**** BEGIN LOGGING AT Tue Mar 09 02:59:57 2021 **** BEGIN LOGGING AT Tue Mar 09 03:54:28 2021 Mar 09 04:10:32 zeddii: hidden bison... I see a movie script Mar 09 05:04:34 /msg NickServ identify 2411 Mar 09 05:07:20 hello what I habe t use gatesgarth or poky-3.2.2 ? Mar 09 09:05:31 jdrol: poky-x.y.z are tags IIRC, so it all depends if you want a "rolling release" of gatesgarth or plan to use dot releases (tags) only Mar 09 09:23:07 qschulz_ what is difference between gatesgarht and 3.2.2 ? Mar 09 09:29:50 @jdrol: gatesgarth is the codename for yocto project version 3.2. the current version is 3.2.2 Mar 09 09:31:07 @jdrol: here http://git.yoctoproject.org/cgit/cgit.cgi/poky/log/?h=gatesgarth you see gatesgarth is a branch in a git repo. http://git.yoctoproject.org/cgit/cgit.cgi/poky/tag/?h=yocto-3.2.2 is a release tag Mar 09 09:33:48 hi Mar 09 09:37:58 RobertBerger, thanks ...and alls this branch are the same in Openembedded git ? Mar 09 09:39:26 @jdrol: in other meta-layers you just need to make sure to be compatible e.g. by using the gatestgarth branch as well. There is not necessarily a 3.2.2 tag, but that's fine. Mar 09 09:40:36 @jdrol: in the specific case I showed above it's all the same git repo: http://git.yoctoproject.org/cgit/cgit.cgi/poky Mar 09 09:40:51 how is one supposed to specify a dependency on a package to provide a user group (i.e. systemd(-journal))? Mar 09 09:41:36 @jdrol: note that there is also a gatesgarth-next branch and that the gatesgarth branch is already on something newer compared to 3.2.2 Mar 09 09:42:57 ok thanks Mar 09 09:43:01 @jdrol: so, as @qschulz_ pointed out it's up to you what you use. You decide on the version e.g. gatesgarth and than the fun starts ;) Mar 09 09:43:18 ok Mar 09 09:51:20 t_unix[m]: you want your package to create a use/group? Mar 09 09:52:39 I've created a useradd package that appends the user to the `systemd-journal` group. But it fails to configure my package for the rootfs, because the group `systemd-journal` is missing. Mar 09 09:53:07 the `systemd-journal` group is provided/instanciated by the package `systemd` Mar 09 09:53:45 t_unix[m]: and a "simple" depends doesn't solve it? Mar 09 09:55:43 unfortunately it does not. RDEPENDS -> building package fails. DEPENDS -> fails while installing (`do_rootfs` of corresponding image) Mar 09 09:59:40 t_unix[m]: i see. Mar 09 10:00:41 I assume I need to inject a specific task dependency, like "systemd:do_rootfs"? Mar 09 10:01:16 the example `useradd-example.bb` does not cover this case. Mar 09 10:04:57 yes, i've also never heard of the requirement so far. Mar 09 10:06:38 but it does make sense, doesn't it? I mean same circumstances apply to bluetooth, wifi, .... ? Mar 09 10:07:55 i'm not sure i can properly comment, sorry. Mar 09 10:09:36 t_unix[m]: it used to be an issue, don't know if it's been fixed or not Mar 09 10:10:47 t_unix[m]: https://bugzilla.yoctoproject.org/show_bug.cgi?id=13419 Mar 09 10:11:09 well question is what happens, if there are two packages that declare the same `GROUPADD_PARAM_${PN} = "-r systemd-journal"` Mar 09 10:11:43 likely return code 9 Mar 09 10:12:51 qschulz_: thanks you! Mar 09 10:13:00 *thank Mar 09 10:57:41 Could anybody share some of their bashrc tweaks done for easy file navigation in a large yocto system ? Mar 09 10:57:55 git grep :-) Mar 09 10:59:15 open all the layers in vscode as a workspace Mar 09 11:00:24 derRichard Thanks:)  was new info to me Mar 09 11:00:34 @rburton Mar 09 11:00:38 JosephAntony: I use fd (find-fd in some distro) extensively, silversearch-ag (ag) as git grep, and just started trying to use fzf Mar 09 11:01:00 @rburton But will vscode be able to index them properly ? Mar 09 11:01:09 