**** BEGIN LOGGING AT Wed Dec 06 03:00:00 2017 Dec 06 04:45:11 New news from stackoverflow: How to install dlib and update gcc on Yocto Dec 06 06:03:40 RP: kergoth: Hi, I was comparing poky/bitbake with standalone bitbake repository (just to prove my point to someone that it's the same) and noticed that there are a few files which exist only in one of them (other than expected .git(ignore)) MANIFEST.in, TODO, classes, conf in standalone and lib/bb/shell.py in poky/bitbake, especially the TODO one looks quite outdated from 2011, would it make sense to Dec 06 06:03:46 synchronize these files one way or another? Dec 06 08:09:23 rburton: your sanity check for "patch" is OK, i've got the error message with patch 2.6.1 Dec 06 08:12:49 oops he's not here Dec 06 09:59:42 rburton: Hi, your sanity check for "patch" is OK, i've got the error message with patch 2.6.1 Dec 06 09:59:49 nayfe: thanks Dec 06 11:20:05 nayfe: likely Dec 06 11:57:59 pohly: is that swtpm stuff in the 2.11 rc? Dec 06 11:58:12 It should be. Dec 06 11:58:50 cool Dec 06 11:59:13 Yes, it is. Dec 06 11:59:16 we've been testing the rc2, annoyingly it breaks other things but good to know this interrim step makes the upgrade easier Dec 06 11:59:25 (current rc patch just drops all the tpm patches...) Dec 06 11:59:39 $ git log --oneline --author=valluri v2.11.0-rc4 Dec 06 11:59:40 c37cacabf2 tpm: Move tpm_cleanup() to right place Dec 06 11:59:40 f4ede81eed tpm: Added support for TPM emulator Dec 06 11:59:40 ... Dec 06 12:00:36 grea Dec 06 12:00:36 t Dec 06 12:11:05 hi, i get QA-Error "already-stripped" when trying to build chromium on rocko branch for a var-som-dual (imx6). I tried using INSANE_SKIP but the binary produced is huge (4GB!) has anybody had a similar problem before? My host system is ubuntu 14.04 Dec 06 12:12:01 fl0v0: 32b host system? Dec 06 12:12:12 thats definitely my guess too Dec 06 12:12:18 can we refuse to build on 32-bit hosts yet Dec 06 12:12:33 seriously might throw a check in the chromium recipe Dec 06 12:12:47 rburton: https://www.youtube.com/watch?v=zn8ndJOIEsI Dec 06 12:14:15 Linux calypso 4.2.0-42-generic #49~14.04.1-Ubuntu SMP Wed Jun 29 20:22:11 UTC 2016 x86_64 x86_64 x86_64 GNU/Linux Dec 06 12:14:25 doesnt seem to be :/ Dec 06 12:14:26 nuts Dec 06 12:14:33 peanuts? Dec 06 12:14:43 fl0v0: run file on the binary that is 4gb and see what that said Dec 06 12:14:57 my hunch is the "is this binary stripped" test is failing because the binary is to large Dec 06 12:15:22 i wonder if my revised elf parser works with a 5gb binary... Dec 06 12:16:20 chrome: ELF 32-bit LSB shared object, ARM, EABI5 version 1 (GNU/Linux), dynamically linked (uses shared libs), stripped Dec 06 12:16:35 lols Dec 06 12:17:02 hehehe Dec 06 12:17:41 * LetoThe2nd hi5es rburton. right guess, wrong conclusion. Dec 06 12:18:18 sry i dont get it Dec 06 12:18:20 i think you might need to force off debugging in the chromium build Dec 06 12:18:56 curious how big is the binary, exactly Dec 06 12:19:43 it says 4.1G (4296523784) Dec 06 12:20:05 the binary is bigger than the address space your target can deal with Dec 06 12:20:16 hooray chromium Dec 06 12:20:26 hmm Dec 06 12:21:10 evil workaround would be to set DEBUG_FLAGS="" in the recipe, to stop it passing -g Dec 06 12:21:33 ok ill try that Dec 06 12:22:00 thank you :) Dec 06 12:26:46 fl0v0: presumably its not easy to share the binary for some testing i'd like to do Dec 06 12:26:57 aehm Dec 06 12:27:25 ill ask :) Dec 06 12:28:01 dont now should i upload to mega or something? I dont have a space here Dec 06 12:28:07 *know Dec 06 12:28:52 compress it to see how much smaller it gets. google drive or something works too. Dec 06 12:29:11 want to compare file vs my new ELF parser on it Dec 06 12:52:24 rburton: we would like to upgrade to rocko, but depend on this patch to get in: https://patchwork.openembedded.org/patch/146248/ Dec 06 12:52:59 I've obly submitted it for master so far, but is there any chance to get it into 2.5.1 once it gets merged ? Dec 06 12:53:12 s/obly/only Dec 06 12:56:52 yeah probably Dec 06 12:59:09 cool, I'll submit it for rocko once it lands in master Dec 06 13:03:06 fl0v0: I think it has a workround for qemu-x86 Dec 06 13:09:01 otavio: thanks Dec 06 13:22:46 rburton: https://drive.google.com/open?id=198eQnw-HHzkcb3YK4bZ9l2U1vi_1-r54 Dec 06 13:22:58 sweet thanks Dec 06 13:54:25 rburton: the evil workaround with