**** BEGIN LOGGING AT Thu Feb 04 02:59:59 2016 **** BEGIN LOGGING AT Thu Feb 04 04:10:39 2016 Feb 04 04:13:45 [about meta-raspberrypi]: hey,guys,trying to build raspberrypi image with poky. but it comes out with many errors as follows: Feb 04 04:14:18 ARNING: Failed to fetch URL http://tukaani.org/xz/xz-5.2.2.tar.gz, attempting MIRRORS if available Feb 04 04:14:21 WARNING: Failed to fetch URL http://downloads.sourceforge.net/expat/expat-2.1.0.tar.gz, attempting MIRRORS if available Feb 04 04:14:24 ERROR: Fetcher failure: Fetch command failed with exit code 4, output: Feb 04 04:14:27 wget: unable to resolve host address 'downloads.sourceforge.net' Feb 04 04:14:29 ERROR: Function failed: Fetcher failure for URL: 'http://downloads.sourceforge.net/expat/expat-2.1.0.tar.gz'. Unable to fetch URL from any source. Feb 04 04:14:32 ERROR: Logfile of failure stored in: /home/jackzhang1992/poky/build/tmp/work/cortexa7hf-neon-vfpv4-poky-linux-gnueabi/expat/2.1.0-r0/temp/log.do_fetch.27687 Feb 04 04:14:35 ERROR: Task 1419 (/home/jackzhang1992/poky/meta/recipes-core/expat/expat_2.1.0.bb, do_fetch) failed with exit code '1' Feb 04 04:14:38 NOTE: Tasks Summary: Attempted 327 tasks of which 0 didn't need to be rerun and 1 failed. Feb 04 04:14:51 anyone who knows sth about it? I have tried this for 3 days Feb 04 08:34:21 good morning Feb 04 08:39:50 mkcmoin Feb 04 08:40:20 can't even make a typo without a typo... Feb 04 08:51:24 nerdboy: :-) Feb 04 09:01:32 is it possible to generate a package.manifest that also has version numbers? Feb 04 09:36:11 so this is my latest try to get telnetd automatically started: http://pastebin.com/SwxP42JL Feb 04 09:36:18 but it still won't do that Feb 04 09:36:32 can anyone point out what I'm doing wrong? Feb 04 09:41:35 hi all Feb 04 09:41:45 I'm currently investigating the use of systemd for our embedded targets. So far so good, it runs fine but... systemd's size is huge (15801 KiB systemd) even with all PACKAGECONFIG disabled (systemd_225.bb) Feb 04 09:41:53 anyone managed to get something smaller? actually I'm mainly interested in the process management (e.g. restart processes whenever they crash, etc) and do not really care about the rest. Feb 04 09:54:56 lostcontrol: maybe use something other than systemd? Feb 04 09:56:46 yes, this is an alternative but I was wondering if it was possible to make it smaller. If not, then this will be definitely a no go and I will look at alternatives Feb 04 09:58:16 e.g. on http://bec-systems.com/site/1098/why-systemd-in-embedded-linux-systems I see that systemd use 1412 KiB which is a LOT less than the 15801 KiB I currently get Feb 04 09:59:32 maybe systemd ate a lot of chocolate and got a lot heavier since 2013 Feb 04 10:08:13 lostcontrol: systemd is just a big crap .. i dont think you can make it smaller Feb 04 10:10:11 lostcontrol: there's lots of binaries in /lib/systemd/ you may try to pick ones that you need Feb 04 10:14:30 hmm, and apparently not everything is controlled though PACKAGECONFIG, not sure why I would want systemd-machined by default Feb 04 10:30:51 @bboozzoo thanks, I will try to kick out some files and see if I still can get something working Feb 04 10:41:10 does a user need to exist in the Yocto environment that I use with chown in do_install_append? Feb 04 10:41:18 hi Feb 04 10:45:17 I guess the answer is an obvious yes Feb 04 10:45:39 iirc it needs to exist in the sysroot as you're running in pseudo context Feb 04 10:55:33 khem: http://errors.yoctoproject.org/Errors/Details/32437/ <— target cflags are leaking into native builds somehow in grub-efi. the compile failed because the host gcc doesn't know what fortify-strong is Feb 04 11:08:36 lostcontrol: took a quick look, it shoudl be possible to shave off a couple of MBs, I'll try to post some patches today or tomorrow Feb 04 11:12:12 thanks bboozzoo :-) I'm trying to add more --disable-xyz to the receipt because looking at the config.log shows that many things are still enabled Feb 04 11:13:36 Hi, how to disable yocto from packaging .so into -dev package? Adding the symlink file to FILE_{PN} doesn't seem to be enough. Feb 04 11:21:51 I've built an image with minimal gui components (xserver, openbox & chromium) . I'm writing systemd service for chromium to start it automatically during boot. I'm able to start it manually using 'systemctl start chromium' but not during bootup. Feb 04 11:22:09 How do I do it? Here is my service file --> https://gist.github.com/2b4c3bc927fc7e1e4f2e Feb 04 11:22:32 rtr_ "systemctl enable chromium"??? Feb 04 11:23:23 disclaimer: I'm new to Poky and systemd (~2-3 days experience) Feb 04 11:23:59 lostcontrol: by the way, it is recipe, not receipt :) Feb 04 11:24:37 rburton: ok, thanks Feb 04 11:24:39 I want it to automatically start during boot Feb 04 11:24:46 I'm new to English also :-D Feb 04 11:25:32 rburton: so USERADD_PACKAGES and USERADD_PARAM_ ought to be fine, I believe. Feb 04 11:26:03 mcfrisk: PN-dev is packages before PN so you'll need to either remove from files_pn-dev, or remove pn-dev from packages, depending on what is best Feb 04 11:27:19 bboozzoo: now 12013 KiB systemd, killed around 3MB and the system is booting :-) Feb 04 11:27:34 rtr_ SYSTEMD_SERVICE = "name_of_the_file"? Feb 04 11:27:57 (After inherit systemd) Feb 04 11:28:33 lostcontrol: when I looked at it about two years ago, it was possible to reduce it to a few megs Feb 04 11:30:05 @lpapp I already added this. SYSTEMD_SERVICE_${PN} = "${PN}.service". Other similar services seems to be working except chrome Feb 04 11:31:28 rtr_: in worst case, you can create the symlink yourself :p Feb 04 11:35:06 @lpapp :) Thing is I could see the chromium.service file in /etc/systemd. Could this be something related to login. Can it be started before login? Feb 04 11:36:50 rburton: the interesting thing is that the sandbox does not show the files with the right owner, but when I install the corresponding package with opkg, the owner is as expected. Feb 04 11:37:02 rburton: it is also possible that I am looking at the wrong directory in the build dir. Feb 04 11:37:12 rtr_: good question Feb 04 11:37:25 rtr_: why Type=fork? Feb 04 11:37:33 isn't there some log somewhere Feb 04 11:37:50 it should be Type=simple Feb 04 11:38:17 do youy want to create a kisok device? Feb 04 11:39:02 @fl0v0 Oh. Will try that. I could see other processes being spawned with chrome and assumed it should be of type 'fork'. Feb 04 11:39:16 Yeah I need kiosk mode Feb 04 11:39:31 i have done