**** BEGIN LOGGING AT Wed Apr 16 02:59:58 2014 Apr 16 06:15:45 what kind of modifications do the yocto kernel carry compared to the mainline kernel? i'm trying to look at the git repo, but there are _alot_ of branches. I assume the ones prefixed with "standard/" are yocto specific branches? and perhaps also the "meta" branch, which seems to contain this fragmented cfg stuff Apr 16 08:44:58 hello, please one quick question - how do I force RPM4 instead of RPM5? TY Apr 16 08:58:10 jeso: we don't support rpm4 at this time Apr 16 08:58:14 morning all Apr 16 08:58:46 bluelightning: so the smart package manager is useless on fedora, is that right? Apr 16 08:59:14 jeso: not that I know of, why would you suspect that? Apr 16 08:59:34 our use of rpm is not dependent on the host system's rpm Apr 16 09:01:39 bluelightning: but can I use it on target system? Apr 16 09:01:50 jeso: yes Apr 16 09:02:03 bluelightning: how? what repo to use? Apr 16 09:02:32 jeso: the idea is you're building your own distribution, so you would normally use your own repository Apr 16 09:03:12 we do provide some test repositories but there is no guarantee their contents will be maintained in a manner that allows them to be used continuously from a target Apr 16 09:05:00 howdy! is there some means to get sources out of a svn repository with username/password *without* having to hardcode them in the recipes? Apr 16 10:33:37 Morning! Apr 16 10:35:55 Just about to start designing a firmware update mechanism for my Yocto project and was just wondering if anything pre-existing in Yocto-land is worth looking at to build on before I re-invent the wheel for the Nth time... Apr 16 10:38:58 there was one posted some months ago, let me grab the link Apr 16 10:39:34 pev: http://comments.gmane.org/gmane.comp.embedded.eldk/2397 Apr 16 10:40:04 would be great to hear how you get on with it, I've not tried it yet myself Apr 16 10:44:15 bluelightning: Thanks - that looks very interesting indeed. It certainly sounds like it manages to tick off a lot of features...! Apr 16 12:12:39 Hi yocto. I am facing a problem with conflicting two packages (libiconv and eglibc). Anyway to prefer a specific package only? Apr 16 12:15:38 in what way are the conflicting? Apr 16 12:15:57 if multiple recipes provide the same package, you can set a preferred provider among those recipes Apr 16 12:19:20 not in multiple recipes. Apr 16 12:19:42 error: file /usr/bin/iconv conflicts between attempted installs of libiconv-1.14-r1.cortexa9_vfp_neon and eglibc-utils-2.18-r1.cortexa9_vfp_neon Apr 16 12:21:40 vicky: you shouldn't be building the libiconv recipe at all with eglibc Apr 16 12:24:14 libiconv is a RDEPENDS needed for another package Apr 16 12:30:48 vicky: yes, but eglibc provides its own libiconv - there is no need to build an additional one Apr 16 12:32:36 vicky: is that an explicit RDEPENDS? if so, why? Apr 16 12:48:56 howdy! is there some means to get sources out of a svn repository with username/password *without* having to hardcode them in the recipes? Apr 16 12:52:51 LetoThe2nd: I saw your question earlier, I don't have a definitive answer - sorry Apr 16 12:52:55 but I suspect the answer is no Apr 16 12:53:23 bluelightning: guessed so. just thought after a bunch of hours one repost is acceptable Apr 16 12:53:40 bluelightning: will have to find another way then :/ Apr 16 12:53:58 I mean you could use variables in the recipe such that the recipe doesn't have the variables for username/password defined, and then set the vars in your config Apr 16 12:54:47 not sure if that helps... Apr 16 12:55:00 bluelightning: its not abad thought indeed, thanks. Apr 16 13:02:19 bluelightning, if a task fails due to a shell command exiting with code 1, how can i determine what shell command it was? the task is huge (due to inheritance) and I don't really know where to begin looking. the logs only state that there was an error in a shell command Apr 16 13:06:36 LetoThe2nd: Have a look at : http://stackoverflow.com/questions/2599281/cant-make-svn-store-password-even-though-the-configuration-is-set-to-allow-it Apr 16 13:06:40 LetoThe2nd: for SVN with username/passwd, can't you use the SVN credentials in $HOME/.svn? Apr 16 13:07:05 well... what pev said ;-) Apr 16 13:07:22 that's what i am doing for some private repo in svn Apr 16 13:07:26 pev: ndec: i'm so totally no svn man, so i don't even know where to look... so thanks! Apr 16 13:08:02 well, the idea is to store the credentials in $HOME, and in the recipe you don't put the information about user/passwd Apr 16 13:08:07 :-D I remember when SVN was new and shiny and an exciting step forward from CVS/RCS :-D Apr 16 13:08:25 Everyones all git-this, git-that these days ;-) Apr 16 13:08:27 ndec: yeah thats about what i need. Apr 16 13:08:45 * LetoThe2nd hands pev an "old man" badge ;) Apr 16 13:08:50 :-P Apr 16 13:09:08 Mate, I may be nearly 40 but I'm still out having more fun than most in their 20's... Apr 16 13:09:11 pev: LetoThe2nd: it may be that we tell svn not to read that file, I'm not sure Apr 16 13:09:19 (to avoid conflicts with user configuration) Apr 16 13:09:29 pev: no offense meant! Apr 16 13:09:38 LetoThe2nd: None taken :-D Apr 16 13:09:53 bluelightning: no, that method is working fine Apr 16 13:10:02 LetoThe2nd: I'd consider