DEBUG_FLAGS = "" worked :) Dec 06 13:54:57 good :) Dec 06 14:13:16 I get this while compiling recipe 'ss' on the do_recipe_prepare_sysroot: Exception: FileExistsError: [Errno 17] File exists: '/srv/builds/default-plus-pyro/build/tmp/sysroots-components/cortexa9hf-neon/canfestival/usr/lib/pkgconfig' -> '/srv/builds/default-plus-pyro/build/tmp/work/cortexa9hf-neon-mx6qdl-poky-linux-gnueabi/sp/99.15.1+hgAUTOINC-3995b7707314-r0/recipe-sysroot/usr/lib/pkgconfig' Dec 06 14:26:01 khem: this patch seems to fix me: https://sourceware.org/ml/libc-alpha/2017-08/msg00588.html Dec 06 15:30:57 Hello, I tested the multiconfig setup on rocko, but I have a problem when using 2 machines with the same architecture (raspberrypi3 and raspberrypi-cm3). It is sometimes trying to bitbake the same package at the same time Dec 06 15:32:45 some packages are declared as 'machine specific'. These will be built for the specific machine and not re-used (other then for that machine) Dec 06 15:33:16 other packages might have a BUG in them, where they are machine specific, but not declared as such. Usually this is due to a reference, in the recipe, to MACHINE, MACHINE_FEATURES or a similar variable that may change from one machine to another Dec 06 15:33:27 the fix in that case is to declare the package to be machine specific Dec 06 15:41:22 and if I want to have it in both? Dec 06 16:17:14 New news from stackoverflow: Intel Edison + Ubilinux + USB to UART = USB Device not created Dec 06 16:43:44 srwalter: that seems a good fix Dec 06 17:04:03 RP, stable/morty-next failed on new cluster with kvm issue on fc25 Dec 06 17:05:59 joshuagl, do you know what might cause "ssh: connect to host 192.168.7.2 port 22: No route to host" ? Dec 06 17:06:21 the qemu died? Dec 06 17:06:42 or, less likely, tap device is hosed Dec 06 17:07:10 k Dec 06 17:07:22 ya when the tap device goes 'nuts'.. either reboot.. or figure out the magic to clear them all and restart.. Dec 06 17:07:28 (I wish there was a YP script that would do that) Dec 06 17:07:38 (I usually reboot out of frustration) Dec 06 17:08:06 if we knew what the magic was, we could create the script. Dec 06 17:08:09 Patches welcome :-) Dec 06 17:08:28 the magic is in the runqemu script.. Dec 06 17:08:37 you have to just start 'shutting down' tap devices.. Dec 06 17:08:49 but the magic is deeeeeeply buried and I usually can't figure it out just reading the script Dec 06 17:09:36 joshuagl, you guys have Harry Potter Dec 06 17:10:48 so would "Couldn't login into serial console as root using blank password" be a problem testimage ? Dec 06 17:40:09 NOTE: Running command: quilt --quiltrc /home/tepperson/gumstix/yocto/build/tmp/sysroots/x86_64-linux/etc/quiltrc pop -a -f Dec 06 17:40:09 NOTE: Exit status of command is 512 Dec 06 17:40:18 what does 512 mean? Dec 06 17:42:35 https://linux.die.net/man/1/quilt doesn't list exit codes so you'll have to look at the quilt source Dec 06 18:24:00 ok i have part of an answer to my bug, quilt says "Patch add-gumstix-board-to-build.patch does not exist; applied empty patch" Dec 06 18:24:59 maybe Dec 06 18:33:01 it seems the patches directory of the git repository im fetching is causing the problem Dec 06 18:43:51 oooh Dec 06 18:44:10 yeah subtle gotcha is that we don't like a repo or tarball to have a patches/ directory in it Dec 06 18:44:29 the moment someone understands the legacy nightmare that is patch.bbclass, we can change where that folder does Dec 06 18:44:44 lets just say it was a bad choice a decade ago when everyone was younger and more innocent Dec 06 18:51:21 rburton: is this documented anywhere? Dec 06 18:55:09 what determines if I can use 'devtool edit-recipe ' on something in a extensible sdk? Dec 06 18:55:56 can I only modify the ones that I added using devtool? Dec 06 19:32:04 Q. How did Yocto Devs roll out an SDK before "populate_sdk" was available? Dec 06 19:35:15 Q. How did Yocto Devs roll out an SDK before "populate_sdk" was available? Dec 06 19:41:25 meta-toolchain? Dec 06 19:41:29 somethin glike that Dec 06 19:41:52 Ah ok Dec 06 19:41:53 Ty Dec 06 19:42:09 You can still do somethin glike that if tyou want Dec 06 19:42:26 Yes I am googling on meta-toolchain. I have to resurrect an old image... Dec 06 19:44:08 yes, meta-toolchain is the way to create a specific targeted SDK.. while populate_sdk is a generic approach Dec 06 19:44:34 (targeted as in has only the