exactly what you want Feb 04 11:39:32 mom Feb 04 11:41:47 http://pastebin.com/bnnAZRGZ Feb 04 11:41:54 this is my service file Feb 04 11:42:44 rburton: we htried FILES_{PN}-dev = "" before adding the .so file link to FILES_{PN} but that isnt working for some reason. Feb 04 11:43:17 @fl0v0 I'll try your config now :) Feb 04 11:43:18 i dont know if you need touch, but i had to use the switches --touch-events=enabled --touch-devices=7 where 7 is the id of the touch device Feb 04 11:43:50 Yeah. I need touch coz I'm eventually going to use a touch screen Feb 04 11:44:07 and if you havent configured anything yet on your system and you are using debug-tweaks i think you need User=root too Feb 04 11:45:14 ah and you should not use --allow-running-insecure-content that was just for development Feb 04 11:55:14 rburton: if I package those loadable modules for apr-util in another package than -dev, I trigger this other QA issue: QA Issue: non -dev/-dbg/nativesdk- package contains symlink .so Feb 04 11:55:34 I should just make the recipe skip this check in order to pass the new one you introduced? Feb 04 11:56:24 @fl0v0 There is some progress. Your service is working :) but it's running before openbox. Feb 04 11:57:06 ah you have to change that back i thinl Feb 04 11:58:09 https://gist.github.com/4d7a31a54066e2981012 Feb 04 11:58:15 my display manager starts as display-manager.service Feb 04 11:58:34 After adding openbox first, I'm facing the error shown in the link Feb 04 11:59:42 hmmmm Feb 04 12:00:04 you already fixed the typo i guess? Feb 04 12:00:28 yeah you did Feb 04 12:03:07 well you have some kind of circular requirements but i do not really understand what excactly Feb 04 12:03:20 Yeah Feb 04 12:05:21 hm you should be able to remove the xserver requirements. openbox is a display manager that needs x by itself, right? Feb 04 12:08:00 Yeah. Just did that. Testing :0 Feb 04 12:11:17 @fl0v0 Woahhhhhh. It's working. Feb 04 12:11:23 :D Feb 04 12:11:48 There is a spelling mistake in openbox.service and unnecessary Requires=multi-user.target Feb 04 12:12:37 @fl0v0 Can't thank you enough. I was debugging this since yesterday Feb 04 12:12:53 you're welcome Feb 04 12:13:07 :) Feb 04 12:30:55 aratiu: is there a reason it ships symlinks? if so then sure, skip it Feb 04 12:38:51 Hi all, I'm looking at how to integrate Toaster into our workflow. Does anyone know if the latest version of Toaster has an analysis-only mode like it had in daisy? Feb 04 12:42:05 Toaster docs for Jethro still mention "analysis mode" and "build mode" but the instructions for setting up a production instance only talk about build mode Feb 04 12:48:29 paulbarker: we no longer have two modes. Toaster is able to pick up builds started from both the webui and the command line Feb 04 12:51:25 paulbarker: you can switch between the 2 ways of starting builds without restarting Toaster. Toaster will show the builds started from CLI inside a "special" project called "Command line builds" Feb 04 12:53:27 paulbarker: and those references to the build mode in the jethro docs should not be there. I'll remove them Feb 04 12:53:38 paulbarker: thanks for spotting that one Feb 04 13:13:03 paulbarker: gah, actually, scratch that. Jethro toaster does have 2 modes :/ The changes to remove the modes went in after the 2.0 release. Sorry Feb 04 13:16:59 belen: I think I missed your first message, IRC dropped and re-connected Feb 04 13:17:44 I basically want to disable the ability to start a build from Toaster Feb 04 13:23:47 belen: Found the messages in the IRC logs now. The "In Progress" docs still list 2 modes Feb 04 13:26:44 hi Feb 04 13:28:37 I made a recipe that replaces the base-files recipe, but this recipe is not my preferred provider, so I want to use this recipe only on one recipe and the other recipes keeping using the base-files recipe Feb 04 13:29:44 But when I do this, the recipe that I install my custom base-files, install both, base-files and my custom base-files, but in the installed packages list, the base-files recipe is not there Feb 04 13:30:20 only the files are installed, but there is no log saying that base-files are installed Feb 04 13:30:44 anyone have ever seen this before? Feb 04 13:49:19 bboozzoo: I'm now down to 8809 KiB (by disabling most if not all configure options). I can probably win 1-2 MB still (e.g. remove some unneeded tools) Feb 04 13:50:54 lostcontrol: is the image booting? Feb 04 13:51:32 yes, otherwise it wouldn't be very useful, right? Feb 04 13:51:59 probably :) Feb 04 13:52:42 there's some weird interaction between systemd-sysusers and the rest of tools (timesyncd, networkd, hostnamed), apparelty sysusers can create (surprise, surprise..) users on demand Feb 04 13:53:15 I --disable-sysusers in the config Feb 04 13:54:02 lostcontrol: but then you either don't use systemd-netorkd (cause it does not start) or set USEADD_PARAM to add systemd-network (I'm actually kind of interested in systemd-networkd) Feb 04 13:54:42 bboozzoo: I --disable-networkd also :-) Feb 04 13:55:12 I found it useful, so it ended up enabled in my config Feb 04 13:56:47 bboozzoo: our application already manages the network interfaces so I really need a very minimal support from systemd Feb 04 14:04:07 rburton: those were not symlinks, the apr-util package triggered a QA check which thought they were Feb 04 14:04:22 I've tried to explain it in the patch message Feb 04 14:05:30 aratiu: could the test be improved? Feb 04 14:06:39 well yes, the so-dev test could check if the .so files are actually symlinks Feb 04 14:06:48 if they are not symlinks, it should not trigger Feb 04 14:07:15 if they aren't symlinks then probably they are loadable modules which are installed in other packages than -dev Feb 04 14:16:56 paulbarker: yes, those will need to be changed at some point Feb 04 14:17:17 paulbarker: the documents, I mean Feb 04 15:17:33 Hmm, the sdk env setup script sets PYTHONHOME regardless of whether nativesdk-python* is installed or not, which breaks the host python if it's not installed Feb 04 15:17:48 seems like that should probably be an environment-setup.d script in python for nativesdk Feb 04 16:04:40 RP: can we get bitbake -e | less fixed? Feb 04 16:08:04 i have to tee it to a file and read it there, and it's been that way for a long time. i should have reported it, i'll admit, keep forgetting about it Feb 04 16:08:28 also, the warnings are colored when stdout isn't a tty, which is another related bug Feb 04 16:11:33 