will work with non-yocto kernels Apr 01 08:54:54 then copied the defconfig over the old one Apr 01 08:55:04 ok then Apr 01 08:55:10 how is it supposed to work? Apr 01 08:55:20 you just list the defconfig in SRC_URI Apr 01 08:55:29 ok Apr 01 08:55:39 and it will copy that over .config? Apr 01 08:55:44 which is what I want Apr 01 08:55:58 Let me try that Apr 01 08:56:05 I have tried menuconfig -> safedefconfig -> linux/linux-fslc/machine/defconfi Apr 01 08:56:09 ^ this seems right Apr 01 08:56:18 yea it works for other settings Apr 01 08:56:26 but not these settings -> then I diffed the defconfig Apr 01 08:56:35 but not these settings -> http://hastebin.com/moxapoxisi.ini Apr 01 08:57:03 in the end these seems to be a process that combines defconfig with fragments Apr 01 08:57:21 no, it's not exactly like that Apr 01 08:58:03 even using linux-yocto, once you opt for a 'defconfig' the whole processing is skipped so you'll miss the fragments mgmt and the cheks Apr 01 08:58:08 inside my workdir the file git/.config is not like any other file anywhere... it must be generated Apr 01 08:58:55 you just have to refresh the metadata ( /machine/defconfig) Apr 01 08:59:26 then check the configure logs Apr 01 08:59:30 ok Apr 01 08:59:38 what is strange Apr 01 08:59:52 is that there is a file git/.config.old which matches my defconfig exactly Apr 01 09:00:58 adding other settings to defconfig the eventually end up in git/.config.... but these ones refuse -> http://hastebin.com/moxapoxisi.ini Apr 01 09:01:50 and those settings are ligit... Apr 01 09:02:41 they have complicated interdependencies among them... but I have made double sure that all are there... I suspect the config fragment combino system cannot handle those interdeps Apr 01 09:03:11 you are not using any fragment management ;) Apr 01 09:03:21 but why does it work then? Apr 01 09:03:33 I am 100% sure they work Apr 01 09:03:48 see, I don't know which kernel you are exactly building... where is the recipe? Apr 01 09:04:58 are you unknowingly using a linux-yocto kernel ? Apr 01 09:05:53 let me get it for you... Apr 01 09:06:58 it is a freescale kernel Apr 01 09:07:49 the point is, the code is in kernel-yocto.bbclass Apr 01 09:08:06 standard kernels do only inherit kernel.bbclass Apr 01 09:08:45 (the inherit is in linux-yocto.inc) Apr 01 09:10:03 https://github.com/Freescale/meta-fsl-arm/blob/dizzy/recipes-kernel/linux/linux-fslc_3.18.bb Apr 01 09:10:18 aah ok Apr 01 09:10:39 that makes sense, I saw kernel-yocto.bbclass and wondered what that was Apr 01 09:10:49 I am on 1.7 Apr 01 09:12:37 I see that it only does a inherit kernel Apr 01 09:12:43 why does the fragments system work then? Apr 01 09:13:11 are you sure it does? Apr 01 09:13:18 let me triple check Apr 01 09:13:36 try i.e. to add an exotic filesystem with a frag Apr 01 09:13:47 I am adding our exotic driver Apr 01 09:13:59 I can't really imagine how it could happen to work Apr 01 09:14:08 :) Apr 01 09:14:14 let me check first... I am 100% sure it works Apr 01 09:14:17 but let me check Apr 01 09:15:39 ok something is off Apr 01 09:16:06 New line of thought... was MACHINE_KERNEL_PR depricated? Apr 01 09:16:10 how do I bumpt the PR? Apr 01 09:16:19 it seems that I am being fed a cached version Apr 01 09:16:37 always do bitbake -c cleansstate virtual/kernel Apr 01 09:16:38 I removed our exotic driver from config fragments and it still remains in .config Apr 01 09:16:42 when doing changes Apr 01 09:16:43 I do a cleansstate Apr 01 09:17:03 ok Apr 01 09:17:09 I am thinking that I should to a PR bump aswell... bitbake is definitely feeding me a cached config Apr 01 09:17:47 since I have removed all references to our exotic driver yet it remains in the .config Apr 01 09:18:13 how do I do