what do you mean by index Mar 09 11:02:06 @rburton I meant when we click on function declarations, will it be able to go directly to the function definitions ? Mar 09 11:02:22 if you open a python file, yes. Mar 09 11:02:38 there's a bitbake extension which sort of works for classes Mar 09 11:03:09 is it possible to teach udhcpc to do link local Mar 09 11:03:42 qschulz_ Thanks :) Mar 09 11:03:46 JosephAntony: like pressing Goto Definiton on 'inherit meson' opens meson.bbclass Mar 09 11:05:34 rburton Okay Got it. Mar 09 11:06:09 Actually question in my mind was related to folder navigation and find and searching files... from large yocto Mar 09 11:07:38 Sharing my tweaks too. Mar 09 11:07:38 alias findrcp='find ./meta-*/meta*/rec* ./meta-*/rec* ./open* -iname ' Mar 09 11:07:48 to find recipes fast. Mar 09 11:07:58 This must be added to bashrc file Mar 09 11:08:09 meh, cmd-p in vscode is faster :) Mar 09 11:08:17 Please share if you have anything like this Mar 09 11:09:05 (i wanted to open ovmf_git.bb, so just did cmd-p ovbb[return]) Mar 09 11:10:40 yeah. Mar 09 11:12:56 my yocto repo is c++ code. Indexing does not seem to work well.. Dont know if I need to enable some settings in vscode Mar 09 11:13:21 @rbu Mar 09 11:15:45 JosephAntony: are we talking about the metadate now, or are we talking about an application that you want to build/develop? Mar 09 11:20:38 LetoThe2nd: highly suspecting it's a case of "one git repo for code and build system" Mar 09 11:20:52 eek Mar 09 11:20:59 No. multiple git repos Mar 09 11:21:33 JosephAntony: "my yocto repo is c++ code" you need to elaborate a bit more then what you mean by that Mar 09 11:22:11 single yocto system with multiple git repos. I have a large build directory. where so many recipes are bitbaked and source content is available. 80 percent of code is in C++. I Mar 09 11:22:55 JosephAntony: are you looking into the build directory? I rarely have the need to do so, any particular usecase? Mar 09 11:23:18 you can always add the build directory to a blocklist in your indexer? Mar 09 11:23:48 how do you understand code flow between libraries ? I grep ''function names'' in build directory to understand the code flow of who calls whom Mar 09 11:25:28 that all doesn't go together. Mar 09 11:26:08 LetoThe2nd I did not understand. maybe I am doing in the wrong way Mar 09 11:26:38 either we are talking about the payload application, or we are talking about the metadata. and as the metadata usually isn't c++, the question and informations are not exactly understandable. Mar 09 11:27:09 we are talking about payload appln now. Mar 09 11:27:37 leave meta-data part. Mar 09 11:28:34 so what is the question then? that you have a payload application written in c++, and you want to be able to browse headers? Mar 09 11:33:04 my exact qn was I have payload app in c++. Since the directory paths are very large and difficult to traverse between directories what were the tweaks applied in command line/ bashrc file. Mar 09 11:34:12 Don't poke at the build directory directly Mar 09 11:34:35 use devtool to put the source/build tree outside of the build directory and open that instead Mar 09 11:35:14 rburton: ++ Mar 09 11:36:49 rburton Okay. had not heard about devtool. Only heard dev-shell Mar 09 11:36:51 thanks Mar 09 11:39:45 another prime example why it is so important to ask as clear and concise as possible. :) Mar 09 11:40:58 okay. Sure.. Will do it next time :) Mar 09 11:45:18 have fun! Mar 09 12:07:29 hello Can I extend a .bbappend one more time ? Mar 09 12:09:10 jdrol: yes, they can be stacked Mar 09 12:10:02 RP : in order of recipes and meta ? Mar 09 12:10:42 jdrol: I don't understand the question Mar 09 