**** BEGIN LOGGING AT Mon Aug 08 02:59:59 2016 Aug 08 04:00:52 Hi, all. I have a question about using armv7 of QEMU on yocto. runqemu-internal have support armv7 of QEMU as qemuarmv7, However, runqemu is such that there is no mechanism to use this. How do I use this? Aug 08 04:01:08 Do I need to incorporate to create a layer with qemuarmv7 machine myself in my environment? Aug 08 07:53:33 mimon: just add qemuarmv7.conf as a new machine config Aug 08 07:55:52 mimon: something like this https://git.linaro.org/openembedded/meta-linaro.git/blob/HEAD:/meta-linaro/conf/machine/genericarmv7a.conf Aug 08 08:02:43 good morning Aug 08 08:03:30 Hi, I have some question regarding static libs and the SDK I provide a static lib as *.a file and want to use it in some other tool, but the lib just shows up in the SDK if I add it via TOOLCHAIN_TARGET_TASK_append in the recipe of the image is this the wanted behaviour? Aug 08 08:39:58 Hi all. Is there a brief documentation, what all the environment variables do that are in the SDK's environment-setup-<...> script? Aug 08 08:42:59 hi all, i need an old version of openssl (0.9.8), how can I add it to my poky korgoth build ? Aug 08 08:58:27 kbo: git log the relevant directory to get the old recipe ? Aug 08 12:49:41 beside a small patch I have to copy some files into the linux kernel before compilation: http://pastebin.com/dUu9CjUb <- is this way corret? Aug 08 12:56:04 Seems like yocto has some special behaviour with new dtb files... Aug 08 13:58:17 HyP3r: KERNEL_DEVICETREE_append = " vf610-colibri-racechip.dtb" Aug 08 13:58:57 and you need to make it build it, I believe some Makefile addition in the same dir Aug 08 14:28:13 jubr: ok, is there way just to recreate the devicetree files and the rootfs? If I run my Image-Creation Recipie, Yocto is compiling everytime a new kernel Aug 08 14:36:13 RP: ping Aug 08 14:39:18 Any recommendations for a 4G LTE-capable USB modem to be used in an embedded device running a Yocto-based Linux distro (kernel = 4.1.13)? I've been trying to get a Netgear 340u working, but it's proving to be... difficult. I'm wondering if there's something else available that Just Works™ on desktop Linux systems (the 340u does not), that I can then 'port' to our Yocto build. Aug 08 14:39:43 The coverage are is North America, preferred carrier T-Mobile (though the carrier doesn't matter much). Aug 08 14:40:22 s/are/area Aug 08 14:42:41 This links hints that the U620L might be a candidate: https://www.raspberrypi.org/forums/viewtopic.php?f=28&t=121326 Aug 08 14:48:11 And there's this: http://www.verizonwireless.com/dam/support/pdf/user_guide/u620-linux-integration-guide-7-17-15.pdf Aug 08 15:29:51 RP I believe I have a workaround for the nfs issues on ubuntu1604. It's now in place. We need to run a build to test. Aug 08 15:30:14 cool Aug 08 15:30:52 rburton, I was about to ping you next. :) Aug 08 15:37:28 rburton, Do you have something to give to the builders? Aug 08 15:37:48 yeah will do fairly soon, shall i just fire when its ready? Aug 08 15:39:29 How can I see why a package has been pulled into a rootfs? If I run bitbake -e lm-core-image-lb it does not list the culprit package, but the resulting rootfs evidently contains it Aug 08 15:39:44 rburton, Yes please. Aug 08 15:41:28 sveinse, try passing "-g -u depexp" to bitbake for the image recipe Aug 08 15:42:26 sveinse, it'll popup a gtk ui for inspecting package dependencies Aug 08 15:43:55 * sveinse notes this one down on his yocto 101's on his wall... Aug 08 15:44:54 needs to be rewritten in gir+gtk3 really, its currently pygtk2 Aug 08 15:50:28 I found my culprit. the automatic rdepends pulled in one of the other packages. -- yet, depexp don't show this... Aug 08 15:50:50 depexp can only show the infromation bitbake has, today. it doesn't examine build output, only metadata Aug 08 15:51:13 halstead: great, thanks. I will defer to rburton Aug 08 15:51:16 kergoth: pong Aug 08 15:52:15 RP: two things, one, is there anything I can help with on the bitbake end of things, any patches you'd like more eyes on or anything? I just got back from a 3 week vacation. two, status of shallow? :) Aug 08 15:54:14 e.g. I have a recipe named sp. It generates sp-mcbapp and sp-rcgui. The package lm-core-image-lb depends on sp-mcbapp. But sp-mcbapp is not in the list to the left. 'sp' is. And the culprit was that sp-mcbapp rdepends on sp-rcgui. And I cannot find this info in depexp Aug 08 15:55:13 kergoth: I believe all this information is available to bitbake at parse-time? Perhaps the implicit rdepends are not? Aug 08 15:55:30 again, bitbake only has the info that's in the recipes Aug 08 15:55:44 automatically generated rdepends based upon shared library dependencies are not in the recipe Aug 08 15:56:12 yeah, and it's this automatic rdepends which caused this fault :P Aug 08 15:57:03 correct, yes, i'm just telling you why -- depexp doesn't examine build output Aug 08 15:57:09 kergoth: I think I asked you about the patch which I was most concerned about. I know there is a bit of a backlog on the list at the moment but I should be able to clear that in one of the builds this week, just been distracted with things. I do have the multi-config stuff