a PR bump? with MACHINE_KERNEL_PR? Apr 01 09:18:25 it shouldn't be needed at this point Apr 01 09:18:52 are you sure you copied back the modified defconfig? Apr 01 09:19:11 yes I did Apr 01 09:19:58 ok wait let me do a new test... Apr 01 09:20:12 I see there are 2 defconfigs under /linux-fslc Apr 01 09:20:22 one is for mxs Apr 01 09:20:45 wait a minute Apr 01 09:20:47 Hi there, I'm trying to compiler valgrind for imx23 but it fails because : incompatible with host arm-angstrom-linux-gnueabi (not in COMPATIBLE_HOST) Apr 01 09:20:52 it might be grabbing the wrong one Apr 01 09:21:39 COMPATIBLE_HOST in recipe is set to : COMPATIBLE_HOST = '(i.86|x86_64|mips|powerpc|powerpc64).*-linux' Apr 01 09:21:52 and there is also COMPATIBLE_HOST_armv7a = 'arm.*-linux' Apr 01 09:22:03 no it is not grabbing those Apr 01 09:22:15 since none of them contain our driver Apr 01 09:22:34 this is a caching issue Apr 01 09:22:50 I removed all mention of our driver and it still ends up in .config Apr 01 09:22:53 even after I do a cleansstate Apr 01 09:23:03 strange...remove tmpdir and let it rebuild from ssstate then Apr 01 09:23:31 that might take a while Apr 01 09:23:51 it's very fast on my laptop... Apr 01 09:23:53 ant_work Apr 01 09:23:56 one last thing.... Apr 01 09:25:12 if I do however make a config fragment and change the driver from built in, to module... Apr 01 09:25:49 maybe I have a btrfs problem here Apr 01 09:26:04 it has been acting up Apr 01 09:26:16 huge iowaits on btrfs-transactio Apr 01 09:27:02 which hangs my entire VM Apr 01 09:30:56 ok then config fragments dont seem to work... you are correct... I thought they worked cause I added one for our exotic driver and removed it from defconfig and it worked... seems like I am getting a cached copy of something else... Apr 01 09:31:24 let me rebuild everything from scratch like you suggested Apr 01 09:33:59 ant_work... ok so fragments don't work... I moved our exotic driver from a fragment to the defconfig and changed it from "y" to "m" and it worked Apr 01 09:34:31 right, that's expected Apr 01 09:34:49 but if I do a savedefconfig and replace the defconfig file with the new one... some settings that have compicated interdeps don't work Apr 01 09:35:37 I think the problem has to do with the fact that http://hastebin.com/moxapoxisi.ini has complicated interdepts and the generator gets confused Apr 01 09:35:45 the .config file still looks generated Apr 01 09:35:57 it is not a exact copy of defconfig? Is that expected? Apr 01 09:36:10 it is the 're-expanded' version Apr 01 09:36:22 I do that daily and have zero issues Apr 01 09:36:31 I stay away from netfilter, though ;) Apr 01 09:36:46 you coukld try that: Apr 01 09:36:58 ok maybe I made a mistake Apr 01 09:38:14 rename the 'right' .config as defconfig and copy it to the metadata. You'll see the size increase but the deps are resolved. You could as well being hitting a kernel bug Apr 01 09:38:44 or I made a huge mistake... let me verify Apr 01 09:38:54 the config fragments confused me Apr 01 09:39:04 I am doing everything from scratch like you suggested Apr 01 09:39:05 let me check Apr 01 09:41:18 apologies ant_work Apr 01 09:41:26 it works now Apr 01 09:42:08 A colleague prepared the kernel and used fragments... which I then immetated... Apr 01 09:42:36 doing everything over with the knowlege that fragments wont work seems to have solved my issues Apr 01 09:44:50 Thanks for your help Apr 01 09:44:54 yw Apr 01 09:48:30 pompomJuice: it's 1st of April...I was tempted to fool you a bit but resisted ;) Apr 01 09:50:01 no worries... it seems I fooled myself there for a while Apr 01 09:50:14 :p Apr 01 09:50:43 btw linux-yocto kernel processing is intriguing Apr 01 09:50:59 you have the fragments (or