**** BEGIN LOGGING AT Wed Jan 30 02:59:57 2019 Jan 30 05:46:36 RP: adelcast here is the error http://ix.io/1zy3 Jan 30 05:54:12 plug in another stick of RAM? Jan 30 05:54:16 * paulg runs. Jan 30 05:57:33 I dont think thats the issue Jan 30 05:57:40 this box has 16GB of ram Jan 30 08:25:08 I'm still getting that ugly "ImportError: No module named attr" while trying to build the wireless-regdb package Jan 30 08:25:15 It's driving me nuts Jan 30 08:26:45 I think the package is just broken Jan 30 08:41:37 https://aur.archlinux.org/packages/wireless-regdb-pentest/?comments=all Jan 30 09:13:33 khem: I suspect memory or some other resource like file handles Jan 30 11:02:59 Hello, ruby gem question: I have implemented a patch for the ruby.bbclass file that allows cross compiling. The problem is the gem needs to build native first, so this step fails. If i remove the patch, the native gem builds, but the packaging step fails due to needing to package for arm and the libs have been native x86 compiled. is there a way to only apply the ruby.bblclass cross compile patch to specific parts of the build process? Jan 30 11:04:19 or a way in the recipe to translate the libs "back" to arm in during the packaging process? Jan 30 11:12:29 would you be able to solve that by changing ruby.bbclass so it can work in target and native, then your foo gem can depend on foo-native Jan 30 11:13:20 I think that is the direction I will have to go Jan 30 11:13:28 or attempt to go Jan 30 11:13:32 easy enough Jan 30 11:14:12 allowing cross compilation should just be 'use the right tools, not gcc'. in native builds, the 'right tool' is $CC which is $BUILD_CC which is gcc Jan 30 11:14:14 so, just works Jan 30 11:31:09 Has someone successfully extracted the Qt5 SDK? Jan 30 11:31:46 It breaks everytime I try Jan 30 12:42:59 hey. I end up with "ssh_allow_empty_password" in images, but can't figure out where it comes from. Is there any way to track down the "setter" of this option? Jan 30 12:48:34 sven^: see your IMAGE_FEATURES Jan 30 12:48:58 sven^: DEBUG_TWEAKS, probably Jan 30 12:49:08 IMAGE_FEATURES += "\ Jan 30 12:49:09 empty-root-password \ Jan 30 12:49:09 package-management \ Jan 30 12:49:09 ssh-server-openssh \ Jan 30 12:49:09 " Jan 30 12:49:12 oops, sorry Jan 30 12:50:48 core-image is defined in poky? Jan 30 12:51:21 we have 5 images. 1 has debug-tweaks, but it's a dev image only used internally. The other images don't have debug-tweaks Jan 30 13:02:52 can anybody help to correctly setup UTF-8 locale in yocto, I seeing LC_ALL: cannot change locale (en_US.UTF-8): No such file or directory Jan 30 13:04:12 josey: https://www.yoctoproject.org/docs/current/ref-manual/ref-manual.html#var-GLIBC_GENERATE_LOCALES should be helpful Jan 30 13:05:25 I have GLIBC_GENERATE_LOCALES = "en_US.UTF-8" Jan 30 13:07:33 I also have IMAGE_LINGUAS = "en-us" Jan 30 13:08:20 This is my local.conf https://gist.github.com/pepone/1ff01e03e1de6b35e4b2eb321e5bdc1a Jan 30 13:15:07 do I need to add libc-locales to DISTRO_FEATURES_LIBC? Jan 30 13:15:34 josey: for some reason lost to history en_US *is* utf8 Jan 30 13:15:50 there's a variable somewhere you can toggle to make it like everyone else Jan 30 13:15:55 been meaning to submit the patch for that... Jan 30 13:16:16 LOCALE_UTF8_IS_DEFAULT iirc Jan 30 13:16:49 set that to 0 in local.conf and i think the naming goes back to what you expect Jan 30 13:17:07 New news from stackoverflow: Undefined references to precompiled opencv 2.4 Jan 30 13:30:25 I simply want to clean the sstate for a specific recipe. *Without* triggering any "look for the current revision of all SRCREV recipes" tasks Jan 30 13:31:07 T_UNIX: bitbake -c clenasstate $YOURFUNKYRECIPE :) Jan 30 13:33:00 that will trigger checks for updates Jan 30 13:33:59 `bb.data_smart.ExpansionError: Failure expanding variable SRCPV, expression was ${@bb.fetch2.get_srcrev(d)} which triggered exception FetchError...