**** BEGIN LOGGING AT Tue Sep 12 03:00:01 2017 Sep 12 07:25:12 hi, I need a way to obtain the list of all bitbake packages that inherit a class Sep 12 07:26:08 grep -r 'inherit.*myclass' seems a bit fragile, and won't catch packages that inherit a subclass of myclass Sep 12 07:51:04 lucaceresoli: bitbake-layers show-recipes -i ? Sep 12 07:53:53 RP: there it is! Thanks! Sep 12 07:56:20 in a bbappend I'd like to override the «PACKAGECONFIG[tiny] = "--with-features=tiny,--with-features=big,,"» to «PACKAGECONFIG[tiny] = "--with-features=tiny,--with-features=normal,,"» Sep 12 07:56:26 is this possible? Sep 12 07:56:27 RP: apparently it was not so easy to find in the docs, but it works fine. Sep 12 08:01:38 BR: actually, as I need to extract the list of those packages in a script, it would be nice if it had a "raw" output mode Sep 12 08:02:20 BR: right now I'm doing bitbake-layers show-recipes -i | grep -v "Parsing recipes" | grep '^[a-z]' Sep 12 08:02:34 BR: and hope the output format won't change in future versions of bitbake Sep 12 08:03:23 RP: but apart from that it's OK Sep 12 08:12:20 lucaceresoli: you could file a bugzilla enhancement request or send a patch! :) Sep 12 08:13:21 sandsmark: you should just be able to set it Sep 12 08:13:49 I tried just setting «PACKAGECONFIG[tiny] = "--with-features=tiny,--with-features=normal,,"» in the .bbappend, but it didn't seem to stick (at least at first glance) Sep 12 08:16:43 Hi, in meta-oracle-java java recipe, there is a RDEPENDS to x11, but in headless environment, it seems to be wrong, no? Sep 12 08:18:41 ah, I'm dumb Sep 12 08:18:42 kinda Sep 12 08:22:25 it worked Sep 12 08:42:23 nefethael: did you mean libxi dependency in oracle-jse.inc ? Sep 12 08:46:48 yes Sep 12 08:47:15 maxin1: yes Sep 12 09:39:32 maxin1: and I'm not sure about the other RDEPENDS as well (gtk+ libxtst mesa... ?) Sep 12 09:41:53 maxin1: other stuff, in oracle-jse-jdk.inc, yocto complains about "ln -sf ${JDK_HOME}/bin/java[c] ${D}${bindir}" being an absolute link Sep 12 09:47:43 nefethael: ok, we should fix it.. Sep 12 09:51:34 maxin1: I'm not really confident about bb recipes for now, otherwise I'd submit patches, damn learning curve :) Sep 12 09:56:18 nefethael: try it :) I think we can go ahead with updating oracle-jse-ejre.inc Sep 12 10:00:44 maxin1: hehe, for rdepends, i'm not sure what to do, maybe move rdepends in specific .inc instead of arms ejre ? Sep 12 10:02:09 maxin1: for link problem, it should be quite easy, i'll fix it if I have time, it's not critical issue Sep 12 10:10:59 nefethael: oracle-jse-ejre inc is used only for headless environments .. Sep 12 10:41:35 hello folks! I am fairly new to yocto i just wanted to check how do we build a sdk which contains all the libs/headers/compiler so that it can be used to cross compile applications Sep 12 10:42:00 #q Sep 12 10:46:30 prabhakarlad: Please refer this document: http://www.yoctoproject.org/docs/2.3.1/sdk-manual/sdk-manual.html#sdk-appendix-obtain Sep 12 10:49:42 prabhakarlad: sed /refer/read/ Sep 12 11:05:49 RP: just applying the .bbappend stuff to the tree - if I want these to only impact on the arm build, I can add _qemuarmv7a (our machine type) to the CONFIGUREOPTS and EXTRA_OECONF assignments, will that also work for do_install and FILES_foo? Sep 12 11:25:14 maxin1: thanks for the link! I am now building the SDK installer :) Sep 12 11:30:54 maxin1: when I build a image (root file system) this has all the headers in it because of which the size is increased can we tell yocto to drop the headers ? Sep 12 11:35:44 don't install them in the first place :) Sep 12 11:35:59 images should only have headers in if you add them to IMAGE_INSTALL, or have dev-pkgs in IMAGE_FEATURES Sep 12 11:36:06 (or if your own recipes are broken) Sep 12 11:45:54 ramcq: that should work, yes Sep 12 11:46:07 ramcq: you'll need to make the anon python code conditional Sep 12 11:46:51 ramcq: its tempting to parametrise the gcc recipes to avoid a lot of this messing around Sep 12 11:46:58 maxin1: RDEPENDS is in global oracle-jse.inc, included by every recipes. We can move it to oracle-jse-jre.inc (not oracle-jse-ejre.inc) ? Sep 12 11:48:00 RP: yeah - I'm also a little confused where in the GCC recipes the tunes get turned in to configure flags, but instinctively it should be possible to make that appliccable to the target recipe as much as it currently applies to like, crosssdk Sep 12 11:49:09 rburton: not sure what exactly you meant I am appending the required packages to IMAGE_INSTALL basically I just need the libs in root file system and not the headers. Sep 12 11:50:29 prabhakarlad: add the libraries to IMAGE_INSTALL, the sdk will contain the headers automatically Sep 12 11:52:09 rburton: root file system will also have those headers ? Sep 12 11:52:15 no Sep 12 11:52:40 assuming your libraries are correctly packages, add the libraries to the image and the sdk will have the headers automatically Sep 12 11:52:59 Anyone seeing / have seen tarball releases from GitHub spontaneously changing content (md5 mismatching)? Sep 12 11:53:16 gunnarx: zip or tar.gz Sep 12 11:53:19 tar.gz Sep 12 11:53:31 if its the archive link then thats not meant to happen Sep 12 11:53:37 url? Sep 12 11:53:45 releases can be statically uploaded, or just a git-archive from a tag Sep 12 11:53:57 of course moving the tag will result in a new md5 Sep 12 11:53:58 I don't mean the yocto releases, it's a recipe in our build Sep 12 11:54:09 yeah i know Sep 12 11:54:18 so I'm just asking if it's been seen on any other GitHub or if it's a one-off Sep 12 11:54:41 archive links which are tar.gz are meant to be static forever and github claim they are, but there's always been thought that they're not Sep 12 11:54:57 which is why i'm asking for the URL and if you have the new/old md5 Sep 12 11:55:06 sure, I'll give the refs Sep 12 11:55:12 could be a one-off Sep 12 11:55:17 (i can restart a thread with the github engineer responsible for) Sep 12 11:55:30 ... for telling me not to worry) Sep 12 11:56:09 It's an erlang package, I've prepared talking to them too. Just wanted to check if this is widespread or not, cause as you indicate I'm not sure how GitHub handles this -- do they generate tarballs on the fly, or have an archive... Sep 12 11:56:14 hang on.. Sep 12 11:57:30 https://github.com/erlang/otp/archive/OTP-18.2.3.tar.gz and this recipe: https://github.com/joaohf/meta-erlang/blob/master/recipes-devtools/erlang/erlang-18.2.3.inc indicates the MD5 it *should* have (or had) Sep 12 11:57:34 *if* its a tarball from a tag then they git-archive and cache Sep 12 11:58:01 * gunnarx checking tags Sep 12 11:59:09 it's from a tag afaict Sep 12 11:59:13 yes it is Sep 12 11:59:18 ramcq: the tune flags don't get built into the compiler, we always pass in the right compiler flags on the cross compiler commandline Sep 12 11:59:36 gunnarx: presumably their cache expires so they could have regenerated the tarball Sep 12 11:59:51 and its an old release so presumably they didn't move the tag Sep 12 12:00:05 rburton: when I am adding libraries to IMAGE_INSTALL, I am still seeing the headers installed [1] https://www.pastiebin.com/59b7cc14d3009 Sep 12 12:00:10 and