the old defconfig if you prefer) Apr 01 09:51:02 plus Apr 01 09:51:27 the audit of it, it is split in hw/non hw, deps are verified, etc. Apr 01 09:51:57 It looks much more compartmentalized than the defconfig way... I like it Apr 01 09:52:19 defconfig is quick&dirty... Apr 01 09:52:21 My colleague who is the expert likes it too it seems Apr 01 09:52:49 most kernel hackers prefer old style defconfig...hard to move them to yocto in my limited experience Apr 01 09:53:16 I don't know that that's really true. Apr 01 09:53:25 I can name 2 or 3 Apr 01 09:53:29 AFAICT it's more that nobody has really tried to push fragments upstream. Apr 01 09:53:47 anyway, I have to bother the maintainers and ask if they finally added a 'vanilla' branch Apr 01 09:53:48 I like the config fragments more... it feels more like the embedded way... Apr 01 09:53:50 unpatched Apr 01 09:54:50 if you plan to push kernel patches upstream you cannot work whith pre-patched sources... Apr 01 09:55:19 I was already bitten once Apr 01 09:56:22 broonie: you are one exception ofc ;) Apr 01 09:57:01 There were actually a bunch of us talking about it at ELC. Apr 01 09:57:07 It'd help with the testing stuff for example. Apr 01 17:39:55 and I am guilty of publishing a repo script with the old epo names :) Apr 01 18:05:37 Crofton|hackfest: www.angstrom-distribution.org is faulty, goes to Error 302 page Apr 01 18:05:45 Crofton|hackfest: without www is fine Apr 01 18:05:51 hmm Apr 01 18:05:54 this i snew Apr 01 18:06:53 ka6sox, ? Apr 01 18:07:35 Crofton|hackfest: also where do our mailing lists run off these days? Apr 01 18:19:51 How is the best way to put a kernel source dependency in a bbappend? Apr 01 18:30:33 XorA, list are on ltg Apr 01 18:30:52 Crofton|hackfest: hmm, admin pages are gone so you cannot actually sign up anymore Apr 01 18:31:12 Crofton|hackfest: and I think I got booted off when moving MX records Apr 01 18:33:55 bother Apr 01 18:34:27 Crofton|hackfest: looks like we need to move them over at some point Apr 01 18:35:08 RP: < otavio> How is the best way to put a kernel source dependency in a bbappend? Apr 01 18:35:22 RP: Do you have any suggestion? Apr 01 18:35:37 I am trying to fix a Qt4 error in fsl-arm Apr 01 18:38:07 Crofton|hackfest: ah you can get to the pages direct if you find them in google Apr 01 18:38:21 Crofton|hackfest: it seems I am still subscribed mailing list is just silent :-D Apr 01 18:44:15 would anyone complain if I send in a patch to drop distcc from the core-sdk packagegroup Apr 01 18:44:54 It annoys me that it starts processes and seems like overkill for core-sdk function Apr 01 18:45:28 distcc seems like a fantastic idea for on-target development, particularly in qemu, since it could call out to the host's cross toolchain.. course not everyone wants it, its true Apr 01 18:48:49 yeah Apr 01 18:48:54 easy to add if you want it Apr 01 18:49:06 we just want to be able to build things if needed Apr 01 18:49:16 but distcc starts processes etc Apr 01 18:49:23 so it is "intrusive" Apr 01 18:49:35 ah, it starts distccd on boot? Apr 01 18:49:46 would be cool if it automatically started it dynamically on distcc execution or something Apr 01 18:50:11 right Apr 01 18:50:54 basically I suspect very few people use it, and it has annoyed at least one of our customers :) Apr 01 18:51:00 wait, isn't distccd the server side piece? couldn't you install the client and not the server, and let the server be on the fast machine? or is distccd needed even when configured to talk to a remote host? Apr 01 18:51:07 * kergoth scratches head Apr 01 18:51:08 dunnoi Apr 01 18:51:15 it just appears in ps Apr 01 18:51:21 and I ahven't used it in ages Apr 01 18:51:26 i think a client configuration makes sense, but a server