**** BEGIN LOGGING AT Mon Jun 26 03:00:04 2017 Jun 26 08:13:12 how do i make a task after do_image is complete, my_task is inside a class. Jun 26 08:16:23 edgar444: you can use bb.build.addtask python API for that Jun 26 08:19:01 ed2: thx for answering, yeah im seeing something like that used by meta/image.bbclass. Is there nice doc for it like: . Either that or just grep source. Jun 26 08:54:43 edgar444: this one? http://www.yoctoproject.org/docs/current/bitbake-user-manual/bitbake-user-manual.html#promoting-a-function-to-a-task Jun 26 08:57:00 ed2: Thats te usual way, i was looking into bb.build.addtask, found it in `/bitbake/lib/bb/build.py` Jun 26 08:57:56 ed2: What im trying to do is create task that creates flashable image out of do_image created rootfs, and kernel. Right now have to do it manually with mkimage Jun 26 08:59:42 edgar444: did you try wic? Jun 26 09:00:26 ed2: seems complicated, and i just want to reuse the already made rootfs, instead of making my own. Jun 26 09:00:34 edgar444: this is exactly what IMAGE_TYPES += 'wic' would do for you generally. Jun 26 09:01:11 ed2: ok I'll look into it. Jun 26 09:01:47 edgar444: what's your target hardware arch? Jun 26 09:01:48 history of failed software in one sentence: "but i just want to" Jun 26 09:02:45 ed2: arm, rootfs is used as initramfs Jun 26 09:05:22 edgar444: you can look at meta-yocto-bsp/conf/machine/beaglebone.conf meta-yocto-bsp/wic/beaglebone.wks in poky repo as a starting point. Jun 26 09:05:58 edgar444: it works just fine for my BeagleBone Black. Jun 26 09:16:48 I get a QA error (expanded-d) when using ${D} in pkg_preinst. What is the reason for that check? Jun 26 09:55:32 edgar444: if you just need an initramfs image you can use IMAGE_FSTYPES += "cpio.gz" and core-image-minimal rootfs. Jun 26 09:56:11 ed2: after core-image-minimal i need to pack everything into a custom uImage with mkimage Jun 26 09:57:47 edgar444: you don't have to if your u-boot understands bootz command with this syntax: bootz ${nb_kerneladdr} ${nb_initrdaddr}:${filesize} ${nb_dtbaddr} Jun 26 09:58:03 edgar444: this works on my BBB just fine. Jun 26 09:58:04 nah i guess edgar444 means a fitimage Jun 26 10:00:33 LetoThe2nd: correct, im using a fitimage. Command: mkimage -f kernel.its kernel-fsl-ls1012a-frdm.itb Jun 26 10:00:56 edgar444: KERNEL_CLASSES += "kernel-fitimage" ? Jun 26 10:01:14 yeah we have proper support for that Jun 26 10:01:16 http://cgit.openembedded.org/openembedded-core/tree/meta/classes/kernel-fitimage.bbclass Jun 26 10:01:22 no need for manual mkimage magic Jun 26 10:01:40 (thx marex-cloud ) Jun 26 10:01:56 edgar444: i also did that, but it turned out to be unnecessary step for me. Jun 26 10:02:32 LetoThe2nd: Reasons to describe your problems on irc! ^^ That looks nice. Jun 26 10:02:45 edgar444: I load initramfs, dtb and kernel from tftp and use bootz to boot the board. Jun 26 10:03:24 edgar444: reasons to actually describe your real problem ("i want to bundle a fit iamge") instead of asking for help on a workaround ("i want to add some custom postprocessing to my rootfs") Jun 26 10:03:34 :-D Jun 26 10:03:57 * rburton wonders why you'd ever want a directory to be mode 0511 Jun 26 10:04:08 or, ----r-x--x Jun 26 10:04:18 rburton: cmon. thats obvious. when ever you want it to be one less than 0512. Jun 26 10:04:32 erm, -r-x--x--x even Jun 26 10:20:15 rburton, RP: any problems with these patches http://lists.openembedded.org/pipermail/openembedded-core/2017-June/138388.html http://lists.openembedded.org/pipermail/openembedded-core/2017-June/138339.html ? Jun 26 10:20:44 rburton: