**** BEGIN LOGGING AT Wed May 05 02:59:56 2021 May 05 03:09:16 RP: I thought about just skipping it if the layer.conf wasn't there, but was trying to leave room for if we did eventually need something specific to dynamic layers. but yah, if a conf/layer.conf isn't there, it really isn't a layer so you can't be building with it anyway, and skipping it wouldn't be an issue. the os.walk just picks it up and triggers on the conf dir to declare it a layer. May 05 03:31:00 RP: Three options in your inbox for the morning, all stacking on the first one I sent. Dynamic layer detection, skip it, log it. May 05 03:41:14 After more tracing, this looks like a failure in utils.py in the copyFile method. The chown is failing on line 893, where it is trying to chown the copied file to root. May 05 03:41:36 This is happening on all intercept scripts. Almost as if fakeroot is broken, or something. May 05 04:08:15 Looks like intercept_scripts directory is listed as a member of PSEUDO_IGNORE_PATHS. May 05 04:50:39 I will report this to the mailing list. I cannot see how this is not a bug. This is the change that is causing the problem: https://git.openembedded.org/openembedded-core/commit/meta/classes/image.bbclass?id=9463be2292b942a1072eea88881b9644e55aadb9 May 05 04:50:56 It direclty conflicts with line 192 here: https://git.openembedded.org/openembedded-core/tree/meta/lib/oe/package_manager/__init__.py#n173 May 05 07:28:13 good morning May 05 09:09:52 Hello chat, i have some trouble yo use Yocto (devshell part) with WSL2, someone here is wotking with WSL2 to help me ? thx ! May 05 09:28:23 hello! I have a problem with sstate server and don't know how to debug. I have 2 clients, they both have same layers and same revisions (I've created a build from scratch) and both use the same crops/poky-container... One client downloads the sstate and succeeds and the other downloads most of them but for one particular recipe (openjre-8) it requires to rebuild. Any idea what to do? May 05 10:05:41 qschulz: michaelo: I've reworked the KCONFIG_MODE documentation patch to include the suggestions. I don't think the mail headers "In-Reply-To" thing worked out as planned, but the patch arrived on the mailinglist. May 05 10:15:01 dwagenk: thanks! It's a bit weird but eh, it made it to the ML which is the most important thing! May 05 10:45:39 Hey guys a May 05 10:46:47 quick question, is there a way to modify device tree from Linux (in user space) , May 05 10:47:20 i need to change some Frequencies in my HW and it takes so long to recompile the kernel each time May 05 10:48:10 quick google search says yes, https://developer.ridgerun.com/wiki/index.php/Edit_device_tree_at_run_time May 05 10:56:12 i k bro, how can i add the dtc fdtget fdtput Packages to my image ? does yocto provides them ? May 05 11:16:54 medaliyou: you don't need to recompile the whole kernel, only the device tree. Use dtc from you computer May 05 11:17:42 put the device tree in the right place (might have to recompile the fitimage if that's what you use, in that case mkimage fitimage.its should work just fine May 05 11:17:49 Im doing it very often tbh May 05 11:18:22 otherwise, device tree overlays loaded at runtime would probably work? May 05 11:21:37 i don't , i just bitbake my_image again and it recompiles the kernel and overlays and dts files May 05 11:22:10 to asnwer your question, you need to add dtc package to your image and I think it'll be enough May 05 11:22:36 dtfget dtfput ? May 05 11:22:37 medaliyou: you don't need yocto to compile the kernel or device tree, or u-boot May 05 11:23:01 https://git.kernel.org/pub/scm/utils/dtc/dtc.git/tree/ has all those tools May 05 11:23:27 only http://cgit.openembedded.org/openembedded-core/tree/meta/recipes-kernel/dtc/dtc.inc?h=master#n26 seems to be put into dtc-misc but no use for you? May 05 11:25:14 or just bitbake dtc recipe with Yocto, then run oe-pkgdata-util find-path 'fdtget' to find out if a package provides