huh i swear i use -e | less all the time Feb 04 16:12:46 i get badly screwed up output Feb 04 16:12:52 haven't had a chance to investigate Feb 04 16:15:19 kergoth: I'm sure we can, I didn't know it was broken:/ Feb 04 16:15:25 khem: ping Feb 04 16:17:09 * armpit tries to wake up Feb 04 16:23:06 hi, I'm having some trouble with crosswalk on the (GDP) Genivi Demo Platform. I had it build successfully but when I run crosswalk I get sandbox errors. I found a message that this was disabled for the linux builds? I'm not sure that crosswalk is getting built with that configuration as a result. Feb 04 16:23:13 @tmpsantos Feb 04 16:28:09 Hello. I'm new to yocto and I'm having trouble adding an out-of-tree kernel module. I can't figure out where to place it Feb 04 16:30:35 dpeterson3: have you look at this http://www.yoctoproject.org/docs/1.6.1/kernel-dev/kernel-dev.html#incorporating-out-of-tree-modules Feb 04 16:30:51 looked* Feb 04 16:32:38 I've been following the guide over there. I can get it to build just fine, but ${D} points to the wrong place when I go to install Feb 04 16:32:58 can you show your recipe Feb 04 16:33:09 sure Feb 04 16:33:58 can file:// be combined with downloadfilename in SRC_URI? Feb 04 16:34:16 might as well try, but thought i'd see if i could save some time :P Feb 04 16:34:29 for a defconfig in this case Feb 04 16:34:41 lpapp, here it is Feb 04 16:34:42 UMMARY = "Build a new kernel module for the rtl8192cu usb dongles. This is supposed to fix many of the problems as it is built against Real$ Feb 04 16:34:42 LICENSE = "GPLv2" Feb 04 16:34:42 #LIC_FILES_CHKSUM = "file://README.md;md5=3f8420c0b5237584b98d02c9425e9f80" Feb 04 16:34:42 LIC_FILES_CHKSUM = "file://README.md;md5=59a6671be9806aa70a5d491280e6c27b" Feb 04 16:34:42 inherit module Feb 04 16:34:43 SRCREV = "${AUTOREV}" Feb 04 16:34:43 SRC_URI = "git://github.com/pvaret/rtl8192cu-fixes.git" Feb 04 16:34:44 S = "${WORKDIR}/git" Feb 04 16:34:44 do_compile() { Feb 04 16:34:45 #unset CFLAGS CPPFLAGS CXXFLAGS LDFLAGS Feb 04 16:34:45 oe_runmake CONFIG_PLATFORM_I386_PC=n KSRC=${STAGING_KERNEL_DIR} KVER=3.18.21 Feb 04 16:34:46 } Feb 04 16:34:46 do_install() { Feb 04 16:35:08 Ulfalizer: you can use file::// in SRC_URI Feb 04 16:35:13 -: Feb 04 16:35:25 my colleague does that for testing from the top of the git tree Feb 04 16:35:32 the echo ${D} part was to tell me where it was trying to install before I found the bitbake -e command Feb 04 16:35:48 (on his local machine so that he can test things before committing broken stuff) Feb 04 16:37:48 lpapp: yeah, i know about file://. it'd be handy if i could rename the file at the same time though. Feb 04 16:38:17 so the file is called "file://something"? Feb 04 16:38:19 accidentally? :) Feb 04 16:39:49 it's a descriptive filename that i'd like to turn into "defconfig", because that will be picked up automatically by kernel.bbclass :) Feb 04 16:40:05 dpeterson3: please do not spam the channel with a bunch of lines, use a pastebin in the future Feb 04 16:40:14 sorry. Also new to irc Feb 04 16:40:49 you can always change the $D in worst case, but I will leave it with more experienced people to comment on this Feb 04 16:41:18 ${D} is not pointing to the wrong place, it's exactly what it should be. what makes you think it's wrong? Feb 04 16:41:25 elaborate, please Feb 04 16:41:51 I guess wrong is not the way to phrase it. It doesn't point where i think it should Feb 04 16:42:17 again, what makes you think it's wrong? Feb 04 16:42:23 dpeterson3: ${D} points to where stuff should be installed. it's more of an "input" (to the recipe) than an output. Feb 04 16:42:27 that tells it where to put stuff Feb 04 16:42:32 it doesn't matter where it points, bitbake will use it to pull files from in future tasks Feb 04 16:42:38 you could put it anywhere, it's irrelevent Feb 04 16:42:58 it's just a root to the files being installed which will be use dto populate the sysroot and create binary packages Feb 04 16:43:02 it's not left as is Feb 04 16:43:11 input vs. output can be a bit confusing in bitbake, especially with S and the like Feb 04 16:43:23 dpeterson3: i'd recommend starting with what you actual problem is Feb 04 16:43:27 not your attempt at analysis of it Feb 04 16:43:31 I have my file stored in ~/arm/yocto/master/poky/meta-gumstix/recipies-kernel/drivers/rtl8192cu-fixes Feb 04 16:43:48 it seems that we just used do_install Feb 04 16:43:59 our contractor then used the install command for the driver wherever he wanted to Feb 04 16:44:12 I do not know whether this is good by quality standards, but it has been working for us Feb 04 16:44:20 when I run bitbake rtl8192cu-fixes, it builds just fine Feb 04 16:44:30 if you aren't calling into a buildsystem to do the install, yes you can install the files yourself in do_install Feb 04 16:44:40 do_install fails Feb 04 16:44:44 $D points to Feb 04 16:44:55 "fails" tells us nothing Feb 04 16:45:00 what exactly is the error or problematic behavior? Feb 04 16:45:36 D seems to point here Feb 04 16:45:37 D="/home/daniel/arm/yocto-workspace/master/build/tmp/work/duovero-poky-linux-gnueabi/rtl8192-fixes/1.0-r0/image" Feb 04 16:45:44 yep, that's exactly where it should point Feb 04 16:45:47 instead of my kernel staging dir Feb 04 16:45:55 you *never* install directly into the sysroot Feb 04 16:45:58 so this is what we do for instance: install ${S}/drivers/gpib/tnt4882/tnt4882.ko ${D}${base_libdir}/modules/${KVER}/gpib Feb 04 16:46:10 bitbake will take the files from ${D} and put them in both binary packages and the sysroot Feb 04 16:46:58 I get that's what its supposed to do, but my kernel seems to get build to D="/home/daniel/arm/yocto-workspace/master/build/tmp/work/duovero-poky-linux-gnueabi/linux-gumstix/3.18-r0/image" Feb 04 16:47:12 yeah, that is fine Feb 04 16:47:18 for the reasons kergoth mentioned. Feb 04 16:47:28 hi Feb 04 16:47:43 someone can help me with a problem? Feb 04 16:47:45 dpeterson3: yes, that's exactly what it should be doing. Feb 04 16:47:51 igor___: if you ask the problem, probably :) Feb 04 16:48:01 thanks Feb 04 16:48:03 I made a recipe that replaces the base-files recipe, but this recipe is not my preferred provider, so I want to use this recipe only on one recipe and the other recipes keeping using the base-files recipe Feb 04 16:48:13 this custom recipe rreplaces, rconflicts and rprovides the base-files Feb 04 16:48:29 but when I install this custom recipe, it install the files of both recipes, my custom and the default