**** BEGIN LOGGING AT Tue Aug 18 02:59:58 2015 Aug 18 06:03:37 Hi all, Aug 18 06:04:41 I'm deploying a customized usb gadget Aug 18 06:05:31 by 2 ways Aug 18 06:20:07 Hi all, I'm deploying a customized usb gadget as following ways: 1. create a new recipes and specify MACHINE_ESSENTIAL_EXTRA_RDEPEND += "mygadget". I expect it is loaded with kernel but the .KO is included only to /lib/modules/3.14.28-1.0.0_ga+g91cf351/extra. I tried with KERNEL_MODULE_AUTOLOAD += "mygadget" but it does not help. 2. Kernel source code is extracted to /build/tmp/work-shared//kernel-source. I intende Aug 18 06:20:39 adget/ but whenever I tried 'bitbake linux-imx -c menuconfig', this source is removed completely. Can you please show me what wrong I have made and what direction should I follow to add a new driver to kernel? Thank you!!! Aug 18 08:26:07 hi guys Aug 18 08:38:02 morning all Aug 18 08:54:41 someone is very agressive about reassigning YP bugs Aug 18 08:55:03 but then doesn't add any comment about why they've reassigned the bug to me, nor change the status Aug 18 09:06:38 joshuagl: is that Stephen Jolley by any chance ? Aug 18 09:07:48 bluelightning: indeed it is Aug 18 09:08:00 reassigned to you as they're fido bugs? Aug 18 09:08:17 no Aug 18 09:08:27 I *think* reassigned to me as the previous commenter set to needinfo Aug 18 09:08:33 but really, it's not info from me they need Aug 18 09:08:38 at least as far as the comments explain Aug 18 09:10:23 joshuagl: right, that would probably be why - if it's needinfo from someone else, by all means reassign to that someone else Aug 18 09:10:50 joshuagl: mostly Stephen makes those changes during bug triage so in theory it's not him making the decision but the triage team Aug 18 09:10:57 FWIW Aug 18 09:11:00 bluelightning: yeah, will do - just wanted to vent a little Aug 18 09:11:08 but sometimes a comment would help I agree Aug 18 09:11:11 well the whole triage team did a bad job in this case then :-p Aug 18 09:12:52 beatings all around :) Aug 18 09:15:07 the beatings will continue until morale improves eh? ;) Aug 18 09:17:08 * joshuagl fetches his bat Aug 18 09:17:59 reading all 1000+ oe-core e-mails from 10 days vacation, done! :) Aug 18 09:18:21 sorry for some out-of-context replies in older threads.. :) Aug 18 09:18:39 going for meta-oe patch queue now Aug 18 09:18:58 JaMa: hi, welcome back - hope you had a good holiday :) Aug 18 09:20:10 yes it was great.. Norway is nice country :) Aug 18 09:20:40 I've not been there, but I had heard the same yes Aug 18 09:31:13 bluelightning, JaMa : hello Aug 18 09:31:24 I also have some queue to send, will do in a few days Aug 18 09:32:17 JaMa: do you see any build failures related to meta-hh / meta-initramfs? Aug 18 09:33:18 I don't but I've only built minimal images Aug 18 09:35:37 ant_work: I don't remember any, but world status doesn't look good for last few months, so who knows what could be hiding there Aug 18 09:36:06 ant_work: and meta-handheld isn't part of my world builds Aug 18 09:36:47 I see Aug 18 09:38:25 I'm wondering about pushing a new linux-yocto bbappend or updating linux-handheld Aug 18 09:38:39 I think I'll do the latter Aug 18 09:39:18 as long as we carry so many patches (collie) Aug 18 10:45:45 Crofton|work: ping? just looking at the OEDEM agenda, I'd like to propose the systemd discussion include whether to stick to the -stable branches or not Aug 18 12:32:32 hi guys Aug 18 12:32:50 i am having trouble with the kernel meta branch changes Aug 18 12:33:23 first it failed to fetch the hash Aug 18 12:33:55 i surveyed and found out that the meta branch was moved... meta --> yocto-kernel-cache Aug 18 12:34:14 digged a little and updated the hash which was causing trouble Aug 18 12:34:33 but now it fails on do_kernel_metadata Aug 18 12:34:40 the error I get is Aug 18 12:34:57 WARNING: addon feature "features/netfilter/netfilter" was not found Aug 18 12:34:57 | WARNING: addon feature "cfg/smp" was not found Aug 18 12:34:57 | WARNING: addon feature "cfg/sound" was not found Aug 18 12:34:57 | ERROR: required features were not found. aborting Aug 18 12:34:57 | ERROR: Could not update standard/common-pc-64/base Aug 18 12:35:14 zeddii: ^ Aug 18 12:36:04 but I confirmed that all these feature/config fragments were available under the kernel-meta tree source that was unpacked Aug 18 12:36:20 can anyone shed some light what i might still be missing Aug 18 12:46:45 bluelightning: sorry for a direct ping but can you please comment? Aug 18 12:47:53 