` Jan 30 13:34:25 setting `BB_NO_NETWORK`will trigger the same :) Jan 30 13:48:53 Is there a way automatically get the host username from a Yocto recipe? Jan 30 13:50:18 @malanecora "host" in the cross canadian convention sense (i.e. MACHINE's user?)? Or host username as in the user that currently executes `bitbake`? Jan 30 13:52:50 is there a way to set timestamp for rootfs ? Jan 30 13:53:21 (for all files) Jan 30 13:55:39 T_UNIX: Host username, who executes bitbake Jan 30 13:56:09 T_UNIX: whoami not available Jan 30 14:21:27 is anyone working on #8743 in terms of updating the documentation? Jan 30 14:21:38 I've spent a day being rather confused... Jan 30 14:29:25 malanecora: I'd assume you could either look at how "host contamination" is implemented or whitelist (via `BB_ENV_EXTRAWHITE`) an environment variable to use. Jan 30 14:40:10 malanecora: why do you care? evil hack would be to run /usr/bin/whoami directly. Jan 30 14:40:21 (we filter it out to avoid host contamination) Jan 30 14:50:46 alimon: have mailed you another example of a different kind of output glitch problem :/ Jan 30 15:05:10 I run into issue of .pyc getting recompiled on the target. currently the main suspect is .pyc's internal mod time which is older than .py file on the file system. does installs in do_install (or who places things on rootfs?) keep build modify time? can it be made to ? Jan 30 15:07:44 Hi, i can’t seem to solve this installed-vs-shipped error. I have tried with and without do_install section. https://gist.github.com/bentech/c341cbd02fb9bfccad1e4bd7c9161ae8 Jan 30 15:09:13 bentech: well, have you read: "Please set FILES such that these items are packaged." Jan 30 15:09:29 bentech: and accordingly inspected what FILES is for that recipe? Jan 30 15:10:18 Is there a way to chattr +i a directory from a recipe please? Jan 30 15:10:33 (after that, modify it to suit your needs= Jan 30 15:12:12 there is REPRODUCIBLE_TIMESTAMP_ROOTFS though, but it's guarded by BUILD_REPRODUCIBLE_BINARIES, which is not yet my concern Jan 30 15:16:30 I have added in FILES_${PN} += "${D}/usr/lib/*" but still get the same message Jan 30 15:18:03 Nevermind Files must be relative to image Jan 30 15:18:16 eeexact, was just about to say that. Jan 30 15:36:34 T_UNIX: rburton_ Ty! That's just what I needed Jan 30 15:41:56 Which is the most elegant way of sharing a file generated by a recipe with another recipe? Jan 30 15:45:11 malanecora: what kind of file? Jan 30 15:47:17 LetoThe2nd: A pub key Jan 30 15:47:35 LetoThe2nd: Generated by wireless-regdb, and needed by CRDA Jan 30 15:47:50 RP: ah file handles is a good pointer Jan 30 15:48:31 RP: I doubt memory since another machine with same specs but using ubuntu 14.04 instead of ubuntu 18.04 works fine to build world for qemumips/musl Jan 30 15:49:01 and using exact distro and commit Jan 30 15:49:13 so it would not be hardware Jan 30 15:49:35 malanecora: have you looked at the latest patch submissions concerning wireless-regdb? Jan 30 15:49:48 malanecora: plus, that sounds like a generic RDEPENDS case Jan 30 15:51:59 LetoThe2nd: As far as I know there