**** BEGIN LOGGING AT Tue May 24 02:59:58 2016 May 24 03:17:53 join #poky May 24 05:38:36 hi is it possible to package extra binaries to the rootfs via bbappend May 24 05:41:10 I tried using do_deploy_append but its not working May 24 06:57:28 I have a recipe with multipe bbappends, is there a way to block other bbappends after a certain bbappend? May 24 07:00:43 RagBal: it may be time to rethink the design at that point ... May 24 07:01:00 maybe make a copy of the recipe and write it the way you want? May 24 07:01:15 jku, agree, but the bbappend is in a layer that I do not maintain and still want to use May 24 07:02:56 Not a favourable choice but if there is no other way, that would be it May 24 07:25:14 hi May 24 07:25:50 it is possible somehow override INITSCRIPT_PARAMS in bbappend (I'm trying to use modutils but in recipe it is S05 and I want S02) May 24 07:34:04 Hey guys, I've just been following the free-electrons course on yocto for the beaglebone. I quite like yocto so I wanted to introduce it at the company I work. I'm trying to get it to work for a da-850-env board of TI. Is their a way to figure out if a "yocto board config" exists for this board? May 24 07:41:14 flitjes: you can check meta-ti layer where is support for many TI boards May 24 08:00:10 marek_, yep! May 24 08:00:32 present, I did it for psplash. It works like a charm. :) May 24 08:02:02 marek_, but you should make some copy of your image before. If you mess in the initscript you won't be able to access your board anymore. May 24 08:02:12 :D Kind of annoying. May 24 08:09:28 present: how you do it? May 24 08:10:02 INITSCRIPT_PARAMS = something new or something else? May 24 08:23:52 nothing more May 24 08:47:28 hi all, I tried adding "DISTRO_FEATURES_remove = "wifi" in my config file, but for some reason, this prevents the kernel from loading May 24 08:47:31 any clues? May 24 08:57:27 bitbake -e | grep '^DISTRO_FEATURES=' May 24 08:57:41 double check your variable first May 24 08:58:53 bitbake -e "your image target" | grep '^DISTRO_FEATURES=' May 24 08:59:03 thanks, let me check May 24 09:04:49 i can see that the variable does not contain wifi May 24 09:05:00 but i noticed, when build, i got the following messages: May 24 09:05:42 https://bpaste.net/show/6e6c762df893 May 24 09:06:19 first time i see these messages, so they are related to removing wifi May 24 09:17:38 I don't play much with those. May 24 09:18:08 greg__, first I think that wifi is not *only* present in DISTRO_FEATURES. May 24 09:18:25 Check also MACHINE_FEATURES... and the others. May 24 09:19:17 Then I think that if some lines should be removed from the kernel I would expect only a "XXX=n" instead of "XXX=y". May 24 09:19:23 And not a complete wipe. May 24 09:19:37 Which is what the messages seem to tell you. May 24 09:20:03 i remove it from MACHINE_FEATURES as well May 24 09:20:58 May be it's safer to only remove it from the DISTRO. May 24 09:21:45 Even I would like also to have a lighter kernel by removing the MACHINE_FEATURES if possible. May 24 09:21:55 i tried that, and got the same issue May 24 09:21:57 Only guessing anyway! :D Ask a pro! :P May 24 09:22:10 thanks anyway :) May 24 09:22:33 Also while you are playing features you should wipeout your whole build. May 24 09:23:34 Keep your build for small patches, recipes,... May 24 09:23:52 thanks, maybe i should do that May 24 09:23:54 For big stuff like features. I would say it's safer to wipeout the build. May 24 09:45:16 Someone can gives details on that: May 24 09:45:21 "Taskhash mismatch 598247d179dcd51262be69e8eee8d700 verses d70156e90e47803e04d4fdf2bfce6119..." May 24 09:45:51 I have to run twice the bitbake of my image. May 24 09:46:16 And I never have the ERROR message by relaunching the exact same bitbake command. May 24 10:20:15 i succeeded in