**** BEGIN LOGGING AT Tue Aug 12 02:59:58 2014 Aug 12 06:31:10 I've configured the qt-creator for wandboard solo i've crosscompiled the sample clock code...but its giving the error can you please see the post http://qt-project.org/forums/viewreply/189080/ and help me Aug 12 07:01:44 I've configured the qt-creator for wandboard solo i've crosscompiled the sample clock code...but its giving the error can you please see the post http://qt-project.org/forums/viewreply/189080/ and help me Aug 12 07:19:54 good morning Aug 12 07:20:41 why does Yocto create broken binaries in my dbg package? http://paste.kde.org/ppcff10i3 Aug 12 07:21:13 the default package ships good binaries for "ldd". Aug 12 07:21:20 but the dbg package is simply too broken. Aug 12 07:24:10 good morning Aug 12 07:25:24 I have one question regarding GStreamer. Does anyone know which package provides playbin2? I installed gstreamer1.0 and gstreamer1.0-plugins-base, but gst-inspect1.0 can't find playbin2... Aug 12 07:34:46 here is the arm-..-objdump -p output for the debug binary generated by Yocto in the package: http://paste.kde.org/pqpblds2p Aug 12 08:29:22 is there a way to get log.do_package more verbose? Aug 12 08:34:16 morning all Aug 12 08:36:32 hello =) Aug 12 08:36:57 its the middle of night, not morning >_< Aug 12 08:37:12 ikey: \o/ Aug 12 08:37:15 o/ Aug 12 08:37:32 ikey: i'm in the office with my laptop and windows laptop trying to hack my way in =)))) Aug 12 08:37:42 ikey: at least i'm on freenode now. Aug 12 08:37:53 heh, if you need a hand drop me a line, more than willing to help :) Aug 12 08:38:52 morning ikey Aug 12 08:38:54 morning :) Aug 12 08:38:57 *night Aug 12 08:40:00 ikey: it is almost lunch time ;) Aug 12 08:40:10 that *would* explain my hunger.. Aug 12 08:41:02 ikey: haha, yeah, that might be so.. Aug 12 09:04:55 lpapp: you can use the --debug flag for bitbake to get more details, also there are multiple log.* files and run.* files you can look at (they should be in the same directory as the log.do_package for your task). But I am no expert on the matter, so if I am saying something wrong, then someone please correct me :) Aug 12 09:05:29 soderstrom: yeah, I was reading log.do_package, but it did not provide any useful information to me, sadly. Aug 12 09:05:37 I was going to read the do_package python implementation. Aug 12 09:05:54 after do_package I have broken binaries, before that I have good binaries. Aug 12 09:06:04 I will try --debug, thanks. Aug 12 09:07:51 lpapp: Also you can check the run.do_package to see the exact commands beeing executed for that task. Aug 12 09:08:17 yeah, I was looking into that, but looked like python. Aug 12 09:08:21 (not a simple log) Aug 12 09:22:18 Is it possible to move the build directory after running bitbake? Aug 12 09:23:13 slips: http://www.yoctoproject.org/docs/current/ref-manual/ref-manual.html#var-B Aug 12 09:23:22 you can set to anything you want, I believe. Aug 12 09:24:06 or the S variable, I do not know. Aug 12 09:24:10 hmm Aug 12 09:24:20 but the safest is to rebuild IMHO Aug 12 09:24:40 I was interested in doing an initial build, and move the build folder to a different computer Aug 12 09:24:44 or I would create a symlink :-) Aug 12 09:25:09 (I'm pitching yocto to some customers, and I don't want to spend an hour of the workshop downloading/building) Aug 12 09:25:29 well, the downloading part I could handle. Aug 12 09:25:36 well, that is what Yocto is about :) Aug 12 09:25:44 building a whole system which can take a lot of time at times. Aug 12 09:26:21 also, an hour building seems to indicate that you do not utilize distributed builds. Aug 12 09:26:29 one hour is way too much. Aug 12 09:27:16 I though one hour was quite good for an initial build Aug 12 09:27:31 although it was core-image-minimal Aug 12 09:28:05 that is very bad IMHO Aug 12 09:28:13 we used to achieve that in 30-40 minutes Aug 12 09:28:20 and we have not utilized all the cores on the network. Aug 12 09:29:00 I would not consider using Yocto without distributed builds to be honest ^^ Aug 12 09:30:15 btw, http://www.yoctoproject.org/docs/1.6.1/ref-manual/ref-manual.html#var-TOPDIR Aug 12 09:30:19 What does it take to set that up? Aug 12 09:32:33 http://www.yoctoproject.org/docs/1.6.1/ref-manual/ref-manual.html#structure-core-script Aug 12 09:34:20 slips: http://www.yoctoproject.org/docs/1.6.1/ref-manual/ref-manual.html#ref-classes-icecc Aug 12 09:35:07 then I only use: INHERIT += "icecc" Aug 12 09:35:07 ICECC_PARALLEL_MAKE = '-j 64' Aug 12 09:37:12 slips: sstate is what you want Aug 12 09:37:47 slips: do a build beforehand, then make sstate-cache available