**** BEGIN LOGGING AT Tue Oct 16 02:59:59 2018 Oct 16 03:00:11 i thought DI was for rootfs Oct 16 03:00:48 di has many options, it can flash each partition based upon above options Oct 16 03:00:55 I guess -s would be for system image Oct 16 03:01:05 ah! http://opensource.rock-chips.com/wiki_Upgradetool Oct 16 03:01:22 illegal hardware instruction sudo ./Linux_Upgrade_Tool_v1.2/upgrade_tool di -s Oct 16 03:02:11 wow, not nice Oct 16 03:02:47 odd, thought, the page i linked says the device must be in maskrom mode to use upgrade_tool, but i'm pretty sure that's not the case Oct 16 03:03:07 maybe you have to be in maskrom mode to use the DI command of upgrade_tool? Oct 16 03:03:13 upgrade_tool uf says Loading firmware... Oct 16 03:03:15 Loading firmware failed! Oct 16 03:03:27 so it does not like gpt image Oct 16 03:04:31 i spent a long time trying to sort all this mess out... which is why i just avoid it all by not bothering with the eMMC Oct 16 03:05:00 they've put so much effort into being special Oct 16 03:05:19 but then don't follow through with open tools or clear instructions Oct 16 03:06:09 i have 3 different binaries called upgrade_tool that i've found in various places on the internet, none two are the same, no idea how they differ Oct 16 03:12:41 hmm Oct 16 03:12:58 so I guess I need to put the board into maskrom mode Oct 16 03:13:05 any tips on how to do it safely Oct 16 03:13:10 I dont have serial access Oct 16 03:13:15 but its connected to monitor Oct 16 03:15:15 tlwoerner: https://nopaste.xyz/?66272efa3471228b#BkYfoZNg49NAtLuDmmWGGL+AFQmtb7zgkf2C/wgIu5Q= Oct 16 03:15:30 may be I can use upgrade_tool EB ? Oct 16 03:20:08 khem: you don't need serial access to put the board into maskrom mode, just a USB OTG cable and upgrade_tool or rkdeveloptool to zero the flash Oct 16 03:20:18 http://en.t-firefly.com/doc/product/info/401.html Oct 16 03:20:45 but you do need a wire to short out two points on the back of the board while you insert the barrel jack Oct 16 03:23:17 khem: however... if you zero out the eMMC on the firefly-rk3288, you need a gpt image which means you have to build u-boot Oct 16 03:23:49 for the last couple months, the latest u-boot hasn't been working on the firefly-rk3288 and nobody has had the time to dig into it to figure out why it stopped working Oct 16 03:24:32 user "" in #linux-rockchip or #u-boot has been working on it off and on, but hasn't found a solution Oct 16 03:25:09 so if you zero out the flash on your firefly-rk3288, you'll need to build an older image, or play around with the u-boot recipe to build an older u-boot Oct 16 03:27:08 master images work fine on the tinkerboard and the vyasa, both of which have rk3288 SoC Oct 16 03:27:39 i haven't had time to play with my vyasa board yet (after ELCE!) but apparently there's a button on the board to put it into maskrom mode (finally!) Oct 16 03:29:32 oh Oct 16 03:29:37 I have older images Oct 16 03:29:41 based on pyro Oct 16 03:30:04 they worked ok so once I figure out way to flash it it might work Oct 16 03:30:19 tlwoerner: I am looking for a non physical way Oct 16 03:30:26 to erase flash Oct 16 03:31:55 what is idbloader.img I see that also in deploy dir Oct 16 03:50:31 odd, i don't have that in my deploy dirs Oct 16 03:50:37 http://opensource.rock-chips.com/wiki_Boot_option#idbloader.img_with_tpl.2Fspl Oct 16 04:22:08 I'm trying to add uclibc-staticdev to my build but get's error nothing provides this package. Is this package not included in Yocto? Oct 16 04:31:46 learningc: uclibc was dropped a couple releases back, try musl instead Oct 16 04:33:40 ok. thanks Oct 16 04:42:37 I get the same kind of error with musl Oct 16 04:42:40 ERROR: Nothing RPROVIDES 'musl-staticdev' Oct 16 04:43:26 What to do? Oct 16 04:43:33 I'm on morty Oct 16 04:49:50 tlwoerner: hi Oct 16 06:30:12 hi, is meta-linaro really required to build for arm with morty? Oct 16 06:30:24 some colleague pulled in, but I do not really like to have one additional layer at all. Oct 16 06:35:19 lpapp: i have been building forever for arm without meta-linaro Oct 16 06:36:15 great Oct 16 06:48:31 do you guys use repo to manage your own stuff Oct 16 06:48:50 openembedded-core and poky layers, your own