**** BEGIN LOGGING AT Mon Nov 28 03:00:00 2016 Nov 28 06:47:32 Good morning everyone! reibax here, from Spain Nov 28 06:47:56 anyone up yet over here? Nov 28 06:50:52 I need a little bit of help understanding sysroot handling in Yocto 2.1 Krogoth Nov 28 07:10:21 Good morning everyone! reibax here, from Spain Nov 28 07:10:26 anyone up yet over here? Nov 28 07:10:33 I need a little bit of help understanding sysroot handling in Yocto 2.1 Krogoth Nov 28 07:11:19 how would be the right way to force a directory to be copied to the target sysroot? Nov 28 07:11:34 *how=what Nov 28 07:18:16 reibax: http://www.yoctoproject.org/docs/latest/ref-manual/ref-manual.html#var-SYSROOT_DIRS Nov 28 07:19:24 Great nrossi!! Thank you very much! I will test that right away! Nov 28 07:26:16 Wait... I think that is something new to 2.3 Nov 28 07:26:22 it's not available in 2.2 Nov 28 07:26:41 I mean Nov 28 07:26:44 new to 2.2 Nov 28 07:26:47 not available in 2.1 Nov 28 07:27:28 I should maybe use an append for do_populate_sysroot then? Nov 28 07:29:28 python sysroot_stage_all () { oe.path.copytree(d.expand(""), d.expand("${SYSROOT_DESTDIR}")) } Nov 28 07:29:36 is the way i have done it in the past Nov 28 07:30:10 thought you will need _append if you are adding additional sysroot content. Nov 28 07:30:26 great, I will try that. Thank you very much! Nov 28 08:51:16 Good morning Nov 28 09:04:34 Good morning aV_V Nov 28 09:09:57 I don't understand why I have generated an nativesdk- package for everything Nov 28 09:10:36 like nativesdk-openssl, nativesdk-opkg... Nov 28 09:12:44 quite surprisingly, the openal-soft recipe (meta-openembedded/meta-multimedia/) does not put any lib stuff in packages here... Nov 28 09:49:52 does anyone know about a recipe using npm-install so that i can use it as an example ? Nov 28 09:50:21 i retrieved the nodejs and nodejs nodejs-contrib meta Nov 28 09:52:03 binarym: what do you mean by "using npm-install"? Nov 28 09:57:07 LetoThe2nd: do you know npm and node.js ecosystme ? Nov 28 09:57:32 most of node packages are installed using npm: node package manager Nov 28 09:58:33 binarym: i do know that. so you are looking for a recipe that gives you an example for something that usually gets pulled through npm using the install command, right? Nov 28 09:58:54 yep Nov 28 09:58:56 right Nov 28 09:58:56 :) Nov 28 09:59:11 binarym: here you are: https://github.com/imyller/meta-nodejs-contrib/blob/master/recipes-extended/forever/forever_0.15.3.bb Nov 28 09:59:39 binarym: please note that it is inherit allarch, which will work only for pure-js packages. Nov 28 10:00:02 oki, looks like a good pointer to begin with, thanks a lot LetoThe2nd Nov 28 10:00:05 binarym: and please note furthermore that this is a very generic approach, which results in a lot of bloat. Nov 28 10:00:21 yep Nov 28 10:00:34 i found this one too http://git.yoctoproject.org/cgit/cgit.cgi/meta-intel-iot-middleware/tree/recipes-devtools/xdk-daemon/xdk-daemon_0.0.37.bb?h=master Nov 28 10:00:51 but i don't like the way it calls npm directly Nov 28 10:00:51 https://wiki.yoctoproject.org/wiki/TipsAndTricks/NPM Nov 28 10:01:05 that xdk hting looks ugly. Nov 28 10:01:18 bluelightning: shouldn't we use oe_npm or something like that ? Nov 28 10:02:06 bluelightning: interesting Nov 28 11:54:16 Can I check what packages will be generated before building a recipe? Nov 28 11:55:11 Hello all, I am trying to build my custom core-minimal-image using yocto (poky-morty-16.0.0). In this I have added my meta layer that have a fragmented kernel configuration file. I have also created a "recipes-kernel/linux-yocto_%.bbappend" and inside that I have mentioned the "SRC_URI" to that fragmented configuration file. Nov 28 