12:11:31 sorry. the first meet is the first applied ? Mar 09 12:13:05 jdrol: all bbappends are concatenated together Mar 09 12:14:17 for example I want to change the busybox config.. I see there is an bbappend Mar 09 12:16:08 you can add your own bbappend ... it will be added to the mix. Mar 09 12:18:44 thanks Mar 09 12:20:19 * RP files another diffoscope performance bug Mar 09 12:21:41 jdrol: They're ordered by the order they're found in BBFILES Mar 09 12:22:23 ok Mar 09 12:33:35 Any make experts around? Do some versions of make add a whitespace for "INSTALL_TARGETS+=" and some not? :/ Mar 09 12:43:29 RP: long shot, but.. do you have any ideas how to tackle a build essentially being "stuck"? situation is docker-in-lxc-on-host, and during somewhat longer builds i keep on seeing that it just "stops", like at task 5xxx of 8yyy. then only the time counters per task go up, no other activity can be seen. Mar 09 12:44:05 LetoThe2nd: can you ssh in? see what its running from ps ax? Mar 09 12:44:23 LetoThe2nd: I think rburton has a script which can tell which tasks didn't run from the logs too Mar 09 12:44:34 LetoThe2nd: which tasks started but didn't finish Mar 09 12:46:07 RP: into the docker container? i probably can attach. Mar 09 12:46:56 htop on the outer lxc is non-revelealing too. Mar 09 12:47:50 LetoThe2nd: the processes do updated their names to match the task they're running Mar 09 12:48:24 RP: ok. once i've got the situation reproduced again i'll try and see. Mar 09 12:49:03 * RP confirms make does behave differently depending on version Mar 09 12:56:26 Hi! I wrote a recipe for my app. It clones a git repository, which unpacks into $WORKDIR/git, so I used S=$WORKDIR/git. I need to apply a patch, which is listed in SRC_URI and works out of the box. Mar 09 12:56:26 Problem: Applying the patch also creates a hidden dir in $S containing the original source. Unfortunately, this is in scope of the app's build process and thus will not compile. Mar 09 12:56:26 Of course, I can manually de-clutter the $S directory before building (git clean -xdf /// rm -rf .pc /// whatever) or I can patch the build process to explicitly exclude the quilt directories, but both of these seem wrong/redundant. Mar 09 12:56:26 What's the correct approach here? Mar 09 13:26:43 jmiehe: OE assumes it can preserve its quilt structure there. It seems odd that something would try and build things in the patches and .pc directories, I'd probably patch the build process Mar 09 13:28:33 so it's fine to patch and build in $WORKDIR/git, and it's not unusual to have the .pc directory in there as well? Mar 09 13:29:45 not unusual at all. Mar 09 13:29:46 I wondered if I should somehow end up with sources in $WORKDIR/git and the quilt structure in (maybe) $WORKDIR/$PN-$PV Mar 09 13:30:24 a build system that rattles through searching for source files that aren't explicitly listed in directories, that's odd. Mar 09 13:32:20 jmiehe: the system isn't built to work like that and you'll have trouble trying Mar 09 13:33:11 jmiehe: normally you'd build in ${B} btw, not ${S} Mar 09 13:33:27 build output is usually separate from the source Mar 09 13:37:48 RP: I didn't touch $B though, I just set $S in the recipe head to be $WORKDIR/git. $B seems to be updated automatically; so I guess I'm building in $B :) Mar 09 13:38:32 tbh, I wasn't even aware of $B, but it makes perfect sense considering CMake exists Mar 09 13:45:47 RP LetoThe2nd: https://github.com/rossburton/ross-tools/blob/master/findfails is close but not quite Mar 09 13:46:02 zeddii: Is https://autobuilder.yoctoproject.org/typhoon/#/builders/48/builds/3176 due to needing a rev bump for genericx86? Mar 09 13:46:18 kmaincent: ^^^ I'm hoping zeddii can help us with that warning Mar 09 13:46:27 kmaincent: probably the same warning in the wic build Mar 09 13:46:40 hm what happens with systemd logs when you turn readonly fs on ? Mar 09 13:46:57 rburton: I always get this mixed up I think! Mar 09 13:52:29 Deleted archived journal /run/log/journal/e8a2662d151f4c67b0428a594bef089d/system@410a63dc3b8f4bb6ad9229edf171f4d8-0000000000039eec-0005bcdff4607250.journal (8.0M) Mar 09 13:52:36 now my system is responsive again Mar 09 13:54:09 so where does it go in a readonly file system? heh Mar 09 13:55:24 RP: hmm. no. I'll fix it, but it's a bit fussy. I'll need to get a bit of time to do it later today. Mar 09 13:57:51 RP: halstead: any known issues with git.yoctoproject.org ? for some reason, I can't push to my kern-tools repo right now. Trying to add the patch submission README, and I'm bouncing on permission denied Mar 09 14:05:55 zeddii: are you using push.yoctoproject.org ? Mar 09 14:06:47 zeddii: I should have caught the warning pre merge, sorry Mar 09 14:07:03 same as I've always been using. ssh://git@git.yoctoproject.org/yocto-kernel-tools and/or push.yoctoproject.org Mar 09 14:07:48 zeddii: should be push if you're pushing as I know there are plans to split for mirroring Mar 09 14:08:07 RP: I'm on it, I'll send a fix in my queue later today. It's something I introduced to fix the boot trace on qemux86, but it is only patched in for the qemu machines. Something is re-using the fragment, and hence is triggering the warning. so I have to detangle it a bit. Mar 09 14:08:13 push doesn't work either Mar 09 14:08:21 FATAL: W any yocto-kernel-tools bruce.ashfield@gmail.com DENIED by fallthru Mar 09 14:09:19 that being said, I have been known to think things are broken in the past and have somehow fat-fingered the push line :D Mar 09 14:09:25 since i quite often type them by hand. Mar 09 14:09:37 but I've checked this a few times, and I'm not seeing where I screwed it up. Mar 09 14:10:34 zeddii: I'm not seeing any change in the server config and I can push to things Mar 09 14:11:05 hmm. same key as all my other repos. I'll ignore it for now and try again later. Mar 09 14:12:02 zeddii: can you send me the output of ssh git@push.yoctoproject.org Mar 09 14:12:09 zeddii: checking the config it says you have access Mar 09 14:12:32 maybe I'm somehow repeaditly screwing up my passprhase. Mar 09 14:12:37 let me consult with ssh-agent Mar 09 14:17:40 RP: zeddii: Duly noted thanks. Mar 09 14:47:28 is there some recipe to get keyboard in french ? Mar 09 14:50:51 rburton: ah thanks Mar 09 14:51:22 jdrol: setxkbmap I guess? Mar 09 14:51:44 tks Mar 09 14:52:23 wouldn't it be enough to grab a sharpie and paint funny dots and dashes over random vowels? Mar 09 14:52:27 * LetoThe2nd ducks and runs Mar 09 14:53:30 LetoThe2nd: said the one with umlaut keys Mar 09 14:55:08 says the one who .. erm... *looks at keyboard* ... nevermind. Mar 09 14:57:25 jdrol: BTW, I use qwerty keyboards with English International with altgr dead key. It's the dream, can still write accent, whatever the language but the keyboard layout is qwerty which is MUCH easier to develop with IMO Mar 09 14:58:01 (did the switch something like 6 years ago, best decision ever) Mar 09 14:58:07 :) Mar 09 14:59:09 one day i'll switch keyboard to some super esorteric layout on chorded hardware, effectively making sure that nobody will ever touch my box again. but OTOH, i regularly work on a laptop. Mar 09 15:15:46 LetoThe2nd: I do enjoy watching Americans try and use a UK keyboard :) Mar 09 15:16:02 I can learn which characters are in their passwords fairly quickly Mar 09 15:18:41 RP: a real classic. thats why it makes sense to construct