RP: I don't see them in master-next and ross/mut Jun 26 10:25:28 now in mut2, i ripped anything wiccy out last week and needed to bring it back in slowly Jun 26 10:41:55 rburton: don't see them in mut2 either. btw, both patches have nothing to do with wic. Jun 26 10:42:04 I'm seeing lz4-native break when configuring (I think after one succesful build). I can send a patch, but am a little surprised no-one else is complaining? Jun 26 10:44:35 ed2: yeah i was agrressive in pruning as the world was blowing up. need to push still :) Jun 26 10:45:30 jku: i guess not many people see rebuilds of stuff that low Jun 26 10:49:30 rburton: Do you still want me to test the video playback stuff with gst-changes btw? Jun 26 10:49:40 jku: please Jun 26 10:49:47 sure Jun 26 10:51:26 * fray is sleepy.. and it's only 12:51pm here.. Jun 26 10:51:30 argh.. must stay awake Jun 26 10:51:46 fray: ah, welcome to the real world! Jun 26 10:51:59 na, I'm on the wrong side of the world Jun 26 10:52:23 hrhr Jun 26 10:52:37 MORNING FRAY Jun 26 10:52:58 only 8 more hours before you can even think about going to sleep Jun 26 10:53:01 lol.. I've been awake (mostly) since 7am Central US time.... and it's 12:52 german time now.. :P Jun 26 10:53:13 rburton, we'll see if I make it.. if I can get to AT least 8pm, I should be much better ;) Jun 26 10:53:59 (my normal alarm goes off, roll over and go back to bed time is in an hour) Jun 26 10:57:46 My least favorite thing about jet lag Jun 26 10:57:58 er about travel Jun 26 11:02:41 Crofton|work yup.. I nodding off.. have to head to meet my coworkers in about 15 minutes.. Jun 26 11:02:50 so I need to be awake enough to get my butt back to the airport.. :) Jun 26 11:03:03 but at least I'm in the land of good beer Jun 26 11:03:21 how long are you there for? Jun 26 11:03:30 through Monday morning Jun 26 11:03:43 fly home just in time for the 4th Jun 26 11:05:35 ah, the get back to airport made me think fast trip Jun 26 11:06:04 ya, I got in at 7:30 this morning, everyone else was scheduled to land between 13.50 and 1420 Jun 26 11:06:15 so I went to the WR office in Ismaning to kill time.. Jun 26 11:06:19 ah so picking people up Jun 26 11:06:22 (got some actual work done too) Jun 26 11:09:40 ohh well time to pack up and get to the station in time for the S8 Jun 26 11:09:42 later Jun 26 11:53:40 Is the data store per recipe? Put another way, can I put stuff into it and pick it up in other recipes / during post processing? Jun 26 11:54:28 hello Jun 26 11:55:20 I get Unable to generate default variables from filename (too many underscores)" for a recipe name having five underscores Jun 26 11:56:01 i've found a code that says that len(parts) should be <=3 Jun 26 11:56:08 MWelchUK: generally, recipes should not be interdependent, except for their provided packages. Jun 26 11:56:11 cornel: underscores are not allowed in recipe names except between the version and the nam Jun 26 11:56:17 (actually says: if len(parts) > 3 ....) Jun 26 11:56:17 *name Jun 26 11:57:24 LetoThe2nd: Yeah, realise that. Trying to build up license information into 1 file (a requirement I have) rather than as separate files as the standard approach does... Jun 26 11:58:36 bluelightning, thank you , that was exactly my question :) why only two underscores ... Jun 26 11:58:47 MWelchUK: huh? the manifest should contain a complete license list for all packages in use. Jun 26 11:59:20 bluelightning, so we can at most have BPN_PV_PR.bb , right? Jun 26 11:59:20 cornel: I honestly don't know why the code allows more than one - I would not recommend it Jun 26 11:59:41 