**** BEGIN LOGGING AT Mon Jun 24 02:59:57 2019 Jun 24 05:59:26 Hi I using yocto Morty (2.2) yocto release where I am building custom images. I am trying to mask set of recipes using BBMASK variable. If the BBMASK entry is entered in conf/local.conf it works fine of running it. But, if entry is made at a layer's conf file (conf/layers.conf) where recipe reside still the image building picks up the Jun 24 05:59:27 recipes which is being masked. I want to have permanent changes in my layer conf file instead of at conf/local.conf. Please give inputs. Thanks Jun 24 06:03:19 Manjunatha: did you try inspecting "bitbake -e" to see how BBMASK ended up after bitbake finished parsing all conf files? Jun 24 06:04:22 If there's a mistake in how things are appended etc you will usually find it by looking at the end result Jun 24 06:04:36 i'd just add that hiding such in a layer.conf file is highly discouraged, as it is not expected there by convention. henvce it will confuse about every user, including your future self. Jun 24 06:32:10 erbo: With O/P of "bitbake -e" doesn't add recipe to be appended when added at layer's layer.conf. With creation of image.bbappend, which appends necessary recipe at the end. With effect of adding at local.conf, which appended at middle of , and later some other recipes are added. Jun 24 06:33:10 From above adding at image.bbappend which still fails in masking the recipes. Jun 24 06:34:44 Manjunatha: erm. you can't do things that affect other recipes inside a recipe, or by appending to it. Jun 24 06:35:18 Manjunatha: so if you add BBMASK to the image, then it will be visible only during the build of that particular image, and all other recipes won't care for it. Jun 24 06:35:40 Manjunatha: bitbake rule #1: recipe data is local, conf data is global. Jun 24 07:00:48 Is there a way to not fail when a meta-layer is not there on a package which is not enabled? ex: I have an application, based on qt5, but even if I don't use it, it fails when meta-qt5 is not added Jun 24 07:03:36 woutervh: you can certainly hack up something, but i'd rather suggest to split your layer then. Jun 24 07:03:55 woutervh: into one that bears the qt dependency, and one that does not bear it Jun 24 07:06:18 good morning Jun 24 07:08:07 Following recipes to be masked for the image build. Is possible to have entries as below in image.bb/image.bbappend, as BBMASK_${PN} BBMASK_$PN} .= "|(recipes-tpm2/tpm2-abrmd/tpm2-abrmd_2.1.1.bb)" BBMASK${PN} .="|(recipes-tpm2/tpm2-tools/tpm2-tools_3.1.4.bb)" BBMASK_${PN} .="|(recipes-tpm2/tpm2-tss/tpm2-tss_2.2.3.bb)" BBMASK${PN} .="|(recipes-t Jun 24 07:08:08 pm2/tpm2-tss-engine/tpm2-tss-engine_1.0.0.bb)"Just want image.bb to ignore these. Jun 24 07:08:27 Manjunatha: erm. you can't do things that affect other recipes inside a recipe, or by appending to it. Jun 24 07:08:30 Manjunatha: so if you add BBMASK to the image, then it will be visible only during the build of that particular image, and all other recipes won't care for it. Jun 24 07:08:34 Manjunatha: bitbake rule #1: recipe data is local, conf data is global. Jun 24 07:08:43 Manjunatha: already told you 30minutes ago. Jun 24 07:16:23 LetoThe2nd, Ok, I see what you mean Jun 24 07:16:59 woutervh: it would be my approach, given the additional fact that nowadays layers can specify what they depend on Jun 24 07:17:55 woutervh: https://www.yoctoproject.org/docs/2.8/ref-manual/ref-manual.html#var-LAYERDEPENDS Jun 24 07:18:48 woutervh: There's also this way, where you only parse some recipes under a certain path if that layer is present: https://github.com/Pelagicore/meta-pelux/blob/7b9afbe64cb26063fe2be4c7cb844cc38b58a571/conf/layer.conf#L18 Jun 24 07:19:03 LetoThe2nd:I understand it. Please let me know any