**** BEGIN LOGGING AT Thu Oct 20 02:59:59 2016 Oct 20 05:43:17 morning guys Oct 20 05:43:49 with http://cgit.openembedded.org/openembedded-core/commit/meta/recipes-connectivity/ppp?h=morty&id=68e917aa778742da104c038a6e1ffa789fe95410 I am seeing broken builds for linux-yocto 4.4 Oct 20 05:44:06 can anyone else comment on this? Oct 20 07:11:43 hi everyone Oct 20 07:11:59 i am getting following while using bitbake Oct 20 07:12:00 http://pastebin.com/4d0QfJMk Oct 20 07:13:01 ERROR: Unable to parse $~/BBB/poky-krogoth/meta/conf/layer.conf: Oct 20 07:14:09 i am new to yocto can you suggest me what action i have to take Oct 20 07:14:21 why is there a $ in front of the path? Don't know if thats the problem, but bitbake can't find the file Oct 20 07:19:30 Thanks fl0v0 Oct 20 09:10:31 hi everybody, I'm currently facing a problem with buildhistory in YP jethro. If "buildhistory" is disabled the rootfs build works fine. As soon as I activate buildhistory (start toaster in my case) do_rootfs fails with the error shown at https://gist.github.com/rleitner/d9b752e7bd89b03cf56d9566555faef3 . Any kind of help is much appreciated :-) Thanks! Oct 20 09:17:48 hi, Yocto somehow managed to generate inconsistent LD and LDFLAGS for my SDK. 1) Why is it so? 2) How can I fix it? Oct 20 09:18:02 export LD="arm-polatis-linux-gnueabi-ld --sysroot=$SDKTARGETSYSROOT" Oct 20 09:18:09 export LDFLAGS="-Wl,-O1 -Wl,--hash-style=gnu -Wl,--as-needed" Oct 20 09:18:35 arm-polatis-linux-gnueabi-ld --help does not show that it has flags like -Wl,-O1. That looks like a gcc linker flag to me rather than ld. Oct 20 09:18:50 this makes it difficult using our SDK, so we would need to fix it ASAP... I am using daisy. Oct 20 10:50:38 Hello everyone Oct 20 10:50:59 I am new to yocto, and I am learning a lot from its documentation Oct 20 10:51:34 I am working on a new recipe, and looking at different examples in order to learn how to propperly do it Oct 20 10:52:00 sadly, as much as I have tried to look for an explanation about the purpose of SSTATE_SCAN_FILES Oct 20 10:52:03 g0hl1n: how old is your yocto? that code doesn't seem to be in the two last releases Oct 20 10:52:27 I have been unable to find any information about that variable Oct 20 10:52:37 can anyone help me understand what it is for? Oct 20 10:54:30 jku_: as mentioned, I'm using jethro (current branch). Oct 20 10:54:55 ah sorry, my bad Oct 20 11:06:42 g0hl1n: see sstate.bbclass, its used to find files that need paths replaced in them Oct 20 11:07:03 in the general case this is magic you don't need to care about. Oct 20 11:52:38 Hi. Can you have a machine specific do_install_append in a bbappend? E.g. do_install_append-mx6 ? Oct 20 11:54:20 arkver: sure, but you have to make it do_install_append_mx6 or the like. see http://www.yoctoproject.org/docs/2.2/ref-manual/ref-manual.html#var-OVERRIDES. Oct 20 11:55:18 that glossary entry (and the ones for CLASSOVERRIDE, MACHINEOVERRIDES, and DISTROOVERIDES) were fleshed out recently Oct 20 11:55:28 Thanks. Also for the link. Should have rtfm'd first. :) Oct 20 11:56:13 it's worth looking at the 2.2 manuals if you're reading an earlier version btw Oct 20 12:00:20 Bookmarked the 2.2 version last time you pointed me at it. :) Oct 20 12:00:51 hehe... didn't remember Oct 20 12:02:38 What is the purpose of SSTATE_SCAN_FILES ? Oct 20 12:09:42 Hello, i builing a really small package witch only copy a script to the target Oct 20 12:09:50 but i have problems with my LIC_FILES_CHKSUM Oct 20 12:09:57 where i have to place this file Oct 20 12:10:05 do i really need one? Oct 20 12:10:27 No Oct 20 12:10:42 you can skip the checking if you set Oct 20 12:10:48 how Oct 20 12:11:13 LICENSE = "CLOSED" Oct 20 12:11:51 But it is good practise to get your licenses right Oct 20 12:13:27 ok Oct 20 12:13:34 and when i define my own license ? Oct 20 12:16:27 when? As soon as possible. I'd say before starting working on the software, but definitetly before releasing it to the user. Oct 20 12:25:41 reibax: There's a hint in the original commit (https://patchwork.openembedded.org/patch/16959) and an example of it's use for apache2 (https://patchwork.openembedded.org/patch/57723/), but