**** BEGIN LOGGING AT Wed Jul 05 03:00:05 2017 Jul 05 07:23:34 I need to have MxDateTime module for python on my Yocto image. When I try to install it. I have the error that i need python.h. How can I install python-dev package in yocto ? Jul 05 08:01:43 I need to have MxDateTime module for python on my Yocto image. When I try to install it. I have the error that i need python.h. How can I install python-dev package in yocto ? Jul 05 08:23:37 hello Jul 05 08:24:14 is anyone using yocto in a CI environment auto-updating the git references in the recipes automatically? Jul 05 08:24:32 at least for a pre-merge gate or something? Jul 05 08:25:12 we are building such a system and I'm wondering if it would make things easier to have all the git references in one central file, and just refer to variables from the recipes Jul 05 08:27:19 robsta: not sure if i understand you correctly. you mean that you want to have a ci running, that you just feed a bunch of git hashes and then it runs the build Jul 05 08:27:38 yes, kind of Jul 05 08:28:07 when a new commit comes into one of the repos holding the code, i need to run a build/test cycle before merging Jul 05 08:28:54 a file-format that is easily machine editable and holds the git revisions for each recipe would help Jul 05 08:29:14 Hey, someone has experience with building python packages on yocto, i got one package (python-kivy) wich has dependencies to python-cython, but wenn i try to build on my machine its always says "Cython is missing, Its required to compile Kivy." , building it outside of yocto is working ... Jul 05 08:31:28 robsta: ah ok. nah, then what we are doing is something different Jul 05 08:31:49 I need to have MxDateTime module for python on my Yocto image. When I try to install it. I have the error that i need python.h. How can I install python-dev package in yocto ? Jul 05 08:31:59 fherr: chances are that the cython detection of this kivy thing is broken for crosscompilation Jul 05 08:33:02 ok, i will look deeper into that Jul 05 08:34:09 fherr: guess you are basing the recipe off something like http://git.openembedded.org/meta-openembedded/tree/meta-python/recipes-extended/python-cson/python-cson_git.bb Jul 05 08:35:37 yes right, and added python-cython to rdepends and denpends Jul 05 08:37:19 LetoThe2nd: do you feel it would be a wild violation of the bitbake concept? Jul 05 08:40:19 robsta: i don't see how it would violate the bibake concept. in fact it would *use* bitbake. you'd have to automatically create some staging area in which the test build is set up and done by bibake Jul 05 08:41:11 LetoThe2nd: yes, ok, well, it would take info away from the recipe and centralise it Jul 05 08:42:49 robsta: well the way how you create your layers and recipe files is totally up to you, i personally don't see a problem Jul 05 08:42:56 cool, thanks Jul 05 08:44:10 robsta: nobody says that the layer has to be a repo of its own. you can certainly have some kind of meta-layer thats just a bunch of magic to create the recipes. it probably will complicate artifact reuse (-> sstate cache) but i don't see no algorithmic argument against it. Jul 05 08:44:37 we have one repo per layer fwiw Jul 05 08:45:07 but i'm looking for a machine-rewritable way of setting git refs Jul 05 08:45:34 one repo per layer is the classic and most wdely used approach of course Jul 05 08:46:07 but nobody says that the layers that bitbake sees have to be maintained by humans. Jul 05 08:51:17 maybe the recipe could default to building git master, but then that could overwritable with a variable from a central conf file Jul 05 08:54:31 robsta: its only software, everything is possible. the question is just, how far you want to go in breaking sstate or reproductibility for your case. Jul 05 08:56:59 but we are already using a variable SRC_URI in a .inc file for each recipe Jul 05 08:57:35 does it not expand the value when checking for sstate cache matches? Jul 05 08:58:59 it should Jul 05 08:59:32 it just means you'd lieft the automated generation out of the recipes into the conf file. Jul 05 09:00:32 shall try Jul 05 09:00:35 cheers LetoThe2nd Jul 05 09:54:25 uhoh robsta is here Jul 05 09:55:01 just like last few months? :) Jul 05 09:55:10 here and asking ;) Jul 05 09:55:21 robsta: for 'latest and greatest' CI style building you can set SRCREV to ${AUTOREV} Jul 05 09:55:27 which goes to the git repo and gets the latest HEAD Jul 05 09:55:53 hm