vendor layers, meta-linaro, etc, whatever we need? Oct 16 06:48:56 or just simply git submodules? What is the standard practice? Oct 16 06:54:59 we use a custom solution, but in general i'd suggest to use kas Oct 16 06:55:31 if kas was already available when we needed something, i would've used it Oct 16 06:56:14 wow, never heard about kas Oct 16 06:57:54 you can certainly use repo too, but repo is more like a upstream source management thing. whereas kas is a one-shot setup solution to reproduce an OE build Oct 16 06:58:01 and i totally prefer the latter. Oct 16 06:58:50 alimon: it supports the idea of override config files so you can add in your own local-config.json which would override Oct 16 07:00:20 LetoThe2nd: perfect, but then can I have another question please, namely... Oct 16 07:01:33 I have this issue with repo, not sure if I had the same issue with kas, that when I pull these repos and their branches together for a complete build, sometimes the build fails... because some repos have issues to come together... now then for example, in my layer, one repo, I fix one thing, and I want to commit from there right away... rather than in another clone separately... but with repo, it is a mess Oct 16 07:01:39 as it detaches each inner repo from their desired branches... Oct 16 07:02:08 kas does not manage your upstream, simple as that. Oct 16 07:02:47 i don't know ad-hoc if it uses detached heads, but even if it does, its trivial to commit anyways. Oct 16 07:03:00 what exactly do you mean by that Oct 16 07:03:00 do I not have to specify where each layer should come from, a github kind of source url of some sort? Oct 16 07:03:24 of course you have to specify that, how would it build? Oct 16 07:03:25 it is not so much trivial... you have to remember to be always in the right branch Oct 16 07:03:37 and we humans well ... fail at that thing Oct 16 07:04:23 is it using containers by the way Oct 16 07:04:24 as i already said, i don't know how the checkout state of kas looks. give it a try and find out. Oct 16 07:04:40 yeah, looks like it uses docker Oct 16 07:04:45 and no, it does not use docker. Oct 16 07:05:01 there is a premade docker container that has kas installed, but thats it. Oct 16 07:05:15 by itself, kas is a standard python thing, can be installed through pip Oct 16 07:06:01 this is not strictly true Oct 16 07:06:10 *sigh* Oct 16 07:06:11 it does use a build environment to avoid host contamination Oct 16 07:06:19 "launch minimal build environment, reducing risk of host contamination" Oct 16 07:06:41 really, i'm totally not into nitpicking. you asked for hints, i gave some. Oct 16 07:06:47 coffe, now. Oct 16 07:10:51 this configuration thing is a bit complex Oct 16 07:11:01 I thought most of these are already in the conf directory of the specific layer Oct 16 07:30:23 LetoThe2nd: so is there an example kas configuration file somewhere? Oct 16 07:31:58 oh, https://github.com/siemens/meta-iot2000/blob/master/meta-iot2000-example/kas.yml Oct 16 07:43:54 good morning Oct 16 07:44:06 tgoodwin: I didn't use an initramfs, just a plain rootfs. the wrong UIDs broke my boot since among other things /bin/mount was owned by 1000 instead of root Oct 16 08:00:15 how can I force yocto and bitbake to put the u-boot binaries again in ./tmp/deploy/images/${machine}? Oct 16 08:17:46 lpapp: I usually run bitbake -c cleansstate u-boot and then bitbake u-boot again Oct 16 08:20:30 lpapp: http://wiki.kaeilos.com/index.php/Bitbake_options Oct 16 08:29:24 rburton: systemtap upgrade isn't so straightforward :( Oct 16 08:43:14 I have an image that pulls in lots of packages, those packages have library-dependencies.. if I populate_sdk for my , will it include those library-dependencies? Oct 16 08:45:20 jofr: yepü Oct 16 08:45:24 LetoThe2nd: kas. Interesting. I'm doing "repo" but only because it was there and I didn't want to roll my own. Thanks! Oct 16 08:45:31 LetoThe2nd: And thanks again ;) Oct 16 08:45:37 jofr: :-) Oct 16 08:48:08 uh, when I build my kernel and later do a -c menuconfig it seems the patches are missing. Is there a way to patch the sources and then run menuconfig? I want to create a new defconfig Oct 16 08:52:05 sven^: I normally do a -c