**** BEGIN LOGGING AT Fri May 22 02:59:58 2015 May 22 07:10:17 /win 1 May 22 07:11:07 * nrossi thinks window switching should be done with keybinds :) May 22 07:16:10 :-/ May 22 07:16:31 * joshuagl thinks he shouldn't be allowed near the computer until he's had coffee May 22 07:41:35 I am trying to build core-image-sato for rapsberry pi using poky master and meta-raspberrypi master and it fails to build libepoxy due to missing include directory for vcos_platform_types.h. This is beacuse that file is in interfaces/vcos/pthreads but when building this include path is not used. I manged to build it by modifiying the src code and appending pthreads/ path in all vcos* files but this is an ugly solution. How do I fix this May 22 08:03:10 good morning May 22 08:16:02 morning all May 22 09:28:30 Hi ...is there anyone here who succeeded in getting Yocto to boot on a Beaglebone Black? May 22 09:29:33 Yocto is a build system, it does not boot. May 22 09:29:54 you probably mean an image generated by Yocto. May 22 09:30:46 well no... I have followed the instructions on the Yocto page to get my Built IMG to Run on a BeagleBone black May 22 09:31:01 But it just makes the lights go on and stay on and then just do nothing May 22 09:31:11 Hi ! What reasons can lead you to the warning "specified values did not make it into the kernel's final configuration"? May 22 09:31:20 I copied eveything where they should be May 22 09:31:21 no what? I cannot follow. Yocto still does not boot. It is just a buildsystem. You need to realize this. May 22 09:31:38 so you followed these instructions? https://www.yoctoproject.org/downloads/bsps/daisy16/beaglebone May 22 09:31:40 I'm trying to add configuration fragments to the kernel configuration, nothing fancy May 22 09:31:48 Ok ... So I have built from yocto an Image for the BeagleBone black May 22 09:31:57 someone running poky on raspberry pi? May 22 09:32:07 lpapp: Yes exactly May 22 09:32:16 ivan_: poky generated image, not poky. Poky is a build system :) May 22 09:32:20 CromFr: did you send a mail to the mailing list CCing Bruce as I suggested? May 22 09:32:23 I followed the guide that you have just linked May 22 09:32:24 ivan_: but yes, many. May 22 09:32:55 do you have problem with the build system generaring error when building? May 22 09:33:15 bluelighting: not yet May 22 09:33:16 ivan_: I think the better question is what your concrete error is :) May 22 09:33:28 I am trying to build core-image-sato for rapsberry pi using poky master and meta-raspberrypi master and it fails to build libepoxy due to missing include directory for vcos_platform_types.h. This is beacuse that file is in interfaces/vcos/pthreads but when building this include path is not used. I manged to build it by modifiying the src code and appending pthreads/ path in all vcos* files but this is an ugly solution. How do I fix thi May 22 09:34:19 ivan_: the real fix is probably to fix the buildsystem for libepoxy. May 22 09:34:32 (btw, your message got truncated at How do I fix thi...) May 22 09:39:12 How do I fix this, is this a bug? May 22 09:39:53 check what buildsystem it uses for starter. May 22 09:40:31 lpapp have you tried to build sato for rpi? May 22 09:41:13 lpapp: I created 2 partitions on my BeagleBone SD Card... 1 512Meg named BOOT and it has MLO and the other one is u-boot.img May 22 09:41:25 ivan_: no, but based on what you say, it is a buildsystem issue. May 22 09:41:35 so if I had to get it working, I would look at that. May 22 09:41:38 in the ROOT which is the 2nd partition and its EX2 ... I have the Core image-sato May 22 09:43:25 lpapp: when you say that it is a build system issue, do you mean poky itself? May 22 09:43:36 could be anything. May 22 09:44:02 I would try to localize it. E.g. is the includepath added directly to the compilation or via the --sysroot option? May 22 09:45:07 the path is added in egl.pc file where it states intefraces/vcos/pthreads but apparently is is not used when building. May 22 09:46:24 if i look at the recepie for libepoxy i can see that it is dependent on virtual/egl May 22 09:46:52 try to amend the compilation command manually. May 22 09:46:59 to