removing wifi, but "cfg80211: Calling CRDA to update world regulatory domain" is still appearing in the shell on bootup May 24 10:20:39 anyway to remove this from the build? May 24 10:34:36 greg__, what you have in IMAGE_INSTALL variable May 24 10:35:13 greg__: changing machine features doesn't change the kernel build, so if you're still building those kernel bits then they'll be loaded if the hardware is present May 24 10:38:38 rburton, I was thinking on some wireless-tools stuff remaining. :) Wrong guess May 24 10:38:41 :) May 24 10:40:05 yeah also entirely possible that some wifi tooling is still going into the images if the images are asking for it May 24 10:40:17 maybe wpasupplicant etc should bail if wifi is disabled May 24 10:41:13 rburton, is rleitner on #yocto sometimes? I would like someone to review the --fbdev option. :D May 24 10:41:23 for psplash May 24 10:50:05 @rburton: IMAGE_INSTALL=" kernel-image kernel-devicetree u-boot-fw-utils" May 24 11:14:35 so I have a dependency on a package, that, in turn, is apparently split into foo, foo-dev, and foo-staticdev RPMs. May 24 11:15:07 right now, I only have "DEPENDS += 'foo'" in my recipe for bar (the package I want to build) May 24 11:16:13 and bitbake does not seem to care to provide (install into sysroot?) the bar*rpm packages prior to attempting to build bar :-/ May 24 11:16:27 thus compiling bar always fails May 24 11:17:41 LeifSo: You mean bitbake is not trying to build foo, when you try to build bar? May 24 11:19:16 mortderire: it does. But it doesn't do so prior May 24 11:20:18 also I was wondering whether I need to specify that foo provides {,-dev,-staticdev} packages and explicitly depend on foo-dev within bar's recipe May 24 11:20:53 LeifSo: you shouldn't need to no. May 24 11:21:47 LeifSo: ah .... wait I have come across this before ... May 24 11:22:47 completly unrelated: *Why* does toaster provide files ending in numbers instead of e.g. NUMBER.log? E.g. Firefox offers to download them, then open. Is there May 24 11:23:09 any workaround? It's pretty annoying May 24 11:23:16 LeifSo: minute May 24 11:23:45 LeifSo: look at the headers, maybe the log download isnt getting a sensible mime type? May 24 11:24:30 LeifSo: You can enforce dependence between reciepes like ... do_backport[depends] = "unifdef-native:do_populate_sysroot" May 24 11:25:07 LeifSo: Above I am saying don't run local task do_backport, before unidefnative:do_populate_sysroot has run. May 24 11:25:31 LeifSo: but rburton is right, you shouldn't need to do this. May 24 11:26:46 LeifSo: the Toaster issue, is this with task logs? or the main build logs? May 24 11:27:12 LeifSo: there is no workaround, but if you open an issue in bugzilla.yoctorproject.org we will fix it May 24 11:27:26 rburton: yeah, if toaster would provide sensible headers, I could assign an application. May 24 11:27:49 belen: task logs May 24 11:28:18 well, text/plain is perfectly reasonable May 24 11:33:26 adding wireless-tools to PACKAGE_EXCLUDE, and im still getting the cfg80211 error message printed to screen May 24 11:39:03 belen: https://bugzilla.yoctoproject.org/show_bug.cgi?id=9670 May 24 11:39:05 Bug 9670: normal, Undecided, ---, toaster, NEW , Toaster does not set a correct MIME type for tasks' logs files May 24 11:39:49 LeifSo: excellent. Thanks! May 24 11:40:24 belen: well I'm thanking you for taking care of this :) May 24 11:42:17 LeifSo: np. We'll assign it to someone tomorrow, and hopefully will be fixed soonish May 24 11:43:32 rburton: yes, text/plain would be fine. Firefox says it identifies it as BIN though :-/ May 24 11:50:09 mortderire: thanks for your help. That's exactly what DEPENDS is supposed to do. Run do_popluate_sysroot before the dependend's package's do_configure May 24 11:57:37 I'm trying to build busybox, which fails