configure, -c devshell, make my_own_defconfig && make menuconfig Oct 16 08:53:15 sven^: Because normally I want to menuconfig from my current defconfig.. not totally from scratch.. and finally I copy the .config out of the build directory and put it in my layer :) Oct 16 08:53:18 I just did bitbake mykernel -c configure -c devshell and when I run make menuconfig my patches are not there Oct 16 08:53:30 options I added are missing Oct 16 08:54:28 Does bitbake accept multiple -c's? Oct 16 08:56:10 not sure. I tried splitting it as well Oct 16 08:56:28 I mean the devshell is executed, but I am not sure if it screws up in between Oct 16 08:56:32 Errr.. sorry. The last bit was incorrect, that's what I do with u-boot. My kernel config is in pairs of .scc and .cfg files. I only use menuconfig to get the names of the CONFIG_'s Oct 16 08:56:47 It shouldn't.. Oct 16 08:57:01 Perhaps I'm missing a step there. I'm actually not patching my kernel at all. Oct 16 08:57:22 And anyone who actually knows more about this, please chime in ;) Oct 16 08:57:23 well, it seems that the order of defconfig is important, so I wanted to use menuconfig to create a proper .config before running savedefconfig Oct 16 08:58:31 o_O I've never known the order to matter...? Oct 16 08:59:05 In any case, -c listtasks .. Oct 16 09:00:13 It's got some specific "kernel_" steps Oct 16 09:00:33 as well as "patch" :p Oct 16 09:01:19 "Also, it should be mentioned that kernel build system keeps very specific order of options in defconfig file, so it's better to avoid to modify it by hand. Instead you should use make savedefconfig rule." Oct 16 09:01:24 taken from some so answer Oct 16 09:01:30 rburton: thoughts on whether we should test ftp urls at all in bitbake any more? I think the firewall is breaking the autobuilder :/ Oct 16 09:01:58 sven^: Interesting! I have never stumbled upon that! :-) Oct 16 09:02:09 https://stackoverflow.com/questions/41885015/what-exactly-does-linux-kernels-make-defconfig-do Oct 16 09:02:58 I am not sure if it's true, but I didn't want to take any risks. There are enough knobs to turn that can lead to failing builds :) Oct 16 09:04:44 it seems that bitbake silently ignores all but the last option given by -c Oo Oct 16 09:04:45 sven^: I know exactly what you mean. :-) Oct 16 09:05:11 sven^: Ahh. Oct 16 09:05:55 at least it seems like this. I get into a devshell as soon as I do -c menuconfig -c devshell Oct 16 09:05:56 sven^: So it's just sequentially overriding. Makes sense. Oct 16 09:06:32 ... without a warning or notice -.0 Oct 16 09:08:28 sven^: Yeah.. But it's probably just doing a range-for on sys.argv, and if it sees a "-c" it grabs argv[current+1]. Perhaps bitbake needs a patch! :-D Oct 16 09:09:10 sven^: I'm also assuming bitbake is written in Python. Which I have actually never checked :p Oct 16 09:10:25 lots of the scripts need a patch :) Oct 16 09:13:39 Now. Has anyone here written a recipe for a package that requires libstdc++fs? I've created a .bbappend for gcc_7.3 where I'm trying to add it, but I'm not sure if I'm going down the right path. Oct 16 09:18:51 oh my god. I think I found my problem. Oct 16 09:19:35 haven't added the new patch to the recipe's files -.- Oct 16 09:21:05 hello, I have a gstreamer-plugin, which is not in bad/ugly/good/base/.... So an own independant gst-plugin, from which I have the source and which is build through meson. the only thing I want to achieve is one .so which is placed in /usr/lib/gstreamer1.0 Oct 16 09:21:25 sven^: And you're a bigger man for admitting it. ;-) Oct 16 09:21:40 I am already pretty tall Oct 16 09:21:42 but ok Oct 16 09:21:46 haha Oct 16 09:21:46 can somebody point me to a simple bb-recipe for such library? Now I get QA error because I get 3 versioned SO's, which are nowhere to be installed Oct 16 09:22:17 RP: it seems that the best idea is to fix the proxies so that codepath gets exercised... Oct 16 09:22:29 sven^: I had a collegue who's name is Sven. And he's tall. You're not him, are you? ;-) Oct 16 09:22:42 depends on who you are :) Oct 16 09:22:57 I have collegues Oct 16 09:23:32 haha Oct 16 09:24:13 This Sven was my colleague at Sennheiser Communications here in Denmark. Oct 16 09:24:23 that's not me then Oct 16 09:24:39 woutervh: 1) a gst module shouldn't be versioned, as libraries are versioned but not modules. 