as rburton says it's not something the new Yocto user would normally need to be concerned about. Oct 20 12:34:09 Thank you very much arkver, and rburton! Oct 20 12:37:02 btw, I think if you set your license to CLOSED you need to whitelist it in the local conf's LICENSE_FLAGS_WHITELIST = "CLOSED_pn" Oct 20 12:39:24 you don't need to whitelist closed Oct 20 12:40:52 ok, odd. I'll retry my build. Oct 20 12:43:43 right enough. Wonder why I ended up with a whitelist in my conf. Can't remember. Oct 20 13:06:17 Hi, I have a little problem with adding user to image. In image.bb I add EXTRA_USERS_PARAMS = "useradd --password `openssl passwd my_password` --non-unique --uid my_UID --gid my_GUI ftp;" because I need another user for ftp, but Sometimes I have error that during adduser ftp user exists. How to fix it ? Oct 20 13:31:51 Is @base_contains('PACKAGECONFIG'... the only way to add to RDEPENDS? Oct 20 13:42:36 can a postinst script define a function? I've tried but it seems bitbake interprets the ending } as the end of the postinst script instead of the end of the function definition. Is there some trick to it? Oct 20 13:43:17 sveinse: the forth argument of the packageconfig flag is something to add to RDEPENDS_${PN}. you only need to do base_contains() if you want to add rdepends to other packages. Oct 20 13:45:10 rburton: yes, but I belive it does not aid if there are multiple PACKAGES, right? Oct 20 13:46:11 ...which is what you're saying essentially. thanks. Oct 20 13:48:13 there is almost always multiple packages but you *need* to say what package rdepends are for, so packageconfig assumes PN Oct 20 13:48:32 Can I add any flag to useradd that if user exist function do nothing and exit with 0 ? Oct 20 14:06:27 Example of what I would like to do: Oct 20 14:07:15 https://www.irccloud.com/pastebin/awhowbLp/ Oct 20 14:08:12 but bitbake errors out: ERROR: ParseError at /.../package.bb:76: unparsed line: 'somecommand || die "failed to do something"' Oct 20 14:10:13 real script is a bit more complicated of course, which is why die() would be useful. Any hints? Maybe I have to install an actual script and run that? Oct 20 14:13:04 Use a python def die : maybe? Oct 20 14:21:10 simonl: i'd file a bug for bitbake but i wouldn't be surprised if using \{ worked for the function definition Oct 20 14:25:30 simonl: This (http://docs.openembedded.org/usermanual/usermanual.html#recipes_syntax) seems to imply you can define your new die function outside the other function and just call it as needed. Oct 20 14:26:41 arkver: I had to read up on python functions and then I stumbled on that myself. Trying it out now. I didn't expect bitbake to be that sophisticated (assuming it works) Oct 20 14:32:06 arkver: hmm it doesn't magically include the definition in the script... I guess for scripts that run during the build process it will be available in the environment, maybe? Oct 20 14:32:51 rburton: \} and \{ in the definition of die() both result in errors still... Oct 20 15:02:51 http://pastebin.com/TbVgtc2t Oct 20 15:02:56 Works for me? Oct 20 15:04:12 Maybe I'm misunderstanding what you're trying to do. :) Oct 20 15:06:30 arkver: doesn't work like that for postinsts though Oct 20 15:06:43 ok Oct 20 15:07:12 (as they just take the pkg_postinst value and put it in the package, no following of functional dependencies afaik) Oct 20 15:08:16 ok, I misunderstood what he was trying to do. Oct 20 15:10:35 So a python func def won't help as this is a shell script and it's (as you say) BB's script parser not handling escaped {} that needs fixed. Oct 20 15:11:35 bitbake will only pick up a } if it's on its own at the beginning of a line. anything not indented at all will be a syntax error, a bitbake function's contents must be indented Oct 20 15:11:42 * kergoth yawns Oct 20 15:20:25 * armpit sees "yawns" are trending Oct 20 15:52:52 How can I fetch from a hg SRC_URI that requires authentication? mercurial stops at "http authorization required but program used in non-interactive mode" Oct 20 15:53:05 any ideas? Oct 20 15:55:03 apparently the yocto/bb env overrides the user's env, and as such any hg settings