base-files Feb 04 16:48:44 what does that even mean? Feb 04 16:48:49 first, you don't install recipes, you install packages Feb 04 16:48:56 second, are you installing it on target or in an image? Feb 04 16:48:59 sorry about the terms Feb 04 16:49:11 its because dont speak english very well Feb 04 16:49:14 third, it' shighly unlikely that installing your package is installing base-files, unless your package depends onb ase-files, more likely the iamge is installing both Feb 04 16:49:39 its a image Feb 04 16:49:45 ok. Then I guess I misunderstand where to put my recipe so it will be installed with the rest of the kernel modules. The specific error I get is nstall: cannot create regular file '/home/daniel/arm/yocto-workspace/master/build/tmp/work/duovero-poky-linux-gnueabi/rtl8192-fixes/1.0-r0/image/lib/modules/3.18.18-custom/kernel/drivers/net/wireless/rtl8192-fixes': No such file or directory Feb 04 16:49:55 how exactly do you expect one package to use one base-files but the rest use another when there's only one filesystem? Feb 04 16:50:28 dpeterson3: sounds like you didn't mkdir -p or install -d to make sure you rdestination path actually exists before trying to copy into it Feb 04 16:50:38 kergoth; my recipe rreplaces base-files, so it would only install one Feb 04 16:51:12 dpeterson3: install -m 0755 -d ${D}${base_libdir}/modules/${KVER}? Feb 04 16:51:22 lpapp: -D Feb 04 16:51:22 when I run bitbake -g, my image dont depends on base-files Feb 04 16:51:32 -d is just to create the dir, -D creates leading dirs when installing a file Feb 04 16:51:43 igor___: most likely another binary package is pulling it in Feb 04 16:51:46 i.e. a packagegroup Feb 04 16:52:44 kergoth: yeah, we do it in two steps Feb 04 16:53:06 perhaps we should do it in one Feb 04 16:53:09 yeah, that's how it's usually done, just saying that commandline will error if he uses it as is :) Feb 04 16:53:16 kergoth: wright, but on the installed_pkgs.txt there is no base-files Feb 04 16:53:31 aah. I'll give that a try. Thank you. I didn't realize that's what the -d option did Feb 04 16:53:31 yeah, we need to make sure that the directory exists and the copy is done into the right directory Feb 04 16:53:46 my package is supposed to replace base-files Feb 04 16:53:47 dpeterson3: either use two install comamnds, one with -d to create the dir, and one to install the file, or use one command and use -D Feb 04 16:53:50 dpeterson3: take your pick Feb 04 16:55:40 igor___: do you really want to replace or can you do .bbappend instead Feb 04 16:56:37 kergoth: yes, but my environment is unusual, I built two images, and one image is copy inside the other image Feb 04 16:57:12 kergoth: so one image needs the default base-files and the other needs my base-files Feb 04 16:57:26 ah, ok, that clears things Feb 04 16:57:32 I was just going to ask you that Feb 04 16:59:35 kergoth: I did this in two recipes: base-files and init-ifupdown Feb 04 16:59:50 kergoth: in the init-ifupdown it worked, but in base-files it is not working Feb 04 17:00:16 igor___: did you use PREFERRED_PROVIDER_virtual/base-files? Feb 04 17:00:46 no Feb 04 17:00:56 I set the PREFERRED_PROVIDER_base-files only Feb 04 17:00:56 well, scrape that Feb 04 17:00:59 to base-files Feb 04 17:01:02 hmm Feb 04 17:01:52 the weid thing is that the logs does not show the base-files been installed Feb 04 17:01:56 I am not sure it is the right way, but one way I can think of is adding another layer with bbappend for base-files and the second image in there... Feb 04 17:02:45 so to build the final image I must build one first, then change the local.conf and build the other one? Feb 04 17:04:05 if base-files hasn't been installed, then i don't see how it's possible for its files to be in your image Feb 04 17:04:10 i'd check your assumptions Feb 04 17:05:27 theorically, if the package was installed on my image it is suposed to be listed on the installed_pkgs.txt file, right? Feb 04 17:06:24 igor___: in buildhistory, you mean installed-packages.txt? Feb 04 17:06:47 i'd suggest reading the do_rootfs log Feb 04 17:06:56 I mean the file that is generated on the work path of the image Feb 04 17:08:44 Ok, I will make some more tests, but on the do_rootfs log does not mention the base-files Feb 04 17:08:50 the "NOTE: Installing the following packages:" Feb 04 17:08:53 there is no base-files Feb 04 17:09:58 and the NOTE: to be installed: there is no base-files neather Feb 04 17:10:42 I will make more searchs, cause its not an usual environment Feb 04 17:11:02 so maybe the yocto is not meant to work with this Feb 04 17:11:15 thankyou guys Feb 04 17:16:02 igor___: another way could be to call it differently and use packagegroup to replace it Feb 04 17:16:33 how so? Feb 04 17:16:48 My custom base-files is on a package group Feb 04 17:16:55 so my image install this package group Feb 04 17:17:29 right, so your problem really is the rootfs creation than rather than image configuration, et al. Feb 04 17:17:33 then* Feb 04 17:17:52 yes Feb 04 17:19:11 I guess that one package group that is default on core-image class is installing the base-files and it is not logged, because it is part of the image build process Feb 04 17:21:19 If I find when it is installing the base-files or who is installing it, it would facilitate everything Feb 04 17:21:59 but I tryied to run bitbake -DDD -v and bitbake -g and I cant find where is it Feb 04 17:24:51 I made another test here and seems that the base-files is not the responsible for the files that are been installed Feb 04 17:25:39 not at all surprising. did you ever read the do_rootfs task log? Feb 04 17:25:57 now i'm at square Feb 04 17:26:06 I read Feb 04 17:27:04 but the problem is that the file I dont want on the image is supposed to be installed by the base-files Feb 04 17:27:24 Hey, I need some help understanding the dependency system that bitbake uses. Feb 04 17:27:28 but, now, I guess there is more packages that install this file Feb 04 17:28:24 wyrm: could you be more precise Feb 04 17:28:26 wyrm: you can run bitbake -g to see the dependence tree Feb 04 17:29:02 wyrm: as lpapp says, you'll need to be more specific. what exactly are you having trouble understanding? Feb 04 17:29:10 ... working on it ... Feb 04 17:29:22 gah, a shell syntax error that breaks the codeparser will prevent you from using bitbake -e to examine the value of the broken function Feb 04 17:29:27 slow connection between work and irc host Feb 04 17:29:37 so it's not possible to see