scripting it via a variable though as it gives you more control in your build for the same reason as bluelightning says... Apr 16 13:10:10 ndec: great :) Apr 16 13:10:29 Having dependencies on user config isn't the best thing, but whinging when someone hasn't set it in local.conf is safer... Apr 16 13:10:35 ndec: ok, it might have been another VCS or maybe my memory is just faulty ;) Apr 16 13:10:38 pev: yeah thats for sure. its just about "getting it to work first, polish later" Apr 16 13:11:49 Sometimes its easier to do the right thing first time round than remember to 'fix' in the future mind :-D Apr 16 13:12:14 Not that my current codebase is littered with "// TODO : ..." or anything.... *ahem* Apr 16 13:12:30 need a pullmoll? Apr 16 15:09:00 hello there Apr 16 15:09:10 hi Xz Apr 16 15:33:32 what is the correct way to tell my local.conf (or other config) to include a specific nativesdk-* package when doing a populate_sdk? It seems to figure out most things it needs automatically but misses the qt5 nativesdk files. Apr 16 15:34:01 take a look at th etop of the bbclass, the vars are there Apr 16 15:34:12 i never remember them off the top of my head, i just re-examine the class to get them Apr 16 15:35:10 but you take one of those and stick it in your local.conf ? Apr 16 15:35:53 from populate_sdk_base.bbclass, TOOLCHAIN_HOST_TASK. so set TOOLCHAIN_HOST_TASK_append = " nativesdk-foo" Apr 16 15:37:02 okay that's what I tried first since the docs seem to hint at this, but it errors out because it can't find a canadian cross build definition Apr 16 15:37:44 the qt5 sdk thing is a big one, many folks are asking about it, and some folks are working on it, see the mailing lists, there's probably info on it there somewhere. i don't really know the status, personally Apr 16 17:26:36 * kergoth mulls over the possibility of asking the UI to prompt for username/password, and adding a github fetcher that uses the API and supports 2-factor auth Apr 16 17:31:37 kergoth: it's possible, but then do we really want interactivity in bitbake's UI? Apr 16 17:32:44 we already behave specially on interactive terminals than non, and i think it's better than git/svn hanging on an input prompt, or just being unable to fetch at all Apr 16 17:32:56 could cache the cookies for the auth Apr 16 17:32:58 * kergoth shrug Apr 16 17:33:52 * nerdboy caches a muffin for later Apr 16 17:35:56 kergoth: well clearly hanging on the prompt is a bug Apr 16 17:36:32 requiring the user to either hardcode ssh into the recipe or use git url replacements to switch from https to ssh for github private repositories is not ideal either Apr 16 17:37:04 the hanging on the prompt is really annoying Apr 16 17:41:00 I'm writing a recipe for a python package out of pypi, is there an easy way to determine its RDEPENDS? Apr 16 17:41:23 read the pypi metadata or setup.py requires Apr 16 17:45:35 Hmm, itd be cool if our fetchers could do mime type verification, with a heuristic default of expected mime type Apr 16 17:46:04 then if you're in a sitaution where a failure to resolve leads to an html doc being downloaded, we'd spot that explicitly, rather than just a checksum mismatch Apr 16 17:47:44 kergoth, I mean more of the std library requires. It seems like the std library was split up into python-core, python-io, python-re, etc. So I'd need to analyze the imports and check them against the manifest.inc? Apr 16 17:48:08 ah, yes, indeed. we're hoping to automate that in the future Apr 16 17:48:29 i have a work-in-progress dependency scanner for python code, but it almost always requires manual checking for cases of complex imports, try/except chains, etc Apr 16 17:49:09 tastycactus: you could try running this against it manually: https://raw.githubusercontent.com/kergoth/meta-package-auto-deps/master/scripts/pythondeps Apr 16 17:49:13 Where can I find that? Maybe I can help progress it.. Apr 16 17:49:24 Thanks Apr 16 17:49:36 https://github.com/kergoth/meta-package-auto-deps is the layer for it Apr 16 17:49:42 patches welcome. there's a todo, iirc. Apr 16 17:50:22 the problem is this manual review and/or excludes. i'm not convinced this should be automatic at package time, it might be better as an opt-in process, when you're developing the recipe Apr 16 17:50:56 I did it as a general framework for handling automatic deps/provides, with the intent that the framework could be used for pkgconfig, python, shlibs, and others Apr 16 17:51:13 haven't had time to work on it recently Apr 16 17:51:29 there's an open yocto issue for it, got postponed to the 1.7 timeframe Apr 16 17:54:46 Yeah, I was thinking just a script that could grep for imports and find the packages in the manifest Apr 16 18:15:04 Hmm, we should look at minimizing some patchsets where possible. e.g. for autoconf, autoreconf-exclude can likely go away in favor of translating the exclusions in autotools.bbclass to empty the env vars. e.g. GETTEXTIZE= Apr 16 18:15:09 for 1.7, that is Apr 16 18:16:47 xmkmf could be killed by an intercept, thoug that wouldn't help the x11 search paths part of that patch Apr 16 18:16:55 still, aligning more closely with upstream would be for the best, i think.. Apr 16 21:24:36 anyone know how to use layerversion in yocto-autobuilder? **** ENDING LOGGING AT Thu Apr 17 02:59:58 2014