passwords that are layout independent :) Mar 09 15:26:57 LetoThe2nd: I see I'm not alone :D Mar 09 15:27:31 though.. that technically makes brute forcing faster... shouldn't have said that Mar 09 15:55:44 YPTM: armin is on Mar 09 16:02:44 zeddii, Ping me when you are able to troubleshoot git pushes. Mar 09 16:12:09 YPTM: Knocked off the call by an audio issue. Back after a reboot. Mar 09 16:14:59 YPTM ross joined, late Mar 09 16:19:54 what is denix's email address? Mar 09 16:25:01 ho, he's Denys I presume? Mar 09 16:25:21 vdl: yeah, look in meta/conf/distro/include/maintainers.inc in oe-core / poky Mar 09 16:26:03 smurray: yeah but I wanted to make sure about the denix -> denys mapping ;-) Mar 09 16:26:37 vdl: heh, if you look for denix in that file in master it'll be clear Mar 09 16:29:02 smurray: ho sorry, I had dunfell checked out, where he wasn't using his @denix address yet. Mar 09 16:33:25 I'm subscribed to poky@, but I need to subscribe to meta-ti@ as well to send a single patch, right? :( Mar 09 17:13:14 Hey folks, do you use any linter in your repos to check your recipes? If yes, which one? I found this one quite useful but would like to know if there are alternatives or industry standards https://pypi.org/project/oelint-adv/ Mar 09 17:17:52 emrius: never heard of that one before, thanks for the link :) Mar 09 17:17:58 afraid to run it on our layers :p Mar 09 17:29:55 looks horrible! Youve been warned Mar 09 17:30:03 I mean the output is massive Mar 09 17:30:08 but was very helpful for me Mar 09 17:30:22 start with a single recipe Mar 09 17:39:00 emrius: already blind Mar 09 17:39:15 emrius: does it not have glob support though? Mar 09 17:39:27 how did you lint a whole layer for example? Mar 09 17:44:49 qschulz: I never did. I just ran it on a couple of recipes and fixed what seemed reasonable. But I'll include it in my CI pipeline asap. Mar 09 18:02:14 my god that is the most pedantic lint tool i've ever seen Mar 09 18:02:24 FILES_${PN}' should not be overriden Mar 09 18:02:27 FALSE Mar 09 18:02:32 INSANE_SKIP should be avoided at any cost Mar 09 18:02:44 sure, but linting on it? Madness. Mar 09 18:03:15 i linted a layer by using $(find) but managed to make it hang in the process Mar 09 18:06:06 It's nice that its super pedantic. Isn't that the sole purpose of a linter? You can disable certain types of warnings as far as I understand Mar 09 18:07:43 a linter is no good if you ignore the warnings because its telling you that you should try to avoid variables Mar 09 18:07:46 yes, i should, but i can't Mar 09 18:09:46 fair enough Mar 09 18:10:24 Well, 10 stars on github looks like a rather new project. Time to get hands dirty :) Mar 09 18:21:25 oh yeah i tried that one once and all it did was make me feel bad Mar 09 18:24:35 RP: There is a libiconv patch on master-next of meta-mingw... do you still need that? Mar 09 18:24:36 :) Mar 09 18:24:51 https://marketplace.visualstudio.com/items?itemName=kweihmann.oelint-vscode Mar 09 18:25:01 ^^ will let you run it in vscode Mar 09 18:25:23 I haven't played with it though Mar 09 18:28:44 JPEW: no Mar 09 18:28:59 JPEW: newer autoconf fixed that issue Mar 09 18:37:54 RP: OK. Pushed a master-next for testing Mar 09 18:38:00 when you have a chance Mar 09 18:43:27 JPEW: https://autobuilder.yoctoproject.org/typhoon/#/builders/89/builds/3185 Mar 09 21:15:35 qschulz: so I'm checking `CONFIG_DEFAULT_DEVICE_TREE` in the u-boot menuconfig, and it says `am335x-evm`. So I'm assuming it will take `./build/tmp/work/pocketbeagle-poky-linux-gnueabi/u-boot/1_2020.07-r0/git/arch/arm/dts/am335x-evm.dts`. Is that correct? I have another one, but if I understand correctly that's a dts for the kernel, not