**** BEGIN LOGGING AT Wed Feb 20 02:59:57 2019 Feb 20 07:54:16 hi Feb 20 07:55:05 I am back as oe/poky for a smal project and need to figure out how to replace a git repo with a .zip file that was delivered to me. I remember doing bitbake and looking at the logs years ago to figure out the magic file name I need to use but that knowledge is long gone. also the mirror howto faq does not engligtent me. Feb 20 07:55:17 can I get some help in getting the magic right Feb 20 07:56:30 (I also vaguely remember doing something like bitbake bla build-env to get a shell for faster iteration Feb 20 07:56:44 keesj: probably you need to create a recipe (.bb) Feb 20 07:58:06 keesj: https://www.yoctoproject.org/docs/current/dev-manual/dev-manual.html#new-recipe-writing-a-new-recipe Feb 20 07:58:40 Ik think I need some help in https://www.yoctoproject.org/docs/2.3/ref-manual/ref-manual.html#source-mirrors Feb 20 08:37:52 Morning guys! Feb 20 08:38:40 Is there a way of bitbaking a specific version of a package directly from command line (without modify any file)? Feb 20 08:39:08 Something like: PREFERRED_VERSION="1.2.3" bitbake package_name Feb 20 08:40:38 malanecora: whats the use case? Feb 20 08:41:09 LetoThe2nd: I just want to build a rpm with an older version Feb 20 08:41:24 LetoThe2nd: For a system with old FW Feb 20 08:41:38 (rocko vs thud) Feb 20 08:42:17 malanecora: hint: about all requests that include "i just want to" are pretty pointless and not well thought through Feb 20 08:42:41 malanecora: in your case, how should it work if in a newer release there are no recipes for the old version anymore? Feb 20 08:42:57 malanecora: the recipes are version-bound, and older versions do not get carried along Feb 20 08:44:58 LetoThe2nd: In this case that expression has a literal meaning Feb 20 08:45:54 Hello, I have a specific problem, but I'm looking for a generic answer. I want to add my custom qt application to my build, in my custom layer. I always have a hard time understanding which part of meta-qt i should copy into my layer Feb 20 08:46:12 malanecora: hehe, about everybody who says "i just want to" means it in the literal way. but it almost never works like that. Feb 20 08:46:28 willie: hum, hopefully none? Feb 20 08:47:03 So, just create a "recipes-qt" and add myapplication.bb Feb 20 08:47:07 LetoThe2nd: Since it is for an old FW based system (and already deployed), I guess I'd to upgrade it Feb 20 08:47:35 willie: your layer basically depends on meta-qt, and your recipe probably does an "inherit qmake5" Feb 20 08:47:46 LetoThe2nd: But as a last resort in the current phase of the project Feb 20 08:47:52 malanecora: thats all nice, but its not how things work. Feb 20 08:48:13 malanecora: they work like: check out the metadata of the deployed state, modify to your needs, then build. Feb 20 08:49:09 willie: so if you're not doing any super tricky and stick to the common ground of qt build processes, that should be fine Feb 20 08:49:28 willie: here's some examples too: https://github.com/meta-qt5/meta-qt5/tree/master/recipes-qt/examples Feb 20 08:49:33 LettoThe2nd: Right now my layer does nothing :> I'm looking in the meta-qt5 trying to figure out how to append my application. Feb 20 08:49:48 LetoThe2nd: That's messy Feb 20 08:50:09 malanecora: no Feb 20 08:50:39 malanecora: messy is "hey i just do this magic incantation of bitbake and then i ship this binary package that nobody can reproduce where it came from" Feb 20 08:50:48 LetoThe2nd: It's not a production pacakge, just monitoring one for the test phase Feb 20 08:52:01 hi. how can I change the 'sysroot' that the compiler uses? Feb 20 08:52:07 malanecora: well you can always use devtool to create ad-hoc packages. but in that case its like, "hey i'm doing this