to workshop machines Aug 12 09:38:35 lpapp: seems like something I should at least set up at our workplace. Dunno about the workshop though Aug 12 09:38:51 slips: well, if you use the same build folder, just copy over and run the environment setup Aug 12 09:38:57 I think it should just work Aug 12 09:39:57 slips: I do not know what kind of workshop it is, so it is hard to say anything concrete :) Aug 12 09:40:00 rburton: can I set up an sstate folder in my local.conf, then build, copy the folder to a USB drive and make participants copy it locally? Aug 12 09:40:02 slips: sstate is enabled by default Aug 12 09:40:11 but yes, that will work Aug 12 09:40:54 slips: why would you put it into local.conf? What is wrong about the default? Aug 12 09:48:23 nevermind, I see the sstate-cache folder is there already Aug 12 09:48:48 I just remembered seeing the option in local.conf, so I thought it had to be enabled somehow. Aug 12 09:49:00 I have an option in local.conf Aug 12 09:49:31 SSTATE_MIRRORS cause I use mirrors. Aug 12 09:58:53 OK,let me see if I got this correct.. it's enough to just copy the sstate-cache from my build-folder to the workshop machines' build folders Aug 12 10:01:36 yeah Aug 12 10:01:59 make sure you rerun the init env script, I would say. Aug 12 10:02:45 noted Aug 12 10:03:04 I just tried copying to a different folder locally Aug 12 10:03:25 rerunning init env and running bitbake core-image-minimal Aug 12 10:03:36 and that seems to work Aug 12 10:04:03 the initial build dir is about 17GB, while the new is just 3.7 GB Aug 12 10:04:40 uh, you do not use rm_work? Aug 12 10:04:57 no Aug 12 10:05:51 soderstrom: I guess what I am looking for is the commands run. Aug 12 10:06:12 log/run.do_foo does not seem to provide that which is crucial for debugging when there is an issue there. Aug 12 10:08:06 lpapp: wow. inheriting rm_work in local.conf is something I should have known of Aug 12 10:08:51 I knew there was a lot of temporary build stuff that was never going to be reused, i just didn't know how to get rid of it Aug 12 10:10:06 I use it all the time by default, and if I need to debug something in the build tree, I comment the line in the local.conf Aug 12 10:10:12 (after debugging I remove the comment) Aug 12 10:11:34 once you get CI into your system, you cannot really afford not using rm_work by default. Aug 12 10:12:19 (if it was to me, I would have made the default behavior) Aug 12 10:12:39 (just like the build history chop) Aug 12 10:17:06 slips: thats why i suggesting sharing sstate, the build directory gets huge and mostly contains built trees you don't care about Aug 12 10:30:35 I've configured the qt-creator for wandboard-solo and getting the make error while crosscompile...can you see the post http://qt-project.org/forums/viewreply/189080/ and please suggest Aug 12 10:48:45 seems, Yocto breaks every single debug binary for me. Aug 12 10:49:03 not only for my software project... I can see this failure on every single binary in -dbg packages. Aug 12 10:51:25 are -dbg packages working for anyone with Yocto at all? :O Aug 12 10:51:59 they did last time I used them... Aug 12 10:52:23 strange, I can reproduce it even on stock packages. Aug 12 10:52:54 for some reason the dynamic section in the binary is badly truncated. Aug 12 10:53:10 leading any -dbg binaries to format/exec error on the target when trying to run it through gdb(server). Aug 12 10:53:34 non -dbg packages are all OK that I have tried. Aug 12 10:53:39 slips: fwiw, i give OE workshope regularly and create the ssate the day before the workshop as you described. it works perfectly fine. Aug 12 10:54:54 ndec: cool. I'll definitely try that Aug 12 10:55:14 I've configured the qt-creator for wandboard-solo and getting the make error while crosscompile...can you see the post http://qt-project.org/forums/viewreply/189080/ and please suggest Aug 12 10:55:34 hi guys, I am build an image...but it seems it got stuck...I build it with -v and -DDD options, but the last debug message I get is "DEBUG: Python function rootfs_runtime_mapping finished" Aug 12 10:56:00 slips: i recommend doing a copy of 1) the downloads folder, 2) the git tree you used. i have seen places with almost no network access. even git clone oe-core would take ages.. Aug 12 10:56:35 the task which never finishes is "Currently 1 running tasks (2592 of 2594): 0: core-image-graphics-1.0-r0 do_rootfs (pid 18838)" Aug 12 10:57:17 Hmm. didn't think of the git tree at all. I'll remember to bring that as well, just in case. Aug 12 10:58:54 slips: i generally use repo + manifest to pull in all trees, and repo can create a mirror with --mirror Aug 12 11:01:41 I wonder if anyone can do me favour and run objdump -p through a debug binary built by Yocto and paste the output... Aug 12 11:01:49 hmm. I've not used repo alot. isn't that just a Git wrapper of some sort? Aug 12 11:02:50 slips: yes. that's that only. Aug 12 11:03:02 slips: it helps you clone a 'forest' of git trees Aug 12 11:03:21 ok Aug 12 11:03:26 which is often useful when using oe/yocto since you ave to pull from many git trees Aug 12 11:04:17 right. and a manifest file is a description of the trees with URI, branch info etc.? Aug 12 11:04:27 to use with repo? Aug 12 11:05:11 slips: yes. e.g. https://git.linaro.org/people/nicolas.dechesne/oemanifest.git Aug 12 11:06:36 slips: also, be aware that when you are doing your workshop, a tiny change in OE-core might impact all tasks signatures and force a rebuild.. so make sure the 'students' use the same recipes as you used when you generated the sstate mirror Aug 12 11:07:11 oh. Aug 12 11:07:50 ndec: out of curiosity, what are the exact requirements for your NOR image? Why can't you build the bootloader separately? Aug 12 11:07:51 I've configured the qt-creator for wandboard-solo and getting the make error while crosscompile...can you see the post http://qt-project.org/forums/viewreply/189080/ and please suggest Aug 12 11:08:07 ndec: I do build 16M images for nor... Aug 12 11:08:08 and repo is handy here again... because you put a commit in the manifest to indicate which version of the tree to use. Aug 12 11:08:19 ndec: how do you normally solve that? fetch a specific commit? Aug 12 11:08:41 ah Aug 12 11:08:45 ant_work: i don't think your msg was for me ;-) Aug 12 11:09:07 hm.. oh, yes..you just answered... sry Aug 12 11:11:02 I'm a bit puzzled by that, maybe he's declaring the partitions with cmdline Aug 12 11:11:10 slips: here is an example of manifest i 'froze' for a workshop last week: http://git.baylibre.com/pub/training/oemanifest/commit/?h=daisy Aug 12 11:12:51 nice. This is exactly what I want Aug 12 11:13:50 thanks alot Aug 12 11:14:07 i guess a have a manpage to read Aug 12 11:14:42 hi Aug 12 11:14:54 although learning repo seems like a manageable task Aug 12 11:15:17 i need help to flash image for odroid board from oe built image Aug 12 11:16:58 is thr anyone who give it a try Aug 12 11:30:59 bluelightning: is the debug binary meant to be a replacement for the main binary or as an additional symbol pack? Aug 12 11:31:12 i.e. whether I should use the main binary as well as the debug for debugging or just the debugging? Aug 12 12:04:37 lpapp: additional symbols Aug 12 12:11:23 bluelightning: strange, the first time I see it used like that, but I guess it is enough to use the debug binary in that case on the host if I run gdbserver on the target, right? Aug 12 12:11:43 (no wonder then I could not run gdbserver through the non-stripped binary file from the -dbg package) Aug 12 12:12:00 anyway, I am still looking for the documentation where it is explained in the YRM since I think it is unusual. Aug 12 12:12:11 I was expecting the debug binary as a full replacement (drop-in). Aug 12 12:12:42 http://www.yoctoproject.org/docs/current/ref-manual/ref-manual.html#usingpoky-debugging Aug 12 12:12:50 I am reading this, but I cannot find this important information anywhere. Aug 12 12:14:53 it doesn't seem all that odd to me... that is how things are done on mainstream Linux distributions Aug 12 12:15:26 I have never seen this concept used before. Aug 12 12:16:07 usually, when I try to debug software (it is rare because I tend to use printf, but there are ocassions when I cannot, e.g. no access to stdin/err), I just build the software with -g -O0, and I get a non-stripped, non-optimized binary. Aug 12 12:16:11 with gdb symbols. Aug 12 12:16:22 so usually strace/gdb/valgrind binary afterwards. Aug 12 12:16:35 I have never seen it happening otherwise, even though I can understand the need for happening it otherwise. Aug 12 12:17:21 (this is the first time I see it, so it was new to me, and as such, the lack of this information - or not appareant -, got me learn it the hard way, but I think it could be put into the documentation explicitly) Aug 12 12:18:44 HI lpapp> Aug 12 12:19:06 hi bryan. Aug 12 12:19:31 last time u helpmed to build image for odroid Aug 12 12:19:45 using open embedded -> yocto - poky Aug 12 12:19:52 its succesful now Aug 12 12:20:12 can u tell me where can i find steps to flash image using sd card Aug 12 12:20:15 for odroid Aug 12 12:20:34 that is yocto agnostic, but I do not know your board. Aug 12 12:20:49 are you using serial flashing, jtag based or something else? Aug 