are not patches in the recipe http://cgit.openembedded.org/meta-openembedded/tree/meta-networking/recipes-connectivity/wireless-regdb/wireless-regdb_2018.05.31.bb?h=master Jan 30 15:52:16 LetoThe2nd: The file is needed in compilation time Jan 30 15:52:28 malanecora: then its a DEPENDS Jan 30 15:52:29 LetoThe2nd: By crda Jan 30 15:53:29 LetoThe2nd: If I add wireless-regdb to crd DEPENDS, where would be the file placed? Jan 30 15:54:13 malanecora: then wireless-regdbs install task would run into the sysroot that crda uses at compile time Jan 30 15:58:31 I am trying to build a rootfs with the systemd service wpa_supplicant@wlan0 enabled by default. However it seems that when the opkg postinst script is run, it is adding a symlink by the name of wpa_supplicant@%i.sevice instead of wpa_supplicant@wlan0.sevice. I have unpacked the ipk file and the post install script is correct. I can also install the package on a running system just fine. Any hints for debuging the package installation pr Jan 30 15:58:42 khem: compare ulimit -a from both Jan 30 15:59:24 LetoThe2nd: I'm not sure if I got it, but I'll take a look over that, thank you man Jan 30 16:01:53 I have even mounting the rootfs to /mnt on my host machine and tried running `systemctl enable --root=/mnt wpa_supplicant@wlan0` and it works fine (although my host machine is using systemd 229 and the target is using 230) Jan 30 16:06:39 i'm creating a kernel module recipe folloking meta-skeleton/recipes-kernel/hello-mod, but instead of socking the source (a single .c file) into the files/ folder i'd like to fetch it from a repository Jan 30 16:08:37 so can i just add something like this: https://paste.fedoraproject.org/paste/ycv9txTW0roNYUccVsuGCA Jan 30 16:09:21 i suspect not... Jan 30 16:09:29 btw, still morty Jan 30 16:09:33 still on morty Jan 30 16:36:23 oh, interesting... apparently the systemctrl that is run to build the offline rootfs is a script mimicing the behavior of systemctrl used at runtime Jan 30 16:36:49 hence the difference in behavior Jan 30 16:37:13 yes, systemctl-native is a fake systemctl Jan 30 16:37:21 if systemctl supports offline mode then we should use that instead Jan 30 16:37:26 when that stuff was written, it didn't Jan 30 16:37:42 hm it has a --root option now Jan 30 16:37:55 a systemd-native recipe that built just systemctl would be interesting Jan 30 16:45:25 how do i form a SRC_URI that contains both files from a files/ folder AND a file from a repo? Jan 30 16:45:49 this doesn't work: https://paste.fedoraproject.org/paste/XFCsrBSW6CFH56xZTFHqDQ Jan 30 16:47:01 it fetched the specified repo folder and gzipped it into downloads, but it's not finding rfcontrol.c there Jan 30 16:47:48 why would it find rfcontrol in the repo? Jan 30 16:47:55 if you've already cloned the repo, you have the file Jan 30 16:48:33 cloned? you mean into downloads? Jan 30 16:48:53 if you specify a relative file: url in SRC_URI (and file://Makefile is relative) then it looks in FILESPATHS, ie files/ and bpn/ under the recipe Jan 30 16:49:19 yes, i'm trying to understand why you want to fetch a file from a repository explicitly when you've just fetched the entire repo Jan 30 16:50:49 it was an attempt, rburton, to get rfcontrol.c from the repo and the others from the files/ folder under the recipe Jan 30 16:50:57 apparenly an erroneous one.. Jan 30 16:51:14 that is my goal Jan 30 16:51:28 you *might* be able to do that with svn if you put .../svn/SmartDisplaySoftware/rfcontrol.c Jan 30 16:51:31 in the url