it and how it's named May 05 11:25:38 alright May 05 11:26:06 and after recompiling the dts May 05 11:26:13 in the userspace May 05 11:26:21 how to provide it again for the kerne l May 05 11:26:23 ? May 05 11:31:19 find where you device tree is stored and change it there May 05 11:31:38 in the example given by JanneKiiskila, the dtb is stored in the /boot partition May 05 11:36:04 how to provide the DTB in rootfs , i have an empty /boot and [find] can't find any dts dtb files May 05 11:39:18 can't help, that's specific to your board. U-Boot (or the bootloader that you're using) needs to load it from somewhere. Find where, load this partition if possible, read the file, modify the file, flash/save, reboot. May 05 11:39:46 U-Boot also has the ability to modify the device tree before it's used by the kernel, fdt tools are available May 05 11:43:36 thank you so much bro May 05 13:21:33 Can anyone give me advice on how to find out why an automatically set linux kernel cmdline var got set the way it was? May 05 13:21:34 https://lists.yoctoproject.org/g/yocto/message/53389 May 05 13:22:12 I think it's the bootloader. May 05 13:24:19 good afternoon, a simple question this time;-)  am compile for an imx6 board, just need to know where the type of image is listed (wic etc ...) as the output does not produce an SD card image which is later needed, as usual I have inherited an incomplete and seriously hacked project May 05 13:44:53 intera_91: `bitbake -e my-image-name | grep -B 10 ^IMAGE_FSTYPES` should give you the answer and also list the classes and bbappends modifying the variable. If you are looking for the options of IMAGE_FSTYPES useable with a certain board I'd do a full text search for IMAGE_FSTYPES through the BSP layer. May 05 13:45:51 dwagenk:  course thank you so much May 05 14:26:06 manuel1985: if it's using the original bootloader from RPi and not U-Boot, I guess you're out of lukc as I'm not sure there are sources available for it? Might be wrong it's a very vague recollection May 05 15:13:21 Can someone point me in the direction of how to make an sdk without the cross toolchain? May 05 15:14:05 I'm using external toolchains with yocto but still want some native tools built by yocto in an "sdk" May 05 15:15:03 rabbit9911: buildtools-tarball would be an example May 05 15:26:01 hmm, qemuppc is spewing irq errors into the logs May 05 15:26:10 related to serial ttys May 05 15:31:12 does anybody use overlayfs ? May 05 15:31:23 I'm trying it simply with: May 05 15:35:05 never mind May 05 15:39:15 try booting the the irqpoll option says the kernel. Except I already have that :/ May 05 15:59:27 zeddii: I'm looking at the qemuppc shutdown issue. It uses a serial command to shut down so if the serial login fails, all bets are off. It is spewing kernel serial irq kernel errors. Any ideas? May 05 16:51:07 RP, try an NMI on zeddii ; ) May 05 17:20:38 zeddii: interestingly, chucking # CONFIG_SERIAL_PMACZILOG_CONSOLE is not set into the config seems to make things a lot happier, no irq tracebacks May 05 17:20:56 zeddii: and looking at how that code works, I'm not entirely surprised May 05 17:21:18 for ppc, I'm tempted by that bandaid May 05 17:42:01 armpit: he's doing well at the masking :) May 05 18:11:46 * zeddii is enjoying an all hands meeting. Merger fun. May 05 18:16:57 zeddii: I don't miss that kind of thing! May 05 18:17:25 zeddii: surely ppc irq handling is much more fun? :) May 05 18:23:03 heh. This is more likely due to the qmp stuff (as you mentioned), and not directly due to the kernel bumps I had recently sent ? May 05 18:23:42 we could try ppc64 and delete ppc32, as well. but I have no objection to trying that option. It wasn't giving us anything we really needed. May 05 18:25:09 zeddii: ppc64 is a different discussion I think :/ May 05 18:26:23 zeddii: I suspect that something in the compiler changes changed the boot timing and that made the serial issue show up more due to bad timing May 