pieces you want the SDK user to have access too.. while the populate_sdk has access to all libraries on the image..) Dec 06 19:45:50 Great. It doesn't have to be targeted, but I guess I have no option as this image is pre-1.2 Dec 06 19:53:01 am I misunderstanding esdks? Dec 06 20:15:44 fischerm: by default - there is an option that lets you edit any recipe Dec 06 20:16:14 fischerm: bear in mind that there isn't a mechanism to keep your metadata changes to existing recipes around (i.e. metadata isn't tracked in git) Dec 06 20:16:21 in the eSDK, that is Dec 06 20:16:30 there's a bug open for that but for now it's not handled Dec 06 20:16:34 brb Dec 06 21:17:59 bluelightning: so what I'm trying is basically I have two packages that developers need to rebuild, so I tried something like 'devtool upgrade -S' or 'devtool edit-recipe' and both failed, with either 'no recipe in workspace', and the upgrade -S fails on unpacking when it tries to attach 'extra' bool to a list of strings Dec 06 21:18:34 bluelightning: https://pastebin.com/jgjkAHg4 Dec 06 21:42:04 halstead: are you breaking the autobuilder? Dec 06 21:42:32 rburton, No. I think joshuagl is working on it. Dec 06 21:42:41 it's me! Dec 06 21:43:28 josh! Dec 06 21:43:39 before you broke it, did my build look green or red? :) Dec 06 21:43:56 both :-) Dec 06 21:44:09 that sounds about right i guess Dec 06 21:53:51 im adding a task to a recipe. i only want the task to run with "bitbake bla -c thetask", i want it to automatically run the tasks run by "bitbake bla" Dec 06 21:55:52 uglyoldbob: you mean you don't only want it to run when called directly? Dec 06 21:56:56 bluelightning: "bitbake bla" should NOT run thetask, "bitbake bla -c thetask" should do everything "bitbake bla" does plus thetask Dec 06 21:56:57 dont use a "before ..." clause in the addtask, then Dec 06 21:57:18 rburton: a.y.io is back up Dec 06 21:57:42 uglyoldbob: right, so you probably want "after do_build" I guess, although that would be unusual Dec 06 22:02:05 bluelightning: ok, will that cause thetask to run by default? Dec 06 22:02:45 uglyoldbob: it will not - only when called directly Dec 06 22:03:00 uglyoldbob: if you did "before do_build" it would Dec 06 22:04:13 if you don't specify -c, it runs the do_build task by default (technically do_${BB_DEFAULT_TASK}). every other task involved in the build is pulled in through dependencies from that. so if you don't make one of those tasks depend on yours, it won't run unless you specify it Dec 06 22:11:52 awesome, thank you very much Dec 06 22:45:50 Dear All, Dec 06 22:46:36 Is it possible to specify list of recipe to not be executed? Dec 06 22:46:36 i.e. MY_LIST_OF_EXCLUDED_RECIPES="sss gff ddd" bitbake core-image-XXX ? Dec 06 22:49:01 lukma: you can exclude *packages* using PACKAGE_EXCLUDE... or on the other end, you can use BBMASK to prevent bitbake from parsing a recipe (specifying a regular expression to match the path) Dec 06 22:49:31 lukma: however it's rarely that simple because you're almost certainly going to run into dependency problems Dec 06 22:49:41 bluelightning: In my case..... e.g. Dec 06 22:49:51 I do need to have only rootfs Dec 06 22:49:57 so linux -> linux-dummy Dec 06 22:50:13 and exclude e.g. bootloader, some modules, etc Dec 06 22:50:22 I was wondering if there is such a switch..... Dec 06 22:50:31 (maybe packages would do the trick - I will check it) Dec 06 22:53:47 lukma: well the bootloader isn't usually a package that goes into the image, that gets built by way of EXTRA_IMAGEDEPENDS typically Dec 06 22:54:39 if you don't already have your own image recipe you are customising, I would strongly suggest making one (copy an existing one and modify if you like) Dec 06 22:54:54 then you have more control over what goes into it Dec 06 23:00:52 I do have misbehaving u-boot-fw-utils.... Dec 06 23:01:06 which is somewhat "special" :) Dec 06 23:01:23 this will be fixed Dec 06 23:16:49 lukma: does it fail with include/timestamp.h:11:10: fatal error: generated/timestamp_autogenerated.h: No such file or directory ? Dec 06 23:24:43 No, It is my "invention" to have both UBOOT_IMAGE + UBOOT_MACHINE defined Dec 06 23:24:53 apparently from 2.4 there is some issue with it Dec 06 23:25:18 It is forbidden to have both defined Dec 06 23:48:42 New news from stackoverflow: Yocto: Create a directory after mount **** ENDING LOGGING AT Thu Dec 07 03:00:01 2017