12 12:20:53 bluelightning: perhaps here, http://www.yoctoproject.org/docs/1.6.1/dev-manual/dev-manual.html#platdev-gdb-remotedebug-launch-gdbserver Aug 12 12:21:17 i am using micro sd card Aug 12 12:21:20 bluelightning: add the client side example, and that will suffice without much additional text, yet crystal-clear then. Aug 12 12:21:36 i did it for beagle board by creating 2 partitions Aug 12 12:21:45 and trying similar for odroid board Aug 12 12:21:52 (at this point, I even need to look up the client side parameters, etc... one-liner example would spare me and others that) Aug 12 12:23:05 bluelightning: I also notice the lack of -dbg package mentioning there, which I think should be vital for gdb. Aug 12 12:23:25 i followed steps like Aug 12 12:23:25 http://elinux.org/BeagleBoardAndOpenEmbeddedGit Aug 12 12:25:45 I only used 'dd' or so a couple of years ago when I flashed the ubuntu-core image onto my beagle board. Aug 12 12:25:54 not sure what the trend is these days. Aug 12 12:26:33 yes we did it same for ubuntu image Aug 12 12:26:36 and it worked Aug 12 12:26:48 we checked that it also create 2 partition Aug 12 12:27:06 for odroid also we used dd Aug 12 12:32:08 i think i can get some help from http://www.funtoo.org/ODROID_U2 Aug 12 12:33:33 bluelightning: https://bugzilla.yoctoproject.org/show_bug.cgi?id=6626 Aug 12 12:33:34 Bug 6626: normal, Undecided, ---, scott.m.rifenbark, NEW , Add an example for debugging software with the corresponding -dbg package Aug 12 12:34:04 i’m trying to add avahi-daemon to my image (it seems to be in packagegroup-base-extended anyway), however, when i finish the image, it tells me Aug 12 12:34:12 “cannot satisfy the following dependencies: avahi-daemon” Aug 12 12:34:36 now when i look closer at it, it seems that the avahi-daemon are in avahi-daemon in the packages-split directory, but they get packages as libdns-sd-daemon Aug 12 12:34:57 which is not i’m not really sure where i am messing up here Aug 12 12:36:30 manuel_: odd, are you changing the avahi recipe in any way? Aug 12 12:36:40 lpapp: it's really not that difficult... you just install the -dbg package, and use gdb as normal; it will automatically find the installed debugging symbols Aug 12 12:36:59 manuel_: i certainly have an avahi-daemon package Aug 12 12:37:08 rburton: just while typing i realized that may be it, i enabled libdns_sd compat which seems to change all the resulting package names? Aug 12 12:37:19 i’m unclear about how the ipk names are created Aug 12 12:37:21 hm shouldn't have like that! Aug 12 12:37:31 it should only rename if a package contains just the library Aug 12 12:37:32 because the packages-split definitely is good Aug 12 12:37:34 * rburton looks Aug 12 12:37:40 the avahi recipe is a bit special Aug 12 12:38:24 manuel_: can i see your change or bbappend? Aug 12 12:38:43 <_valle_> Hello! I have followed the following guide, https://www.yoctoproject.org/downloads/bsps/daisy16/beaglebone, to setup an sd card for beglebone black. Aug 12 12:38:51 EXTRA_OECONF += “ —enable-compat-libdns_sd” Aug 12 12:38:58 hat’s it Aug 12 12:39:06 i need the dns_sd.h to compile another util Aug 12 12:39:11 <_valle_> However after Starting kernel ... the output goes blank Aug 12 12:40:10 so i see a couple of options, understanding why most packages are renamed to libdns-sd instead of avahi, making it so only the compat wrapper is renamed as a package Aug 12 12:40:18 <_valle_> Here is the output: http://pastebin.com/Yji77Zn0 Aug 12 12:40:26 another would be to install avahi as normal, make a new recipe that will only build and install the dns_sd wrapper (seems dirty) Aug 12 12:40:38 or rewrok the avahi recipe to actually use the new package names for everything Aug 12 12:40:47 (i’m very new to yocto, bear with me) Aug 12 12:41:04 manuel_: oh eek Aug 12 12:41:09 what's going on here Aug 12 12:41:26 yeah this is messed up :) Aug 12 12:41:27 rburton: in a weird way it’s satisfying to hear someone else going “wat?” Aug 12 12:41:40 manuel_: quick fix, add a new package for it Aug 12 12:41:59 rburton: package as in a package line inside the recipe? Aug 12 12:42:00 or a new recipe? Aug 12 12:42:01 ah, right Aug 12 12:42:15 so, there isn't a package called "avahi" Aug 12 12:42:26 but the avahi package gets created as this library isn't packaged anywhere else Aug 12 12:42:35 it contains just a library so renames to libdns_sd Aug 12 12:42:46 which causes everything else to be renamed, so avahi-daemon becomes libdns-sd-daemon Aug 12 12:42:51 ok Aug 12 12:43:01 i need a bit more informationa bout the packaging step that’s still a bit fuzzy to me Aug 12 12:43:19 so, you