with some other tar version that actually modifies some bytes!? oh no Sep 12 12:00:36 if that's true it is no bueno, GitHub... Sep 12 12:00:38 gunnarx: presumably you're seeing a md5 of 7fe... Sep 12 12:01:14 on the new one yes. I have the old one in our cache, with the 2367 md5 Sep 12 12:02:19 I found this on a new build machine where the download cache had been wiped Sep 12 12:02:41 now where was that discussion with github... Sep 12 12:03:09 that's why I came here :) I'm hoping someone has some insight on this... Sep 12 12:03:32 there's long been suspicion that the tarballs if regenerated change Sep 12 12:03:55 maybe not if you just run git-archive repeatedly, but a new gzip could change Sep 12 12:04:11 BKM has been to not use the archive links, but do a git clone if the maintainers don't upload static tarballs Sep 12 12:04:16 RP: OHHHHHHHHHHHHHHHHhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhh Sep 12 12:04:24 RP: that makes so much more sense now Sep 12 12:04:40 so there isn't any thing that turns tune flags into gcc config options Sep 12 12:04:49 you just build an arch compiler and drive the rest with CFLAGS of doom Sep 12 12:04:52 right Sep 12 12:05:05 rburton, so there's a way to see on GitHub if it's a static tarball or a generated one? Sep 12 12:05:37 Hello Sep 12 12:05:45 gunnarx: https://github.com/erlang/otp/releases?after=OTP-18.3 shows just a tag sha and zip and tar.gz links, which implies its generated Sep 12 12:05:51 gunnarx: are you on twitter? Sep 12 12:06:05 I'm created a recipe that provides a lib let's say XX.so.1.3.0; I made another recipe that depends & rdepends on it Sep 12 12:06:10 not really. I've got an account somewhere :) Sep 12 12:06:22 when I'm building the image, the rootfs task fails saying it cannot find libXX.so Sep 12 12:06:31 though all the .so are provided in XX and XX-dev Sep 12 12:06:48 I don't really understand why it's failing to find that file Sep 12 12:07:06 You proposing I should tweet out nasty things about GitHub on this topic? :) Sep 12 12:07:23 gunnarx: no i was going to cc you on my message to the tech lead ;) Sep 12 12:07:40 gunnarx: https://twitter.com/bkeepers/status/645951267961896960 Sep 12 12:07:43 I'd like to avoid that - we have a good relationship with GitHub. But this is not a good thing. Sep 12 12:07:45 ramcq: our cross gcc explodes if it doesn't get certain flags so for our build environment, we usually know if the flags go missing Sep 12 12:08:45 gunnarx: my recommendation is to change the src_uri to use git: if they won't upload proper tarballs somewhere Sep 12 12:09:03 ok cool, thanks rburton. I'll try to follow the twitter thread. Sep 12 12:09:14 any thoughts ? Sep 12 12:11:43 rburton: query regarding those verilog/fpga recipes. Just wanted to query why you think it would be ideal as a separate layer? since its only a few recipes. Just on the fence as to whether to create/put them in a new layer Sep 12 12:12:05 nrossi: because i dislike meta-oe as a dumping ground Sep 12 12:12:35 gunnarx: can you get a timestamp from the tarball on your mirror? Sep 12 12:12:53 yeah, let me check. Should be on a few different build agents actually Sep 12 12:18:17 rburton: fair enough, that is one of the reasons why i was not sure where it should go. Khem suggested oe-core, but i guess its not quite core enough to really go there either. I wanted to avoid making a 'fpga' layer since it very close to BSP, but i guess doing it as a language layer (like go/rust/etc) would be much saner (e.g. meta-hdl) Sep 12 12:19:56 gunnarx: github tech lead said mail support@, just have. Sep 12 