not sure we want that Jul 05 09:56:04 you can do that through a conf file, so your layers can have SRCREV=12345 for the known good, but the CI system can override with latest HEAD Jul 05 09:56:06 rburton: thats what i would have mentioned if he had said "sure, i totally want to break reproductibility" Jul 05 09:56:26 definitely putting AUTOREV in an actual recipe is a capital punishment Jul 05 09:56:55 if you have the hashes in advance you can override srcrev the same way Jul 05 09:57:13 rburton: do you have a recipe/project online where i can look at it? Jul 05 09:57:33 just set SRCREV_pn-recipename = "the hash" in a configuration file somewhere, ie your CIs local.conf Jul 05 09:58:03 http://git.yoctoproject.org/cgit/cgit.cgi/poky/tree/meta-poky/conf/distro/include/poky-floating-revisions.inc Jul 05 09:58:30 thats part of poky-bleeding, a fairly unknown poky variation that does builds of git tip Jul 05 10:00:08 rburton: still not 100% sold because we want to build a branch in CI and merge if tests good Jul 05 10:00:30 that would make a point for doing it with actual hashes Jul 05 10:00:39 then you may need to rewrite the SRC_URI anyway as there is hash/branch validation Jul 05 10:01:04 oh that example was using autorev, but if you replace autorev with a real hash then its the same Jul 05 10:01:10 it was an example of how to do what you were asking Jul 05 10:01:31 *but* if you want to build a different branch then you'll hit the branch/hash validation, which checks that the sha you say is actually on the branch you asked for Jul 05 10:04:11 is it really SRC_URI and SRCREV? Jul 05 10:05:47 yes Jul 05 10:07:01 ah you can set nobranch=1 in the SRC_URI if you want to disable branch validation, you could do that in the layer always and then just fiddle SRCREV to change what you build Jul 05 10:12:43 thanks rburton guess i have to ask you a bit more about it some time Jul 05 10:12:49 sure Jul 05 10:12:55 you know where to find me ;) Jul 05 10:37:46 ed2: did you see my mail last night? Jul 05 10:37:57 oh you replied sorry :) Jul 05 11:08:21 rburton: yes, I did and answered. Can you show me the link to the build? Jul 05 11:08:29 rburton: I'll try to reproduce it. Jul 05 11:15:04 Hello guys ! Does anybody know how to include a package which is RPROVIDE by recipea.bb as a RDEPENDS of recipeb.bb ? Jul 05 11:15:43 I've tried RDEPENDS="package_name" in recipeb but bitbake doesn't like it Jul 05 11:16:12 fberg: note that should be RDEPENDS_${PN} += "pkgname" Jul 05 11:18:43 ty zero_note Jul 05 11:21:11 yw Jul 05 11:36:38 bug 11751 really shows how you never know with performance. Patch that switched to multi-threaded xz has been flagged for perf regression... Jul 05 11:36:39 Bug https://bugzilla.yoctoproject.org/show_bug.cgi?id=11751 normal, Undecided, ---, leonardo.sandoval.gonzalez, NEW , Performance regressions detected on core-image-sato and virtual/kernel at 2.4M1 rc1 Jul 05 12:04:43 rburton: is it this build? https://autobuilder.yoctoproject.org/main/builders/nightly-arm-lsb/builds/1131/steps/BuildImages_1/logs/stdio Jul 05 12:05:00 rburton: mips and arm builds were failing, right? Jul 05 12:05:07 no https://autobuilder.yocto.io/builders/nightly-arm/builds/389 Jul 05 12:05:45 rburton: btw, the log shows incorrect environment: IMAGE_FSTYPES=tar.gz Jul 05 12:08:16 hm Jul 05 12:10:10 the selftest failed a few wic tests too https://autobuilder.yocto.io/builders/nightly-oe-selftest/builds/362/steps/Running%20oe-selftest/logs/stdio Jul 05 12:22:20 rburton: looks like you took more than this patchset into ross/mut: http://lists.openembedded.org/pipermail/openembedded-core/2017-June/138802.html Jul 05 12:22:37 yeah Jul 05 12:22:41 going to rip out and try again :) Jul 05 12:23:36 rburton: that explains failing oe-selftest, but doesn't explain image_tar failure. I'm trying to reproduce it. Jul 05 12:34:53 Hi everyone, do you know if it is possible to make a fetchall for a populate_sdk task? Jul 05 12:37:57 what are you trying to do? Jul 05 12:46:18 i would like to populate sdk with no network Jul 05 12:47:16 i'd be a bit surprised if just fetchall on the image didn't get enough Jul 05 12:49:11 ok, i may make a mistake but i just tried it, it was missing some stuff, thank you Jul 05 12:51:39 just one example, it was missing chrpath Jul 05 12:54:17 you could try fetchall on buildtools-tarball too Jul 05 12:54:58 rburton: ok thx Jul 05 13:17:18 is there some documentation for opkg.cnf