configure and make the source, after the do_configure went through Aug 12 12:43:25 do_compile finishes it’s thing, all good too Aug 12 12:43:41 now comes the do_install step, which will in this case do a make install after the prefix was set by do_configure Aug 12 12:44:01 this will copy a pletora of files into a target root directory inside the build dir Aug 12 12:44:35 now comes the step that will use the PACKAGES and the FILES_package variables to create individual ipks Aug 12 12:45:00 so what you mean is that any file that is left out of these FILES_something will get packaged into a package with the same name as the recipe Aug 12 12:45:36 because the only thing that is in that package is libdns_sd.a, it will rename the package from avahi to libdns_sd, which, for some glitch in the packaging code somewhere, will lead it to rename all the other packages to Aug 12 12:45:36 manuel_: http://pastebin.com/zJqQjU1v works Aug 12 12:46:05 manuel_: the default value for FILES_$PN includes lots of directories. the glitch is intentional. Aug 12 12:46:24 as avahi.bb explicitly doesn't put anything into avahi itself, we should set FILES_avahi to "" to stop this happening Aug 12 12:46:30 the solution being PACKAGES += “libdns-sd” FILES_libdns-sd += libdns_sd.so Aug 12 12:47:06 cool, thank you (that diff paste is broken btw) Aug 12 12:47:09 thank you very much Aug 12 12:47:13 did my mental workthrough make sense? Aug 12 12:47:18 yes, create a new package and put the libraries in it Aug 12 12:47:26 yeah you were pretty close :) Aug 12 12:47:41 cool Aug 12 12:47:43 oh yeah, broken paste Aug 12 12:47:49 (no worries) Aug 12 12:48:08 ok, let’s hope i’ll be able to build a decent toolchain now (i’m trying populate_sdk, no idea where this is going to go) Aug 12 12:48:13 i’ll be back to bother you people probably Aug 12 12:48:19 try http://pastebin.com/s6vuJJDd Aug 12 12:48:24 QXcbConnection: Could not connect to display pl help Aug 12 12:49:02 bluelightning: I am not using gdb, but gdbserver as written. Aug 12 12:49:53 bluelightning: but even then someone wrote me to use some options -s and -e... so options or not? It is confusing. Aug 12 12:50:00 rburton: already did. another question relating to that, not everyone wans the compat-libdns_sd stuff (although i see it making sense to have it on per default). how would i make a switch to toggle that on and off Aug 12 12:50:17 lpapp: I can't say I've ever used gdbserver myself Aug 12 12:50:24 i have some customization for the qt5 recipes which have a bunch of PACKAGECONFIG_GL PACKAGECONFIG_DISTRO etc… variables that you can overwrite in your bbappend, but that strikes me as quite custom Aug 12 12:50:40 lpapp: I don't disagree we need to document this better, but someone has to actually provide the correct background material so that we can Aug 12 12:51:06 manuel_: ideally PACKAGECONFIG, see the documentation for that Aug 12 12:51:51 bluelightning: well, I am trying to provide the reasons for my confusion. Aug 12 12:51:54 thanks Aug 12 12:52:00 to help to clarify these in the document. Aug 12 12:52:14 1) -dbg is additional info, not replacement Aug 12 12:52:22 2) Do I need to use options or will it just work off-hand? Aug 12 12:52:26 3) What about gdbserver? Aug 12 12:53:05 4) What to do when changing the debug location Aug 12 12:54:20 5) Perhaps even a FAQ for the format/exec error: the user is trying to run gdb over the debug file, but that is not how it is supposed to work. Aug 12 12:55:44 it's unlikely to be a FAQ, you're the first person to hit this to my knowledge Aug 12 12:58:29 lpapp: my last use of it was last year but all went smoothly after reading some docs around Aug 12 12:58:49 lpapp: happily once seen th eerror I debugged just using od Aug 12 12:59:08 ant_work: actually, no one could help here before my realization and investigation. Aug 12 12:59:15 (it was the bx instruction for armv4) Aug 12 12:59:26 if it was so total obvious, one could have already told me yesterday, so it is apparently not so obvious Aug 12 12:59:42 otherwise I would not have wasted one day as a reasonably experienced embedded developer. Aug 12 13:00:03 I do use gdb very seldmo, I don't remember offhand the steps Aug 12 13:00:17 and trust me or not, I have read the documentation at least twice or thrice. Aug 12 13:01:11 can't remember exactly which examples I've followed, sorry Aug 12 13:03:36 no worries, once I figure out all the doubts, I will propose some documentation extension :-) Aug 12 13:04:52 it should be copy/paste finger-exercise for developers to debug their software. Aug 12 13:43:45 so i have this meta-toolchain-qt5 recipe that creates