why the hell it's failing, particularly if there are bitbake variable expansions involved Feb 04 17:29:40 also, slow connection between brain and keyboard. Feb 04 17:29:41 * kergoth grumbles Feb 04 17:29:44 so, yeha Feb 04 17:29:56 So, I have this program that requires alsa and fftw to build, but I can't figure out the right incantations to put in REQUIRE_${PN} . Feb 04 17:31:08 you mean R/DEPENDS? Feb 04 17:31:21 there is no REQUIRE variable Feb 04 17:31:27 see the yocto documentation at yoctoproject.org Feb 04 17:31:28 to build you want DEPENDS Feb 04 17:31:49 yeah, I could not find REQUIRE in the glossary either. Feb 04 17:31:52 DEPENDS="alsa-lib fftw" Feb 04 17:33:50 If I issue the command `bitbake myrecipe`, it grinds for a while, and then crashes with the message "fatal error: alsa/asoundlib.h: No such file or directory". Feb 04 17:33:54 there is a way to know which recipe installed a specific file? Feb 04 17:34:04 and yes, I meant DEPENDS Feb 04 17:34:07 sorry Feb 04 17:34:16 ANYWAY Feb 04 17:34:21 wyrm: have you tried what rburton wrote? Feb 04 17:34:44 stand by Feb 04 17:34:44 igor___: yes, it depends on how you want to know it Feb 04 17:35:03 $ oe-pkgdata-util find-path */asoundlib.h Feb 04 17:35:03 alsa-lib-dev: /usr/include/alsa/asoundlib.h Feb 04 17:35:03 alsa-lib-dev: /usr/include/sys/asoundlib.h Feb 04 17:35:18 wyrm: did you depend on alsa-lib? Feb 04 17:35:34 At that point, .... Feb 04 17:36:16 I clearly should have typed all of this up beforehand and just dumped it into the channel en masse. Feb 04 17:36:26 lpapp: I'm looking on the rootfs path on the work path of my image Feb 04 17:36:47 wyrm: if you've got a DEPENDS=alsa-lib, then the makefiles or whatever are likely broken Feb 04 17:36:47 so the file is var/log that is a symbolic link to volatile/log Feb 04 17:36:59 igor___: I would use what rburton wrote, and if you are on the image, then opkg files foo and similar dpkg/rpm magics can do it Feb 04 17:37:09 actually, opkg search Feb 04 17:37:32 lpapp: thank you sir Feb 04 17:37:42 So, yeah. At that point, if I look for asoundlib.h ... Feb 04 17:37:54 Y'know what... I'll just type this up and come back. Feb 04 17:38:14 wyrm: does the compiler know about the include path? Feb 04 18:41:51 rburton: hmmm thats a latent issue my change is unearthing thats good Feb 04 18:41:59 kergoth: pong Feb 04 18:57:22 khem: noor mentioned you had some pending plymouth+systemd stuff, status? Feb 04 18:57:39 kergoth: I dont remember Feb 04 18:57:44 k Feb 04 18:57:56 I have plymouth itslelf which I submitted long ago Feb 04 18:58:00 but let me check Feb 04 18:58:10 and see if we stil carry systemd units Feb 04 18:58:52 thanks Feb 04 19:05:19 Hi, Im having trouble adding github git repositories to a yocto recipe, is there anybody who has some experience with git repo's from github? Feb 04 19:05:48 github git repos are used all over the metadata Feb 04 19:05:53 grep for some examples Feb 04 19:22:59 fenrig: plus you could describe your trouble details Feb 04 19:23:53 yeah Im having some troubles with github, but then again I have to run yocto on a vm so maybe its my vm that is to blame :D I will get back to you if I can pinpoint the trouble more Feb 04 19:26:18 lpapp: I may have gotten it to work, but its a large repo so I will have to wait a bit :D thx f Feb 04 19:52:37 Why do I have to supply a git src with w SRV_REV ? Feb 04 19:53:19 fenrig, so it knows what to check out? Feb 04 19:53:48 Crofton|work: but what if HEAD was sufficient? Feb 04 19:53:48 fenrig: if you want to track a branch HEAD, set SRCREV = "${AUTOREV}" Feb 04 19:53:59 ah okay Feb 04 19:54:13 but remember to put SRCPV in PV Feb 04 19:54:26 we really need to fix it so do_fetch depends on the value of srcrev when there's a git repo regardless of PV Feb 04 19:54:29 imo anyway Feb 04 19:55:30 kergoth: well its for a linux kernel source so I dont think that pv really applies here Feb 04 19:56:00 if you don't add SRCPV to PV, it will *not* rebuild automatically when someone checks into the branch Feb 04 19:56:06 if you're fine with manual rebuilds, then don't Feb 04 19:56:37 kergoth: okay got it, yeah Im choosing for manual rebuilds for this one :) Feb 04 19:57:45 does bitbake keep the git repo locally or does it clone fully every rebuild Feb 04 19:58:48 fenrig: as with any other sources, it's downloaded to DL_DIR and used from there Feb 04 19:59:08 kergoth: so full clone every rebuild then Feb 04 19:59:13 it'll only fetch from upstream if the srcrev you want isn't available in the local git repo, or you're using autorev, in which case it'll update the existing repo to get the rev you need Feb 04 19:59:15 huh? Feb 04 19:59:25 no, just like every other source, it's kept and used from DL_DIR Feb 04 19:59:27 as i just said Feb 04 19:59:37 downloads are only wiped if you remove DL_DIR or just -c cleanall Feb 04 19:59:49 kergoth: yeah well then changing the srcrev makes it clone again, its now fully downloading the repo again Feb 04 19:59:54 no Feb 04 19:59:58 i said it *fetches* Feb 04 19:59:59 not clones Feb 04 20:00:04 it runs git fetch to update the existing repo Feb 04 20:00:16 that is, it git fetches, not bitbake fetches :) Feb 04 20:00:29 it clones once, that's it Feb 04 20:00:36 kergoth: Im in the download dir, and the size was around 1gb and its 150mb (and counting) Feb 04 20:00:44 at the moment Feb 04 20:02:29 OOOH when the recipe fails then it deletes the dl dir Feb 04 20:02:53 if the clone fails, sure, an incomplete git clone is useless, git can't resume Feb 04 20:03:21 to clarify, it didn't delete the dl dir, just the clone Feb 04 20:03:47 kergoth: yeah well it failed because it couldnt find my particular srcrev :D Feb 04 20:05:06 hmm, interesting, afaik that shouldn't have resulted in wiping the existing git repo, just a fetch to try to download it, but it would have failed at parse time up front before even fetching the thing Feb 04 20:05:25 although, if srcpv isn't in pv, maybe it's not expanded at parse time Feb 04 20:05:34 *works through the code in his head* Feb 04 20:26:05 Typed up in advance, as promised: Feb 04 20:26:24 (recap) Hey, I need some help understanding the dependency system that bitbake uses. Feb 04 20:26:37 (recap) I have this program that requires alsa and fftw to build, but I can't figure out the right incantations to put in DEPENDS_${PN} . Feb 04 20:26:45 If I issue the command `bitbake myrecipe`, it grinds for a while, and then crashes at myrecipe with the message "fatal error: alsa/asoundlib.h: No such file or directory". Feb 04 20:26:53 At that point, if I look for alsa/asoundlib.h, I find it under "build/tmp/sysroots/x86_64-linux/usr/include/alsa/asoundlib.h" and two places under "build/tmp/work/x86_64-linux/.../alsa/asoundlib.h". Feb 04 20:27:01 I also find various instances of the "sys/asoundlib.h" deprecated compatibility header, and "build/tmp/work/cortexa9hf-vfp-neon-mx6qdl-poky-linux-gnueabi/alsa-lib/1.0.29-r0/alsa-lib-1.0.29/include/asoundlib.h". Feb 04 20:27:46 Note that the last path does not have the "alsa/" component of "alsa/asoundlib.h", which may or may not be germane to the problem. Feb 04 20:28:01 Then, if I issue the command `bitbake --continue myrecipe`, it errors immediately on myrecipe, then builds a whole bunch of other stuff. Feb 04 20:28:09 At this point, I also have "build/tmp/sysroots/nitrogen6x/usr/include/alsa/asoundlib.h", and several places under both "build/tmp/work/x86_64-linux" and "build/tmp/work/cortexa9hf-vfp-neon-mx6qdl-poky-linux-gnueabi/". Feb 04 20:28:17 There are also several copies of "fftw3.h" under "build/tmp/work/cortexa9hf-vfp-neon-poky-linux-gnueabi/fftw/3.3.4-r0/", which shall become relevant in a moment. Feb 04 20:28:20 there is no DEPENDS_${PN}. Feb 04 20:28:26 Now, when I run `bitbake --continue myrecipe`, it errors immediately on myrecipe, with the new message, "fatal error: fftw3.h: No such file or directory". Feb 04 20:28:32 there's RDEPENDS_, which are the runtime dependencies that go on target in the binary pcakges Feb 04 20:28:37 and there's "DEPENDS", which is build time Feb 04 20:28:43 DEPENDS_${PN} will have no effect at all Feb 04 20:28:53 as was mentioned earlier, set DEPENDS = "onerecipe anotherrecipe" Feb 04 20:29:02 Oh! Feb 04 20:29:39 https://www.yoctoproject.org/docs/1.8/mega-manual/mega-manual.html#new-recipe-configuring-the-recipe, etc Feb 04 20:29:58 http://www.aosabook.org/en/yocto.html may also be of interest Feb 04 20:30:17 particularly the section on the bitbake scheduler Feb 04 20:30:32 I thought you were questioning my brain-fart where I typed REQUIRE_${PN}. Feb 04 20:30:49 But the whole thing was wrong. _${PN}, too. Feb 04 20:33:21 I'll take a closer look at those links, thanks. Feb 04 21:16:53 Ever present dumb question of the day: what combination of depends/rdepends, etc for to use in a package that depends on GTK to build some utility, but you'll never actually deploy said utility, and thus don't need gtk? Feb 04 21:17:17 Obviously, the other option is just to patch it all out Feb 04 21:17:24 just add it to DEPENDS Feb 04 21:17:38 it'll only get added to rdepends by the shlibs code if one of the binaries you package links against the shared library Feb 04 21:18:35 ah, deps handline "magic" RDEPENDS. Feb 04 21:18:37 fun Feb 04 22:10:20 RP: there ? Feb 04 22:10:41 RP: I want to ignore 15 recipes when doing world builds Feb 04 22:10:50 in OE-Core Feb 04 22:11:00 so we get clean world builds for musl Feb 04 22:11:08 and use that as baseline Feb 04 22:11:21 do we have some where 1 place to do so ? Feb 04 22:11:34 or do I have to blacklist each one of them Feb 04 22:17:19 khem: it appears that your fortify changes are causing more packages to break. doing a world build here and so far three recipes are failing which previously have worked on the AB Feb 04 22:17:24 wouldn't EXCLUDE_FROM_WORLD suffice, or are deps on them also a problem? Feb 04 22:17:30 * kergoth yawns Feb 04 22:17:36 ie shared/newt.o: relocation R_X86_64_PC32 against undefined symbol `SLtt_Screen_Rows@@SLANG2' can not be used when making a shared object; recompile with -fPIC Feb 04 22:17:40 kergoth: yes that would work Feb 04 22:17:51 khem`: fyi, meta-musl/meta-clang have issues with current master, version bumps Feb 04 22:17:57 also there's an exclude from world include file but surely you want to only exclude if musl Feb 04 22:18:11 kergoth: dont use meta-musl Feb 04 22:18:17 meta-clang Feb 04 22:18:19 ah, is it all in oe-core now? Feb 04 22:18:24 wasn't sure what the merge status was Feb 04 22:18:27 meta-clang issues should be reported Feb 04 22:18:29 k Feb 04 22:18:36 just starting to play around :) Feb 04 22:18:37 kergoth: khem has been patchbombing for a while now :) Feb 04 22:18:48 yeah, i've seen it, but wasn't sure how much was actually in master :) Feb 04 22:19:14 EXCLUDE_FROM_WORLD_libc-musl = "1" Feb 04 22:19:19 would work isnt it ? Feb 04 22:19:50 kergoth: right now there are 20 packages in OE-core which dont build with musl Feb 04 22:20:06 see http://errors.yoctoproject.org/Errors/Build/12230/ Feb 04 22:20:18 every thing else is ok all images build Feb 04 22:20:21 cool Feb 04 22:20:29 and they boot and work too :) Feb 04 22:20:36 for all arches Feb 04 22:20:41 khem`: there's a poky-world-exclude.inc in meta-yocto for excluding on the autobuilder Feb 04 22:20:50 except mips64 which I havent tried Feb 04 22:20:53 i ran into this with clang-native: | CMake Error at cmake/modules/CheckAtomic.cmake:36 (message): Feb 04 22:20:54 | Host compiler appears to require libatomic, but cannot find it. Feb 04 22:21:03 assuming i should upgrade my build machine? :) Feb 04 22:21:05 or rather, build chroot Feb 04 22:21:12 kergoth: is it Centos Feb 04 22:21:21 nope, u12.04 i think Feb 04 22:21:30 rburton: I am working on fixing efi issue Feb 04 22:21:50 rburton: it seem SECURITY_LDFLAGS are leaking into all sort of LDFLAGS Feb 04 22:21:51 i'll try installing libatomic in tha chroot Feb 04 22:22:01 kergoth: 12.04 hmm Feb 04 22:22:14 i should probably switch to 14 anyway, but :) Feb 04 22:22:20 I use 14.04 oldest distro in my world Feb 04 22:22:34 for my regular stuff archlinux Feb 04 22:22:38 so I am never old Feb 04 22:22:39 :) Feb 04 22:23:13 rburton: once my patch to switch poky-tiny is accepted then we shuld do that Feb 04 22:23:25 khem`: pango on the AB i could explain and fix, grub-efi-native less so Feb 04 22:23:33 khem`: there is conf/distro/include/world-broken.inc, I'm kind sad to add to it after we got it empty! Feb 04 22:23:55 khem`: you should ask RP what his acceptance critera for the poky-tiny change is, i've had it in mut for a while :) Feb 04 22:23:56 RP: I promise to empty it out again Feb 04 22:23:59 khem`: EXCLUDE_FROM_WORLD_pn-X_libc-musl = Feb 04 22:24:05 khem`: i need to get an arch machine going again, that used to be my goto distro, but never caught up after the switch to systemd :) Feb 04 22:24:18 rburton: I just never saw it on the list, keep meaning to merge it tbh Feb 04 22:24:21 kergoth: use antergos if you like installers Feb 04 22:24:36 for tiny, it appears to work fine Feb 04 22:24:38 khem`: please just use the msl override Feb 04 22:24:44 i just wrote my own scripts for it ages ago Feb 04 22:24:44 https://github.com/kergoth/kergoth-arch-setup Feb 04 22:24:47 RP: I am using poky-tiny with musl Feb 04 22:24:51 just need that nightly-musl target on the ab Feb 04 22:24:57 but i can't recall the voodoo for the fs/partition setup anymore, used to use btrfs Feb 04 22:25:00 gotten rusty :) Feb 04 22:25:07 oh its there Feb 04 22:25:14 lets give it a spin and see what happens Feb 04 22:25:22 yes Feb 04 22:25:35 khem`: https://autobuilder.yoctoproject.org/main/builders/nightly-musl/builds/1 Feb 04 22:25:37 musl has does tonnes of good to packages Feb 04 22:25:42 a lot of rust got fixed Feb 04 22:26:08 rburton: gr8 add core-image-weston as well please Feb 04 22:26:21 if possible Feb 04 22:26:36 easy Feb 04 22:26:40 * khem` add to bookmarks Feb 04 22:27:46 RP: adding to EXCLUDE_WORLD will help keep the current quality Feb 04 22:27:56 while we fix the remaining 20 packages Feb 04 22:28:34 kergoth: I think arch handled the transition to systemd quite smoothly Feb 04 22:29:44 khem`: I agree as long as we use he musl override Feb 04 22:29:56 we say TARGET_LDFLAGS_append = " ${SECURITY_LDFLAGS}" Feb 04 22:30:00 then in native.bbclass we say Feb 04 22:30:10 TARGET_LDFLAGS = "${BUILD_LDFLAGS}" Feb 04 22:30:29 khem`: Add a SECURITY_FLAGS_class-native = "" ? Feb 04 22:30:33 RP: yes musl override will be there Feb 04 22:30:45 RP: yes tryin that out now Feb 04 22:32:00 khem: oooooooh Feb 04 22:32:31 question is do we want to make native and nativesdk use it as well Feb 04 22:32:36 its not a bad thing Feb 04 22:32:42 from security perspecitve Feb 04 22:32:47 not native, it failed all over the AB as older builders don't have fortify-strong Feb 04 22:32:49 but tools will be a bit slow :) Feb 04 22:33:04 we control the cross tools, but not host Feb 04 22:33:42 I can remove -fstack-protector from LDFLAGS_remove_class-native = "-fstack-protector-strong" Feb 04 22:33:48 RP: thinking we should make bitbake -e bypass codeparsing / signature generation. i had a shell syntax error that made pysh choke, but couldn't use bitbake -e to examine the value to figure out why :) Feb 04 22:34:11 question is should we care for native and nativesdk tools Feb 04 22:34:18 we can do it eitherway Feb 04 22:34:25 kergoth: I don't disagree. Patches welcome... Feb 04 22:34:29 kergoth: I'm drowning in bugs :( Feb 04 22:34:37 CFLAGS have same problem Feb 04 22:34:40 np, wasn't asking you to do it, just making sure it made sense Feb 04 22:35:24 khem`: lets start with complete prune to fix things and then add some next? Feb 04 22:35:31 ok Feb 04 22:35:51 khem`: I think we should just rely on the host to use whichwever flags it thinks make sense there Feb 04 22:36:00 for native anyway Feb 04 22:36:07 we control what we build for nativesdk Feb 04 22:38:15 khem`: so basically you have this: SECURITY_FLAGS_class-native = "" Feb 04 22:38:15 SECURITY_LDFLAGS_class-native = "" Feb 04 22:38:31 * rburton is about to kick a world locally before going to bed and doesn't want to rebuild it all again tomorrow :) Feb 04 22:39:33 rburton: testing this https://gist.github.com/cdcbfcac6185165d08af Feb 04 22:39:57 hm, what about nativesdk Feb 04 22:40:09 gah, i need to rewrite toolchain-shar-*, that's some ugly shell with bad filename handling Feb 04 22:40:12 maybe replicate the appends for nativesdk for clarity Feb 04 22:40:36 rburton: we disable it for everything but target Feb 04 22:40:43 in that patch Feb 04 22:40:49 yeah Feb 04 22:41:05 I think thats what we really want Feb 04 22:41:20 otherwise we disable it for native nativesdk cross crosssdk Feb 04 22:41:31 may be cross-canadian Feb 04 22:42:43 rburton: I will bundle it into existing pu Feb 04 22:42:56 will that be more convenient ? Feb 04 22:43:21 grub-efi-native works now Feb 04 22:43:33 on my old ubuntu box Feb 04 22:44:19 khem`: it was a short branch so that's fine, but also post it to the list pls Feb 04 22:44:30 sure Feb 04 22:44:57 right, bed. g'night all. Feb 04 22:45:04 night Feb 04 22:49:37 rburson updated kraj/pu Feb 04 22:50:31 logged off, as your inability to complete the nick shows ;) Feb 04 22:50:39 * kergoth bangs head against desk Feb 04 22:55:47 * armpit can hear it from here Feb 04 22:56:56 RP: do you see any imrovements I need to do for http://git.openembedded.org/openembedded-core-contrib/commit/?h=kraj/master&id=15832ff80d377300ce323920f1f11a9a6e0ded8d Feb 04 22:57:20 and for http://git.openembedded.org/openembedded-core-contrib/commit/?h=kraj/master&id=e810abe120b42e1622068e0b3b02cd405eaf765d Feb 04 22:58:50 khem`: Personally, I prefer whitespace changes to be separate but I have meant to merge it Feb 04 22:59:05 Oh yes Feb 04 22:59:26 I think that keeps it clean Feb 04 22:59:28 logically Feb 04 22:59:51 usually I do it when I see it as a side when fixing something else Feb 04 23:00:34 but it doesnt hurt to do it as a separate commit Feb 04 23:04:41 RP: instead of adding it to world-broken.inc Feb 04 23:04:54 I wonder what it would take to make all nativesdk stuff build static against musl :) would really shrink the sdk down Feb 04 23:04:56 I think its better to add them to BLACKLIST Feb 04 23:05:12 or non-static, for that matter Feb 04 23:05:22 * kergoth wonder sif you can just set TCLIBC_class-nativesdk Feb 04 23:05:37 hmm Feb 04 23:05:39 kergoth: Its an interesting thought Feb 04 23:06:06 i know static is not even an option with glibc due ot how large the binaries are, and nss, but musl might be an option Feb 04 23:06:13 * kergoth adds to todo to play with Feb 04 23:06:57 kergoth: once musl is nailed I will like to bring clang to core in 2.2 :) Feb 04 23:07:00 static would ease our relocation headaches, though Feb 04 23:07:06 khem`: ooh, that'd be nice :) Feb 04 23:07:21 right now we can build SDKs which offer both Feb 04 23:07:26 gcc and clang Feb 04 23:07:33 I bet anti-gpl and/or pro-bsd folks are going to love you Feb 04 23:07:42 i wonder if elftoolchain has a linker yet Feb 04 23:07:49 I personally dont care Feb 04 23:07:58 I just like to have options Feb 04 23:08:05 * kergoth nods Feb 04 23:08:12 i don't care fo rmy own stuff, but some companies are *really* paranoid Feb 04 23:08:21 sometimes they should be, but not always :) Feb 04 23:08:26 yeah I know Feb 04 23:09:07 so far my benchmarks show clang is very impressive given its age Feb 04 