since the cflags do not include the ${sysroot}/usr/include Is that on purpose? May 24 11:58:47 is there a simple way to exclude a kernel module? May 24 11:59:25 greg__: well you could opt to not build it May 24 11:59:55 would something like: KERNEL_MODULE_AUTOLOAD_remove = "cfg80211" work? May 24 12:00:22 simonl: how can i can i prevent it from being built? May 24 12:00:27 greg__: Then it would still be installed, but not added to an explicit autoload file May 24 12:00:57 it might still be autoloaded by "automatic autoload" May 24 12:01:13 greg__: That's in the kernel config file May 24 12:02:10 Or if you need it to be built and available in a feed but not installed along with the other modules, something like RDEPENDS_kernel-modules_remove = "mymodule" May 24 12:02:18 greg__: ^ May 24 12:03:04 That will remove it from the package that pulls in all the built modules except if there is some other dependency on the specific package May 24 12:04:13 * simonl switches over to speak about his own problem May 24 12:05:16 I have a cmake recipe that fails to find git when the recipe runs. I can run cmake from a devshell, and then the problem isn't there. Any suggestions for debugging that? May 24 12:05:24 thanks simonl :) May 24 12:06:52 greg__: yw May 24 12:14:43 simonl, I would say you did not provide enough information in the SRC_URI. Like protocol= May 24 12:15:35 You have commands inside CMakeLists.txt that pull source code?! May 24 12:15:37 present: it happens later. There's a thing in the CMakeLists.txt files that tries to find git during configure and fails May 24 12:15:47 present: yeah... :/ May 24 12:16:18 Then I'll say you will need to patch the CMakelists.txt. May 24 12:17:10 present: it should only need to find the executable, for the configuration that needs to run nothing should actually be checked out. May 24 12:17:17 You should go directly to the ../tmp/... recipe directory to see what it's trying to do in the the build. May 24 12:17:19 but the strange thing is it used to work with svn May 24 12:17:52 use bitbake -c May 24 12:18:06 to run each steps one by one May 24 12:18:23 unpack configure compile May 24 12:18:50 You will have one log file for each step. You will find exactly what you need there. May 24 12:21:11 You might use hard executable path inside the CMakeLists.txt. May 24 12:21:29 present: I'll have a look, thanks! May 24 12:45:26 RP: High. the change to LC_ALL="en_US.UTF-8" has a funny side-effect. My u-boot started to fail with split-strip QA issues. The detected binary file in /boot behaves now differently :) May 24 12:45:41 s/High/Hi/ ... :} May 24 12:47:19 zecke: Hi. We did use LC_ALL=C previously so that is somewhat surprising... May 24 12:47:39 zecke: do you have en_US.UTF-8 installed? May 24 12:47:45 (I need to finish that sanity check) May 24 12:49:36 rburton: how to check? cat /etc/locale.gen | grep "^en_US"? LC_ALL=.. lcoale? May 24 12:52:32 "LC_ALL=en_US.UTF-8 ls" and see what happens May 24 12:52:36 rburton: RP: http://paste.lisp.org/display/316708. May 24 12:53:10 haha May 24 12:53:14 yes. :( May 24 12:53:27 funny in a oh fffffsss sort of way May 24 12:53:43 time to forcibly set LC_ALL=C when calling time? May 24 12:53:45 erm, file? May 24 12:55:14 rburton: calling time is very much an en_GB thing... May 24 12:56:01 rburton: or handling the output differently? I mean it is certainly no ELF file.. 16 -kernel module doesn't seem to be set anymore? May 24 12:56:27 s/output/error/... *sigh* May 24 12:57:04 Hi. I'm writing my own buildbot steps to build yocto starting from 'repo sync'. I'm failing during final bitbake step --> bitbake core-image-minimal. Log shows this --> OE-core's config sanity checker detected a potential misconfiguration.Please use a umask which allows a+rx and u+rwx May 24 12:57:16 rburton: The function