a toolchain with the host and native qt, but i also have the populate_sdk toolchain for my image. what would be the best way to combine both? Aug 12 13:43:59 i’m still abit unsure about that part and the sysroots and what populate_sdk itself does Aug 12 13:51:23 is there an option for bitbake to print out the OS commands run? Aug 12 13:53:12 heh, http://git.yoctoproject.org/cgit/cgit.cgi/poky/tree/meta/classes/package.bbclass#n1405 -> " 2>/dev/null" Aug 12 13:53:21 (why not os.devnull ^) Aug 12 13:55:26 what I do not get is this: the gdb documentation ( https://sourceware.org/gdb/current/onlinedocs/gdb/Separate-Debug-Files.html ) says: - /usr/bin/.debug/ls.debug Aug 12 13:55:35 yet, yocto is generating /usr/bin/.debug/ls Aug 12 13:55:41 (how is this supposed to work?) Aug 12 14:51:20 YPTM: Good Morning. Markus is on the call Aug 12 15:01:18 YPTM: Company: WIND RIVER SYS Ready-Access Number: 8007302996/9139049836 Access Code: 2705751 Aug 12 15:01:24 YPTM: Stephen joined Aug 12 15:01:32 * RP joined Aug 12 15:01:34 YPTM: Tom Z on the call Aug 12 15:02:02 YPTM: Joe MacDonald on the call Aug 12 15:02:02 YPTM: Nitin is on the call Aug 12 15:02:03 YPTM: Saul is on Aug 12 15:02:06 YPTM: Michael on the call. Aug 12 15:02:37 YPTM: Alexandru Vaduva here Aug 12 15:03:19 YPTM: Denys is on Aug 12 15:03:22 YPTM: Jeff Polk is on Aug 12 15:03:39 * Jefro is on YPTM Aug 12 15:03:49 YPTM: Mark Hatle is now on Aug 12 15:03:59 YPTM ross will be on shortly Aug 12 15:04:53 YPTM: Paul Eggleton is on Aug 12 15:04:59 YPTM: Belen in on Aug 12 15:05:59 Alex D is here Aug 12 15:16:23 YPTM is over Aug 12 15:19:45 is it possible to modify a bbclass in oe-core without having to rebuild all recipes afterward? Aug 12 15:31:18 Anarky: modifying a class will cause a rebuild of anything that inherits that class, and anything that depends on the things that inherited that class, assuming the changes to the class affect the metadata of the tasks being run. adding 'foo="bar"' to a random class won't cause a rebuild of anything, for example, unless foo is used in a task directly or indirectly Aug 12 15:34:40 kergoth: I'm planning to modify a function in package.bbclass; can I desactivate temporarily this inherit system? Aug 12 15:36:23 you can use bitbake -b to make it just run exactly what you tell it Aug 12 15:41:34 hmm, interesting gdb-cross does not have any content other than temp? Aug 12 15:42:10 ls ./tmp/work/armv5te-foo-linux-gnueabi/gdb-cross/7.5.1-r0.0/ -> build-arm-foo-linux-gnueabi pseudo temp Aug 12 15:42:16 the first two are empty unlike "temp". Aug 12 18:32:56 I'm having trouble with alsa or my lack of knowledge about audio on linux. My Yocto image works, XBMC runs fine and XMMS2 runs fine but when i play audio with both it results in noise/static until i restart. I can use only one output device (spdif, i can't use hdmi) and i think i need some kind of mixer. Anyone who can give me a hint how to play audio from two sources at the same time? Aug 12 18:52:57 pulseaudio can do mixing Aug 12 19:53:57 I am new to Yocto and using a freescale imx6 solo lite. I have modified u-boot-fslc and tested that it works. There seems to be quite a bit of documentation, but figuring out the whole flow for contributing back isn't all that clear to me. Aug 12 20:13:46 rburton: thanks, i will try to install pulseaudio Aug 12 20:25:59 The board I am working on is similar to one of the reference freescale designs but doesn't use sd card and has less periph (no audio, headless, no video, no wifi). Aug 12 20:27:08 uses emmc and has a different ram part. Aug 13 00:19:50 does anyone know if a recipe for OpenJDK exists? Aug 13 00:19:52 I can't find one Aug 13 00:26:11 ah, found it Aug 13 00:48:37 maxtothemax: yes it does. Look at meta-java Aug 13 00:49:15 and when you need to look for layers or recipes go to http://layers.openembedded.org/layerindex/branch/master/layers/ Aug 13 00:49:18 and search Aug 13 01:00:05 i don’t really understand the relationship between meta-toolchain-* recipes and -c populate_sdk on my image Aug 13 01:00:20 will meta-toolchain-* package my whole sysroot? because i seem to be missing includes Aug 13 01:00:44 or is there a way to have meta-toolchain-qt5 (in my case) package the sysroot too Aug 13 01:05:02 oh i think i need to add my packages to TOOLCHAIN_HOST_TASK Aug 13 01:12:19 yeah, I found it Aug 13 01:22:23 manuel_ 'meta-toolchain' provides a very basic toolchain environment, but very few (if any) libraries beyond the minimum.. Aug 13 01:22:34 you can think of this as a very targeted SDK with a limited purpose.. Aug 13 01:22:50 bitbake -c populate_sdk on the otherhand is a "generic" SDK