for dev, and what happens around me, i don't give a f***" Feb 20 08:52:24 malanecora: in that case, look at devool -h and devtool modify Feb 20 08:53:11 LetoThe2nd: I'll take a look over! Ty Feb 20 08:53:43 LetoThe2nd: So basically, the .inc file "creates" a lib/program and as long as the .bb file has "inherit xx" it uses said program? Trying to understand the basics Feb 20 08:54:03 willie: erm, no Feb 20 08:54:09 willie: look at https://github.com/meta-qt5/meta-qt5/blob/master/recipes-qt/examples/qt5ledscreen_1.0.bb Feb 20 08:54:42 willie: this is what your recipe should look like for a generic qml-based application Feb 20 08:54:56 willie: you can the DEPENDS (for building) and RDEPENDS (for running) Feb 20 08:56:08 willie: actually, it seems to be missing the .inc, indeed. Feb 20 08:59:43 hello! what do you suggest for managing resolv.conf when using systemd + musl ? Feb 20 09:00:37 LetoThe2nd: I guess, I'll try to build it first and then understand the core conecpt Feb 20 09:01:45 willie: yeah, get started from the examples, then copy one over and modify it for your use case Feb 20 09:14:11 as systemd-resolved is not available, ... how can I pass dhcp's dns to resolv.conf ? Feb 20 09:32:22 RP: yes, I saw the gtk-native thing, and was pulling my hair yesterday trying to reproduce it. I took the exact same commit id from master-next, and it would still build without issues :-( Feb 20 09:55:47 kanavin: bugs like this are a pain :( Feb 20 09:56:26 adelcast: maybe you can clarify for me? using opkg -o and -d , what are the intended use-cases ? and also i found that post-install scripts don't honour the -o option. is this considered a bug? Feb 20 09:57:27 RP: yes, I guess I would need the trees of glib-native and gtk3-native from the AB, but let's get the current batch merged, it's looking good so far :) Feb 20 09:58:12 kanavin: There is a subset of -next ready to merge when I get a few mins and yes, hoping this next batch looks better too Feb 20 09:58:29 the one that is running currently, I mean Feb 20 09:58:41 so far, no issues, other than mingw Feb 20 10:48:01 kanavin: oe-selftest is a mess in it :( Feb 20 10:48:13 kanavin: I don't understand why that failed :( Feb 20 10:48:27 hash mismatches and a blocked bitbake server :; Feb 20 10:51:48 RP: yep, I have no idea either. maybe a random fluke, that disappears with a re-run :( Feb 20 10:54:28 kanavin: trying a rerun, lets see Feb 20 11:23:10 RP: this time it's perfect :-/ Feb 20 11:23:48 kanavin: indeed :( Feb 20 11:32:18 kanavin: nice batch merged with a few more waiting in -next for mailing list review or tweaks Feb 20 12:20:33 RP: cheers :) let me rebase my branch and see what's left Feb 20 12:33:03 RP: I rebased my stuff, here's what is left http://git.yoctoproject.org/cgit.cgi/poky-contrib/log/?h=akanavin/package-version-updates Feb 20 12:33:17 RP: I just checked again, gtk+3-native builds without any problems Feb 20 12:34:02 RP: so the only way out is if you try this locally, or give it again to the AB, so we can preserve the failed workdirs Feb 20 12:34:05 ndec: ping again (advocacy-subscription) Feb 20 12:34:22 RP: glib will drop autotools (already dropped in master, in fact), so we have to do the meson transition Feb 20 12:46:08 LetoThe2nd: I can build and run any example from meta-qt but my custom application (in my custom layer) is not detected, i even tried to add it in local.conf. Feb 20 12:47:27 willie: have you added your layer to bblayers.conf? Feb 20 12:47:49 LetoThe2nd: I can see that all the example applications are installed from /sources/meta-fsl-bsp-release/imx/meta-sdk/recipes-images/fsl-image-validation-imx.bb Feb 20 12:48:10 LetoThe2nd: Yes, i have added custom u-boot, psplash etc so the layer is included Feb 20 12:48:36 