Jan 30 16:51:39 as svn itself can fetch single files Jan 30 16:51:47 whether the fetcher supports that, no idea. Jan 30 16:52:07 (or whatever the correct path is) Jan 30 16:52:17 i tried that and the fetcher complained "attempt to fetch file instead of a directory" or somesuch Jan 30 16:52:22 yeah fair enough Jan 30 16:52:54 well remove file://rfcontrol.c and you'll have it in the workdir anyway from the checkout Jan 30 16:53:13 you can make the fetcher drop the other files inthe same place if you want Jan 30 16:53:35 put the svn one first, then file://Makefile;subdir=${WORKDIR}/wherever-the-checkout-went/ Jan 30 16:53:41 will put makefile from the layer into the path you said Jan 30 16:54:05 you can then set S to that path and the standard build rules work too Jan 30 16:55:44 doh! Jan 30 16:56:00 simply dropping file://rfcontrol.c worked! Jan 30 16:56:05 (as you said rburton_) Jan 30 16:56:49 the confusion might be from there being two steps: fetch will checkout the repo to DL_DIR, and unpack will checkout from there to WORKDIR Jan 30 16:57:15 for the central cache and all that Jan 30 16:57:38 svn might just copy but that's the design Jan 30 17:00:59 do the files "file://afile.ext" get copied by the unpacker to the WORKDIR too? Jan 30 17:02:28 rburton: I'm not sure I'm quite to the level of being able to change Yocto over to using the real systemctrl, but I'm working on debugging the fake one right now and getting a patch out at least. Jan 30 17:02:44 baby steps Jan 30 17:07:14 yates: yes Jan 30 19:21:07 Which mailing list should send patches to meta-poky? Jan 30 19:22:18 Err, which to mailing list should I send patches for meta-poky? I don't think the mailing lists themselves will be send many patches :) Jan 30 19:29:03 JPEW: poky@yoctoproject.org Jan 30 19:29:10 Thanks Jan 30 20:51:49 My eudev warns about a missing "kvm" group. eudev/permissions.rules in OE-Core master contains a rule referring to "kvm", but the resulying /etc/groups from base-passwd doesn't contain one. I see others have patched in missing groups, would it be acceptable to add a "kvm" group via a patch here ? Jan 30 21:02:59 is there a variable that defines gui configuration directories such as /usr/share/applications at do_install() time? Jan 30 21:03:44 or at least /usr/share? Jan 30 21:04:47 share is datadir. not sure if applications has its own or not Jan 30 21:05:26 /usr/share will work, thanks kergoth Jan 30 21:06:11 http://git.yoctoproject.org/cgit.cgi/poky/plain/meta/conf/bitbake.conf?h=bernard Jan 30 21:07:20 yates: ${datadir}/applications is the official definition Jan 30 21:07:55 ok Jan 30 21:08:01 you can look those paths up in the relevant spec and see they're defined as DATADIR/applications Jan 30 21:14:10 possibly naive question: is it necessary to mkdir the standard directories such as ${bindir}, ${datadir}, etc under ${D} at do_install() time? Jan 30 21:16:56 if you're going to put files there, yes Jan 30 21:16:59 D is empty when do_install starts Jan 30 21:17:36 i see - good to know! Jan 30 21:20:34 similar for FILES_{PN}: that is a list of files and directories, but do i need to specify all the standard directories like ${bindir} in the list? Jan 30 21:22:01 i noticed that in one of my existing recipes i specify a file ${bindir}/${PN} without specifying the bindir path and it worked, but can i assume this for all "standard" directories? (e.g., /usr/share"? Jan 30 21:25:10 yates: the default is in bitbake.conf Jan 30 21:25:34 i'd definitely recommend reading