**** BEGIN LOGGING AT Thu Mar 24 02:59:58 2016 Mar 24 07:48:30 Hello. Is there an easy way to mask systemd services from recipes. Tried to ln -s /dev/null /etc/systemd/foo.service, but got a permission denied. Is the post installation script the only way? Mar 24 07:56:24 find the issue. I was doing the symbolic link incorrectly. Mar 24 08:53:42 Good morning everyone Mar 24 08:53:55 We have multiple recipes that fetch from the same SVN repository Mar 24 08:54:19 That sometimes leads to problems because one of the fetch tasks has the svn working copy in DL_DIR locked Mar 24 08:54:49 Is there a recommended way to avoid the issue (e.g. specifying a separate download cache directory for each of the recipes)? Mar 24 08:57:06 I think you can specify a name for the download in SRC_URI, check the bitbake documentation Mar 24 08:57:16 i'm not sure what it does for svn specifically Mar 24 09:00:11 Yeah, there's downloadfilename, but that's only used in wget.py and sftp.py, so I doubt that does what I need :( Mar 24 09:00:42 maybe patch bitbake ? :P Mar 24 09:01:08 I wonder if you could set DL_DIR inside a recipe... would that break stuff ? Mar 24 09:03:19 That's a very good question Mar 24 09:04:09 There's also SVNDIR, maybe I can set that per-recipe... Mar 24 09:08:56 we should be writing a lockfile that prevents concurrent fetches of the same file in DL_DIR Mar 24 09:09:01 if that's not working for SVN that's a bug Mar 24 09:10:01 And the path_spec property of the svn fetcher seems to allow me to change the directory where the initial checkout happens, but the updating code using ud.moddir which doesn't respect path_spec, so that code path seems to be broken, too. Mar 24 09:10:40 bluelightning: it's not only concurrent fetches, the tuple (fetch, unpack) needs to be run under lock, otherwise a parallel job might copy different files than it expected Mar 24 09:11:35 that shouldn't be necessary... unpack can't start without fetch having succeeded Mar 24 09:11:41 recipe A fetching r123 from $url, recipe B fetching r456 from the same $url, A:do_fetch, B:do_fetch, A:do_unpack, B:do_unpack, both end up with r456 in $S Mar 24 09:12:42 you've observed that exact scenario? Mar 24 09:12:52 I think I'm observing that right now, yes. Mar 24 09:12:59 hmm ok Mar 24 09:13:19 I'm not familiar with the specifics of the svn fetcher... to be honest I try to steer clear of svn as much as possible Mar 24 09:13:35 sometimes there's no choice of course Mar 24 09:13:56 The symptom is that the source code for my recipe A is empty, possibly because B:do_fetch and A:do_unpack run in parallel and the first thing svn does is remove the files, then re-add them? Mar 24 09:14:26 I'd really hope it's not doing that because it would be somewhat wasteful Mar 24 09:14:48 but then I can't be sure Mar 24 09:15:38 I could just put SVNDIR .= "/${PN}" to fix this for now, I guess. Mar 24 09:30:26 hi.. in my recipe i am going to install a shared lib to /opt/sys/lib .. but bitbake/yocto does not create the symlink (soname) for this lib -- but it does when installing to /usr/lib -- so how to tell bitbake/yocto to the same for /opt/sys/lib? Mar 24 09:32:57 hello Mar 24 09:42:49 ping bluelightning: in my recipe i am going to install a shared lib to /opt/sys/lib .. but bitbake/yocto does not create the symlink (soname) for this lib -- but it does when installing to /usr/lib -- so how to tell bitbake/yocto to the same for /opt/sys/lib? Mar 24 09:43:23 yocto doesn't create the symlink, afaik. Mar 24 09:43:25 btooth: we don't create those symlinks, whatever build system you are using does that Mar 24 09:43:29 ldconfig may, if you run that. Mar 24 09:43:58 btooth: e.g. libtool Mar 24 09:44:20 it seems bizarre that destination path would have any effect on that Mar 24 09:44:32 but in package-split i find my lib (libmylib.so.0.0.0) and symlink libmylib.so.0 -- that symlink is somehow created by yocto Mar 24 09:44:50 when installing to /usr/lib Mar 24 09:45:21 as I said, "yocto" isn't creating that symlink Mar 24 09:45:38 how is your lib built? Mar 24 09:46:16 inherit pkgconfig autotools-brokensep Mar 24 09:47:03 So it seems one can also not set SVNDIR on a per-recipe basis either :/ Mar 24 09:47:56 btooth: so the library is built using autoconf/automake/libtool then? Mar 24 09:48:15 thats all, my recipe does not more for installing into /usr/lib, for installing to /opt/sys/lib i overwrite do_install .. but i think i am missing then something in my do_install Mar 24 09:48:16 btooth: I'm asking more about what's in the source tree, not the recipe Mar 24 09:48:42 bluelightning: yes Mar 24 09:49:40 if you want to change the destination library