2) just add ${libdir}/gstreamer1.0 to FILES_${PN} Oct 16 09:26:34 rburton, I agree they shouldn't be versioned, but it's yocto which does it, that's why I was asking for an example bb which does not do that. And for 2, will try that Oct 16 09:26:43 woutervh: no, it's not Oct 16 09:27:22 the meson.build file will be saying 'build a library' not 'build a module' Oct 16 09:27:41 ah, ok, like that Oct 16 09:28:19 I think you're right, cause I now added this (2), and now it doesn't complain about the .so anymore, only about the versioned ones Oct 16 09:28:29 rule of thumb: yocto just moves files from place to place, and tells other things to do stuff Oct 16 09:28:35 so compiling is just "run make" Oct 16 09:28:46 if the makefile creates versioned libraries, then its the makefile's problem Oct 16 09:30:05 can I add a rule in the recipe to ignore these versions? And only get the .so? without having a qa issue? So I don't have to change the sourcetree? Oct 16 09:31:10 much easier to fix the build Oct 16 09:31:17 https://mesonbuild.com/Reference-manual.html#shared_module Oct 16 09:35:59 rburton, well, I don't know. Cause the source is not mine, and also used by other build systems. I prefer not touching it if I can fix it locally. I can also create a patch, but then I can have issues in the future Oct 16 09:36:39 if you fix it once upstream then you never have to deal with it again. if you work around it locally when upstream does fix, your workaround breaks. Oct 16 09:36:42 what recipe is this? Oct 16 09:37:13 i'll say again: if this really is a gstreamer plugin then versioning is just wrong Oct 16 09:37:29 plugin: https://github.com/woutervanh/gst-plugins-rtp Oct 16 09:38:47 bb what I'm currently working on: https://gist.github.com/woutervanh/d270df647be0849fa7edeacb11bcaba8 Oct 16 09:39:25 but the plugin itself is forked from another one, of which I don't know how they are using it... And I don't want to force them to do it differently then they do know Oct 16 09:39:26 :) Oct 16 09:40:08 it's a bug in their build system Oct 16 09:40:41 probably due to passing version to the shared_library call Oct 16 09:41:50 "version a string specifying the version of this shared library, such as 1.1.0. On Linux and OS X, this is used to set the shared library version in the filename, such as libfoo.so.1.1.0 and libfoo.1.1.0.dylib. If this is not specified, soversion is used instead (see above)." Oct 16 09:42:27 yes, they are indeed passing "version: pkg_version" to therir shared_library call Oct 16 09:42:48 but again, maybe they intended to do it that way, that I don't know... Oct 16 09:42:54 which is wrong. delete it. demonstrate that 1) the build works better and 2) the plugin still works. then send that one-liner upstream. Oct 16 09:43:35 Ok, I can try. Will see if they accept it upstream Oct 16 09:43:37 thanks Oct 16 09:45:33 but I keep the shared_library? Or I change it to shared_module? Do the other arguments still apply? Oct 16 09:49:32 the rest is good Oct 16 09:49:42 module is relatively new, keep as shared_library Oct 16 10:11:43 how to generate only -native package? Oct 16 10:11:53 ... not generate... provide Oct 16 10:22:04 rburton, I have an issue again, and I think I had it before, but in one way or another was able to fix it in the past (I think by intervening directly in the tmp folder) I think it's again related to this meson build. Would you mind to have a look? Oct 16 10:22:11 I updated my gist: https://gist.github.com/woutervanh/d270df647be0849fa7edeacb11bcaba8 Oct 16 10:22:17 it's the log at the bottom Oct 16 10:32:05 kergoth: good news! the boost people have written another build system http://stefan.seefeld.name/faber/ Oct 16 10:34:17 woutervh: looks like the meson.build is broken and is passing an empty filename to gcc Oct 16 10:35:02 yes, see '-fPIC' '-pthread' '' '-MMD' '-MQ' Oct 16 10:35:19 you'll have to figure out where that is coming from Oct 16 10:35:40 if something else that uses meson works then its most likely the meson.build in this project Oct 16 10:37:04 ak77: inherit native Oct 16 10:37:48 jofr: thank you. but