cornel: PR won't be extracted from the filename automatically like that Jun 26 11:59:46 PN and PV yes Jun 26 11:59:47 oh Jun 26 12:00:07 MWelchUK: tmp/deploy/licenses/$YOURIMAGENAME/license.manifest - there you go. Jun 26 12:00:08 I'd also prefer not to have the *whole* source file that LIC_FILES_CHKSUM specifies. Jun 26 12:00:51 to be more precise, its $YOURIMAGENAME-$YOURARCH-$YOURBUILDTIMESTAMP Jun 26 12:01:11 LetoThe2nd: I need the text of the license as well. Jun 26 12:02:23 As in, the license text, rather than just broadly which license it is. Jun 26 12:03:28 MWelchUK: so you want something that basically takes thins manifest, and then inserts the whole text of the license where the license short name is now. hm, no idea if that already exists. Jun 26 12:03:30 For example, kernel-image doesn't actually include the preamble added by Linus, so need more than just the manifest. Jun 26 12:04:17 LetoThe2nd: Yes, I also need to get it on to the target image. Jun 26 12:04:36 Which isn't too bad if I do it during post-processing. Jun 26 12:04:53 MWelchUK: maybe you can leverage the buildinfo stuff, its does quite similar things. Jun 26 12:05:34 Yeah - thanks for pointing me towards the license manifest, it clearly gets a good percentage of what I need... Jun 26 12:06:28 MWelchUK: good luck. Jun 26 12:06:33 bluelightning, thank you very much Jun 26 12:06:39 FWIW, there's also LICENSE_CREATE_PACKAGE, but that splits everything out as separate files. Jun 26 12:09:20 it's a bit of a hack, but you could generate license packages, include them in the rootfs and use a rootfs postcommand to post-process the image to include a combined file (and remove the individual license files)? Jun 26 12:14:10 in fact the manifest is written as a ROOTFS_POSTPROCESS_COMMAND, could probably leverage code in license.bbclass to write a merged license file somewhere into the rootfs Jun 26 12:15:44 MWelchUK: ^^ Jun 26 12:41:30 LetoThe2nd: I still went my way instead! :) http://dpaste.com/3ASB77H Jun 26 12:42:39 edgar444: feel free to use whatever fits your case :) Jun 26 12:48:22 rburton: FYI, gstreamer 1.12 branch works with meta-intel using gst-play but not with with gtk-play. A bit strange, maxin promised to have a closer look Jun 26 12:48:39 huh Jun 26 12:48:47 would be interesting to compare with 1.10 Jun 26 12:49:01 rburton: our gtk-play of course comes from the deprecated repo... Jun 26 12:49:16 and the new one only has meson as build system :) Jun 26 12:49:57 lol Jun 26 12:50:03 yeah best get that into core Jun 26 12:50:14 g-i breaks with meson right now Jun 26 12:56:13 joshuagl: Thanks will look into it. Leaning away from LICENSE_CREATE_PACKAGE as I'd still have to hack that to not provide the whole file listed in LIC_FILES_CHKSUM Jun 26 13:09:20 Hi guys, I doing some kind of recipes "porting" from yocto 1.8 to yocto 2.2.1 and I am facing some strange problems with all of my recipes: basically I have lots of warnings on do_package_qa task about missing RDEPENDS on core-related libraries, eg: all of my packages depends on libc.so.6 or libgcc_s.so.1, but "no providers found in RDEPENDS_${PN}" Jun 26 13:10:57 are you adding those libc.so.6 dependencies yourself? Jun 26 13:11:10 if so, don't do that. let do_package find them. Jun 26 13:12:35 zero_note: I faced something similar when i had to port from 2.2.1 (Ubuntu 16.04) to 1.8 (Ubuntu 14.04). My mistake was using libraries that i had built with a different compiler. Jun 26 13:12:36 rburton: no, I am not addind those depends Jun 26 13:15:12 zrrbite: and it was my first thought, but none of my libraries were