doesn't Apr 01 18:51:27 yeah, me neither Apr 01 18:51:28 if I need speed, I use an sdk Apr 01 18:51:36 probably not since i had my zaurus :) Apr 01 18:51:40 rofl Apr 01 18:51:41 yep Apr 01 18:51:53 though aboriginal linux automatically configures its qemu instance to use distcc to call the crosscompiler on the host for builds Apr 01 18:52:07 which does make a nice little "native" development environment for that arch Apr 01 18:52:23 could see having runqemu do that with our qemu machines Apr 01 18:53:12 I still think it'd be cool to build a lightweight http server and use it to arrange automatic use of DEPLOY_DIR_{IPK,RPM,DEB} as a feed. runqemu, opkg install something-i-forgot-to-put-in-IMAGE_INSTALL Apr 01 18:53:33 * kergoth wanders off to walk up to the coffee shop Apr 01 19:01:17 kergoth: +1 Apr 01 19:19:21 so *that's* what happens when he runs out of corner cases... Apr 01 19:20:38 * nerdboy sneaks elc san diego on kergoth's calendar while he's distracted Apr 01 19:36:29 hah Apr 01 20:30:42 otavio: did you figure it out? Apr 01 20:40:58 RP: yes. I did an annonymous Apr 01 20:41:16 http://privatepaste.com/857a015cf7 Apr 01 20:44:06 otavio: do_configure[depends] = "X" X = "" X_mx5 = ' virtual/kernel:do_shared_workdir' Apr 01 20:44:19 otavio: something like that would like also work assuming mx5/6 are overrides Apr 01 20:44:25 otavio: but what you have is also ok Apr 01 20:49:00 Ahhh I see Apr 01 20:49:01 :) Apr 01 20:50:29 otavio: = "${X}" obviously :/ Apr 01 21:02:45 yes :) Apr 01 21:17:45 any ideas on this cmake error Apr 01 21:17:54 http://pastebin.com/xg63UCg5 Apr 01 21:31:34 man, i hate when i'm stuck waiting on a from-scratch build to progress in my tasks Apr 01 21:31:58 still a lot of overhead in oe/yocto development due to build time, even with sstate, since it's still rather brittle in some ways Apr 01 21:33:04 heh, waiting on rm -rf sstate-cache now :) Apr 01 21:35:30 Certainly helps having a 6 core i7 :) Apr 01 21:45:57 kergoth: ask bluelightning about my "Yocto blues" song Apr 01 21:46:12 "I just changed my sstate" Apr 01 21:46:18 "I've got the Yocto Blues" Apr 01 21:46:20 etc Apr 01 21:46:34 looking for more verses :) Apr 01 21:59:18 When trying to build a makefile project, how do I get it to execute the commands from the externalsrc directory? Apr 01 22:00:11 Do I need SRC_URI=file:// etc to make it pull all the files into a separate build directory? Apr 01 22:01:07 Eep, I hope I'm asking this in the right place. This seemed like a bitbake thing, but perhaps this is a yocto thing. Apr 01 22:04:50 yocto is an umbrella project which includes poky, which is an integration of bitbake, oe-core, and meta-yocto, among other components. saying something is a bitbake thing but not a yocto thing really doesn't make sense :) Apr 01 22:05:56 pumpernickel: I think you want EXTERNALSRC_BUILD_pn- = "/path/to/src/dir" Apr 01 22:06:31 by default it assumes a separate build dir is OK Apr 01 22:11:50 I'm getting a populate_sdk error with dizzy: https://gist.github.com/cbrake/9d311852ae3b33345beb Apr 01 22:12:11 workstation OS is debian wheezy Apr 01 22:12:25 gcc (Debian 4.7.2-5) 4.7.2 Apr 01 22:17:56 kergoth: thank you for the correction. I was reading only today about the differences between them (and how angstrom is still not in yocto). Apr 01 22:18:21 I was thinking that bitbake may have been under oe-core, and yocto "downstream" Apr 01 22:18:25 I misunderstood. Apr 01 22:18:35 poky is one distro, angstrom is another. they have quite different goals and priorities, but they both use bitbake and oe-core Apr 01 22:18:41 the difference is which layers they include Apr 01 22:18:48 Yeah, they are more like siblings. Apr 01 22:18:59 And angstrom seems to be more focused on binaries, rather