then better to have xy-native.bb file because it wont add -native Oct 16 10:37:53 jofr: yes? Oct 16 10:38:12 Yes. Oct 16 10:38:51 or at least that's what I do. I don't know if it has just a naming-convention or if the "-native" part of the name actually has any side-effects... Oct 16 10:39:10 if you inherit native directly its expected that you call the recipe foo-native Oct 16 10:39:30 the alternative is to call the recipe foo, and use BBCLASSEXTEND to magically generate foo-native Oct 16 10:39:39 rburton: But is it a convention or is there something that checks or in some way enforces it? Oct 16 10:39:55 Ohh. Nice. Oct 16 10:40:11 if you have a recipe called foo that isn't native then you'll break lots of code which checks for nativeness by looking at the recipe name Oct 16 10:40:26 using class extending is preferred if you have a recipe which you need for target, native and or nativesdk Oct 16 10:40:33 But would you do that (the BBCLASSEXTEND version) for something that you will never install unto a target? Oct 16 10:40:41 s/unto/onto/ Oct 16 10:40:50 well, is it possible that it makes sense? or in a sdk? Oct 16 10:41:04 if its a tool that has no use outside of native then sure, foo-native and inherit native Oct 16 10:41:11 plenty of those in oe-core Oct 16 10:41:13 I only have one such recipe, and for that particular one: No. Oct 16 10:41:54 zagor: Ah, I see. Thanks for clarifying :) Oct 16 10:43:05 rburton: Thanks for the tip and clarification. :-) Oct 16 10:44:21 rburton: yes, thank you Oct 16 10:49:01 how could I see which recipe/package placed a file in recipe_sysroot_native ? npm of wrong version is installed in recipe_sysroot_native, but as far as i can tell, nodejs-npm is not being installed (which would provide such a file) Oct 16 10:49:27 (checked dep. graph and manifest file) Oct 16 11:06:09 I have found the guilty one, nodejs-native, although it defines nodejs-npm(-native) package that should contain npm executable Oct 16 11:06:30 rburton, ah, thanks, missed that one, I'll have a look Oct 16 11:10:41 RP: https://github.com/boostorg/python/commit/d4d41d94aecc9f8098aabd3587fcb95458451f71#diff-42dd6ec1330a7c47aaccf2ab2b8f1e02 Oct 16 11:16:59 rburton: so the other application needs updating? Oct 16 11:17:23 rburton, thanks, you rock ;-) Oct 16 11:18:59 RP: yes Oct 16 11:19:07 fedora/debian don't have this as they haven't updated Oct 16 11:19:57 rburton: makes sense, thanks for digging into it :) Oct 16 11:33:01 New news from stackoverflow: How to solve the problem running ufw-init? Oct 16 11:44:54 of course, DEPENDS doesn't work through packages, but with do_sysroot_populate... how to prevent specific files do be used in do_syroot_populate ? Oct 16 11:48:06 to describe my issue a little bit more: my recipe depends on nodejs-native, but nodejs' do_populate_sysroot provides old npm, I want to provide my own with npm-native,... there's a clash because sysroot's /usr/bin/npm is already there Oct 16 11:58:41 ak77: sounds like node needs to be split so npm can be shipped/upgraded independently Oct 16 12:00:20 rburton: currently, I added --without-npm to nodejs configure... but looking how to install new version of npm.. without npm :) Oct 16 12:03:04 rburton: does alternative mechanism work for native as well ? Oct 16 12:03:05 New news from stackoverflow: linux extended partition types Oct 16 12:03:08 ak77: no Oct 16 12:13:30 I have a question of an application I'm building. I have as common 2 shared libraries, and 3 static libraries, all with their own git repo. Then I have one main application, also with it's own git. Now I want to build the application using yocto Oct 16 12:13:58 the shared libraries are easy to add, but how do you do that with static libraries? Oct 16 12:14:03 I just need them at compile time... Oct 16 12:14:44 rburton: Just understood why _remove is behaving oddly, there are two bugs :/ Oct 16 12:14:57 it is at least deterministic Oct 16 12:28:43 Why "git-replacement-native" but not just "git-native"? Oct 16 12:30:06 rburton: should TEST = "A B" TEST_remove = "A" give " B" ? Oct 16 12:47:52 yes Oct 16 12:49:10 rburton: currently it doesn't and to do that we have to rewrite the tests but I think we may need to Oct 