flying around somewhere? Jul 05 13:29:46 hi guys, is bitbake-env something currently under development? Jul 05 13:31:02 * rburton wonder what bitbake-env is Jul 05 13:34:10 rburton: I was refreshing some bitbake commands, reading http://elinux.org/Bitbake_Cheat_Sheet , at the bottom of the page you can find it under "Additional bitbake-related commands" (not sure how much reliable it is) Jul 05 13:34:49 and kergoth should know something about :) Jul 05 13:35:14 depexp is now taskexp Jul 05 13:35:32 well the bitbake-env link doesnt work so i guess thats a no? Jul 05 13:36:30 ok, found my answer https://gist.github.com/kergoth/4133101 Jul 05 13:37:04 rburton: indeed, thx for checking Jul 05 14:02:57 kergoth: I'm tempted to add that to bitbake itself, maybe as "bitbake-var" as users probably see this as variables, not an environment. Any thoughts on that? Jul 05 14:32:17 I'm trying to make a recipe for a package that uses cmake. Unfortunately this project builds in the source path and then directly links to a library in the build output. Jul 05 14:33:38 I've patched the cmakelists so that it now builds in the build directory, but I'm still getting an rpath error during packaging Jul 05 14:33:48 nathani_: look at the autotools-brokensep class Jul 05 14:34:27 its not optimal, but often its easier to just build in-tree instead of doing the big patching thing. Jul 05 14:36:21 LetoThe2nd: thanks for the advice Jul 05 14:38:01 IIRC its not much more than B = S Jul 05 14:38:32 yeah, i see that :) Jul 05 14:41:51 still getting rpath error patebin.com/rqC351qW Jul 05 14:45:32 I can see the -rpath,/foo/foo/foo in the link.txt, but am unsure how to track this down as I don't have much experience with cmake Jul 05 14:46:26 or do I just -DCMAKE_SKIP_RPATH and everything will be fine? Jul 05 14:47:14 nathani_: does your CMakeLists.txt have an install rule for the binary? Jul 05 14:47:41 Or are you, by chance, writing your own do_install() that just copies the file from the build directory? Jul 05 14:48:39 no, cmakelists installs Jul 05 14:59:38 rburton: I can't reproduce it :( Jul 05 15:09:48 zero_note: bitbake-env went away in favor of a subcommand (show) of the 'bb' experimental sub-command based utility, which is currently broken and not yet fixed due to changes in tinfoil (bitbake's library used for external utilities). also neither of them supported the new variable tracking mechanism. but yes, i think Jul 05 15:10:59 RP: it's not a bad idea. I'd still like to prototype integration of a sub-command based cli ui, but that can always be a future step, it's clear bitbake-env/bb-show provides value on its own, bitbake -e | grep is common and quite ugly Jul 05 15:16:36 kergoth: ok, thanks for the explanation Jul 05 15:17:58 kergoth: agreed on a cli, I still have shell.py around as a partial inspiration for that and this wouldn't preclude that Jul 05 15:19:24 zero_note, RP: I was never 100% happy with the interface of that tool, for what it's worth. for scripting, you really also want the ability to see *just the value* of the variable, rather than having to sed out the quotes and whatnot, but then what's the behavior if you set that and pass multiple variables? could just error, but that doesn't seem super clean.. also it would need variable tracking support added, and desperately needs the ability to Jul 05 15:19:24 accept FOO[bar] syntax to see specific flags rather than having to look at all flags for a specified variable.. Jul 05 15:20:30 imho, bitbake -e recipe | grep become usable only after several times of usage, when one knows how to customize the grep reg expression Jul 05 15:20:32 kergoth: some kind of stripped output or value only shouldn't be too difficult and yes, those other things are valid issues Jul 05 15:21:06 zero_note: yeah. you can write a shell function to semi-automate it, but only for the common case, it doesn't scale well, and is rather brittle Jul 05 15:25:33 would also be nice to be able to get the unexpanded value without having to grep out the commented section, and a better way to extract the varaible tracking data for a particular variable in a more script-friendly format.. Jul 05 15:25:54 being limited to the shell-like output bitbake -e provides isn't ideal — there's a reason we did it that way, but.. Jul 05 15:26:12 at any rate, i do think we should integrate such a script, but i'd like to see a discussion about the details Jul 05 15:27:46 