does not seem to apply, such as auth settings Oct 20 15:55:21 How is this managed with git? Oct 20 15:57:18 sveinse: https://lists.yoctoproject.org/pipermail/yocto/2014-April/019136.html Oct 20 15:58:44 moto-timo: thanks. yet, I don't really want to hardcode any username. this is up to the user running the build, not the recipe Oct 20 16:03:46 sveninse: set an environment variable that can be passed in on the command line? Oct 20 16:04:35 embed username/password in SRC_URI as variables that your user has to set in local.conf? Oct 20 16:04:54 ie ${CORPORATE_USERNAME} etc Oct 20 16:05:50 and SRC_URI=hg://${CORPORATE_USERNAME}:${CORPORATE_PASSWORD}…. might work Oct 20 16:06:08 rburton: even better :) Oct 20 16:08:07 yes, probably. but *gross*, this has horrible security Oct 20 16:11:07 then use kerberos or ssh keys that operate outside of OE Oct 20 16:11:52 sveinse: so write a function to extract the password from a secure keystore and store it in secure memory (not the bitbake data store) until the fetch happens and then somehow pass it to the fetcher without it appearing in ps or hg's memory Oct 20 16:12:41 we've got a method that works with the GIT_ASKPASS, I don't know if hg does that or not.. Oct 20 16:13:01 basically the easy fix is to make it not bitbake's problem - kerberos / askpass / ssh / etc Oct 20 16:13:06 I've been working to get that support open sourced.. somewhat for more eyes to review it, but cause I think it would be generally useful.. Oct 20 16:13:13 ASKPASS really is the right approach from what ic an see here.. Oct 20 16:13:23 wget was enhanced to support askpass as well.. Oct 20 16:13:35 (not sure if that version of wget is in YP 2.2 though, but it is upstream) Oct 20 16:15:24 actually it looks like we have already open sourced the code.. Oct 20 16:15:35 https://github.com/WindRiver-OpenSourceLabs/anspass Oct 20 16:15:59 program uses the ASKPASS mechanism and a key/value store to handle things.. Oct 20 16:16:21 it has a master password for the store (optional) to assist in securing things, but nothing is left in plain text (say like a .netrc would be) Oct 20 16:16:48 hg is normally (via ~/.hgrc) to use mercurial-keyring, a gnome-keyring-ish thing. And the repo I'm trying to access from my recipe, has already been put into the keyring. So the problem is that hg is invoked in an environment not being able to use to or access the user's setting or auth Oct 20 16:17:19 then you may hvae to change the setting for hg in the bitbake.conf file (or add one) so the right environment is present Oct 20 16:17:56 I dont really see the different from hg's keyring and askpass. They both need interactivity for asking for password and/or some user environment to store their previous approved auths. Oct 20 16:18:24 anspass asks for an unlock password -before- you run bitbake.. and then you use the replied token to set an environment variable for everything to talk.. Oct 20 16:18:35 anspass does require wget, git, and other things to pass that specific token in the environment Oct 20 16:18:55 keyring can do the same thing, afaik. the python keyring module is quite capable, and can interact with any number of password stores / keyrings Oct 20 16:19:02 for 'hg', there is probably little to no difference in behavior.. but it become specific to that one item Oct 20 16:19:04 right, so that is how it is done for git auth in yocto? Oct 20 16:19:05 https://pypi.python.org/pypi/keyring Oct 20 16:19:28 anspass uses the ASKPASS API, which works with wget, curl, git, and other common systems.. so it's more then just one particular use Oct 20 16:19:43 YP does not do anything specific for git authentication Oct 20 16:20:16 sveinse: if hg's keyring thingy needs magic env vars add their name to BB_ENV_EXTRAWHITE in your environment so they get preserved Oct 20 16:20:42 sveinse: (see scripts/oe-buildenv-internal which already preserves the ssh agent vars) Oct 20 16:22:56 rburton: I'll have to check how hg handles auth, but I'll look into it. Don't know if its env (doubt it) or via some python.... ahh, yocto builds its own mercurial client. Thus it needs the keyring extension installed to be able to handle