than building? Apr 01 22:19:27 yocto was created largely to let various companies, who were all revinventing this particular wheel, collaborate on a baseline, and they can all build on that Apr 01 22:19:49 somewhat, yeah, angstrom prioritizes binary packaging and feeds and package upgrades Apr 01 22:20:06 bluelightning: I had tried just EXTERNALSRC_BUILD = "path" Apr 01 22:21:55 Actually, I guess the mistake I was making was that I'm thinking of poky as yocto. Apr 01 22:22:08 When poky is just one "distro", correct? Apr 01 22:22:56 yeah, poky is both yocto's reference distribution, and the repository is an integration of components to provide embedded build tools Apr 01 22:23:09 the yocto project includes other components, not just poky, as well Apr 01 22:24:34 I understand that part, I just slip because I'm still new to it. Apr 01 22:24:45 So bitbake is theoretically part of the foundation, right? Apr 01 22:25:05 Open-embedded and bitbake being the pretty core parts? Apr 01 22:25:53 yep Apr 01 22:26:04 bitbake being the build tool, and oe-core being the core metadata that bitbake uses to build Apr 01 22:39:55 Still having problems with it running the build from the directory containing the makefile. Apr 01 22:42:12 I suppose tomorrow when more people are around, I'll post my recipe (I can't imagine someone could help me much without seeing what I'm attempting.) Apr 01 22:42:33 Unless someone could point me to an example of a Makefile project using externalsrc. Apr 01 22:55:20 pumpernickel: shouldn't matter whether it's a makefile or any other build system Apr 01 22:59:53 The problem I was having was that I couldn't get it to execute from my directory. Apr 01 23:00:03 I now believe I understand the problem. Apr 01 23:01:00 I'm a complete newbie when it comes to this, but I had a do_compile clause in my recipe. Apr 01 23:01:23 I was trying to pass variables to my makefile, to provide the parameters like CC, CXX, etc. Apr 01 23:06:00 Previously, I didn't know about the output things like temp/run.do_compile.xxxx Apr 01 23:06:40 Now that I can see what it is actually running for the different phases, the shroud has been lifted somewhat. Apr 01 23:23:18 bluelightning: Also, thanks! That syntax was the syntax I needed to get it to execute from there. Apr 02 01:02:52 damn, i wanted to know if he liked caraway seeds... Apr 02 01:06:07 nerdboy: now you will never know! :) Apr 02 01:07:18 heh, he thinks he won't be back... Apr 02 01:07:51 * nerdboy will now cross fingers and reboot Apr 02 01:08:07 also wikipedia informs me that its caraway fruits not caraway seeds :| Apr 02 01:08:41 really? Apr 02 01:08:59 http://en.wikipedia.org/wiki/Caraway Apr 02 01:09:08 the little weird things often found in rye bread? Apr 02 01:09:24 although they seem to be at odds with themselves -> http://en.wikipedia.org/wiki/Caraway_(disambiguation) Apr 02 01:09:29 and sometimes pumpernickel even Apr 02 01:10:35 hmm, looks like a new kernel... Apr 02 01:10:53 smells like a new kernel... Apr 02 01:11:05 and tastes like a new kernel Apr 02 01:11:21 * nrossi smashs kernel Apr 02 01:11:22 Linux arm 3.14.35-ti-r56 #2 SMP PREEMPT Tue Mar 31 23:19:08 PDT 2015 armv7l armv7l armv7l GNU/Linux Apr 02 01:11:28 and breaks like a kernel? :) Apr 02 01:11:56 hopefully not, if it does i'm mailing you the pieces Apr 02 01:12:19 * nrossi puts the kernel in a dangerous situation Apr 02 01:12:22 47,000 times... Apr 02 01:12:27 yep it panics like a new kernel :) Apr 02 01:13:04 next graphics blobs Apr 02 01:13:39 * nerdboy just noticed funky linaro toolchain sig Apr 02 01:14:27 needed actual deb env for robert's package script... Apr 02 01:15:10 and now i have shiny new initrd even **** ENDING LOGGING AT Thu Apr 02 02:59:58 2015