pre-built with something different toolchain Jun 26 13:15:22 Hmm, allright. Jun 26 13:17:05 and this happen even with Qt stuff, because the majority of my projects use qmake and Qt stuff in general (eg. no providers for libQtCore.so.5 found) Jun 26 13:19:00 recipe specific sysroot? Jun 26 13:19:15 Which host OS are you using? I had to switch from Ubuntu 16.04 to 14.04 - If you've installed 2.2.1 on an older version of ubuntu there will be QA warnings. Jun 26 13:20:02 zero_note, RSS will expose missing recipe DEPENDS Jun 26 13:22:08 zrrbite: right now I have both configurations running on ubuntu 14.04.5 Jun 26 13:22:34 but, from what I read here http://www.yoctoproject.org/docs/2.2.1/ref-manual/ref-manual.html#detailed-supported-distros, both 14.04 and 16.04 should be supported Jun 26 13:22:37 Ok, i think that's your problem. Check your logs - You'll find a warning. Jun 26 13:23:07 Crofton|work: sorry, what do you mean with RSS? Jun 26 13:23:20 recipe specific sysroot Jun 26 13:24:05 (Atleast, i did - Moving to 16.04 fixed my dependency issues) Jun 26 13:24:53 Crofton: he mentioned 2.2.1 so RRS should not be involved, right? Jun 26 13:25:04 RSS Jun 26 13:25:35 Crofton|work: I think RSS should be done for 2.3 Jun 26 13:25:42 lol, I can't track the number to name mapping Jun 26 13:25:56 the numbers have no meaning in my head Jun 26 13:25:56 Crofton|work: lol, no prob Jun 26 13:26:10 I jsut figured he would jump as far as possible Jun 26 13:27:59 zrrbite: anyway I should have an ubuntu 16.04 virtual machine somewhere, I will test your suggestion in the next hours Jun 26 13:28:43 Fingers crossed! Jun 26 13:32:33 zero_note: You could just be missing gcc_multilib (it has libc6 as dependency). Jun 26 13:42:43 looks like I'm getting hit by a problem with sstate sharing (with morty): many build artifacts seems to hardcode various paths, including sysroot Jun 26 13:43:29 that causes kinda funky behaviours... is that a known issue ? Jun 26 13:47:27 yann: example? hardcoded paths like that should be sorted on extract from sstate Jun 26 14:15:16 hello guys !! Does anybody know how to copy a whole directory to the ${D} folder within a recipe ? I've tried install -m 0644 ${S}/MyDir/* ${D}/MyDir, but acctually copies only the files without the subdirectories Jun 26 14:18:19 kappe: there's probably a cleaner way, but you could just for loop the wild card Jun 26 14:18:56 kappe: also look at -d & -D args for install Jun 26 14:19:57 thank you CTtpollard Jun 26 14:20:05 I'll look now Jun 26 14:27:48 Are there no dbus-c++ recipes anymore? Jun 26 14:30:26 ah. blacklisted... Jun 26 14:35:10 CTtpollard, please can you give me there reference to the install command ? I still can't find it Jun 26 14:36:33 kappe: cp -r? Jun 26 14:42:16 ty rburton Jun 26 14:43:21 not quite the same :) Jun 26 14:43:35 I have heard something Jun 26 14:43:52 but still haven't understood very well the difference Jun 26 14:43:58 I'm a total newbie Jun 26 14:45:12 I just have a folder (with subfolders) representing a website, I jsut want to copy that directory to the image/ folder Jun 26 14:46:34 An alternative is to use: find ${WORKDIR}/git -not -path '*/\.