willie: what do you mean exactly then by "is not detected"? Feb 20 12:50:02 LetoThe2nd: in meta-custom/recipes-qt/myapplication.bb then in conf/local.conf i add IMAGE_INSTALL_append = " myapplication" Feb 20 12:51:01 willie: because "myapplication.bb" does not match the canonical rules Feb 20 12:51:22 its either myapplication_1.0.bb, or myapplication_git.bb Feb 20 12:51:25 correction, it is myapplication_1.0.0.bb Feb 20 12:51:56 then its still to check if that recipe actually provide a package called "myapplication". Feb 20 12:52:03 can you put the recipe on a pastebin? Feb 20 12:53:35 and, does bitbake myapplication build your stuff? Feb 20 12:54:02 https://pastebin.com/kbk9Zv1R Feb 20 12:54:34 I'm not sure, adding psplash and other programs was a bit more straight forward, now i tried to copy how the qt examples were installed Feb 20 12:55:30 generally the recipe doesn't look bad. what happens upon "bitbake ked-qml-qt-project"? Feb 20 12:56:00 or is that only the recipe-internal name? Feb 20 12:56:23 https://pastebin.com/rDrw7Gwr Feb 20 12:56:46 what do you mean by recipe-internal? Feb 20 12:56:51 nevermind Feb 20 12:57:35 and the file *really* is located at meta-yourlayer/recipes-qt/ked-qml-qt-project/ked-qml-qt-project_1.0.0.bb? Feb 20 12:57:48 do not leave out any hierarchy levels! Feb 20 12:58:18 ehm, its in a subdirectory "files" Feb 20 12:58:24 möööööp Feb 20 12:58:46 (and do not add any hierarchy levels, too!) Feb 20 12:59:09 kanavin: did you check in combination with virtgl? Feb 20 12:59:32 you can see where bitbake goes searching if you look at the regexp thats in meta-yourlayer/conf/layer.conf Feb 20 13:00:23 I did try a master-next commit yesterday, but it shouldn't matter as remaining virgl patches are all tweaking qemu or something else unrelated Feb 20 13:00:44 RP: I can try once more to rebase on master-next Feb 20 13:01:29 LetoThe2nd: I assumed adding a files directory was some sort of yocto standard, Still getting the same error tho. Feb 20 13:01:44 Letothe2nd: that is good advice, I'll look that up Feb 20 13:02:04 willie: a files directory is standard. but its for files that are *consumed* by the recipe. Feb 20 13:02:09 kanavin: I've fired a master-next with your changes and the updated mesa ones Feb 20 13:02:19 willie: i guess that your path is still off in some way, then. Feb 20 13:02:31 willie: maybe you can pastebin a tree of your layer Feb 20 13:02:40 RP: thanks, if it fails, I'd like to look at the workdirs of glib and gtk Feb 20 13:02:57 kanavin: right, we should be able to sort that Feb 20 13:04:46 LetoThe2nd: https://pastebin.com/VKw2ubFz , it is a bit messy and the custom MACHINE is not used in any way Feb 20 13:05:34 willie: like i said, you left out a hierarchy layer Feb 20 13:05:48 12:57 < LetoThe2nd> and the file *really* is located at meta-yourlayer/recipes-qt/ked-qml-qt-project/ked-qml-qt-project_1.0.0.bb? Feb 20 13:05:51 12:57 < LetoThe2nd> do not leave out any hierarchy levels! Feb 20 13:06:20 ohhhhh Feb 20 13:08:30 thanks ALOT Feb 20 13:08:38 :-) Feb 20 13:13:29 RP: it failed again in the same way (e.g. poky-tiny) Feb 20 13:26:13 kanavin: do you have autobuilder ssh access? Feb 20 13:26:48 RP: I don't think so, I only acces auh machine. By the way the failure seems specific to centos 7. Feb 20 13:27:19 kanavin: ok, so which files do you need? Feb 20 13:28:28 RP: /home/pokybuild/yocto-worker/poky-tiny/build/build/tmp/work/x86_64-linux/gtk+3-native/ and glib-2.0-native/ Feb 20 13:30:43 kanavin: https://autobuilder.yocto.io/pub/non-release/20190220-9/akanavin.tgz (189MB) Feb 20 13:31:24 RP: thanks, I hope I can sort it via the tarball Feb 20 13:31:47 kanavin: if not let me know. We should sort you worker access at some point too Feb 20 13:37:19 RP: for some reason +x