12:20:05 i'll let you know if anything happens Sep 12 12:20:15 yeah, hang on I'm collecting up better info. I already had a draft text Sep 12 12:20:28 I'll pastebin it and reference it for your. Sep 12 12:24:39 gunnarx: rburton: I've seen this at least for 10 different archives over last couple years, that's why I always recommend to use git:// fetcher Sep 12 12:26:25 gunnarx: rburton: when I've compared the contents of the tarballs it was identical (and the tag seemed old enough) so I don't assume it was moved, just regenerated Sep 12 12:32:21 gunnarx: rburton: e.g. https://github.com/williamh/dotconf/releases/tag/v1.3 used to have b395edef5ac957a931f09e618f58f077 md5 and then a81eee25aa15dde23d69381571031277 Sep 12 12:32:48 JaMa, no need to do that I have done it also. Sep 12 12:32:58 gunnarx: ? Sep 12 12:33:34 sorry, thought you had compared the content of "my" tarballs, misunderstanding? Sep 12 12:34:48 I don't have old version of your tarball Sep 12 12:53:23 Hi, I'm working on a project with bitbake from scratch - is there a way to control the dependencies better than defining rules like do_build[depends]="dependency:do_populate_filesystem" ? Sep 12 12:55:00 JaMa, no, of course you don't, never mind my rambling Sep 12 13:15:50 heyho, I have a small third party library I want to use in a recipe. The source doesnt have any build system support. Sep 12 13:16:30 "${CC} -Iinclude src/csvparser.c -fPIC -shared -Wl,-soname,libcsvparser.so.1 -o libcsvparser.so.1.0.0" produces QA issue: -dev package contains non-symlink .so: cccsvparser-dev Sep 12 13:16:37 Any idea how I can solve that? Sep 12 13:22:04 spooky_d: that's exactly how to do it. there's syntactic sugar in DEPENDS but that may not work how you want Sep 12 13:28:13 Ok, so I'll have to find a smart way to automatize this for my needs Sep 12 13:28:17 thank you for confirmation! Sep 12 13:28:40 wanted to make sure I don't reimplement the wheel (again) Sep 12 13:29:21 spooky_d: DEPENDS is magic that basically sets this:do_configure -> other:do_populate_sysroot. should be simple enough to replicate in a core class... Sep 12 13:33:45 spooky_d: good news: I just figured out how the DEPENDS magic actually works, and its trivial for you to use. Sep 12 13:34:19 spooky_d: do_prepare_recipe_sysroot[deptask] = "do_populate_sysroot" <-- do_prepare_recipe_sysroot has a dependency to do_populate_sysroot automatically added for every entry in DEPENDS Sep 12 13:34:48 spooky_d: so you just need do_build[deptask] = "do_populate_filesystem" and then set DEPENDS as appropriate Sep 12 13:35:10 (do the deptask thing in a common class once, the recipes just need DEPENDS= Sep 12 13:36:16 gunnarx: when github reply to my mail i'll forward your pastebin Sep 12 13:39:12 Uh, that's awesome, thanks! :) Sep 12 13:47:07 no ideas? Sep 12 13:54:42 FabKna, make the .so being a symlink to the actual library? Sep 12 13:55:04 foo.so points to foo.so.1 that points to foo.so.1.2 Sep 12 13:55:07 done Sep 12 13:55:29 ok ty, will try that. Sep 12 13:58:00 yw Sep 12 13:59:51 one more question: I have foo.so and foo.so.1.0.0 after compiling. Which of them I have to install? Sep 12 13:59:58 Sorry, im new to versioning Sep 12 14:00:13 packages-split does this for you Sep 12 14:00:32 .so goes in dev (the symlink is used by gcc) and the library goes to foo, used at runtime Sep 12 14:02:06 so in dev is only a symlink? Sep 12 14:02:42 and headers Sep 12 14:08:02 ok, that worked. However, I get "QA Issue: No GNU_HASH in the elf binary". Sep 12 14:08:11 Relates to https://lists.yoctoproject.org/pipermail/yocto/2012-August/008898.html