other way to mask these recipes during this particular image (image.bb) build other than masking from local.conf globally. My problem is, the common couple of PROVIDES from different recipes conflicts giving error as "Multiple provides" from 2 different recipes. On dependent recipe is tpm2-tss_2.2 Jun 24 07:19:04 .3.bb which is newly added which I want image.bb to avoid it. I want to have permanent changes rather than at local.conf. Jun 24 07:19:12 It's a bit more hackish, but works well if you want to stick with one layer Jun 24 07:20:24 Manjunatha: you could provide template configs with BBMASK correctly set up with your layer, and then set TEMPLATECONF when sourcing the oe-init-build-env script Jun 24 07:22:46 Thanks for inputs. Iam looking this. Jun 24 07:27:16 erbo, cool, that was exactly what I was thinking about Jun 24 07:34:14 woutervh: be aware that such will confuse the s**t out of anybody who tries to use it, including your future self. nobody expects functionality to be hidden in layer.conf Jun 24 07:34:33 Manjunatha: sounds like you actually want a virtual provider Jun 24 09:29:04 hello all Jun 24 09:29:31 is it possible to specify that BB_GENERATE_MIRROR_TARBALLS only saves the last tree, instead of the entire historie's trees? Jun 24 09:31:23 if i want to distribute those mirrors for people to recompile my system, I'm supplying lots of unused commit history. so therefore I want to truncate commit history to the last commit Jun 24 09:31:38 (without changing the commit ID of course, so that recipe files continue to work!) Jun 24 09:37:38 litb: You could try BB_GENERATE_SHALLOW_TARBALLS. I've not tested it myself though Jun 24 10:01:46 Hi guys, could someone help me to update a 'chromium' recipe to build my image? F.e., currently, my 'yocto' based on a 'pyro' branch. I have added the 'meta-browser' layer which has a recipe with the chromium v52.. But, problem is that that version is too old, besides, I can't upgrade the 'yocto' layers too. So, I have planns to use the 'chromium' from the 'master' branch of meta-browser layer. So, I have switched the meta-browser branch to master. Also I Jun 24 10:01:46 have created my own recipes directory /recipes-browser/chromium, where I have copied the 'chromium' recipe from the newest 'meta-freescale' layer branch (not from old 'pyro' branch). But a problem is that that 'meta-freescale' layer does not contain the .bbappend file. So, I have created the 'chromium-x11_75.03770.80.bbappend' file which respects the chromium-x11_75.03770.80.bb' file from the 'meta-browser' layer. But, when I run bitbake, it say that: Jun 24 10:01:49 ERROR: No recipes available for: ...//recipes-browser/chromium/chromium-x11_75.03770.80.bbappend... Jun 24 10:03:03 why it happens? Jun 24 10:04:09 are you sure you have the meta-browser layer in your bblayers? Jun 24 10:05:52 qschulz: Yes, the command: bitbake-layers show-layers show the 'meta-browser' layer. Jun 24 10:07:00 When I did remove my .bbappend file, then bb say that ERROR: Nothing RPROVIDES 'chromium' Jun 24 10:09:00 paulbarker, thanks, it works! and I can override it on a per-recipe base. Jun 24 10:18:34 Why bitbake don't see the 'chromium' recipe? Even the 'meta-browser' layer has been added.. Jun 24 10:25:31 Hmm. it does work when I add to the IMAGE_INSTALL: chromium-x11 instead of chromium.. Jun 24 10:29:51 chromium-ozone-wayland keeps on giving me: collect2: fatal error: ld terminated with signal 9 [Killed] Jun 24 10:29:54 what does it mean? Jun 24 10:50:26 woutervh: The linker crashed. Check you kernel & system logs to see if there's any info there, you might be running out of RAM or something Jun 24 10:58:35 anyone in here use fortran with OE? Jun 24 11:10:31 RP: recipes that throw skiprecipe