*' -exec 'install -m 0755 "{}" ${D}/' \; Jun 26 14:47:11 but then the error is due to the missing subfolder (also the install -d dir must be automated) Jun 26 14:52:52 yeah you'll need to create the target first Jun 26 14:53:23 honestly mkdir -p ${D}/var/www, cp -a ${S}/ ${D}/var/www/ will work fine Jun 26 14:53:29 erm, cp -r not -a Jun 26 14:53:30 -a would be bad Jun 26 14:53:57 wasn't there something on the list lately? Jun 26 14:54:38 ok rburton ty Jun 26 14:55:24 recipes: Replace "cp -a" with "cp -R --no-dereference --preserve=mode, links" Jun 26 14:55:35 rburton: that was the one, exactly! Jun 26 14:55:40 -r is -R, maybe -R is POSIX Jun 26 14:55:54 for a load of files, cp -r will work Jun 26 14:58:02 Can I ask for a suggestion ? Jun 26 14:58:11 never eat yellow snow Jun 26 14:58:37 unless you colored it yourself with vodka-orange Jun 26 14:58:38 I have to compile for a board Jun 26 14:58:43 LetoThe2nd: good call Jun 26 14:58:55 rburton: :-) Jun 26 14:59:31 which differs by some file configurations Jun 26 15:00:02 it's better to create three separate machine that install the proper file set due the value of the machine name Jun 26 15:00:31 or copy the whole three sets and replace them with a script ? Jun 26 15:00:40 better eg easier Jun 26 15:00:53 rburton: g++ -v shows an include search path referring to the sysroot where the first builds were done Jun 26 15:01:10 yann: thats fine we override the sysroot in builds anyway Jun 26 15:01:37 kappe: are the boards physically identical? if the hardware is the same then you're describing three different images. Jun 26 15:02:27 yes, same boards with same hardware Jun 26 15:02:28 you could have a machine for each board, as most packages are built for the tune they'll be shared, then you can have machine-specific packages which have the right values for each machine. but if they're literally the same why not do the same in images. Jun 26 15:03:38 bi thought abput different images because for example I need three boards with three different wifi ssid AP Jun 26 15:03:42 for example Jun 26 15:04:24 so basically software/hardaware are the same, they differ only in configuration Jun 26 15:04:30 i really wouldn't go as far as new machines to change a configuration file Jun 26 15:04:38 ok Jun 26 15:04:45 you'll build a kernel per machine you see Jun 26 15:05:16 conf-a.bb, conf-b.bb, conf-b.cc, and pour it into three images. Jun 26 15:05:17 don't really know, I've started working with yocto during the last week Jun 26 15:05:26 * LetoThe2nd agrees with rburton one more time. Jun 26 15:05:32 LetoThe2nd: invoice me later Jun 26 15:06:03 * LetoThe2nd you are aware what "the voice" is in the context of my nickname? ;-) Jun 26 15:06:18 ty guys Jun 26 15:06:35 in fact my original problem is that mesa build fails, not finding , so I fired up a devshell, pasted all the env exports from run.do_compile, and went to run the same g++ command with just -v added - the result is the same so I'm at least a bit confident that i'm doing the right thing, and -v discloses that buggy search path, and no .../include/c++/... at all, which seems to hint for a relation between the 2 issues Jun 26 15:07:02 LetoThe2nd: ooh its a dune reference. havent' read that for yeeeaars Jun 26 15:07:58 rburton: http://dune.wikia.com/wiki/Voice Jun 26 15:12:28 Hello all, I have a question about building yocto using sstate_cache. It seems that qt is being built every time, even though it was already built before. It seems that all other packages are not rebuilt except this one Jun 26 15:12:39 Would anyone know what could be the root cause? Jun 26 16:21:58 rburton:^^ don't know if you noticed the details above:) Jun 26 16:42:21 zrrbite: same problems on ubuntu 16.04 :( Jun 26 17:46:30 Anyone know how I can get populate_sdk to include gdbus-codegen in my SDK? Jun 26 17:57:02 gerharddc: TOOLCHAIN_HOST_TASK_append = nativesdk- in Jun 26 17:58:37 Thanks, I'm running TOOLCHAIN_HOST_TASK += nativesdk-glib-2.0-codegen right now and will have to see if it works as when I previously