prove that if you add the includepath it does work. May 22 09:46:59 how? May 22 09:47:04 copy/paste. May 22 09:47:26 ahaa okay May 22 10:00:37 Sigh.... no guide works for this beaglebone Poky stuff May 22 10:40:24 bluelightning: is the prefix variable documented somewhere that Yocto seems to set to /usr? May 22 10:40:32 it screws up some software trying to work with that. May 22 10:40:53 not our software, so I am not in a position to fix that, but if Yocto documented this, I could argue with that to them to fix it. Although they may still reject it. May 22 10:41:06 currently I have to monkey patch their buildsystem to "unset prefix". May 22 10:41:22 I am not sure why Yocto uses that variable though. May 22 10:42:04 what is the main use case for it that could not be solved better? May 22 11:03:22 lpapp: http://cgit.openembedded.org/openembedded-core/tree/meta/conf/bitbake.conf#n15 May 22 11:04:50 some distro used to collapse /usr/into / May 22 11:05:04 lpapp: When I try and use Toaster is says I should download dependancies with pip en VirtualEnv May 22 11:05:13 But I cant install either of the programs with apt-get ? May 22 11:05:13 lpapp: http://cgit.openembedded.org/meta-micro/tree/conf/distro/micro.conf#n91 May 22 11:07:48 ant_work: yeah, although unset is not the same as "" May 22 11:08:10 ant_work: still, why is it set? May 22 11:09:49 .what do you mean? prefix is used on configure May 22 11:14:34 ah, okay, so it is standard pratice. May 22 11:14:53 in which case, probably the third-party software that we use ought to use something else for their software. May 22 11:14:57 foo-prefix or whatever. May 22 11:15:04 thanks. May 22 11:17:18 yes, probably you should configure the extra with --prefix= May 22 11:17:46 hmm? May 22 11:18:26 the 3d-party soft should be configured May 22 11:18:48 accordingly to OE/Yocto May 22 11:21:12 not according to Yocto, but common sense, yes :) May 22 11:21:17 to be compatible with autotools. May 22 11:21:35 when I do USERADD_PARAM_${PN} = ... how can I execute some custom command with the freshly created users? May 22 11:21:43 so I want to make sure to run commands AFTER the user is added May 22 11:21:48 I want to modify a user's profile for instance. May 22 12:51:31 lpapp: there isn't a facility for that that I know of May 22 12:51:52 bluelightning: ok, thanks. May 22 12:52:14 bluelightning: I put the whole profile stuff on hold for now ... and just prepend the PATH in the same line where the application is started. May 22 12:52:36 that means, it is more difficult to run manually than via the initscript system, but that is it for now. May 22 13:03:34 Hey which setting do I change to NOT have a Live image built by bitbake? May 22 13:06:58 Hello guys, Im working on a Cyclone 5 devleopment kit and would like to change the kernel. Currently Im on 3.10 and want to try out higher versions. The information I could gather online was to set a variable called Preffered_kernel in local.conf with your desired kernel name May 22 13:07:13 However it doesn't seem to be working, could anyone help me regarding this? May 22 13:07:55 cart_man: live image? May 22 13:08:15 IMAGE_FSTYPES -> remove live? May 22 13:08:32 I am not sure I understand what you mean by live image May 22 13:09:12 you do not have to. May 22 13:09:37 "foo:" means on IRC that it is addressed to user foo. May 22 13:12:04 Hello ! May 22 13:12:15 mimetonbo: http://www.macieira.org/blog/2012/05/doesnt-work-doesnt-work/ May 22 13:12:27 mimetonbo: what exactly does not seem to be working? May 22 13:13:14 btw, I put it into the distro and machine config, but local config might be also good. May 22 13:13:43 Guys, i modified the mplayer2 recipe on the fido branch to enable the directFB support, i would like to share it for any purpose but i don't even know if people would be interested or just how to do it , do i have to post it on the mailing list on contact somebody in particular ? Thanks for your answers :) May 22 13:14:21 mimetonbo: PREFERRED_PROVIDER_virtual/kernel is presumably what you are looking for, that will select a different kernel recipe May 22 13:15:41 Happycat: http://cgit.openembedded.org/meta-openembedded/tree/meta-multimedia/README May 22 13:15:51 Happycat: I suspect OE would take the enhancement if it was made into a PACKAGECONFIG option May 22 13:16:21 I would make it optional. May 22 13:16:23 Happycat: ideally it would be great if you could make a patch and send it to the openembedded-devel mailing list (since it's for meta-oe) - see http://www.openembedded.org/wiki/How_to_submit_a_patch_to_OpenEmbedded May 22 13:18:01 Ok i'll read that , thanks ! i already made a patch but i am patching the recipe , another solution is to make it a .bbappend, is that what you mean by PACKAGECONFIG option ? i started working with yocto several days ago i don't know a lot about it sorry :/ May 22 13:18:47 http://www.yoctoproject.org/docs/current/ref-manual/ref-manual.html#var-PACKAGECONFIG May 22 13:19:30 thanks, well i missed a lot of things on internet obviously , i'm gonna read all of it, thanks a lot for your help May 22 13:29:46 Hmm on last question , i was reading the PACKAGECONFIG option description but i don't see anything saying that i can add something to the do_configure function , i have to set a dependecy , and the --enable-directfb flag that's totally possible but i also have to add a line to the do_configure method , am i missing something ? May 22 13:31:08 oh my bad the line i have to add can be unconditional , and the option would only be activated by the PACKAGECONFIG May 22 14:08:24 bluelightning: I am lost for a second... if the -dev package carries the SO, how can I install the application to be run on the target without the -dev package? May 22 14:09:20 .so* May 22 14:09:40 I thought the -dev packages are never necessary to be installed. May 22 14:09:43 lpapp: the -dev package does not contain .so*, just .so May 22 14:09:46 by default May 22 14:09:55 oh, star meant correction. May 22 14:09:57 :-) May 22 14:10:10 the .so file is meant to be a symlink, not an actual binary May 22 14:10:15 sure May 22 14:10:37 and applications shouldn't be linking to it, they should be linking to the versioned file May 22 14:10:49 the .so symlink is there for development purposes May 22 14:10:56 hence why it goes into the -dev package May 22 14:11:56 ah, ok, that is how I remembered, too. In that case, I do not know why I am getting this: /usr/bin/foo: error while loading shared libraries: libmylib.so: cannot open shared object file: No such file or directory May 22 14:12:12 it is looking for libmylib.so, not libmylib.so.X.Y May 22 14:12:29 it's probably not correctly linked, then May 22 14:12:52 hmm, good question why not. May 22 14:13:11 I just use the regular -lmylib May 22 14:14:22 The lib is installed in /usr/lib fwiw. May 22 14:14:27 so all looks ok to me. May 22 15:00:32 hello guys I have one curiosity: May 22 15:00:34 log_check: NOTE: warning: libstdc++-staticdev-4.9.1-r0.0@ppce500mc is already installed May 22 15:00:51 but there does not seem to be a duplicate May 22 15:02:04 any ideas of what the cause of such a bug should be May 22 15:42:32 kergoth: FWIW, http://git.yoctoproject.org/cgit.cgi/poky-contrib/log/?h=rpurdie/noupdatedata4 basically works. Its a bit slower than the current data store but no update_data is needed May 22 15:42:58 bluelightning: you might be interested in the above too. Obviously needs a bit more cleanup to the commit messages and so on May 22 15:46:05 looks promising, will take a look May 22 16:01:45 moin May 22 16:36:22 kergoth: just realised the data store no longer uses COW with these changes. VariableHistory still does but not DataSmart... May 22 17:35:40 RP: I am not able to reproduce the qemux86-64 failures w.r.t. iptables coreutils and elfutils May 22 17:35:42 locally May 22 17:45:26 are these still happening on AB May 22 17:46:27 this is what I was looking at http://errors.yoctoproject.org/Errors/Search/?items=10&query=a2fbb1b6b6e564b3f9d40550c0b408978f420039&limit=100 May 22 18:37:46 so, oprofile gets rebuilt between machines with different kernel - should it be marked as machine-specific or added to safe deps/abi list? May 22 19:04:29 denix, use perf :) May 22 19:11:54 Crofton|work: heh :) May 22 19:12:21 I hav ea patch I'm going to send to