05 18:27:35 I can trivially roll it into the qemuppc config and send a SRCREV bump, or we could carry it out of the kernel as an addon cfg. Either or works. May 05 18:27:57 zeddii: well, I have the cfg appended locally but I know how you hate it when I do that :) May 05 18:28:36 :D I can do that and send you a SRCREV bump for the kernel meta. easy enough. May 05 18:28:40 zeddii: I'm leaning to putting it in the ppc config and bumping and then we can see if the issue gets any better May 05 18:29:26 zeddii: I've noticed other issues in the code too and we can improve the logging a bit May 05 18:30:49 * zeddii nods May 05 18:32:06 was an interesting stroll through it, when I was looking :D May 05 18:35:09 * paulg envisons zeddii is now covered in burrs and hayseed and mud. May 05 18:39:53 RP: change is done, but I want to run one configure of that kernel to confirm. My sstate is out of date for ppc32, so I'll have to watch it churn for a few. May 05 18:43:01 zeddii: no problem, thanks May 05 19:02:39 zeddii: you mean _CONSOLE in that log message right? May 05 19:03:15 yah. not sure what happened to that cut and paste. May 05 19:03:28 zeddii: I'll tweak :) May 05 19:03:30 +# CONFIG_SERIAL_PMACZILOG_CONSOLE is not set May 05 19:03:45 zeddii: I thought you would mean that but wanted to check :) May 05 19:03:58 worth checking, stranger things have happened :D May 05 19:04:04 zeddii: disabling serial entirely would be a different problem May 05 19:04:14 ssh only! May 05 19:04:15 :D May 05 19:04:22 * RP is having a day like that May 05 19:04:30 netconsole and netpoll. I remember those days. May 05 19:20:02 Is there such thing as nativesdk--src? I want to ship the source in case I have to debug some of the tools. May 05 19:26:40 RP: This commit (https://git.openembedded.org/openembedded-core/commit/meta/classes/image.bbclass?id=9463be2292b942a1072eea88881b9644e55aadb9&h=master) appears to have broken this line of code (https://git.openembedded.org/openembedded-core/tree/meta/lib/oe/package_manager/__init__.py?h=hardknott#n192). Should a bug be filed? May 05 19:28:39 resoum: I did start to look at that as I think you mailed somewhere about it. I don't understand why that is breaking for you but not the autobuilder or anyone else May 05 19:29:21 resoum: We obviously need to work out what is wrong but it isn't making any sense to me May 05 19:32:50 resoum: we don't really care about ownership here so ideally we'd use a different copy function which doesn't copy the ownership over May 05 19:33:28 resoum: sadly adding a "no owner" mode to bb.utils.copyfile will complicate it and non of the shutil copy functions look more helpful May 05 19:37:12 resoum: can you confirm exactly what the error message you saw was please? May 05 19:37:47 RP: Here it is precisely cut and pasted: https://gist.github.com/chuckwolber/8d5fd45f6410c9a5775dbbf3bae69628 May 05 19:38:26 I split the chown and chmod into separate functions _just_ to be sure, and it definitely is a chown, so that is what pointed me at the pseudo failure. May 05 19:40:09 resoum: Why is nobody else seeing that though? :/ May 05 19:40:40 resoum: (and thanks, that paste does confirm a couple of things) May 05 19:40:53 RP: He is seeing it too: https://gist.github.com/rpcme/808f79afd42abdc0d3245acc7063f491 May 05 19:41:19 resoum: well, ok, why are some seeing this but I'm not and the autobuilder isn't either May 05 19:41:32 Those messages were Warnings BTW. But the udev-hwdb actually failed to run and broke the build. That is because the chmod was _after_ the chown, so the postinst_intercept script could not execute. May 05 19:41:48 RP: yup, I am confused by that too. May 05 19:42:10 The chain of logic is inescapable to me. If you use PackageManager in your image build, you _will_ hit this so far as I can tell. May 05 19:42:22 And I really do not know how to _not_ use PackageManager. May 05 19:42:54 resoum: the autobuilder builds images using that, as do I locally, as