11:55:15 When I run "bitbake core-image-minimal" and in the final image when I inspect the "/proc/config.gz" the all kernel parameters are not being configured in it. How can I add my custom kernel configurations in the yocto build. Nov 28 12:01:46 A little help or pointer will be very helpfull. I have added couple of updates/fixes but none of work for me. Nov 28 12:02:54 Robbin_Parrish: not having experience with config fragments myself i can only give some rough ideas... but have you checked that your append actually gets triggered? Nov 28 12:03:28 for one there ist bitbake -e, and there is also something that shows all appends, just don't remember at the moment Nov 28 12:07:10 LetoThe2nd: Thankyou very much for the reply. I really appreciate it. And yes I have ran the "bitbake show-appends" and its shows me my custom appended file. Even some configuration kernel variables are shows in the "/proc/config.gz" but some doesnt. I have also monitored the ".config" and ".config.old" file from the "${B}" build directory during the image build and the ".config" files doent shows that variables in it but t Nov 28 12:08:23 no idea then, sorry - we're just using monolithic configurations here :) Nov 28 12:13:02 LetoThe2nd: Ok no problem. I will keep debugging it untill I make it work. Hoping I will get some help around here from the Gurus. Nov 28 12:13:18 LetoThe2nd: Thanks again. Nov 28 12:13:28 good luck Nov 28 12:16:18 Robbin_Parrish: Have a look at an existing layer that appends configs, http://git.yoctoproject.org/cgit/cgit.cgi/meta-virtualization/tree/recipes-kernel/linux/linux-yocto_4.8.bbappend Nov 28 12:19:59 hello, might be off-topic, but: is evdevtouch good on qt 5.5 with single touch resistive touchscreen? Nov 28 12:20:10 or I should only use tslib? Nov 28 12:21:41 nrossi: Thanks for the reply. FYI I am only using ".cfg" and not the ".scc" file as features. I did also tried with features file but it too didnt work. Nov 28 12:22:27 nrossi: Does that mean I should also include my ".cfg" file to the "KERNEL_FEATURES_append "? Nov 28 12:24:43 Robbin_Parrish: no, SRC_URI. Did you check the .meta logs to see if your configs are just conflicted and just not being enabled cause of config deps. Have a look in "tmp-glibc/work-shared//linux-yocto//linux--standard-build/kernel-source/.meta/cfg" Nov 28 12:25:28 mismatch-all.txt is generally useful. Nov 28 12:28:24 nrossi: FYI the path is being shows as "tmp/work-shared/qemux86/kernel-source/.kernel-meta/cfg/" and here my custom configuration file looks like missing. Let me check the "mismatch-all.txt" file. Nov 28 12:29:01 thats fine, the one i linked i just a symlink to the shared one Nov 28 12:29:41 the merge_config... file will also be useful, it will tell you if your .cfg file is being processed Nov 28 12:32:24 nrossi: I have checked the "mismatch-all.txt" file and it shows me my custom configuration file. The path reflect to "tmp/work-shared/qemux86/kernel-source/.kernel-meta/configs/standard/". Is it normal? Let me open the "merge_config" .. Nov 28 12:33:01 nrossi: this above path shows me my custom configuration file and its there with all the parameters. Nov 28 12:36:19 my main problem is as follows: how do I invert y axis for a qt 5.5 app that uses tslib? Nov 28 12:36:26 Robbin_Parrish: If you have entries in the mismatch-all.txt file, this means there are issues with setting the configs you want set. Nov 28 12:36:40 nrossi: the both build and audit logfile shows me my custom configurations being merged to the kernel. If it is then why the couple of other parameters are not being merged correctly? Nov 28 12:39:54 Robbin_Parrish: Post your merge_config_audit.log and mismatch-all.txt so we can see what is going