is called isELF and it creates an unskippable QA error with split-strip. I don't know the history of it but it does sound a bit odd for an outsider May 24 12:57:35 But when I do 'bitbake core-image-minimal' manually, it works May 24 12:57:35 is there a way to dump the final recipe as compiled (i.e. appended) by bitbake? May 24 12:57:53 zecke: i'm guessing the problem is that file is aborting and so returning an error code so isELF aborts May 24 12:58:27 kteza1: that would be because buildbot is running in a different environment to your shell. set the umask somewhere, either in the buildbot init script or the step that invokes bitbake. May 24 12:58:52 LeifSo: bitbake -e is close to what you want i guess May 24 12:59:40 zecke: nice of file to abort whilst assuming that a set of bytes is utf8 May 24 13:00:08 do I need to explicitly add CFLAGS for the sysroot or is that supposed to happen automagically? May 24 13:00:41 zecke: there does appear to be a bug in file here which is then causing other things to break :( May 24 13:01:12 LeifSo: the sysroot flags are in the CFLAGS CC etc variables, so if they're being passed to the build by the recipe then it should work May 24 13:01:17 rburton: My breakpoint in abort/vasprintf is not hit. It seems to normally exit. But if we want to determine if it is an elf file and file doesn't say it is, isn't that good enough to return 0? May 24 13:01:17 As rburton says, we may have to force C locale for file :/ May 24 13:01:28 at least its isolated May 24 13:02:10 In my company build I just nuke /boot/* of u-boot and I am on Debian7 (which is officially not supported anymore, yeah these users and their secrets) May 24 13:02:47 So I have my "workaround", if you want to modify/fix then I am happy to test, otherwise feel free to move on. :) May 24 13:02:50 * zecke gets some food May 24 13:02:51 zecke: is there a copy of the failing binary somewhere we could experiment with? May 24 13:02:59 JaMa, a patch for EGL has been proposed directly to qtbase. https://codereview.qt-project.org/#/c/160149 May 24 13:03:12 zecke: I'm curious how widespread the problem is May 24 13:03:29 RP: please msg me your preferred mail address for the binary May 24 13:03:40 present: I've seen it, thanks May 24 13:04:40 JaMa, this kind of line "QMakeVar set QMAKE_LIBS_EGL "$QMAKE_LIBS_EGL" does not seem to work. May 24 13:05:04 rburton: what's the sysroot variable? May 24 13:05:05 Or at least it does not set the value inside the pro file. May 24 13:05:27 JaMa, anychance it could be related to Yocto? May 24 13:05:47 ${STAGING_DIR_TARGET} ? May 24 13:05:58 LeifSo: usually passed via ${CC} (see bitbake.conf) May 24 13:08:02 rburton: apparently busybox unsets all of that :-/ May 24 13:09:41 present: you mean that the value isn't expanded before being set in .pro? May 24 13:10:38 rburton: or rather unsets all, but the cross compile env May 24 13:10:45 present: see "Add linux-oe-g++ platform" commit about the Eval functions if it helps in any way May 24 13:13:02 JaMa, that's it. The work done by QMakeVar is not used in the *.pro files. May 24 13:14:33 In this case we would not need to swap pkgconfig and mkspecs variables as they would have the same proper content. May 24 13:22:45 @rburton Thanks. That worked :). Create a slave with umask 022 May 24 13:22:53 Created* May 24 14:13:49 Hi everyone. May 24 14:13:52 I need to update the bootargs from my bootloader with the following command: "global linux.bootargs.video="video=HDMI-A-1:1280x1024-32" May 24 14:13:56 I would like to incorporate this to my yocto configuration. May 24 14:14:01 I can't see a way to create a recipe to do such a thing. May 24 14:14:04 Does anyone have an idea? May 24 14:36:24 I tryed to add a recipe for my barebox May 24 14:36:35 but no luck May 24 19:38:36 hi guys -- I know I can