nearly ready for review and that will likely need some eyes for review Aug 08 15:57:10 tools like oe-pkgdata-util or the buildhistory have that info Aug 08 15:57:12 ...now I'm off on exploring why the package got the unexpected rdepends in the first place Aug 08 15:57:32 kergoth: with the shallow stuff, I just keep forgetting to go back and think about it :( Aug 08 15:57:40 RP: okay, will keep an eye out for it, thanks Aug 08 15:58:55 kergoth: cool, thanks. Learning something new every day. oe-pkgdata-util seems neat Aug 08 15:59:50 it's real handy, yeah. particularly being able to search packages by target path, i.e. tell me what includes /usr/bin/foo Aug 08 16:01:13 I've become a master of grep and find to do exactly this by hand :o Aug 08 16:01:35 with varying precision.... Aug 08 16:02:04 What task defines the rdepends of a package? Aug 08 16:02:29 the dynamic bits? do_package. see PACKAGEFUNCS in package.bbclass Aug 08 16:07:05 kergoth: btw, are there any plans on integrating that logviewer you made (right?) into upstream poky/bb? Aug 08 16:07:42 not sure what you're referring to Aug 08 16:08:02 okay, maybe I mixed things up Aug 08 16:08:44 rburton: https://patchwork.openembedded.org/patch/128261/ is not staged in ross/mut any issues ? Aug 08 16:09:20 rburton: meta-raspberrypi is held on this patch for lot of patches especially kernel upgrades and VC4 support patches Aug 08 16:09:35 sorry, looking now Aug 08 16:09:40 thanks Aug 08 16:11:18 hm branches without cover letters are a pain to find the rest of the series Aug 08 16:11:24 got a link to 2/3 and 3/3? Aug 08 16:11:53 2/3 and 3/3 are merged Aug 08 16:12:52 cool Aug 08 16:12:55 well its in mut now then Aug 08 16:13:03 sorry Aug 08 16:13:08 kergoth: I was thinking of https://github.com/kergoth/bb Aug 08 16:13:32 rburton: thanks. Aug 08 16:14:19 rburton: I see there you also have 2.27 staged which has two issues as far as see. libunwind/gold and another issues I see is webkitgtk/mips Aug 08 16:14:31 yeah staged for giggles Aug 08 16:14:41 connman too Aug 08 16:24:21 Is there a way to setup a check which fails if my two packages depend on each other? In my case, if they do, due to shlib auto deps, this is an error in my package splitting Aug 08 16:26:55 Something like RDEPENDS_${PN}-pkg1_NEVER += "${PN}-pkg2" Aug 08 16:28:01 sveinse: you have foo.so and bar.so coming from same recipe and going into seprate output packages ? Aug 08 16:28:06 and depend on each other ? Aug 08 16:30:32 khem, slightly more complex: foo depend on foo.so going into package a, and bar depend on bar.so going into package b. If I did a mistake in package splitting and placed foo.so into b, a will automatically depend on b. I'd rather have the recipe fail Aug 08 16:33:01 RP: understandable, hoping to get it merged for this release, though, if possible. on the plus side, the impact when it's not enabled is nil, so there isn't much of a worst case, at least :) Aug 08 16:33:18 sveinse: this will be detected if a doesnt have RDEPENDS_{PN} on b Aug 08 16:33:45 rdep QA will catch it hopefully Aug 08 16:34:14 khem: Detected, yes. and automatically added. No stopping or major complaining afaics Aug 08 16:35:20 sveinse: you could always add a custom package sanity check for that recipe, via the mechanism used by insane.bbclass Aug 08 16:35:37 kergoth: thanks, I'll look at that Aug 08 20:00:45 folks, I need an advice to use readonly filesystem Aug 08 20:01:10 set read-only-rootfs image feature? Aug 08 20:01:13 (see the documentation) Aug 08 20:01:16 I'm using yocto 1.7 with initramfs, so I have 2 partitions, the readonly is squash and other one is rw Aug 08 20:02:16 rburton, the problem is to use a rw configuration file Aug 08 20:02:35 today I'm linking all the conf files to an rw partition Aug 08 20:04:55 the var IMAGE_FEATURES = "read-only-rootfs" is already configured Aug 08 20:07:48 the connman only use the /var/lib/connman Aug 08 20:10:13 the tzdata use the /etc , alsa I can pass the configuration file at the initialization Aug 08 20:11:09 there's some layer or distro to help to use readonly + rw configuration files? Aug 08 20:19:35 rburton, I found VOLATILE_BINDS , I will try to use Aug 08 23:21:24 rburton: do you know if there's any particular reason why useradd.bbclass does DEPENDS_append = "${USERADDDEPENDS}", USERADDDEPENDS = " base-files ...", USERADDDEPENDS_class-cross = "", USERADDDEPENDS_class-native = "", etc., instead of just DEPENDS_append_class-target = " base-files ..."? Aug 08 23:23:52 not off the top of my head Aug 08 23:24:09 ok... maybe it's some legacy thing Aug 08 23:24:46 the variable doesn't get altered by another class Aug 08 23:24:49 so yeah patch welcome :) Aug 08 23:25:49 i'll prepare one. bunch of other ones like that too in useradd.bbclass. Aug 08 23:28:11 it may be that back when that was implemented, _class-target didn't exist (or was very new) since there was a time when we used _virtclass- instead which didn't support target Aug 08 23:28:38 alright Aug 09 00:02:52 http://pastie.org/10933207 quick review? Aug 09 00:04:06 SSTATECLEANFUNCS could be appended rather than set as well. only useradd.bbclass seems to use it. **** ENDING LOGGING AT Tue Aug 09 02:59:58 2016