for u-boot: Mar 09 21:15:36 `./build/tmp/work-shared/pocketbeagle/kernel-source/arch/arm/boot/dts/am335x-evm.dts` Mar 09 21:17:08 The problem I have is that I am trying to set usb to "peripheral" mode, because `dm tree` shows that the driver loaded is `ti-musb-host` and I want my pocketbeagle to be in usb gadget mode in u-boot, for fastboot. And to do that, I added `dr_mode = "peripheral"` to this dts. But it does not change the loaded driver. Mar 09 21:17:45 I am not sure if I'm rebuilding correctly, though. After I change the .dts, I run `bitbake u-boot`, then `bitbake -c clean ` and `bitbake `. Shouldn't that use the modifier dts for u-boot then? Mar 09 21:31:36 RP: I've a strange thing here which might be prserv/hashserv/sstate related. libtool-native is pulled from our sstate-mirror and will make srecord-native fail to compile. E.g.: Mar 09 21:32:18 libtool-native pulled from sstate looks like: Mar 09 21:32:29 https://www.irccloud.com/pastebin/Hyjmfftp/ Mar 09 21:32:49 but the host we're building on has /usr/lib64/gcc/x86_64-suse-linux/7/ Mar 09 21:34:56 dl9pf: Are you using uninative? Mar 09 21:35:12 yes Mar 09 21:35:57 Hmm, I suppose it gets that path from the host GCC Mar 09 21:36:16 it looks like sys_lib_search_path_spec is not adapted but 2 lines below sys_lib_dlsearch_path_spec is adapted. Mar 09 21:37:01 https://www.irccloud.com/pastebin/nLTikkI6/ Mar 09 21:38:11 so some bits are adapted to the machine that we're running on right now, but the sys_lib_search-path_spec seems to be from the host when sstate was build back then. Mar 09 21:39:01 dl9pf: There are quite a few patches dealing with sys_lib_dlsearch_path Mar 09 21:39:36 so in my case things end-up linking using the 'wrong' values in sys_lib_search_path_spec which obviously fails Mar 09 21:40:02 dl9pf: Ya. It also looks like libtool.bb should be handling that in do_install Mar 09 21:40:44 I'll follow that lead ... let's see Mar 09 21:44:02 libtool.bb does a do_install sed calls, but not libtool-native Mar 09 21:52:05 ah, its compiler_lib_search_dirs as well Mar 09 21:54:29 how does the current TOPDIR get in there at all Mar 09 22:00:12 looks to me the fixpath and fixpath.cmd in sysroots-components/x86-64/libtool-native are not catching these Mar 09 22:08:34 dl9pf: we assume libtool-native isn't host specific and can be used on different OSs :/ Mar 09 22:09:07 compiler_lib_search-dirs is really different Mar 09 22:09:15 dl9pf: knowing what generic values to use there per host is hard :/ Mar 09 22:09:25 indeed Mar 09 22:09:35 dl9pf: A quick fix might be to use the same tricky we use with pseudo-native Mar 09 22:09:51 but /usr/lib/something vs /usr/lib64/something is a difference Mar 09 22:10:09 looking ... Mar 09 22:10:24 I'd warn that whilst that works, it was partly the trigger of the runqueue issue I spent ages trying to find and it does nasty things with the task tree when an early native isn't present Mar 09 22:10:45 I think modern runqueue could handle it much better but I'd not done any work on that yet Mar 09 22:15:24 can't find what you mean in pseudo.bb Mar 09 22:23:03 compiler_lib_search_dirs needs to be mangled for libtool-native when installed through sstate to the output of g++ --print-search-dirs | grep "libraries:" Mar 09 22:37:14 I'm a bit confused: the only way I find for bitbake to build u-boot.img and put it into build/tmp/deploy/images// is by running bitbake -c cleansstate u-boot, but then I cannot change the `.dts`. How would I do that? Is there another `-c` that I can use? `-c do_compile` does not install it... Mar 09 22:44:13 RP: I reworked the webkit/riscv patch, so please drop v1 and use v2 Mar 09 22:45:10 jonesv[m]: You can try '-c clean' but I