are silently removed from world builds right Jun 24 11:10:46 a note, but nothing fatal Jun 24 11:25:07 paulbarker, indeed, out of memory. I now added some swap space. Hope it gets through Jun 24 12:05:50 hi! has anybody been successful building cmake-native on a recent fedora machine? The host glibc version is higher than the toolchains. That somehow interferes with the bootstrapping(configure stage). Iam just trying something out at home as a side project, so i dont want to reinstall ubuntu :/ Jun 24 12:05:51 "poky/build/tmp/sysroots-uninative/x86_64-linux/usr/lib/libstdc++.so.6: version `GLIBCXX_3.4.26' not found (required by poky/build/tmp/work/x86_64-linux/cmake-native/3.14.1-r0/build/Bootstrap.cmk/cmake)" Jun 24 12:09:12 there's been an update to uninative in the last few days to fix a problem with latest fedora Jun 24 12:09:18 so maybe try updating? Jun 24 12:11:59 rburton: thank you! will try that Jun 24 13:05:30 rburton: yes. You might want virtual/gzip for what I think you're trying Jun 24 13:21:09 Does anyone know of an existing external Qt5 plugin built using Yocto? I am looking for a BitBake recipe to build an external (platform device integration) plugin. Jun 24 14:06:50 ^ I figured it out. Jun 24 14:41:47 likewise: cool, please share the answer for reference as this chan is archived :) Jun 24 14:42:44 New news from stackoverflow: How to access bitbake config files variables in CMake? Jun 24 15:03:26 What generates the qemuboot.conf file? Jun 24 15:04:58 nm, think I found it Jun 24 15:05:05 jonmason: qemuboot.bbclass Jun 24 15:05:11 yup Jun 24 15:05:25 jonmason: just looking at your patch, I don't think you quite understood what I meant Jun 24 15:05:47 jonmason: also found the core-image-cmdline-full failure is a problem with the actual image :( Jun 24 15:05:56 RP: I think I'll have the virtio serial stuff working sometime today, but looks like I'll have to rewrite a chunk of the runqemu script Jun 24 15:05:56 basically the race I worried about is real Jun 24 15:06:42 RP: the "test_" was causing it to be called like a test, right? Jun 24 15:07:00 jonmason: yes, that bit is fine, I hadn't meant to add the systemd test though Jun 24 15:07:04 and while I was messing around, I thought I'd cover all of the bases Jun 24 15:07:27 jonmason: I guess its a reasonable test of systemd though Jun 24 15:07:47 jonmason: I wasn't keen on the code duplication :/ Jun 24 15:07:57 yeah, its not very elegant Jun 24 15:08:20 I just couldn't think of a better way to do it Jun 24 15:08:31 jonmason: I have a local patch which handles rsyslog more neatly but I then found the image really is broken Jun 24 15:08:55 the problem is that there are a number of the syslogs have the same binary names, except for rsyslog and systemd Jun 24 15:14:56 anyone have a technique to generate a list of files from a package list? Jun 24 15:17:21 Specifically, the files that will be installed in the rootfs? Jun 24 15:21:34 on warrior I'm hit by curl-native (used as recipe dependency) trying to use ca-certificates.crt from .../curl-native/7.64.0-r0.shadow/recipe-sysroot-native/etc/ssl/certs/. When it is called by a third-party script it's cumbersome to add --cacert by hand. Would a patch to curl to use a wrapper script to add --cacert be considered a good idea ? Jun 24 15:31:22 Hi guys, why bitbake does not say that I have a two providers of 'foo' recipe? e.g. the meta-oe layer contains an own foo-v1 recipe.. But I too have added the new foo-v99 recipe (which has a newest version)... But the bitbake takes the newest recipe foo-v99... Why the bitbake does not warning me about the two recipes? Jun 24 15:33:11 darknighte: a bit of sh using oe-pkgdata-util should be able to knock that up easily Jun 24 15:33:36 