bitbake.conf, it covers both the default values of variables liek FILES as well as the dfault values of all the target variables, flags, etc Jan 30 21:25:55 gotcha Jan 30 21:26:14 c-x-f ... Jan 30 21:26:15 s/target/target path/ Jan 30 21:28:15 nope, no need to specify those! Jan 30 21:28:36 there's only one thing i like better than yocto Jan 30 21:28:46 ...emacs! Jan 30 21:34:46 what do you guys use? kergoth? LetoThe2nd? RP? rburton_? Jan 30 21:34:54 derRichard: ? Jan 30 21:35:21 I practically live in vim Jan 30 21:35:58 elisp is to emacs as "what" is to vim? Jan 30 21:35:59 well, used to use nano for many years. but for large files the syntax highlighting is slow, therefor i switched to vim Jan 30 21:36:06 yates: you don't want to know ;-) Jan 30 21:36:18 RP: xemacs? :) Jan 30 21:36:26 but i use at most 5 features Jan 30 21:36:37 yates: combinations of gedit and mcedit Jan 30 21:36:47 RP: yay, mcedit! Jan 30 21:37:00 derRichard: I'm a big mc fan Jan 30 21:37:16 yates: that's vim's biggest weakness, viml kind of sucks. thankfully you can write scripts in lua/python/ruby too, though few do for portability reasons :\ Jan 30 21:37:23 RP: i fear i'm too young for mc since i never used norton commander on dos Jan 30 21:37:31 I'm playing around with vscode a lot lately locally, doessns't help when i'm doing remote development over ssh/mosh though Jan 30 21:37:52 derRichard: I didn't use norton commander but I did use Xtree and Xtree Gold extensively Jan 30 21:38:18 derRichard: mc is as close as I can get on linux Jan 30 21:38:23 i tried to use mc, but after getting used to managing files at a commandline the hard way a file manager feels weird and awkward Jan 30 21:38:44 kergoth: I do both depending on which is easier Jan 30 21:39:03 there are a number of interessting new console file managers recently Jan 30 21:39:08 call me a looser, but i like kde's dolphin :) Jan 30 21:39:16 kergoth: mc is nice to visualise things and operating on blocks. You can type shell straight from mc Jan 30 21:39:32 kergoth: there are? I'll have to have a look Jan 30 21:39:38 kergoth: I miss things Xtree used to do Jan 30 21:39:42 ranger, nnn, etc Jan 30 21:39:56 (https://github.com/ranger/ranger, https://github.com/jarun/nnn) Jan 30 21:39:59 nnn is more recent Jan 30 21:40:09 IF i were to consider switching (which i'm not) i would consider microsoft's new (?) Visual Studio Code Jan 30 21:40:18 which is cross-platform Jan 30 21:40:18 i subscirbe to changelog nightly to get daily views into trending and mosst starred github repos Jan 30 21:40:23 find a lot of cool tools that way Jan 30 21:40:37 yates: yeah, a friend of mine is using it on linux to develop python Jan 30 21:40:39 yates: yeah, vscode is surprisingly capable and well done Jan 30 21:40:43 it works really great Jan 30 21:40:52 really vibrant extension ecosystem, too Jan 30 21:41:03 but i still like "bash" as ide more :D Jan 30 21:41:50 https://marketplace.visualstudio.com/items?itemName=auchenberg.vscode-browser-preview for example looks awesome for the webdev folk Jan 30 21:42:01 (a very recent release) Jan 30 21:42:57 RP: nnn seems quite powerful from my admittedly newb perspective. might have to try it out Jan 30 21:43:57 kergoth: yes, just playing. I can see it being quite interesting once you understand the key bindings Jan 30 22:47:34 So.. No one but me who has seen this "missing kvm group" eudev message on boot ? Jan 30 22:52:24 * kroon reboots to see what Arch Linux does **** ENDING LOGGING AT Thu Jan 31 02:59:57 2019