is not set on the offending binary :-/ Feb 20 13:37:27 RP: and this is specific to centos Feb 20 13:37:41 kanavin: how 'interesting' :/ Feb 20 13:37:59 race of some kind? Feb 20 13:39:42 kanavin: with it being native, pseudo should be disabled and out the picture too Feb 20 13:40:14 RP: nope, tarball extraction I believe - the +x bit presence/absence goes all the way to the original file from the glib tarball Feb 20 13:41:37 kanavin: how is it working anywhere then?! Feb 20 13:41:50 RP: still looking Feb 20 13:43:33 RP: diff --git a/gio/gdbus-2.0/codegen/gdbus-codegen.in b/gio/gdbus-2.0/codegen/gdbus-codegen.in Feb 20 13:43:33 index 8050981..e693ef3 100644 Feb 20 13:43:33 --- a/gio/gdbus-2.0/codegen/gdbus-codegen.in Feb 20 13:43:33 +++ b/gio/gdbus-2.0/codegen/gdbus-codegen.in Feb 20 13:43:40 100644 Feb 20 13:44:10 apparently some version of whatever we use to patch ignore this, others do not? Feb 20 13:46:03 kanavin: I'm getting quite worried about the seemingly odd behaviours of different versions of patch :( Feb 20 13:46:58 you'd hope unix had stability in tools like patch, but clearly not Feb 20 13:47:21 RP: internals of patch are quite awful too. 80s-style C. Feb 20 13:47:32 global variables everywhere etc. Feb 20 13:47:43 kanavin: not surprised. saddened maybe. Feb 20 13:49:58 kanavin: and people complain we don't rely on distro tools Feb 20 13:50:08 kanavin: nicely tracked down :) Feb 20 13:51:13 RP: cheers, I hope this is the correct fix :) patches sent Feb 20 13:53:57 RP: will go for a bike ride now Feb 20 13:57:45 kanavin: thanks, have fun! Feb 20 13:57:49 short interruption: has the cc softlink been removed lately? if yes, for which reason? Feb 20 13:58:00 kanavin: raining here so I can't avoid work :) Feb 20 13:58:50 RP: it's probably same rain front that arrives tonight to berlin Feb 20 14:00:55 kanavin: I'll stop that build and start a new one with the updated patches (and throw libdrm into the mix) Feb 20 14:13:19 JPEW: I'm ready to merge the diffutils upgrade to core, just not sure about mingw Feb 20 14:30:49 Hi, I have some python recipes that were working great on sum, but after upgrading to thud I get the following error when building: ERROR: Do not try to fetch `pytest-runner' for building. Please add its native recipe to DEPENDS. Feb 20 14:30:53 Any hints? Feb 20 14:47:03 RP: I tried it last night, and it doesn't look like it is just a trivial rename of the bbappend :( Feb 20 14:53:56 JPEW: I had a closer look at it and yes, doesn't look simple :* Feb 20 14:53:58 :( Feb 20 14:54:50 RP: I get this: ./../diffutils-3.7/gnulib-tests/setlocale.c:36:10: fatal error: localename.h: No such file or directory Feb 20 14:55:09 Trying to figure out what changed in diffutils that would cause that Feb 20 15:03:09 JPEW: update of gnulib? :/ Feb 20 15:03:41 RP: Thats my guess Feb 20 15:04:09 JPEW: other option may be to take 2.6 into mingw and pin it for mingw Feb 20 15:04:18 3.6 Feb 20 15:04:38 Ya, we might have to do that Feb 20 15:17:14 kanavin: http://errors.yoctoproject.org/Errors/Details/229403/ looks like a race :( Feb 20 15:17:58 hello! how does one approach branding? would "some variable" give distinction based on which different set of packages would be installed ? Feb 20 15:18:18 different images for different brands? different distros? so many possibilities ?! :) Feb 20 15:18:53 packages to be installed comes from the image recipe and nowhere else.. Feb 20 15:19:15 the configuration of the operating system would be distro.. but a change in distro means recompiling everything.. Feb 20 15:19:33 for branding, I'd suggest doing that in the image.. and picking either raw graphic changes or a different package that provides the graphics Feb 20 15:19:47 fray: a change in distro doesn't