16 12:49:30 it doesn't?! Oct 16 13:00:18 oh do you mean the leading space Oct 16 13:00:27 remove does listification doesn't it, so it drops the whitespace Oct 16 13:06:20 RP: https://www.yoctoproject.org/docs/current/bitbake-user-manual/bitbake-user-manual.html#removing-override-style-syntax Oct 16 13:16:51 Is there a way to add tasks that execute after all other tasks (each)? And is there a way to setup a task that will be run regardless if the previous (before) task fails? Oct 16 13:19:11 My use case is to build a system for collecting log files from specific recipe tasks. Either I implement it in a bbclass that adds log collection tasks (that must run even if the task fails in order to collect the log from it), or I build the framework for this outside of bitbake Oct 16 13:23:41 do_package_qa: QA Issue: package contains bad RPATH /buld/external/lib in file /packages-split//usr/bin [rpaths] .. I'm a bit lost Oct 16 13:31:28 sveinse: a class that hooks onto build completed? Oct 16 13:31:48 ie https://github.com/rossburton/meta-ross/blob/master/classes/timer.bbclass Oct 16 13:33:40 rburton: does the bb.event.BuildCompleted fire even if the build stops/fails? Oct 16 13:33:45 yes Oct 16 13:33:50 thanks Oct 16 13:57:59 I'm writing a recipe that compiles and installs a CMake-based package. But it complains with the message I wrote above: do_package_qa: QA Issue: package contains bad RPATH /buld/external/lib in file /packages-split//usr/bin [rpaths] .. Oct 16 13:59:54 Is there a correct way for a user recipe to insert a file into /etc/sudoers.d that will not cause a packaging conflict when using RPM packages? The user recipe inherits from adduser and includes only its home and sysconfdir/sudoers.d/ in FILES_${PN}. When making an image, the packager throws a conflict that sudo also provides the "sudoers.d" directory. Oct 16 14:00:07 The executable is a statically linked one. I can manually scp it from the build-directory to the target and run it. My recipe has a do_install() that just installs the executable to ${D}${bindir}and then has FILES_${PN} that lists ${bindir} and ${bindir}/ Oct 16 14:01:28 The "external/lib" folder is one of the artifacts when building the package, but it then statically links that library. i.e. this library should just be discarded and never has to go to the target.. Oct 16 14:04:12 Never mind, I just found a comment from bluelightning on SO about DIRFILES = "1". It would be nice if this were in the megamanual. Oct 16 14:15:49 Seem to have solved it by adding an: chrpath -d ${B}/ to my do_install Oct 16 15:28:32 .. why does bash install a .pc at all, and why is CC included in it? Oct 16 15:43:08 Hi all, I had a quick question about the new python manifest file in Sumo (python2-manifest.json and python3-manifest.json). Oct 16 15:43:32 I'd like to be able to customize that file in my layer, but it looks like the python recipes (such as meta-devtools/python/python_2.7.14.bb) pull it in by referencing THISDIR in an anonmymous Python function. Is there any official way to customize that file other than editing it in the meta layer itself? The reference manual only talks about how you can edit the file directly. Oct 16 16:16:00 ernstp: thanks, think we may need to edit that... Oct 16 16:16:17 kergoth: would you agree we should preserve whitespace in remove operations? Oct 16 16:28:44 kergoth: http://git.yoctoproject.org/cgit.cgi/poky/commit/?h=master-next&id=10f651b7aa4c45a109d7ad490b7b40da1825dfa9 and http://git.yoctoproject.org/cgit.cgi/poky/commit/?h=master-next&id=0baeadf4c65da7b70da27dca3d98da7209c19748 Oct 16 16:33:59 New news from stackoverflow: linux extended partition types [on hold] Oct 16 16:40:27 RP: thanks for those bitbake fixes, small nitpick in "data: Ensure task checksums account for remove data" is typo comibed Oct 16 16:42:50 RP: I'd agree with that, at least until we get proper typing in bitbake itself, otherwise we have checksums changing when we shouldn't Oct 16 16:42:58 looks good to me Oct 16 16:43:39 kergoth: right, I wrestled with it and without typing we can't really do anything else, thanks Oct 16 16:43:52 np Oct 16 16:44:29 i'm not sure we can even do typing without changing the file format — the