that matches your root filesystem Aug 13 01:23:15 So, if you want to provide your developers with access to all of the development components for the rootfs, use the -c populate_sdk.. Aug 13 01:23:36 if you want to provide your developers with a limited set of APIs, use the meta-toolchaiN (or equivalent) and add only the items they should have access to Aug 13 01:47:18 fray__: you know developers who know their required dependencies? where do these people work? Aug 13 01:48:38 system developers should have a good understanding of what they want to export to the application developers.. Aug 13 01:48:47 for those who don't know what to export, I'd recomment the full filesystem.. Aug 13 01:49:00 populate_sdk ftw Aug 13 01:49:00 for those who have planned only a limited set of APIs, then the manual meta-toolchain approach is for them Aug 13 01:49:24 there is a reason I implemented populate_sdk 3 years ago.. since that is how most people work Aug 13 01:49:26 * nerdboy finds out about new deps when the nightly breaks Aug 13 01:49:55 then i get "oh yeah, forgot about that..." Aug 13 01:50:20 in the automotive world.. those guys are -very- strict about APIs.. Aug 13 01:50:30 so they want the target toolchain approach, not rootfs approach Aug 13 01:50:46 seriously, the flexibility is there for limited-to-full image Aug 13 01:51:06 yes Aug 13 01:51:11 mostly outside of "strict" i would use populate_sdk Aug 13 01:51:51 yes, that is what I recommend for peopel to use Aug 13 01:52:45 you automotive/aerospace/launch guys and your "strict" Aug 13 01:53:18 * nerdboy was the launch range IV&V guy Aug 13 01:54:04 woulda been nice to see good use of OE on some of their embedded/infra stuff Aug 13 01:54:08 the other place I've seen strict is on certain consumer devices -- tablets and settop boxes where they allow applications.. Aug 13 01:54:20 but that isn't nearly as common as the MILS/Aero/Automotive Aug 13 01:54:29 timing system, range infrastructure, comm Aug 13 01:55:30 even the new range safety standard doesn't include the updated software piece Aug 13 01:55:58 nasa lost funding for the third volume i guess Aug 13 01:56:37 so it's still do178b/spark if you really want strict as in correctness Aug 13 01:57:11 or wail on it with your own formal verification stuff Aug 13 01:59:25 :) Aug 13 02:06:20 fray_: the populate-sdk doesn’t provide the nativesdk packages, can i just add those to my image? Aug 13 02:06:26 i’m trying to bbappend RDEPENDS_ to my meta-toolchain-qt5 but that seems to not work Aug 13 02:06:32 although i just bitbake -c cleaned Aug 13 02:07:04 those can be added to the variable mentioned before.. Aug 13 02:07:19 TOOLCHAIN_HOST_TASK ?= "nativesdk-packagegroup-sdk-host packagegroup-cross-canadian-${MACHINE}" Aug 13 02:07:34 RDEPENDS won't do it (AFAIK).. Aug 13 02:07:47 ok, i looked inside the packagegroup to see how it works under the hood and it seems to use the RDEPENDS Aug 13 02:07:49 ok Aug 13 02:07:56 there are two key variables in the SDK.. TOOLCHAIN_HOST_TASK and TOOLCHAIN_TARGET_TASK.. (the later is automatically filled out when using the -c populate_sdk) Aug 13 02:08:01 that makes sense though because the RDEPENDS wa in the package group Aug 13 02:08:23 ya, packagegroups are brought in my the explicit list (either PACKAGE_INSTALL or TOOLCHAIN_TARGET_TASK) and then rdepend to bring in additional stuff.. Aug 13 02:08:50 to add single packages, you can just add (_append) to the TOOLCHAIN_TARGET/HOST_TASK and it will be specific to the system or recipe Aug 13 02:08:57 cool, i’ll add that stuff to TOOLCHAIN_TARGET_TASK, try it out, then make a proper packagegroup for my needs Aug 13 02:09:26 TOOLCHAIN_TARGET_TASK is things that are -target- based that go into the sysroot.. TOOLCHAIN_HOST_TASK are things installed in the host side.. i.e. cross compiler or similar tooling Aug 13 02:09:26 other related question though, if i add the nativesdk packages to my image.bb, will it get put into the toolchain when populate_sdk ? Aug 13 02:09:39 (nativesdk packages are TOOLCHAIN_HOST_TASK) Aug 13 02:09:57 your image.bb would need a TOOLCHAIN_HOST_TASK_append = "..." Aug 13 02:10:08 that is perfectly legal to have in an image and it'll just be ignored if building a rootfs Aug 13 02:10:27 cool Aug 13 02:10:55 the meta-toolchain is basically just a recipe missing all the image stuff Aug 13 02:11:07 i’m still thinking of them a separate mechanism but they’re not Aug 13 02:11:38 ya, they are the same mechanism, just different ways to express it Aug 13 02:11:46 thank you for writing that btw Aug 13 02:12:04 * fray_ notes meta-toolchain existed long before he became involved