kuzulis: because two recipes with the same name and different versions isn't considered a problem Jun 24 15:36:16 rburton: ok, thanks Jun 24 15:40:36 rburton: so, I even can upgrade, e.g. the GCC, just to copy a recipe from the newest layer to the old layer? Jun 24 15:41:25 upgrading gcc is more than just copying a recipe but in the general case yes Jun 24 15:41:38 ok, thx Jun 24 15:41:42 obviously, it's your responsibility to also copy across any relevant changes Jun 24 15:42:01 you can't just cp gcc/ from master to a four year old release branch Jun 24 15:42:24 ok Jun 24 15:43:19 but, at least between GCC 6.4 and 7.3 is it possible? Jun 24 15:43:40 depends if gcc 7.3 introduced compile warnings and errors that 6.4 didn't so now e.g. glibc doesn't build Jun 24 15:44:00 ahh.. clear Jun 24 15:48:19 hi - is it possible to disable a postinstall-intercept hook? I don't have any gui so gdk-pixbuf is not built, but update_icon_cache requires it so the rootfs build is broken (distro based on thud) Jun 24 15:48:41 can't find anything looking at the manual, but maybe I'm missing something obvious Jun 24 15:56:34 bluca: what is calling update-icon-cache? Jun 24 15:56:53 (that missing dependency is fixed in master, iirc) Jun 24 16:01:46 if I understand correctly it's called because it's a postinst-intercepts script Jun 24 16:02:17 so do_rootfs calls it and all the others Jun 24 16:02:56 do you remember vaguely what the change was/which directory/commit title so I can look it up? Jun 24 16:13:23 bluca: sure but that only runs if a specific recipe asks for it Jun 24 16:13:39 bluca: so your image with no gui is pulling in a package that wants to update the icon cache Jun 24 16:13:50 therefore, maybe you can just remove that instead Jun 24 16:13:54 ok - any idea how to find out which? it's not in any of the locally defined package lists or dependencies Jun 24 16:14:22 so it's probably a dep of a dep Jun 24 16:14:52 tried running bitbake -g but gdk-pixbuf was not mentioned Jun 24 16:18:27 it won't be, it would be an icon theme most likely Jun 24 16:23:10 grepping for icon/theme/desktop in the recipes_depends.dot returns nothing, any idea what they could be called? any convention? Jun 24 16:24:07 so the only thing to use update_icon_cache is gtk-icon-cache.bbclass, so $ bitbake-layers show-recipes -i gtk-icon-cache Jun 24 16:30:19 rburton: thanks. I forgot about oe-pkgdata-util. Jun 24 16:31:16 thanks - that command lists: Jun 24 16:31:19 adwaita-icon-theme atkmm blueman connman-gnome esound fltk gcr geany-plugins gimp gnome-themes-standard pavucontrol vte9 Jun 24 16:31:48 but none of these are in the recipes-depends.dot unfortunately Jun 24 16:33:29 khem: what do you think about bumping the glibc 2.28 srcrev to tip of the branch for thud? Jun 24 16:33:58 looking at a number of CVEs fixed in the branch that we don't have patches for yet Jun 24 17:12:35 rburton: I dont see an issue, there are only bug fixes there Jun 24 17:39:43 khem: gwan do it Jun 24 18:26:23 rburton: w.r.t. efibootmgr I built with clang Jun 24 18:41:38 ah maybe thats the difference Jun 24 18:41:52 i do wonder *why* though Jun 24 18:50:29 rburton: it seems certainly a bug in efibootmgr detection mechanism for efivar lib+headser Jun 24 19:31:09 khem: i really would appreciate a glibc bump for the stable branches to pick up all the cve pieces in there :) Jun 24 21:36:29 rburton, seems reasonable Jun 24 21:44:11 <__ad> hi, in config i have a PREFERRED_VERSION_linux-xxx="4.9.75" . Is there a way to change version building, on the fly, without changing the config file ? Jun 25 02:09:16 __ad:yes you can add another override something like PREFERRED_VERSION_linx-xxx_forcevariable = "x.y.z **** ENDING LOGGING AT Tue Jun 25 03:00:38 2019