tried something similar, it failed at the end. Jun 26 18:01:05 Thanks rburton, it worked brialliantly! Jun 26 18:07:12 rburton: I am trying to reproduce the PIE failures and failing to do so Jun 26 18:07:29 rburton: is there any special nuance I need to be aware of ? Jun 26 18:07:41 my distro config is pretty close to poky Jun 26 18:08:29 may be its time to merge glibc 2.26 pre-release patches that I am carrying in my tree. The freeze is due this friday anyway Jun 26 18:19:13 khem_: i'll replicate the failure and get recent logs Jun 26 19:00:24 rburton: in fact, looking closer at the g++ invocation, the --sysroot (and -fdebug-prefix) passed to it refer to the old builddir Jun 26 19:01:50 came there from LLVM_CXXFLAGS Jun 26 19:02:22 yann: sounds like a bug in meta-clang or whatever then Jun 26 19:05:08 meta-oe or meta-amd, then Jun 26 19:08:56 ah, they stopped including the oe base recipe, so most probably meta-amd alone, in fact Jun 26 19:17:46 rburton: which mechanism is usually used to fix such things ? Looks like with no override flags on llvm-config cmdline, we would need to sed its output in the wrapper script ? Jun 26 19:35:43 oh, they fixed it already in a later version Jun 26 20:16:31 hello! i'm trying to read environment variables from bitbake/lib/bb/cooker.py but can't? like running `PAGER=cat bitbake -e` with lines like `import os; print('PAGER' in os.environ)` still returns false Jun 26 20:16:58 (pager? re: bug 9203) Jun 26 20:16:59 Bug https://bugzilla.yoctoproject.org/show_bug.cgi?id=9203 enhancement, Low, Future, garbados, NEW , Automatically paginate bitbake -e output Jun 26 20:19:20 garbados: yes, we wipe almost everything from the environment Jun 26 20:19:43 grep for BB_ORIGENV Jun 26 20:21:50 rburton, thanks for the tip. i'm looking at bitbake/lib/bb/fetch2/__init__.py now and not really understanding how i'm supposed to pass environment variables to the bitbake -e command Jun 26 20:22:14 exactly as you did, PAGER=cat bitbake -e Jun 26 20:22:22 that sets PAGER to cat for that call Jun 26 20:22:39 bitbake then needs to go via BB_ORIGENV as os.environ() is almost empty Jun 26 20:22:50 (it is intentionally stripped of most things) Jun 26 20:24:15 rburton, can you describe that more for me? i don't understand how to "go via BB_ORIGENV" Jun 26 20:24:25 meta/classes/testimage.bbclass: display = d.getVar("BB_ORIGENV").getVar("DISPLAY") Jun 26 20:27:14 what is "d"? i'm looking at lib/bb/fetch2/__init__.py and seeing calls like `d.getvar` but i'm not seeing where d gets defined or imported Jun 26 20:29:18 its global in almost every context in bitbake Jun 26 20:29:35 garbados: It's an instance of data_smart from lib/bb/ Jun 26 20:30:26 NameError: name 'd' is not defined Jun 26 20:30:30 :| Jun 26 20:31:05 garbados: It's available in bitbake-flavoured python, but if you want to use it in self-written python functions, you'll have to pass it Jun 26 20:31:32 i'm modifying lib/bb/cooker.py Jun 26 20:31:40 bitbake/lib/bb/cooker.py Jun 26 20:32:49 Isn't it self.data in there? Jun 26 20:33:42 there is it Jun 26 20:33:46 neverpanic, thanks Jun 26 20:34:00 Hi there, I am facing a problem with a foreign sdk installer, built with yocto 2.2. I wanted to open a socket, but includes which wants to include , but which is not available for cortex9hf, but why? Jun 26 21:17:01 Hello! I just built a firmware from the pyro branch and I noticed these errors after the boot: https://pastebin.com/5hDa5kf9. I understand it means the key is missing, but isn’t the build procedure supposed to create those? **** ENDING LOGGING AT Tue Jun 27 03:00:07 2017