semantics of when conversions happen to/from string is non-trivial Oct 16 16:44:44 * kergoth shrugs Oct 16 16:44:45 * kergoth yawns Oct 16 16:45:59 JaMa: thanks, tweaked Oct 16 16:46:20 kergoth: its the performance overhead that scares me :/ Oct 16 16:46:42 oh yeah, that'll be a tough one to deal with Oct 16 17:34:12 New news from stackoverflow: *.bb file error: parser error when using if else condition in yocto sumo Oct 16 19:29:41 What is the best way to check why a preservice bumped a pr? Oct 16 21:34:56 New news from stackoverflow: Error on building Openembedded for Intel Quark Oct 16 23:05:15 New news from stackoverflow: How to clone a private git repo from within a BitBake recipe? Oct 17 01:04:21 Say you have a SRC_URI_machinename override in a recipe, and you are making an append. Is there a way to remove or mask that override from the append? Oct 17 01:05:36 Specifically, having the machine variable override makes the devtool flow more manually intensive and error prone. Oct 17 01:33:09 https://www.libssh.org/2018/10/16/libssh-0-8-4-and-0-7-6-security-and-bugfix-release/ <-- possibly one of the most painful security bugs i've seen recently Oct 17 01:46:10 NU-Slacker: Please correct me if I'm misunderstanding your situation, you're trying to have different appends depending on the machine? Oct 17 01:46:25 sounds like they want to undo a machine override Oct 17 01:47:43 Yes to both. Oct 17 01:48:27 The base recipe is inside a vendor BSP layer and I'm making a board-specific layer Oct 17 01:49:27 ahhh okay, The solution that comes to mind is to have a separate directory within your `files` search path for each machine. Oct 17 01:49:55 I'm not sure if it will fully "undo" the machine override though Oct 17 01:50:02 The annoyance comes with devtool updating SRC_URI but not SRC_URI_machinename when updating or finishing. Oct 17 01:50:48 devtool is never going to be a complete replacement for manual recipe/append changes. overrides add a great deal of complexity to source modification Oct 17 01:51:03 So you want to append the file for all machines, but one machine (your specific board) has a special file that needs to be appended? Oct 17 01:52:14 kergoth: Understood that devtool cant automate everything. Just checking to see if Im not missing an easy solution. Oct 17 01:57:39 robbaweba: Pretty much, except that the base recipe from the vendor BSP layer has the override. I want to make what I have in the append apply to all. With the current situation I would need to set SRC_URI and SRC_URI_machinename in the bbappend to work Oct 17 01:59:01 I guess I could just create a new machine and that would "fix" it ¯\_(ツ)_/¯ Oct 17 02:01:37 NU-Slacker: ahhh i see. In that case, I'm not sure how to cheat the machine override system. Sounds like a manual job :/ Oct 17 02:01:59 NU-Slacker: although I've never used devtool, so I'm not quite sure what I'm mising lol Oct 17 02:03:56 robbawebba:I had never used devtool either until relatively recently. And of course the first recipe I tried it on had this odd case making it fail. Oct 17 02:05:21 It definitely removes a good portion of the more monotonous tasks. Oct 17 02:19:58 NU-Slacker: I'll try and give it a show sometime soon! Oct 17 02:20:35 Unrelated, I've got a question about switching between two kernel versions for the same image, distro, and machine. Oct 17 02:23:29 I have two separate recipes for the kernel (4.9 and 4.19). our custom machine.conf file declares PREFERRED_PROVIDER_virtual/kernel, and that's the mechanism we're using to switch between kernel versions. We modify PREFERRED_PROVODER_virtual/kernel in our custom machine.conf file when we want to switch kernel versions. Oct 17 02:24:28 Is this the correct way to switch between kernels when all else is held constant? Or is there a better way to switch kernel versions? Oct 17 02:30:20 robbaweba: FWIW that is the mechanism I've used in the past. Though we usually set it in local.conf (generated by a source controlled setup script) Oct 17 02:57:20 NU-Slacker: How could I configure the PREFERRED_PROVIDER_virtual/kernel from local.conf? Do I have to use late assignment ( ?= ) in the machine.conf ? **** ENDING LOGGING AT Wed Oct 17 02:59:59 2018