Sep 12 14:08:33 so in dev package is not good? Sep 12 14:09:20 FabKna, you shoudln't override CFLAGS/CXXFLAGS/CPPFLAGS/LDFLAGS and so on Sep 12 14:09:22 check the Makefile Sep 12 14:13:54 I have no Makefile Sep 12 14:32:05 TARGET_CC_ARCH += "${LDFLAGS}" worked :) Sep 12 14:42:41 Hi all, I need your advice for the best way to remove a file from the rootfs installed by a package in dependency Sep 12 14:43:22 hostapd install /etc/init.D/S20hostapd and /etc/hostapd.conf Sep 12 14:44:09 do you want to edit the conf file? Sep 12 14:44:21 and my recipe overwrite /etc/hostapd.conf (fixed by using DDEPENDS += "hostapd") and has its own init script S80caminspector Sep 12 14:44:35 I do this with a bbappend Sep 12 14:44:40 I would remove S20hostapd Sep 12 14:46:21 ok, you mean a hostapd.bbappend with rm /etc/S20hostapd in do_install , right ? Sep 12 14:48:43 You've quit, but tks anyway Sep 12 14:49:13 janho: why do you want to remove hostapd autolaunch ? Sep 12 14:51:09 because I get a rootfs from someone who already has its own init scripts. Sep 12 14:51:49 and I install this rootfs in one step Sep 12 14:52:21 not really a rootfs, sorry , but a lot of files : configuration , executables and medias Sep 12 14:52:46 janho: and that script can't be modified ? Sep 12 14:53:03 I prefer not Sep 12 14:54:10 If I modify its scripts , I'll need to modify them everytime he delivers me a new pack of files. Sep 12 14:55:39 janho: for me it's a bit counterintuitive that hostapd is not launch by its own launch script, FabKna solution is ok, but for me, i'd choose to make a do_install_append() with some sed or custom script Sep 12 14:59:33 nefethael, ok. However what is difference with the bbappend ? Don't I need to write a hostapd.bbappend file to write do_install_append inside ? Sep 12 15:00:10 I have my own layer and image for this target Sep 12 15:01:49 janho: for me, it would be a caminspector.bb modifying S80caminspector Sep 12 15:03:35 It won't be enough, because S20hostapd is definitively not the way I need to launch hostapd. I think I 'll try to remove the file in a hostapd.bbappend file. tks Sep 12 15:26:25 I was wondering how people do to add "image-specific" configuration files ? ROOTFS_POSTPROCESS_COMMAND? I see this patch can do what I need, but what is it the best solution ? https://patchwork.openembedded.org/patch/138100/ Sep 12 15:27:34 nefethael: using that class seems like it would be fine Sep 12 15:35:09 rburton: thanks, I'll give it a try :) Sep 12 15:53:58 wohoo, my trainee's job is actually used :) Sep 12 15:54:15 (we use it a lot internally, but it's great to see it used elsewhere... Sep 12 16:21:36 boucman_work: talking about ? :) Sep 12 17:37:52 Can I delete the git2 directory to free up space? Sep 12 17:38:28 if you don't mind it re-downloading the git repositoires later if needed, sure Sep 12 17:38:39 [build]/downloads/git2 Sep 12 17:39:03 kergoth: ok, i think i may move to another machine, so should be fine Sep 12 18:31:01 fray / zeddii: question for you about LSB and linux-libc-headers Sep 12 18:44:43 LocutusOfBorg: Are you here? :) Sep 12 18:48:03 I have a recipe created a shared library from source with ${CC}. I now try to add pkgconfig. To do so I added a .pc file and inherit pkgconfig. However, my packages-split does not containt a pkgconfig folder / file. Did I forgot sth.? Sep 12 19:21:34 FabKna: did you actually install it? Sep 12 19:22:00 inherit pkgconfig is only for recipes that *use* pkgconfig to find libraries Sep 12 19:23:02 ahh ok, I understand. I have done it now without inherit pkgconfig. So I created dir ${libdir}/pkgconfig and installed my .pc file. Sep 12 19:23:08 It seems working Sep 12 19:25:05 Is there a way to publish recipes for openembedded so that other can profit from it too? Sep 12 19:25:28 either submit them to a suitable layer, or make your own layer and submit it to layers.openembedded.org Sep 12 19:25:44 k ty Sep 12 19:25:48 ie if you wrote a recipe for a python module, adding it to meta-python would be good Sep 12 20:55:55 halstead: btw, the issue last week with git mirror was due to us using secondary yocto-project.org address since the beginning of time... it worked for 5-6 years until last week. Sep 12 20:56:16 halstead: I know that preferred primary address is yoctoproject.org and it works fine, I don't remember why we used the other secondary one back then... Sep 12 20:58:19 denix, Thank you very much for the report. Can you confirm it's using git.yoctoproject.org and not plain yoctoproject.org ? Sep 12 21:01:54 halstead: looks like git remote is set to just git@yoctoproject.org:meta-ti - should we change to git@git.yoctoproject.org? Sep 12 21:03:12 denix, Yes please. The subdomain is tested to work the other may not always work in the future. Sep 12 21:03:38 denix, Is it difficult to update? Sep 12 21:04:38 halstead: no, not difficult - will change shortly Sep 12 21:05:55 denix, We can eventually speed up read-only git ops if we copy what kernel.org does with a dedicated push url. Sep 12 21:06:39 halstead: I see. so that would be on a dedicated git.* subdomain? Sep 12 21:06:59 denix, kernel.org uses ra.kernel.org as the push url. Sep 12 21:07:41 denix, We now have subdomains for push.yoctoproject.org and ra.yoctoproject.org. If you are updating maybe we can take the opportunity to use one of those. Sep 12 21:08:16 denix, Does dig ra.yoctoproject.org give you back an A record? Sep 12 21:08:24 halstead: sure, please let me know which one to use to push to Sep 12 21:08:50 halstead: looks like it Sep 12 21:09:22 denix, You can be the first. Yes let use git@ra.yoctoproject.org. You should be set for another 5 years. Sep 12 21:09:38 I have a draft about it for the list. Sep 12 21:10:13 :) Sep 12 21:10:19 halstead: thanks, will do! Sep 12 21:10:39 Thank you denix :) Sep 12 22:09:45 hello everyone Sep 12 22:20:11 I have a cyber security Virtual Appliance , which has some python web project located at /usr/local/project_name/ Sep 12 22:20:30 this location has the web server script and i can see it running by listing the processes Sep 12 22:20:41 but when i change directory into that location , i can't Sep 12 22:21:07 how can disallow even root user from accessing a specific directory but let the kernel has access to it only Sep 12 22:21:25 i really don't know how to explain that but i am sure of what i saw :) Sep 12 22:21:59 sorry /opt/project_name Sep 12 22:22:26 how can i do the same Linux for my Next project Sep 12 22:23:06 might be in a mount namespace, check the mount namespaces of the webserver process Sep 12 22:25:25 neverpanic: how can i check the mount namespace Sep 12 22:29:18 snouto: https://unix.stackexchange.com/questions/113530/how-to-find-out-namespace-of-a-particular-process Sep 12 22:30:50 neverpanic: Thank you Sep 12 22:53:58 zeddi: what's the 'taging' file you added to 4.4 in 9e5bc767? **** BEGIN LOGGING AT Tue Sep 12 23:44:57 2017 Sep 12 23:53:05 https://www.armis.com/blueborne/ Sep 12 23:53:42 go get your phone updates Sep 13 01:48:52 hi, is there any one know something about Sierra Wireless stuff? Sep 13 01:49:04 I'd like to build image without legato stuff **** ENDING LOGGING AT Wed Sep 13 03:00:00 2017