pre-recorded auth. I'll follow this lead... Oct 20 16:23:43 you can prevent YP from building it's own hg client by doing an assumed_provided Oct 20 16:23:55 it builds it because we have found a lot of hosts do not have hg clients by default Oct 20 16:23:56 sveinse: i'd be surprised if you couldn't add mercurial-native to ASSUME_PROVIDED to make it use the host one Oct 20 16:24:16 (just remember to wipe the sysroot once you do that so it doesn't continue to use the one in there) Oct 20 16:24:36 ASSUME_PROVIDED+="mercurial-native" in local.conf, right? Oct 20 16:24:47 yeah Oct 20 16:24:53 thanks guys! Oct 20 16:30:41 so i was thinking about creating a Planet Yocto or similar, if anyone actively blogs about relevant stuff feel free to ping me your URL Oct 20 16:31:15 i really should start blogging about relevant stuff, but don't yet :) Oct 20 16:31:18 * kergoth ponders Oct 20 16:31:29 yeah same here Oct 20 16:33:11 basically rebirthing planet linuxtogo but with people actually still here Oct 20 16:33:31 sveinse: suppose you could probably also add python-keyring-native as a dep of mercurial-native to ensure it's in the native python sysroot for it to use Oct 20 16:34:46 additionally, mercurial must also access ~/.hgrc to be able to read identity settings Oct 20 16:44:35 update: the yocto mercurial client /does/ indeed read ~/.hgrc, as the env HOME is defined prior to running hg clone. And it seems to not protest at the enabling of the mercurialkeyring setting, which in the hgrc file. So next step is to figure what mercurial keyring indeed needs Oct 20 16:45:12 must have some magic env var Oct 20 18:57:24 Hello! Why there is no support for using built-in Buildbot steps in buildset-config file? (ex. ShellCommand). This would give us the full power of Buildbot and Autobuilder's elegance making the latter powerful general purpose build system. Oct 20 18:57:24 Hello! Why there is no support for using built-in Buildbot steps in buildset-config file? (ex. ShellCommand). This would give us the full power of Buildbot and Autobuilder's elegance making the latter powerful general purpose build system. Oct 20 18:57:46 I have no clue what you're even talking about Oct 20 18:57:52 maybe someone else does Oct 20 18:58:37 yocto's buildsystem uses bitbake, which is a tool. how you happen to run it from your autobuilder of choice is on you Oct 20 19:02:14 I'm talking about yocto-autobuilder which uses Buildbot under the hood, but doesn't support Buildbot's built-in classes inside its scripts Oct 20 19:04:24 if you're reporting an issue with a project, i.e. yocto-autobuilder, it helps to actually mention it :) Oct 20 19:04:31 i'll let the autobuilder folks respond Oct 20 19:23:24 I think I found it. hg needs the env var DBUS_SESSION_BUS_ADDRESS to be able to get to the proper keyring. I've found hg.py in the sources, but from what file is this called? I want to see if there are any vars that I can configure env not to erase, or as a last resort, modify the actual env command. Oct 20 19:43:24 sveinse: see runfetchcmd() in fetch2/__init__.py Oct 20 20:15:29 Can I get a variable from env and set as a variable in local.conf? Oct 20 20:18:36 runfetchcmd() has a exportvars list which is seems to be what I need. However this list fetches its contents from d.getVar(), which is not the environment. So what would be the best way to copy env var DBUS_SESSION_BUS_ADDRESS into a bb variable? Oct 20 20:18:49 ^ kergoth Oct 20 20:24:32 sveinse: cherry-pick the commit from bitbake master that adds DBUS_SESSION_BUS_ADDRESS? :) Oct 20 20:25:12 whoever wrote 20ad1ea87712d042bd5d89ce1957793f7ff71da0 was clearly awesome Oct 20 20:34:09 rburton: I wonder who that is.... Oct 20 20:35:02 rburton: I actually made exactly the same patch, just missing the "or d.getVar("BB_ORIGENV", False).getVar(var, True)" statement :D Oct 20 20:35:04 thanks! Oct 20 20:51:17 sveinse: you'll need that bit Oct 20 20:54:09 rburton: and now it works! Oct 20 20:54:16 * sveinse very happy Oct 20 20:54:49 (this is now you leave for the evening, NOT saying to yourself "oh, by the way, let me test one more thing") **** ENDING LOGGING AT Fri Oct 21 02:59:58 2016