the upside to the current behavior is the output is nearly in bitbake metadata format. that's a possibility, too, for the variable tracking — it could optionally emit the multiple bitbake format lines that led to the final value, rather than the final assignment itself Jul 05 15:27:48 * kergoth shrugs :) Jul 05 15:28:14 I used bb-show often enough that I ran into the weaknesses as often as its benefits Jul 05 15:29:29 I'd be glad to see a discussion about too, as junior dev I don't think I can give any help on the dev side, but as yocto user maybe could be worth Jul 05 15:29:29 kergoth: real world usage is invaluable... Jul 05 15:29:57 zero_note: that's a valuable perspective, and we'd appreciate it. It's not always easy to see things from a beginner perspective once your'e not a beginner anymore Jul 05 15:30:06 i know I suck at it Jul 05 15:30:45 kergoth: I used to be better at it :( Jul 05 15:32:43 I recently realised that it's one of the many things that requires time to stop and think Jul 05 15:33:01 * RP was just thinking lack of time doesn't help Jul 05 15:33:54 I took time to think on the train this morning, it was wonderful. Made me remember why I like this career Jul 05 15:36:42 http://www.useronboard.com/features-vs-benefits/ comes to mind Jul 05 15:36:55 particularly http://www.useronboard.com/imgs/posts/mario-water.png Jul 05 15:40:05 kergoth: from my point of view, in 90% of cases my yocto problems comes when I don't understand completely how yocto works in its single components, eg: how and when some variables being expanded, or how the sstate really works Jul 05 15:40:19 stuff like this Jul 05 15:40:20 that's fair, there's a lot going on Jul 05 15:41:34 and it these scenarios "grepping" around becomes my (slow but) favourite tool Jul 05 15:41:44 *in Jul 05 16:21:10 rburton: patchtest will start catching (finally) metadata issues, line checksums changed without reason on the commit description, all thanks to tinfoil Jul 05 16:21:43 rburton: just need to update the guest machine used to exec patchtest. Hopefully early next week Jul 05 17:43:17 RP: do we care to update gcc6 to 6.4 in master ? Jul 05 19:15:00 RP: considering my tests here, it seems multiconfig has problems when you build two different tunes for same arch in same tmpdir (e.g armv5 and armv7a) Jul 05 19:15:13 RP: different arches seems fine Jul 05 20:00:49 folks Jul 05 20:01:52 I can't seem to select SystemD as my runtime manager? Jul 05 20:01:53 https://pastebin.com/hybkrXjd Jul 05 20:01:56 Any clues? Jul 05 20:29:47 Does anyone know what has happened to Jeffro? I see him off line but in vacation planner he does not appear. Jul 05 20:33:50 mortderire: is that from your local.conf? Jul 05 20:55:44 otavio: I'll have to see if I can reproduce. There is something odd going on... Jul 05 20:56:06 RP: likely Jul 05 20:56:22 RP: I am building qemuarm and imx6qsabresd for test Jul 05 20:56:55 otavio: I don't have imx6qsabresd handy but I guess beaglebone (armv7) should do it Jul 05 20:57:04 I can quickly try that after meetings Jul 05 20:57:18 RP: I guess so Jul 05 20:57:47 RP: right; it is not urgent but getting multiconfig working should speed up builds and tests Jul 05 21:37:14 rburton: I sent a stable backport for mesa 17.1.4 that fixes an issue for etnaviv. Jul 05 21:37:25 Anyone know of a utility to flatten a bitbake file by expanding all the includes and requires in place? Jul 05 21:39:05 jpew: bitbake-layers has it, IIRC Jul 05 21:40:41 k, thanks! Jul 05 21:41:49 jpew: also bitbake -e target can do the job Jul 05 21:43:25 jpew: but it has much more that the recipe itself Jul 05 21:46:21 otavio: I only get one set of sigs for beaglebone and qemuarm Jul 05 21:49:38 otavio: meta-freescale doesn't work with master? Are you testing against pyro? Jul 05 21:51:31 otavio: so using meta-freescale, imx6qsabresd and qemuarm also gives the same sigs. Jul 05 21:51:44 otavio: what else are you including layer wise? a specific distro? Jul 05 22:01:53 Ya, I was looking to flatten my machine.conf file so I could send it to a 3rd party easily, but I realized that I only included things from poky, which they should have anyway, so it was unnecessary. bitbake -e wasn't quite what I was looking for :) Jul 06 00:55:44 RP: you will like this patch http://git.openembedded.org/openembedded-core-contrib/commit/?h=kraj/master&id=9374be8eeb45060785834b4a2e9916c0ea8383f5 **** ENDING LOGGING AT Thu Jul 06 03:00:02 2017