don't know enough to really help you further Mar 09 22:46:31 jonesv: apply your change as a patch Mar 09 22:46:37 Thanks Spooster . I have tried too, but then it does not reinstall it. The only one that does it is `-c cleansstate` Mar 09 22:46:56 khem: aha, so I `.bbappend` u-boot and patch it? Mar 09 22:48:08 right dts files can be patched Mar 09 22:48:30 and just listing the patch in the FILES in your .bbappend should apply it Mar 09 23:27:29 khem: thanks, updated -next Mar 09 23:49:02 dl9pf: mangling it to the right value is definitely the preferred solution Mar 09 23:49:34 dl9pf: we should be able to arrange to do that in do_populate_sysroot_setscene maybe as a postfunc? Mar 09 23:49:43 or maybe just appending Mar 09 23:52:31 dl9pf: looks like I hardcoded the tweak I was thinking of related to pseudo-native sstate in sstate.bbclass Mar 09 23:52:34 if d.getVar("PN") == "pseudo-native": Mar 09 23:52:34 d.appendVar('SSTATE_PKGARCH', '_${ORIGNATIVELSBSTRING}') Mar 09 23:53:00 I hardcoded it deliberately to stop anyone using it very easily as it has nasty side effects I mentioned Mar 10 00:09:25 khem: Spooster: same when patched: when I then go into u-boot and check with `fdt print`, my usb0 is said to be `dr_mode = "otg"`, though my dts says `dr_mode = "peripheral";` ☹️ Mar 10 00:11:02 you might have to check if your patched dts is really what is used Mar 10 00:11:24 and you also want to perhaps cleansstate and rebuild image Mar 10 00:11:26 Though in `bitbake -c menuconfig u-boot`, `Default Device Tree for DT Control` is set to "am335x-evm", and I do patch `./build/tmp/work/pocketbeagle-poky-linux-gnueabi/u-boot/1_2020.07-r0/git/arch/arm/dts/am335x-evm.dts` Mar 10 00:12:18 khem: I have been doing some cleansstate and all for 3 days now, I guess I should now try to make sure the the dts I am patching is what is being used. But I don't know how 😕 Mar 10 00:13:00 I just assumed it was this one because of what I see in menuconfig ("Default Device Tree for DT Control"), but I don't know if that's correct Mar 10 00:13:04 jonesv[m]: 'bitbake-layers show-appends | grep u-boot' to make sure your .bbappend is actually being picked up Mar 10 00:13:57 I recently had some trouble where I has to push mine around until bitbake was happy where I placed the properly named file Mar 10 00:14:39 chances are, either you need to name it u-boot_%.bbappend or move it around until it's picked up... then you can debug why the patch isn't applying from there Mar 10 00:15:15 * jonesv[m] sent a long message: < https://matrix.org/_matrix/media/r0/download/matrix.org/SRWbGKsGqwqDmUPJHmpWioBc/message.txt > Mar 10 00:16:52 or not at all? Mar 10 00:17:38 Like I can tell that my patch is being applied, because if I mess up with it, `bitbake u-boot` fails to patch. And also I can see that the file has been patched Mar 10 00:18:28 But the thing is that this `.dts` file being patch does not tell me that the file is actually being used by u-boot. It just tells me that my `.bbappend` patches the file I want to patch. How can I know this is the file I should be patching? Mar 10 00:21:55 But when I think about it, I am trying hard to get my usb0 in peripheral mode, but it is already in otg mode, so maybe that's not my issue at all, and I just got away from my initial fastboot problem 🤨 Mar 10 00:22:25 I guess I need to give up here Mar 10 00:44:29 zeddii: https://autobuilder.yoctoproject.org/typhoon/#/builders/62/builds/3180 :/ Mar 10 00:51:19 edgerouter no less! Mar 10 00:51:47 I did have a conflict there, but it built. I'll fire up a new one and re-submit the kernel SRCREV bumps for the yocto BSPs. Mar 10 01:04:18 zeddii: thanks **** ENDING LOGGING AT Wed Mar 10 02:59:56 2021