necessarily mean recompiling Feb 20 15:20:04 RP: go on... :) Feb 20 15:20:09 true.. as long as distro name and other things used elsewhere don't change.. but it almost always does.. :) Feb 20 15:20:19 ok. Feb 20 15:20:40 what if distro features change Feb 20 15:20:47 right, depends what the change is and whether that means things need to rebuild Feb 20 15:21:08 changing distro features should always be done in a distro -- unless you are willing to manage a -local to you- infrastructure.. which is not community best practices Feb 20 15:25:32 kanavin: also wondering about https://autobuilder.yoctoproject.org/typhoon/#/builders/47/builds/309/steps/7/logs/warnings Feb 20 15:29:53 How does licensing work with custom qt applications? I'm getting "Recipe file does not have license file information (LIC_FILES_CHKSUM) [license-checksum]". I tried adding "MIT" with the COPYING file in the same directory but it still complains Feb 20 15:41:19 Piraty: the -o option is used by OE to tell opkg to use the root folder as the opkg root. That way all the files are installed under $OFFLINEROOT\. Regarding post-inst using an offline root, you need to pas the --force-postinstall flag to tell opkg to run postinsts with offline roots (by default postinsts are not run if -o is used) Feb 20 15:42:04 if you check opkg test suite (make check on the opkg repo), you'll see that it creates packages and feeds, then installs them using an offline root. That way, each test gets its own sandbox, its very handy Feb 20 15:42:36 yeah i use --force-postinstall already. what i meant is postinstall commands still use / instead of what i give as -o Feb 20 15:43:00 adelcast: thansk for pointing me to test suite, i have something similar in mind- Feb 20 15:43:15 and yet, what is -d useful for in contrast to -o ? Feb 20 15:57:51 Piraty: I pretty much never use -d, since all the use cases I can think of can be satisfied with an offline root Feb 20 15:58:08 alright Feb 20 15:58:11 the idea, I believe, is that using a .conf, you register directories with certain names Feb 20 15:58:33 so you can have "option dest test /tmp/mytest" Feb 20 15:59:02 i found some postinstall snippets that make use of var you have to pass via env, set to the same as -o in order to not perform stuff in / but in instead Feb 20 15:59:09 then via -d you can switch where you install by referencing the directory by the name test, so "-d test" Feb 20 16:00:30 let me check the code real quick, but yeah, if I remember correctly OE used to use the $D variable to determine if you were running on the target or not Feb 20 16:00:51 then the postinst will switch on that variable to defer execution to the first boot Feb 20 16:01:33 so, some postinst execution will be defered until first boot....I believe that mechanism changed recently Feb 20 16:05:25 and yes, you are right, the postinsts don't run on the offline root, that's why OE has that mechanism: https://www.yoctoproject.org/docs/2.6/dev-manual/dev-manual.html#new-recipe-post-installation-scripts Feb 20 16:10:30 RP: Hah! Got diffutils 3.7 compiling for mingw Feb 20 16:10:42 Let me run the SDK test and I'll push up the patch Feb 20 16:23:08 JPEW: ah, great! Feb 20 16:25:01 New news from stackoverflow: python in embedded linux system will not print unicode Feb 20 16:33:18 hello, i have a question about yocto und proftpd. I want to use the compiled proftpd server with tls/ssl. I found out that i need to compile proftpd with the mod_tls Module for this. Does anybody knows how I can achieve this with Yocto? Feb 20 17:10:51 RP: Oddly, nativesdk-diffutils doesn't appear to be in the SDK by default Feb 20 17:57:15 JPEW: I'd have thought nativesdk-quilt would be what would pull it in Feb 20 17:57:35 JPEW: seems odd to build it but not need