wrong Nov 28 12:42:21 nrossi: FYI looking back to the "mismatch.txt" its shows the same value being applied by the two or more configuration files from the features files, its shows like "Requested value" and "Actual value:" Nov 28 12:43:30 Robbin_Parrish: Yep, its listing all the configs that were requested by don't appear in the .config, and showing the difference Nov 28 12:43:38 nrossi: its shows that couple of my custom configurations parameters as "is redefined by fragment" for my file. Nov 28 14:12:25 It would be useful to hear how other people manage their build folders. It takes me quite a while to delete my old ones when I get short on disk space. Nov 28 14:12:56 use rm_work if you find yourself running out of space, and getting a new disk isn't an option Nov 28 14:13:39 and putting the build dir on a separate partition means you can delete them with a mkfs Nov 28 14:13:41 What are recommended HDD sizes for work? Nov 28 14:13:47 well depends on what youre building Nov 28 14:13:59 if you build qt5 and webkit and chromum, then a lot more than a console based system Nov 28 14:14:10 i have a 2tb work disk Nov 28 14:14:17 I use a 1 TB HDD. I build qt5, but not chromium. Nov 28 14:14:24 but many people just use a 30gb tmpfs and rm_work Nov 28 14:14:57 30 GB tmpfs? Nov 28 14:16:49 I am still learning linux stuff in general so I am not really sure what you mean. Could you elaborate? Nov 28 14:17:48 create a tmpfs somewhere with a nice large size, assuming you have plenty of ram Nov 28 14:18:24 oh, so you mean a ramdisk Nov 28 14:20:06 Yes, except the filesystem itself is called tmpfs today on Linux Nov 28 14:22:06 rburton, JoiF understood, thank you Nov 28 14:22:17 Or.. the new implementation is called tmpfs Nov 28 14:22:46 The main difference being that tmpfs allows you to put size constraints while ramdisk didn't; it grew until you ran out of RAM Nov 28 14:22:49 anyone using SSDs for their builds? would one get severe disk wear this way? Nov 28 14:23:17 s/ramdisk/ramfs Nov 28 14:23:20 I guess it depends on how frequently you are planning to do builds Nov 28 14:23:41 but the build process in general is pretty hard on reads/writes Nov 28 14:23:47 I guess it also depends on which SSDs you choose Nov 28 14:24:23 All of this might also depend on how you manage your expectations.. ;) Nov 28 14:24:36 But yes, we're using SSDs in all our build slaves and in most developer build machines, too, and we're doing multiple builds a day on developer machines and tens of builds on the buildslaves every day Nov 28 14:24:53 I'm happy with my speed building Yocto on an regular spinning-cylinder hdd Nov 28 14:25:13 but to have a 30 GB tmpfs I actually need to have more than 32 GB of RAM on my machine, I presume? Nov 28 14:25:23 SSDs provide quite a speed boost for Yocto builds when compared to an HDD Nov 28 14:25:28 eduardas_m: yes Nov 28 14:26:55 neverpanic: depends how much RAM you have ;-) Nov 28 14:26:55 tmpfs can do swapping, but then it becomes absolutely pointless to use it in the first place. ;) Nov 28 14:27:15 neverpanic, I currently only have 16 GB RAM :( Nov 28 14:27:30 how much ram does it need though? what is "enough"? Nov 28 14:29:55 My rule of thumb is 2G per core. Plus whatever you want to put in a ramdisk, of course. Nov 28 14:30:24 neverpanic: I have core-image-sato building with my branch now btw Nov 28 14:32:15 my rule of thumb is: get as much ram as possible in any way. Nov 28 14:33:16 But this is actually quite interesting. Perhaps I shouldn't settle for doing my builds with on the HDD anymore. Nov 28 14:33:37 ive a hdd, its fine. lots of ram is always good Nov 28 14:33:39 hmmm... is 64gb enough to house yocto+compilind? Nov 28 14:33:52 RP: Nice! Nov 