I've no idea, I was pinging someone who should know Aug 18 12:48:19 I'm a lot less knowledgeable about our kernel build process than I ought to be I'm afraid Aug 18 12:50:53 bluelightning: thanks anyways :) are there any documentation updates? Is such a failure expected with the updates? Aug 18 12:53:06 abelal: I doubt it is expected Aug 18 12:53:33 abelal: unfortunately documentation for changes like this tends to lag behind... it's one of the challenges of keeping up with master unfortunately Aug 18 12:54:21 bluelightning: ah, yeah that is fun as well at times Aug 18 15:28:43 probably I'm missing some documentation which points this out, but is there an easy way to generate a dot file that shows the dependencies for software which is built for the target? I don't care what gets built on the machine doing the build, I only want to see a graphical representation of the target dependencies. Or is there an easy way to manipulate the output of "bitbake -g" to strip off the host things? (removing the "-cross" and "-nativ Aug 18 15:28:43 e" items out of pn-depends.dot still seems to show me host-side things) Aug 18 15:29:58 pn is build time by definition. the PN variable is the name of the recipe Aug 18 15:31:52 kergoth, am I looking at the wrong dot file? Aug 18 15:34:25 yes. Aug 18 15:34:34 recipes emit binary packages. binary packages are installed in images. Aug 18 15:34:36 see package-depends.dot Aug 18 15:35:06 bradfa: for the full picture within an image, enable buildhistory and see the graphs it produces for the image (after building the image) Aug 18 15:35:37 good idea Aug 18 15:36:23 kergoth, ok, but I see a ton of packages in package-depends.dot which are being built for the host only, such as python when I 'bitbake core-image-minimal' so I felt that was also not really a good place to look Aug 18 15:36:27 bluelightning, ok, I'll take a look Aug 18 15:39:26 bradfa: package-depends.dot will list all the packages built for that target, yes, not just packages installed in the image, for the latter see buildhistory as bluelightning points out Aug 18 15:40:39 kergoth, bluelightning, ok, generating build history now, I think this will solve my problem, thanks! :) Aug 18 16:01:12 should i be seeing this in armv5 in a softfloat build? Aug 18 16:01:32 Tag_ABI_HardFP_use: SP and DP Aug 18 16:01:46 that doesn't look right... Aug 18 16:02:16 build config output looks correct Aug 18 16:02:19 TUNE_FEATURES = "arm armv5 thumb dsp arm926ejs" Aug 18 16:02:20 TARGET_FPU = "soft Aug 18 16:02:51 however, as soon as the kernel mounts root i get a panic Aug 18 16:03:12 one that looks just like a float ABI mismatch Aug 18 16:03:40 otavio: got a clue? Aug 18 16:08:08 there is a way to include selected package from coreutils leaving there rest out? Aug 18 16:24:18 kergoth: I am looking for ways to enable a knob into meta-clang Aug 18 16:24:27 where I can say disable clang for this recipe Aug 18 16:24:35 at global metadata level Aug 18 16:24:41 current logic is inverged Aug 18 16:24:44 inverted Aug 18 16:24:55 where you select a package to be compiled with clang Aug 18 16:25:04 I would like to make it otherway Aug 18 16:25:29 see https://github.com/kraj/meta-clang/blob/master/conf/clang.conf Aug 18 16:28:34 there's a side-effect somewhere in armv5 tuning Aug 18 16:29:27 it actually built some weird but incompatible FP thing Aug 18 16:39:47 bluelightning, the buildhistory does exactly what I want, thanks! :) Aug 18 16:44:11 nerdboy: armv5 doesnt have FP in general Aug 18 16:44:19 i know Aug 18 16:44:24 unless you have some of those special phillips parts Aug 18 16:44:28 it should be softfp Aug 18 16:44:52 yes Aug 18 16:44:59 which is why it's incorrect Aug 18 16:45:07 whats going on ? Aug 18 16:45:27 i tried building for imx28 and it blows up Aug 18 16:46:17 TARGET_FPU = "soft" <= it lies Aug 18 16:49:46 i'm rebuilding default config to see if TUNE_FEATURES_append = " arm926ejs" is what triggered it Aug 18 16:50:03 so the most interesting result so far from my profiling and performance tweaking on pseudo is the discovery that roughly 80% of the cost of pseudo is outside the actual operations and/or IPC. Aug 18 16:50:05 whoops. Aug 18 16:50:15 should still be fine afaik Aug 18 16:51:22 where/how can i see exactly what flags are being used? Aug 18 18:39:17 RP: if i have a clonedir which needs_update, and the update fails (can't contact upstream), will the fetcher try to fetch a new mirror tarball via MIRRORS, or are mirror tarballs only