remoe oprofile from tools-profile May 22 19:12:34 partly becuae it leaves an open port (oprofile-ui) May 22 19:12:36 Crofton|work: good that you don't suggest using openwrt... :) May 22 20:01:44 Hmm, should enhance `bitbake-layers add-layer` to warn if the added layer depends on layers which aren't in bblayers.conf yet May 22 20:04:42 denix: machine specific if it has kernel module May 22 20:10:10 hmm, shouldn't the create-recipe script in scripts/ be obsolete in favor of `recipetool create` at this point? May 22 20:14:19 khem`: it doesn't have a kernel module - http://cgit.openembedded.org/openembedded-core/tree/meta/recipes-kernel/oprofile/oprofile.inc May 22 20:37:51 and one more question about sstate... May 22 20:40:52 when EXTRA_OECONF points to staging dir, it changes between machines due to different sysroots and forces package rebuilds... am I missing a fix or is it expected? May 22 20:41:49 that doesn't make a great deal of sense. sstate includes *unexpanded* versions of the variables May 22 20:42:14 e.g. it includes ${STAGING_DIR_HOST} which includes ${STAGING_DIR} which is ${TMPDIR}/sysroots, and TMPDIR is realtive to TOPDIR, and TOPDIR is excluded from checksums May 22 20:42:28 so the full absolute path will *not* be part of the checksums unless you forced early expansion with := May 22 20:43:34 presumably overriding TMPDIR to some other path could cause reuse problems since i think TOPDIR is excluded but not TMPDIR, but would have to doublecheck BB_HASHBASE_WHITELIST May 22 20:43:35 kergoth: was this *unexpanded* condition always the case? or was it implemented recently? May 22 20:43:41 that's always been the case May 22 20:43:56 the version richard pulled in has always been based on unexpanded variables May 22 20:44:20 my initial implementation, which his was based on but still differs substantially from, used expanded forms, which had some advantages, but increased complexity May 22 20:45:24 kergoth: ok, thanks. I'm seeing expanded EXTRA_OECONF when running bitbake-diffsigs - Variable EXTRA_OECONF value changed from ... May 22 20:45:35 then something is forcing it to be expanded early May 22 20:45:41 e.g. with := or a getVar() at parse time May 22 20:45:48 well, getVar+setVar, i should say May 22 20:45:51 kergoth: yeah, let me check that May 22 20:47:43 the downside to unexpanded is the checksums end up caring not just about what things were set to, but also how they were set May 22 20:48:01 e.g. there's no real reason for hte checksum to be different between FOO = "${BAR}" and FOO = "bar" if BAR is bar May 22 20:48:16 the end result is the same, after all May 22 20:48:19 oh well May 22 20:48:26 kergoth: yep! both packages with different EXTRA_OECONF try to mangle the value in .bbappend and hence use := May 22 20:48:50 sadly common, most will use :=+oe_filter_out before using anonymous python to remove/alter the value May 22 20:49:14 _remove isn't really an option for bbappends, not if you want to play well with others May 22 20:49:24 best to convert that to anonymous python May 22 20:49:51 kergoth: thanks! May 22 20:49:54 np May 22 20:50:16 kergoth: can you comment on oprofile question from above? :) May 22 20:50:36 also sstate related... May 22 20:51:20 if the resulting package isn't machine specific, andn has no kernel modules, i'd be inclined to say it should be added to the safe deps for its dependency on the kernel, but i haven't played with oprofile much May 22 20:52:50 kergoth: yeah, me neither. I'm inclined to add it to safe deps, but looks like it pokes into kernel internals, so I'm a bit worried if the ABI is not entirely safe May 22 20:52:58 * kergoth nods May 22 20:54:23 oprofile been around in oe-core for long time, but looks like not many use it these days, as I'd expect it to be fixed one way or another... May 22 20:55:18 could always try marking it safe and using it on a few machines to check sanity May 22 20:55:22 not full proof, but better than nothing May 22 21:00:15 kergoth: btw, doing getVar/setVar in __anonymous will also expand it, isn't it? May 22 21:16:56 denix: if it's getting the expanded version, yes **** ENDING LOGGING AT Sat May 23 02:59:58 2015