28 14:33:57 I actually have a 512GB SSD via NVMe in my PC, but I'm not using it. Like.. at all. It's not even mounted. Nov 28 14:33:59 JoiF: My point is that its complicated. If you have enough ram, hdd verses ssd is much less of a win Nov 28 14:34:16 RP: Was my stuff useful at all? It seemed like you had most of the stuff figured out already. Nov 28 14:34:24 Yeah. I got 32GB of ram Nov 28 14:34:37 We block sync calls during the build so we never really flush much to the disk if the disk cache is big enough Nov 28 14:34:39 JoiF: once you really have enough ram, the discussion is moot. caching will kick in so heavily that the storage backend literally has no effect anymore. Nov 28 14:34:58 neverpanic: it was a useful sanity check, its interesting that we ended up with roughly the same approach Nov 28 14:35:21 neverpanic: my branch isn't without its issues, I need to write up something for the mailing list about that Nov 28 14:37:35 graphiqs: depends - a console only image with busybox only will probably live happily in 64GB Nov 28 14:43:25 nah, thinking about x11 + qt5 stuff Nov 28 14:50:00 anyone else here had issues with "imx-gpu-viv" before? Nov 28 14:51:57 RP: about blocking sync calls, how does that work? We've been hitting the problem mentioned in http://blog.ronnyegner-consulting.de/2011/10/13/info-task-blocked-for-more-than-120-seconds/ . I was assuming it was related to sync'ing, but maybe I'm wrong. Nov 28 14:52:15 mario-goulart: pseudo intercepts and drops them Nov 28 14:52:30 ah, ok. Nov 28 14:53:08 mario-goulart: there is pretty much never a case where we need a sync call in a build Nov 28 14:54:18 RP: indeed Nov 28 14:54:37 If I would want to append QtSerialBus module to the Qt SDK, should I added on TOOLCHAIN_HOST_TASK, TOOLCHAIN_TARGET_TASK or both? Nov 28 15:17:15 is it right to append packagegroups recipes? Nov 28 15:23:35 Depends on what you're trying to do, but it's a valid use-case, yes Nov 28 15:25:24 you should be careful not to overload its purpose, however. i.e. base should stay base Nov 28 15:26:15 I only want to add a package Nov 28 15:27:06 like this: RDEPENDS_${PN} += "package" Nov 28 15:31:48 it really depends on what you're adding and why, whether it's the best approach or not Nov 28 15:32:02 like neverpanic said, depends on what you're trying to do :) Nov 28 15:32:36 I tried, it seems to work :S Nov 28 15:32:41 no issues Nov 28 15:33:28 as I said before, I'm trying to append QtSerialBus module to the sdk, because by default it isnt there Nov 28 15:35:01 that doesnt' seem unreasonable Nov 28 15:37:34 nice Nov 28 15:53:40 kergoth: I'm curious if you've noticed the performance issue I've mentioned on the bitbake list (and sent a patch for) Nov 28 15:53:56 kergoth: I'm struggling to figure out what exactly is triggering it... Nov 28 16:58:14 halstead: Hi, Any estimated time for the OE-patchwork git-hook to be in place? Will a commit be needed to have a first sync? Nov 28 16:58:30 lamego, Should be today. Nov 28 16:59:04 lamego, Do you need it updating the staging pw as well? Nov 28 17:00:09 halstead: No. We have access to the staging-pw, where we actually test changes before pushing to prod. Nov 28 17:02:39 I need to bisect my kernel. I'm using linux-yocto-4.4 with a custom config, but no patches. How can I keep bitbake from resetting the git checkout state? Nov 28 17:03:33 you can force compile Nov 28 17:09:56 Will building core-image-mydevice use the kernel source state? Nov 28 17:15:42 force compile the kernel, then bitbake your image Nov 28 17:15:47 lamego, The hook isn't working yet but I just updated to production HEAD. Nov 28 17:15:48 Ok Nov 28 17:15:52 There's probably a cleaner way, that's how I do it though Nov 28 