in YP.. Aug 13 02:12:15 it just wasn't how I wanted to work with the SDK.. thus the change Aug 13 02:12:17 for contributing to that Aug 13 02:12:22 :} Aug 13 02:12:44 yup.. different usage patterns.. OE/YP is great at have a million ways to do something.. problem is making sure those million ways are known Aug 13 02:13:11 source is easy to read Aug 13 02:13:23 it’s been surprisingly easy to get some complicated stuff done Aug 13 02:13:41 but the whole “build stuff that will take hours, fix a var, build stuff that will take hours, fix a var” cycle is tedious Aug 13 02:14:51 this is where you convince your boss that a faster machine will pay for itself in developer time.. Aug 13 02:14:59 I'm lucky enough to have such a boss.. Aug 13 02:15:16 developer pay/hour * time saved by faster hardware < cost of new hardware Aug 13 02:15:19 what are you building on Aug 13 02:15:39 offically, dual 8-core Xeon (w/ HT) and 128 GB of ram and RAID SATA disks.. Aug 13 02:15:41 as that is actually what i requested today Aug 13 02:16:02 unofficially, I have a 'target' which I've turned into a builder.. dual 12 core Xeon (w/ HT) and 64 GB of ram, and RAID SATA disks.. Aug 13 02:16:07 32 cores thus Aug 13 02:16:18 the later is about 50% faster (on master builds) then the 8-core.. Aug 13 02:16:20 target meaning (new to the lingo) Aug 13 02:16:33 both around about the same GHz range.. 2.4-2.6 if I remember correctly.. Aug 13 02:16:54 target, the system was purchased to be a target of the oe-core building, not as a builder.. Aug 13 02:17:17 (it's a high-end Intel Xeon system used in data centers for network routing and monitoring... I found a 'better' purpose as a builder) Aug 13 02:17:34 but since it's not an official developer system, if another person in the company needs it as a target system I have to send it to them Aug 13 02:18:11 Only word of caution.. there is a kernel bug when you get over about 32 (w/ HT) virtual cores.. and buildstats doesn't work right all the time.. Aug 13 02:18:24 what does a target do? just sit and run tests once a build has been created? Aug 13 02:18:27 there is a DMA buffer that fills up in the kernel that reports "out of resources" and causes the buildstats module to crash Aug 13 02:18:58 we have a variety of targets.. they're used for development (bootloader, kernel, and userspace), as well as regression testing and such.. Aug 13 02:19:08 gotya Aug 13 02:19:29 this one was originally assigned to me for kernel and userspace development.. but I finished that work a month or so ago.. and nobody else has requested it.. so I installed CentOS and turned it into a builder... :) Aug 13 02:19:37 it's -really- fast with 24/48 cores Aug 13 02:19:47 that sounds tempting Aug 13 02:20:03 nice, the TOOLCHAIN_TARGET_TASK stuff worked Aug 13 02:20:14 we have some actual developer machines that are comparable (Dell rack mount systems).. but they're "shared" machines.. any developer is allowed to access them.. Aug 13 02:20:25 we have a few large core machines dedicated to toolchain development as well.. Aug 13 02:20:40 all depends on the workload.. for system development cores, RAM and GHz all matter.. Aug 13 02:20:41 who is we btw? Aug 13 02:20:46 I work for Wind River Aug 13 02:20:57 cool Aug 13 02:21:18 for kernel developers the GHz matter a lot less (in my experience) but you still want lots of cores.. Aug 13 02:21:31 for application developers you likely want faster CPUs and less cores.. (not always) Aug 13 02:21:57 for 'test' systems (systems running tests) large cores really aren't important.. as they are usually directing remote systems to do things Aug 13 02:23:51 a few cores wouldn’t hurt our application build machine Aug 13 02:24:03 i just sent out a request for some > 24 cores machine Aug 13 02:24:32 thank you again Aug 13 02:24:38 i’m off to bed Aug 13 02:24:41 ya.. the 12 core Xeons are expensive.. but they really do hel Aug 13 02:24:47 ave a good evening! Aug 13 02:25:03 (BTW on my 'target' I can do a full core-image-sato style build in less then 2 hours) Aug 13 02:27:44 that’s enticing Aug 13 02:35:31 I have haswell based box Intel(R) Core(TM) i7-4770 CPU @ 3.40GHz with 16G DDR3 and my core-image-sato takes < 100 mins Aug 13 02:35:46 so fray_ something is wrong with your big honking machines Aug 13 02:36:26 Dell XPS 8700 from costco under 1000 bucks gets the job done Aug 13 02:41:56 :) Aug 13 02:58:07 * nerdboy relaxes to shark week on discovery channel... Aug 13 02:58:57 khem, I haven't timed it in a while.. Aug 13 02:59:14 and I was including the populate_sdk time as well, which takes a while longer **** ENDING LOGGING AT Wed Aug 13 02:59:58 2014