overide a variable in a package from local.conf using VAR_pn-PN and especially use that to inject an _append. Is there a variation on this to append to a function in a bbclass. Specifically a image_types bbclass? May 24 19:42:46 overrides work fine with functions. from bitbake's perspective the only difference between a function and a non-function variable is a 'func' flag is set May 24 19:42:59 of course, you can't define a function in a .conf, so you'd have to do i tin a bbclass, bbappend, or .inc May 24 19:44:03 interesting May 24 19:46:20 so i'll need to make a 'new' bbclass which inherits from the image_types one I want to modify and point the local.conf to that new bbclass... May 24 19:48:03 that'd work, yeah. create a new class, add 'inherit image_types' to it, add that class to IMAGE_CLASSES May 24 19:48:25 i've done exactly that from time to time, actually May 24 19:51:03 I'm still thinking (perhaps in vain) to find a way to do it without deploying more files into my layers. But that sounds relatively painless. May 24 19:51:50 if I could *just* ( :) ) add one command to the end of the function in that bbclass I would be done... May 24 19:52:47 might be doable from local.conf using non-function syntax, but your options would be limited, i.e. for that purpose you can't really inject extra newlines due to the limitations of the FOO="bar" syntax May 24 19:53:11 hmmm . could I declare the _append function in an .inc and include it from local.conf? May 24 19:53:16 yep May 24 19:53:36 or add a new class and add it to INHERIT instead. but i don't really see how a .inc is any better than a class, it's still a new file May 24 19:53:48 good point May 24 19:53:55 TOPDIR is already in BBPATH, you can add a class in your build dir if you don't want it source controlled May 24 19:53:58 mkdir classes May 24 19:54:06 neat! May 24 19:54:28 so I could put it in build/classes/ here May 24 19:54:30 i do that from time to time while creating it, then move it to an appropriate layer when done May 24 19:54:32 yep May 24 19:55:01 i woudln't necessarily advise it, the more you put into your local setup the harder is it to reproduce and control May 24 19:55:05 but it's your call May 24 19:55:27 nice trick -- this is getting more manageable. Your point regarding RCS is 100% on. May 24 19:55:34 no disagreements May 24 20:00:03 tangential question: anyone know why meta-fsl-arm image_types bbclass has sync && sync after their dd commands? May 24 20:00:11 one sync I wouldn't question... May 24 20:02:35 according to http://pubs.opengroup.org/onlinepubs/007908799/xsh/sync.html, the sync() call can return before the data has actually been written.. May 24 20:02:36 https://unix.stackexchange.com/questions/5260/is-there-truth-to-the-philosophy-that-you-should-sync-sync-sync-sync#5261 May 24 20:02:39 * kergoth shrugs May 24 20:02:47 maybe check the git log for the class? May 24 20:03:50 :) wow ok. fun links May 24 20:05:23 years and years ago i developed the habit to sync;sync;sync; if i was paranoid and was about to hard poweroff rather than shut down safely, but that's a pretty rare case, and no idea where i got it from. does seem to be in the unix folklore May 24 20:06:01 :) in light of the material in those links May 24 20:11:32 kergoth: my professor at uni got me into the sync;sync;sync practise and he had some mental rationale about caching layers May 24 20:11:50 definitely folklore and i wouldn't be surprised if its in the jargon file May 24 20:17:24 +1 to the triple sync May 24 20:18:39 https://utcc.utoronto.ca/~cks/space/blog/unix/TheLegendOfSync :) May 24 20:21:29 Ah yes, I remember doing the triple sync (from the console) on the old PDP-11/70 running BSD Unix 2.x. On that system, it may have actually needed the delays.... **** ENDING LOGGING AT Wed May 25 02:59:58 2016