23:09:26 it reduces the compile time by almost 30% Feb 04 23:09:32 wow Feb 04 23:09:38 i love how nice clang's error messages are Feb 04 23:09:42 really pleasant Feb 04 23:10:01 * kergoth wonders how tough it'd be to make bitbake use remake instead of make :) Feb 04 23:10:06 so if you use musl+clang you can build same image with number of components in probably 40% less time Feb 04 23:10:11 guessing circular deps would be the main issue Feb 04 23:10:15 khem`: that's impressive Feb 04 23:10:28 musl has untangled the include mayhem Feb 04 23:10:45 makes sense Feb 04 23:10:45 and clang is just fast Feb 04 23:10:52 * kergoth nods Feb 04 23:11:00 by modern design Feb 04 23:11:19 I'll have to take a look at the recipes, I started looking at creating them, but didn't know enough about the build process to untangle it :) Feb 04 23:11:23 (clang that is) Feb 04 23:11:30 once I am out of trenches I am going to do some runtime benchmarking Feb 04 23:12:26 clang itself needs a lot longer to compile :) Feb 04 23:12:29 than say gcc Feb 04 23:12:39 but banking on sstate here Feb 04 23:12:43 at least it's a fairly one time cost Feb 04 23:12:45 * kergoth nods Feb 04 23:13:19 clang recipes are there in github.com/kraj/meta-clang Feb 04 23:13:47 right now it plugs in as default compiler if you use it Feb 04 23:13:48 * kergoth nods Feb 04 23:13:54 saw that, the TOOLCHAIN bits Feb 04 23:14:05 a world build I see 6 packages fail Feb 04 23:14:15 so not a bad start Feb 04 23:14:43 there are some rogue packages which just wont budge Feb 04 23:14:48 like kernel ;) Feb 04 23:14:52 i feel like it should leverage TCMODE, but tcmode is global, not per-recipe, so i can see why you had to roll your own switching mechanism Feb 04 23:14:58 * kergoth nods Feb 04 23:15:13 yeah earlier I had it reversed Feb 04 23:15:26 but then users said that is not what they expect Feb 04 23:15:29 when they use meta-clang Feb 04 23:15:40 they want to use it as default and use gcc sparingly Feb 04 23:15:49 so I chnaged it Feb 04 23:16:08 TCMODE is a good idea Feb 04 23:16:33 i'm not sure i agree, i like it when just including a layer makes its stuff *available*, but then i opt-in to make use of it Feb 04 23:16:43 but then again i did what you did with meta-sourcery out of necessity, so .. :) Feb 04 23:16:54 there are exceptions to every rule, so.. Feb 04 23:17:01 right Feb 04 23:17:53 now that i'm not overriding tuning files, i should change the meta-sourcery behavior. before i coudlnt' conditionally override a .conf/.inc :) Feb 04 23:18:09 don't need to now that i just muck with TUNE_CCARGS instead Feb 04 23:18:48 yeah Feb 04 23:19:17 btw. I got SD card image generated by using wic but it did not boot well Feb 04 23:19:21 on bon Feb 04 23:19:22 e Feb 04 23:19:29 we should really have some guidelines on lnayer creation about stuff like this. folks often want to have some feature, but dont' know how/where to make it available / configure it Feb 04 23:19:34 s/ln/l/ Feb 04 23:19:51 so they end up making it just mangle stuff on layer inclusion or even worse Feb 04 23:20:09 khem`: huh. we're using it for boneblack anyway.. i ahven't tested regular bone though Feb 04 23:20:16 i should play with it, dig my bone out of a drawer Feb 04 23:20:26 I dont have black Feb 04 23:20:55 i wouldn't either except work sent it to me to mess with, i should really send all my old boards back, it's taking up room in my house Feb 04 23:21:06 actually i should send back my ancient decrepit work laptop i don't use too :) Feb 04 23:22:39 I have pi2 and minnow2 Feb 04 23:22:48 or minnowmax as they call it Feb 04 23:22:52 nice Feb 04 23:23:06 as my primary boards Feb 04 23:23:12 then emulators Feb 04 23:23:43 Hmm, we should add an option to make buildhsitory do its git stuff in the background and then report back to the user's tty when it completes Feb 04 23:23:44 I also have an atmel just for armv5t stuff Feb 04 23:23:51 would really speed up my test cycles Feb 04 23:24:00 I agree Feb 04 23:24:10 i hit ^C^C and .. wait....; Feb 04 23:24:22 eventually can vim.. Feb 04 23:24:24 heh Feb 05 01:35:32 Task 1577 (/home/julian/poky/meta/recipes-sato/matchbox-terminal/matchbox-terminal_git.bb, do_unpack) failed with exit code '1' Feb 05 01:36:04 I get 16 error like above, anyone could provide any idea ?? Feb 05 01:38:01 I use ubuntu 14.04, branch is dizzy, MACHINE is "qemux86-64" Feb 05 01:38:47 anyone here ?? Feb 05 01:39:12 sad...... Feb 05 01:41:53 that doesn't actually tell us anything Feb 05 01:42:07 see the log file for the task, which those errors refer to Feb 05 01:46:03 pkgconfig-native is broken on master Feb 05 01:47:22 er, let's internal glib is broken Feb 05 01:47:27 *say even Feb 05 01:50:54 er~~ sorry I am very new to yocto. Feb 05 01:53:12 OK, I should find a way to post log on the web then come back here. Feb 05 02:04:11 Lanser: google for pastebin, there are lots of them Feb 05 02:04:53 khem: is there a way, after the fact, to get a list of the recipes that built with gcc in my build? perhaps it'd be wortha dding a log for that like qa.log or something Feb 05 02:05:00 khem: (re: clang) Feb 05 02:07:49 * nerdboy thinks building from the release tag should actually work Feb 05 02:14:57 Hi kergoth: http://pastebin.com/xFBMGrCZ Feb 05 02:16:41 What I did before bitbake is: sudo apt-get install gawk wget git-core diffstat unzip texinfo gcc-multilib build-essential chrpath socat libsdl1.2-dev xterm Feb 05 02:17:11 2. git clone http://git.yoctoproject.org/git/poky and switch to dizzy branch Feb 05 02:18:33 3. modify ~/poky/build$ gedit ./conf/local.conf -> add MACHINE ?= "qemux86-64", remove #MACHINE ??= "qemux86" Feb 05 02:18:43 4. bitbake -k core-image-sato Feb 05 02:19:19 Then error happen, if no error, I expect to run: runqemu qemux86-64 Feb 05 02:24:38 Here is the log file: https://www.dropbox.com/s/av8qql18rn6wyz3/yocto_build_error.log?dl=0 Feb 05 02:29:44 libffi-native also broken? Feb 05 02:30:14 master seems a lot chunkier than usual... Feb 05 02:30:55 * nerdboy has a working hack for pkg-config-native tho Feb 05 02:30:57 Hmmm... Feb 05 02:31:46 /usr/bin/ld: src/x86/.libs/ffi.o: relocation R_386_GOTOFF against undefined hidden symbol `ffi_closure_REGISTER' can not be used when making a shared object Feb 05 02:37:03 Hmm, binutils-crosssdk's suffix doesn't include SDK_OS, so a multi-sdkmachine build between mingw and linux blows up Feb 05 02:37:21 should probably include the full sys it's targeting **** ENDING LOGGING AT Fri Feb 05 02:59:59 2016