17:17:28 halstead: Cool. So most likely the hook will be ready for next commit, right? (expected by mid-week, maybe) Nov 28 17:17:44 lamego, Yes. Nov 28 17:18:02 halstead: thanks! Nov 28 17:19:09 lamego, We may hit some trouble the first time though. Ping me for testing when the time comes. Nov 28 17:19:44 halstead: Got it. I'll ping you. Nov 28 17:46:25 looking for a python expert... classes/linux-raspberrypi-base.bbclass in meta-raspberrypi is trying to figure out the version of kernel being used Nov 28 17:46:58 i'm trying to apply the rt patch to the kernel, so the min_ver is coming out "32-rt43" instead of "32" Nov 28 17:47:58 min_ver is being defined on line 18 http://git.yoctoproject.org/cgit/cgit.cgi/meta-raspberrypi/tree/classes/linux-raspberrypi-base.bbclass Nov 28 17:48:19 how do i get the ver.split() function to ignore the "-rt43"? Nov 28 17:50:16 tlwoerner: create your own .bbclass that would implement the get_dts() and split_overlays() functions and do what you want on it Nov 28 17:50:22 i think this is the cleanest way Nov 28 17:50:34 but i'm a newbie here Nov 28 17:50:40 so i may be wrong Nov 28 17:52:22 tlwoerner: re.split(r'[.-]+', '4.2.3-rt34', maxsplit=3) maybe? Nov 28 17:55:02 binarym: i think it would be idea if the same bbclass could handle both rt and non-rt kernels Nov 28 17:55:15 neverpanic: nice, i think that'll work Nov 28 19:04:20 neverpanic: yes, that works. thanks! Nov 28 19:48:49 bluelightning: thoughts on adding an argument to newappend to create the new append in a layer-specific area based on the layer the recipe file is in? would have to warn if the bblayers doesn't include those paths, though Nov 28 19:48:56 seems to be a fairly common convention, now Nov 28 19:58:54 kergoth: sounds like a reasonable feature yes Nov 28 19:59:16 k, thanks, will add to my list Nov 28 20:31:56 hmm, poky.conf could probably still use a bit of paring down / breaking up Nov 28 20:35:41 the MIRRORS change in poky.conf is redundant, those are already in mirrors.bbclass. PREMIRRORS is fine Nov 28 20:35:55 kergoth: chop chop patches welcome :) Nov 28 20:36:25 mel is still using poky at the moment, need to switch back to oe-core, so was checking to see ho wmuch copying and pasting i'd need to do :) Nov 28 20:37:03 * kergoth makes a list Nov 28 20:37:55 rburton: think there'd be interest in submitting something along the lines of ptest but for examples/samples/demos? add packageconfigs, distro feature, separate package, and consistent paths on disk? Nov 28 20:38:20 right now they're a bit all over the place Nov 28 20:51:22 FYI I have a patch to tidy up that bit of poky.conf Nov 28 20:51:31 I should really get around to sending it... Nov 28 20:51:40 I'm getting as bad as rburton ;) Nov 28 21:16:01 here's the status/background for tinfoil2, in case anyone's interested: https://wiki.yoctoproject.org/wiki/Tinfoil2 Nov 28 21:16:13 (yes, I'll be updating bug 5470 when I have pushed my latest branch) Nov 28 21:16:14 Bug https://bugzilla.yoctoproject.org/show_bug.cgi?id=5470 normal, High, 2.3 M2, anibal.limon, IN PROGRESS IMPLEMENTATION , bitbake-layers cannot be used with memory resident bitbake running Nov 28 21:32:59 can someone point out to me where i can find an explanation of all the different kinds of assignment operators found in bitbake (+=,:=?=,??=,=,..)? Nov 28 21:35:20 fmeerkoetter: http://www.yoctoproject.org/docs/current/bitbake-user-manual/bitbake-user-manual.html Nov 28 21:35:44 fmeerkoetter: specifically http://www.yoctoproject.org/docs/current/bitbake-user-manual/bitbake-user-manual.html#basic-syntax Nov 28 21:35:53 bluelightning: haha Nov 28 21:36:36 bluelightning: thanks. looks obvious now :-) **** ENDING LOGGING AT Tue Nov 29 03:00:00 2016