path wouldn't you change libdir? Mar 24 09:50:18 Has someone here built meta-fsl-ppc for T1040D4RDP successfully with jethro? Mar 24 09:50:55 bluelightning: sounds good to me, how to do this? libdir = /opt/sys/lib? Mar 24 09:51:28 btooth: quotes around the value, but yes Mar 24 09:51:37 btooth: this is within the recipe of course Mar 24 09:56:22 btooth: you overwrite do_install? Do you call "make install" in your overwritten copy? Mar 24 10:01:18 bluelightning: if i do: libdir = "/opt/sys/lib/" then do_configure fails.. Mar 24 10:02:01 neverpanic: i tried: make install DESTDIR=${D}/opt/sys/lib/ -- but that installs then to /opt/sys/lib/usr/lib Mar 24 10:03:02 i don't really whant to overwrite do_install, i want yocto to install to /opt/sys/lib instead /usr/lib Mar 24 10:09:36 btooth: yes, because DESTDIR isn't prefix. Mar 24 10:09:45 maybe you want to configure your package with --prefix=/opt/sys? Mar 24 10:13:44 Hi all Mar 24 10:14:30 I suppose it's a rather simple question but i just can't get my head wrapped arround it: How do i implement a recipe in such way, that it installs a unit file, if systemd is used and a 'legacy' init script when sysvinit is used? Mar 24 10:15:33 Also i'm curious since it's possible to use systemd for the main image and sysvinit for the rescue image how that would work out. I guess i could use bb.utils.contains('DISTRO_FEATURES', 'systemd', ....) but would that not cause any problems with the scenario i just described? Mar 24 10:46:36 neverpanic: thanks! that's it Mar 24 11:58:35 While trying to run ptest for lttng-tools, I can see lot of logs of PASS. The final log that I see is: "PASS: 290 - Kill session daemon". And after that no output is seen on the console Mar 24 11:59:33 when I try to rerun ptest for lttng-tools, I see final output as "PASS: 1 - Start lttng-relayd (opt: -o /tmp/tmp.mTXKVGVtZZ)" Mar 24 12:00:09 This message comes after PASS: 290 - Kill session daemon Mar 24 12:00:38 The lttng-tools version is 2.6 Mar 24 12:01:28 any pointers would be really helpful, to know why the tests doesn't proceed ahead. Mar 24 12:05:09 someone know how smart choose between two rpms that provides the same package? Mar 24 12:06:02 i'm trying to understand why a image is installing a custom package that provides initscripts instead the original initscripts Mar 24 12:06:15 my preferred_provider is initscripts Mar 24 12:06:59 but other package installs the custom initscripts and the image is installing the custom initscripts Mar 24 12:13:07 <_4urele_> igor there are priorities between channels Mar 24 12:13:14 <_4urele_> igor1, there are priorities between channels Mar 24 12:18:15 <_4urele_> igor1, as far as I know if two different packages provides the same file you will face some errors (at compilation time if both packages are part of the image, and at installation time with the rpm) Mar 24 12:18:46 <_4urele_> igor1, I had those kind of problems but maybe it was a mistake from my side Mar 24 12:20:41 _4urele_: my cursom initscripts only RPROVIDES, so it does not provide at all Mar 24 12:21:34 _4urele_: my environment is kind of different because I build two images with one is inside the other Mar 24 12:22:07 so one image installs custom initscripts and the other would install default initscripts, cause my PREFERRED_PROVIDER is initscripts Mar 24 12:22:13 but it is not happening Mar 24 12:22:25 I made the same thing with base-files and it worked Mar 24 12:22:37 I am getting this problem only with initscripts Mar 24 12:23:29 I tried to debug bitbake inside the code and at the end I saw that smart chooses the packages it will install Mar 24 12:24:19 it only have a list of pkgs, that is the packagegroups that have the dependencies, so packagegroup-core-boot depends of initscripts but smart is installing my custom initscripts instead Mar 24 12:24:29 and I can't wonder why Mar 24 12:27:34 <_4urele_> igor1, sry now I understand... but I can't help further Mar 24 12:27:49 _4urele_: thank you sir Mar 24 12:28:54 _4urele_: but this channels priorities can be something, I will learn a bit more about smart to try to understand it Mar 24 12:28:57 thank you Mar 24 12:30:21 <_4urele_> igor1, I used channels priorities to install packages directly on the target (maybe check if your packages are on different channels) Mar 24 14:09:36 error.yp.org is getting slugging has anyone else noticed Mar 24 14:09:57 trying to open grub_cv_target_cc_soft_float= Mar 24 14:10:00 err Mar 24 14:10:08 open http://errors.yoctoproject.org/Errors/Details/57093/ Mar 24 14:10:20 takes long Mar 24 14:12:30 I see the same here Mar 24 14:57:19 Yes, error.yp.org seems very slow. Already experienced