it Feb 20 17:57:42 Right Feb 20 18:07:40 RP: Hmm, not sure. Anyway, the patch is in master-next Feb 20 18:09:15 RP: Looks like the fault was either than the gnulib tests are broken on MinGW, or they don't cross compile. Either way, we don't need the tests so I disabled them Feb 20 19:48:08 Question for the group: When I attempt to run "devtool modify u-boot-variscite", devtool fails to create the workspace because "A branch named 'devtool-override-mender-uboot' already exists." Feb 20 19:49:11 I'm able to get it to create the workspace if I specify --no-overrides to "devtool modify" but that seems to change the behavior of my build Feb 20 19:49:42 * armpit if I add sshd to core-image-minima I can run most tests Feb 20 19:49:44 (e.g. i get different errors than the ones i was trying to use devtool to solve) Feb 20 20:19:45 jtrimbl3: that's a bug I'm afraid :( Feb 20 20:20:00 for some reason it is considering the same override twice Feb 20 20:36:50 bluelightning: Where would I report said bug? Feb 20 20:37:03 jtrimbl3: bugzilla.yoctoproject.org Feb 20 20:41:58 ok. any idea how to work around it? this is literally the first time i've ever used devtool Feb 20 20:59:15 jtrimbl3: the particular recipe you are working with is "special" in that it presumably sets up multiple variants which are active in your build Feb 20 21:21:33 JPEW: sounds good. I'll merge the upgrade to master since we're ready Feb 20 21:23:11 RP: Cool. I found a few other issues that I'm fixing, so I'll be putting those on master-next also. Feb 20 21:24:26 JPEW: just pushed it Feb 20 21:53:36 khem: are you looking at the gtk+ issue or do you need me to? Feb 20 23:18:42 armpit: there are a few recipe upgrades in -next for recipes you maintain fwiw. I pulled them out the AUH Feb 20 23:26:47 RP, ok Feb 20 23:27:25 * armpit has mixed feels over the AUH auto merge idea Feb 20 23:29:41 armpit: I have as well. I'm testing those, haven't posted them yet Feb 20 23:30:09 armpit: I would like to free the maintainers to spend more time on the problematic ones Feb 20 23:33:34 My biggest issue is missing Change log info that may be important, but then again, most folks don't do that Feb 20 23:33:55 helps identify if a package can be backported to a stable branch too Feb 20 23:34:55 armpit: right :/ I'm torn Feb 20 23:35:20 armpit: I'm also conscious that if we can get up to date, it will be easier to maintain that Feb 20 23:36:30 yes and no... if the package is Latest and greatest but is not an LTS version, maintaining takes more work Feb 20 23:37:15 LTS versions if they exist are just bugfixes Feb 20 23:37:38 armpit: LTS are few and far between for the bulk of the recipes Feb 20 23:38:41 yeah... Feb 20 23:44:19 RP, I noticed more recipes moving away from autotools to meson ( for example). is that a direction we want more recipes to go if they support meson or cmake? Feb 20 23:44:24 armpit: anyhow, if you want me to hold any of them let me know. I'll continue to ponder and see what the results are like tomorrow Feb 20 23:44:32 armpit: meson is a lot faster for configure Feb 20 23:44:53 RP, package updates are fine as they are Feb 20 23:44:59 armpit: we're moving where upstream moves Feb 20 23:46:11 I think a lot of the x11 recipes I maintain are meson capable.. been thinking of moving them Feb 20 23:46:58 armpit: it'd be interesting to see the effect on buildtime Feb 20 23:50:05 BTW, the curl update fixes these cves: Feb 20 23:50:09 o CVE-2018-16890: NTLM type-2 out-of-bounds buffer read [67] Feb 20 23:50:10 o CVE-2019-3822: NTLMv2 type-3 header stack buffer overflow [68] Feb 20 23:50:10 o CVE-2019-3823: SMTP end-of-response out-of-bounds read [66] Feb 21 01:52:01 RP: I will once the OE servers are free **** ENDING LOGGING AT Thu Feb 21 02:59:57 2019