fetched when clonedir doesn't exist using PREMIRRORS? Aug 18 18:49:39 as in while it's building... Aug 18 19:14:00 can yocto build working compilers to run on the target? Aug 18 19:16:59 nerdboy: was there no existing tune that configures the way you want? Generally you want to set DEFAULTTUNE to an appropriate tune, never manually manipulate TUNE_FEATURES. Aug 18 19:17:11 barfod: yes Aug 18 19:19:07 hmm, the buildroot project does not seem interested in supporting that Aug 18 19:19:14 seems a bit odd Aug 18 19:19:28 so i figured you guys did not either Aug 18 19:19:31 thanks kergoth Aug 18 19:20:04 packagegroup-core-buildessential is a package that pulls in the core on-target development packages. gcc, make, libtool, pkgconfig, binutils, autoconf, automake, gettext, libstdc++ Aug 18 19:20:15 there are other similar bits, and i think an image or two Aug 18 19:20:19 np Aug 18 19:21:18 there's self-hosted stuff too, to do bitbake/oe/yocto builds on target where appropriate. those are used to build our old build appliance virtual machine Aug 18 19:22:01 well that is prtty awesome. i am writing a toolchain myself right now. but i will certainly be looking mroe at yocto in the future if i need to whip out some targets Aug 18 19:22:29 i was quite frustrated by needing to get my compilers statically linked from special sources to work on buildroot Aug 18 19:22:47 thanks again. bbl Aug 18 19:24:09 np Aug 18 19:25:13 it yelled at me for setting default_tune Aug 18 20:28:23 i would like to make a list of default packages in an image recipe in a way that it's easy to disable them Aug 18 20:29:40 so that by default it would include, say, gcc, gdb and python-modules but these could be turned off one by one from local.conf Aug 18 20:32:31 look at how IMAGE_FEATURES and EXTRA_IMAGE_FEATURES are implemented. Aug 18 20:35:13 or just add them to IMAGE_INSTALL and instruct the user to use IMAGE_FEATURES_remove or IMAGE_FEATURES_remove_pn-yourimage from local.conf. or add a variable defined with ?= that's referenced in IMAGE_INSTALL Aug 18 20:35:16 lots of ways to go Aug 18 21:30:05 Huh, thought this was a bug in the shallow support, but looks like it's just a bug in the fetcher in general Aug 18 21:39:36 ensure that there's an http git mirror in PREMIRRORS. rm -rf tmp downloads; echo 'BB_GENERATE_MIRROR_TARBALLS="1"' >conf/auto.conf; bitbake testrecipe:do_fetch; rm -rf downloads/git2 tmp; bitbake testrecipe:do_fetch -> fails saying no checksum is defined in the recipe for the mirror tarball. Aug 18 21:41:08 I expect the mirrors shouldn't check checksums if the main fetcher doesn't support them, as the user will never define them in that case Aug 18 21:55:51 * kergoth preps fix for that for upstream submission Aug 18 22:00:31 haha. if you use an absolute file uri for a git repository, since it replaces all /'s with .'s, including the leading one in the absolute path, it ends up hidden in ${DL_DIR}/git2/ :) Aug 18 22:49:32 erk, i just noticed another questionable behavior Aug 18 22:49:45 let's say you have a giant git clone in your ${DL_DIR}/git2/, e.g. a kernel Aug 18 22:49:52 now lets say it's out of date, and upstream is temporarily down Aug 18 22:50:21 the attempt to update the repo will fail, and it'll *remove* the entire existing dir in ${DL_DIR}/git2/ entirely Aug 18 22:50:35 so you're stuck re-cloning if no mirror tarball ends up fetchable Aug 18 22:51:32 I can see why it was done that way, to avoid incomplete fetches, and in the case of git, to allow the unpacked mirror tarball to not step on an existing repo, but in this case it wasn't an incomplete original fetch, it's a failed *update*, we don't need to wipe it out unless we need to, in that case Aug 18 22:51:46 * kergoth checks when m.clean() is run Aug 18 22:53:33 hmmmm Aug 18 23:28:06 * kergoth argues with bb.fetch Aug 18 23:38:30 * paulg_ wonders if anyone has fought with the fetcher and felt like they came out ahead. :-/ Aug 18 23:41:49 I doubt it, I've found 4-5 potential bugs in the fetch code today Aug 18 23:48:41 there we go, one wasn't a bug, and i have local fixes for 2, so only 2 left to deal with, and those 2 are both pretty minor Aug 18 23:48:47 * kergoth digs further Aug 19 00:05:56 * paulg_ ties a rope to a tree and throws the other end to kergoth just in case he can't find his way out. Aug 19 01:30:34 * kergoth ponders Aug 19 02:54:47 normally he comes out on the other side of the earth Aug 19 02:54:55 with bus fare even **** ENDING LOGGING AT Wed Aug 19 02:59:58 2015