that yesterday but thought it was due to my upstream. Mar 24 14:57:27 Is anyone here who has sucessfully compiled openssl_1.0.1i.bb from meta-fsl-ppc with yocto-jethro? Mar 24 15:00:20 is there some nice way to disable the download cache for just a single recipe? Mar 24 15:00:40 to repair a broken cache entry on a remote machine that's hard (for me anyway) to access :P Mar 24 15:01:13 just disable the use of it preferably. would be nice if it was repopulated. Mar 24 15:02:18 bitbake brokenrecipe -ccleanall Mar 24 15:02:30 will delete it from the cache Mar 24 15:02:35 bitbake -f -c fetch package should refetch the package Mar 24 15:03:38 what exactly are oyu calling 'the download cache'? DL_DIR or PREMIRRORS/MIRRORS? the former better fits the desdription imo, but your statements make it sound more like the latter. best to use terminology used by the project.. Mar 24 15:03:44 * kergoth yawns Mar 24 15:10:08 DL_DIR... seems yocto has issues with concurrent fetches from the same SVN repository, if the original patch is to be trusted Mar 24 15:11:08 and yeah, sorry for not providing context. long day. :/ Mar 24 15:11:19 i usually snark at other people for not doing that Mar 24 17:49:09 http://marguskoiduste.eu/images/the_general_problem-974.png Mar 24 18:15:44 Hmm, need to make the source preparation bits in devtool a bit more generic. has hardcoded kernel bits Mar 24 18:30:24 How can I clean a previous build completely? E.g. which dirs in build should remain behind and not be deleted? Mar 24 18:31:03 just wipe tmp Mar 24 18:33:40 I just learned that swiching PACKAGE_CLASS from rpm to deb half ways into a compilation doesn't work very well... Mar 24 18:34:00 thanks, kergoth Mar 24 18:34:20 np Mar 24 18:38:00 Hmm, it seems the sstate-cache dir also had to be deleted Mar 24 18:43:06 In my core-image-minimal.bb file can I inherit two things in a single line "inherit core-image populate_sdk_qt5" or should I make two separate "inherit" lines? Mar 24 18:45:28 did you look at any other recipes? Mar 24 18:45:36 half the recipes in oe-core inherit more than one thing on a single line Mar 24 18:45:39 grep is your friend :) Mar 24 18:45:56 I looked are recipes for minimal and base Mar 24 18:46:03 Ill check it out now Mar 24 18:46:19 more recipes than just images exist Mar 24 18:47:07 Just looked at also Mar 24 18:47:09 inline is fine Mar 24 18:47:10 thanks Mar 24 18:47:15 alsa* Mar 24 18:47:49 huh, oddly the bitbake manual doesn't mention that it can inherit multiple with one inherit directive, that's surprising. https://www.yoctoproject.org/docs/2.0/bitbake-user-manual/bitbake-user-manual.html#inherit-directive Mar 24 18:48:35 Not that I saw Mar 24 18:52:44 riz__: mind opening a documentation bug at the yocto bugzilla about it, so others don't run into such confusion? Mar 24 18:53:13 ok. no problem Mar 24 18:53:35 thanks Mar 24 18:54:56 Hmm, it'd be interesting if devtool modify -x / extract made it easier to resolve rejects with patches, i.e. if it applied all the patches at once using git-am, so you could hop into the tree and resolve it and am --resolved Mar 24 18:55:21 * kergoth adds to the eventually-look-into-this todo list Mar 24 18:56:46 kergoth: for the bitbake user manual? Mar 24 18:57:15 the bug? yeah, it should be against hte bitbake docs i expect. though it wouldn't hurt to cover it in the main yocto docs somewhere Mar 24 18:58:51 someone knows how smart choose between two rpms that provides the same package? Mar 24 18:59:20 that's redundant. an rpm is a package Mar 24 18:59:40 ok Mar 24 19:00:54 yes, but he can provides more than a package Mar 24 19:01:05 if you run rpm -q --provides Mar 24 19:01:47 ah. if it's like other package managers, i'd guess when encountering two providers it'd error out and force you to specify which by package name, but i don't know about smart specifically, one of the rpm experts in the channel will probably know better than i on that :) Mar 24 19:02:13 nice, thanks Mar 24 19:02:51 I'm trying to understand why the smart is installing a custom package that a made instead the default package Mar 24 19:03:26 my PREFERRED_PROVIDER_initscript = "initscripts", and the smart is installing the afw-initscripts, that RPROVIDES initscripts Mar 24 19:06:50 Somehow when running bitbake core-image-minimal -c populate_sdk, it is giving me errors about do_compile in qtsvg. I have not even included qtsvg at all. I did include it in local.conf at one point but now I removed it. Mar 24 19:22:07 I tried cleanall and that didnt work Mar 24 21:12:08 what.. how did i not know about quilt push --merge? Mar 24 21:58:34 howdy kergoth **** ENDING LOGGING AT Fri Mar 25 02:59:59 2016