so many others. Why is this the first time we've seen the reports? :/ May 05 19:43:28 RP: Yup, very weird. I looked for code paths that would avoid it. Maybe a different package type superclass is bypassing the PackageManager constructor? May 05 19:43:51 RP: That was the only thing I could think of. We use ipk, but deb and RPM all looked the same to me. May 05 19:43:59 resoum: most images would need those scripts May 05 19:44:16 resoum: To fix this, we need to understand why its only happening in some cases :/ May 05 19:44:27 RP: Yup. And how that copyfile works without pseudo support, is way beyond me. May 05 19:44:42 RP: Unless others are chowning to !root? May 05 19:48:34 resoum: could you also share your full PSEUDO_IGNORE_PATHS please? May 05 19:49:11 resoum: bitbake initramfs-image -e | grep PSEUDO_IGNORE_PATHS= May 05 19:53:21 resoum: I confirmed that locally its using the UID of my local user, not root May 05 19:54:49 resoum: and it does that since there is an entry in PSEUDO_IGNORE_PATHS matching XXX/scripts May 05 20:00:54 RP: Sure thing. Standby while I spin up a devshell May 05 21:02:13 what does the sysroot-native command "spp" do? May 05 21:09:15 * yates checks his internet connection May 05 21:12:00 in run.do_kernel_metadata: bsp_definition=$(spp ${includes} --find -DKMACHINE=qemucsky -DKTYPE=standard) May 05 21:14:50 has the yocto project been discontinued? May 05 21:14:58 it's awfully quiet around here... May 05 21:18:12 rburton: do you know? May 05 21:21:22 RP: PM'd you a gist with the build environment. May 05 21:33:28 yates: It comes from this package: https://github.com/openembedded/openembedded-core/blob/master/meta/recipes-kernel/kern-tools/kern-tools-native_git.bb May 05 22:05:17 resoum: thanks. For the logs, it looks like the source files are actually owned by root. That isn't a common situation. We can fix this by using shutil.copy() instead of bb.utils.copyfile() May 05 22:05:58 +1 May 05 22:17:23 hi all -- how do you guys deal with python packages depending on the compilation of protobuf .proto files? May 05 22:30:40 moto-timo: sorry to ask directly, are you aware of any protobuf-distutils use? I need this .proto compilation step and I want to make sure that packaging protobuf-distutils first is the good way to do May 05 22:32:50 v2d: i have never needed to do that, but there is a recipe http://layers.openembedded.org/layerindex/recipe/72577/ May 05 22:33:17 v2d: perhaps someone else has direct knowledge May 05 22:35:08 resoum: ok thank you. May 05 22:41:06 RP: You might find the patch I just sent interesting.... we are invoking touch *a lot* when reproducible builds are enabled May 05 22:42:44 JPEW: nice! :) May 05 22:43:31 Might help some of the AB-int problems? May 05 22:44:57 JPEW: I suspect we have other much more problematic fork bombs in the system (e.g. configure) but can't hurt! May 05 23:06:47 moto-timo: if I'm not mistaken, protobuf-distutils is different from python3-protobuf, it seems to be a distutils extension, not the protobuf library May 05 23:07:16 Cannot help you there. Zero knowledge. May 05 23:19:06 You might be able to use devtool add and https://pypi.org/project/protobuf-distutils/1.0/ to get started. May 05 23:27:02 moto-timo: does OE provide support for distutils extensions (aka setuptools plugins)? May 05 23:27:24 I've not yet figured out how to plug such extension with the setuptools3 bbclass May 05 23:53:22 v2d: I don't see why it wouldn't just be (1) create a python3-protobuf-distutils recipe (2) BBCLASSEXTEND = "native" (3) DEPENDS += "python3-protobuf-distutils-native" May 05 23:53:47 v2d: but that is honestly the limit of my knowledge here... I just know how I expect computers to work and they usually do May 06 00:48:45 is it possible to embed newlines in module_conf diriectives used by kernel-module-split.bbclass ? May 06 00:49:07 (if not we're resorting to an anonymous python function) **** ENDING LOGGING AT Thu May 06 02:59:56 2021