**** BEGIN LOGGING AT Thu Oct 29 02:59:57 2020 Oct 29 03:24:02 RP: poky/bitbake/lib/bb/cache.py:546, I don't think this path is ever taken otherwise it would crash? Oct 29 03:25:34 kiwi_29: other than what's in the YP docs and my container presentations, I think there have been a couple of presentations on using it for builds with of both host & VM images. There are couple of presentations at the YP Summit tomorrow that I believe will cover it Oct 29 03:27:02 RP: configData isn't define in the scope. Another thing. `def parse` only takes filename and appends. Oct 29 03:28:16 kiwi_29: there are a few other people that will likely be on channel tomorrow during the day in the US that I think have mentioned they use multiconfig to build many images in one build out a single tree Oct 29 03:31:18 RP: My real question is: Is this dead code? This `else` branch is probably never taken for some reason. Oct 29 03:33:36 Is there a readme/wiki to setup a "good" debugging experience with bitbake? I'm trying with VSCode, but it doesn't look like it can follow subprocesses. Oct 29 03:39:16 smurray how to join the yocto summit tomorrow? Oct 29 03:41:20 kiwi_29: info is here, I think you need to go through the ELC event page to register: https://wiki.yoctoproject.org/wiki/Yocto_Project_Summit_2020 Oct 29 07:56:06 yo dudX Oct 29 07:56:38 i'll look at the upcoming summit at https://www.twitch.tv/theyoctojester at 9:00 over a mug of coffee, join us! Oct 29 07:56:49 e.g., in 4 minutes :) Oct 29 08:20:23 Hi, I'm trying to extend the core-image-minimal, so that I can run "bitbake core-image-minimal" with my own modifications. Is it possible to add image features from another layer? Does anyone know a tutorial? Oct 29 08:21:03 guest1243: write your own image, you can start by copy-pasting core-image-minimal Oct 29 08:21:03 guest1243: create your own image Oct 29 08:21:32 guest1243: tutorial is here: https://youtu.be/nqHylLP2NmA Oct 29 08:21:41 So it is not recommended/possible to use .bbappend files to extend an existing image? Oct 29 08:21:50 strongly discouraged. Oct 29 08:21:59 possible, yes. Oct 29 08:22:05 core-image-* are reference images Oct 29 08:22:17 just copy them if they're good starting points Oct 29 08:22:34 Okay, understood Oct 29 08:23:15 its all in the video. :) Oct 29 08:49:40 good morning Oct 29 08:53:59 Hi! Do you know that effect with the console during booting of Linux when suddenly all output is just cryptic character salad? As if the encoding is wrong. As soon as I restart my serial client program everything is OK again Oct 29 08:55:30 Sometimes it fixes by itself, which looks like this: Oct 29 08:57:24 [ 0.122948] 003: CPU3: Booted secondary processor 0x0000000003 [0x410fd034] 2À™Ü‚rŠš¢¢ÊÊê‚‚‚ÒšÕ¥¢õÑ…±z™¢‚ɽ•ÍͽÉ́ÓS'SSS ‚‚ÒUÉ«VW—®Ö.'HYYºY,W–'¤&¦K¬Z]Q¦˜ Support[ 0.145265] 000: CPU features: detected: CRC32 instructions Oct 29 08:58:27 fbre: very likely bad hardware or cpu registers settings Oct 29 08:58:59 fbre: i have seen serial-usb converters being massively confused too, where a connection reset would fix it at best, and only unplugging both sides at worst. Oct 29 08:59:04 it all happens with a git version of dunfell on an imx8 eval board Oct 29 08:59:43 fbre: time for a bisect then... if you have a known good working state from earlier. Oct 29 09:01:25 hmm... ok thanx. I just thought it may be a well known issue somehow Oct 29 09:04:05 probably a problem in the serial-usb stuff. I did not see that at all with previous yocto branches (sumo, zeus, warrior) Oct 29 09:11:19 roussinm: It probably isn't used due to the access patterns bitbake as to cached data? Oct 29 09:24:05 fbre: let me guess - you use FTDI-based usb-serial converter? I've seen those issue a lot with FTDI chips on a lot of different SOCs (not only imx8) and various yocto branches Oct 29 09:24:42 try to replace the converter to another one, that is what usually helped me Oct 29 09:45:25 zandrey_: Not knowingly. Now I searched the kernel config for FTDI and found USB_FTDI_ELAN =n and USB_SERIAL_FTDI_SIO =n Oct 29 09:45:57 zandrey_: Not sure what my current serial-usb converter is Oct 29 09:52:59 Where can I find the choice of serial to USB converter in bitbake virtual/kernel -c menuconfig? Oct 29 09:58:14 fbre: i was talking from the other side: the converter you use on the desktop to connect to the HW - that one is usually gets messed up. Oct 29 09:58:57 zandrey_: I use TeraTerm on Windows 10 Oct 29 09:59:30 zandrey_: Which actually never has such problems Oct 29 10:01:00 fbre: Bingo! :D TeraTerm is the one I had most of those corrupted strings with! Oct 29 10:02:24 zandrey_: OK, so you mean a different serial client should help. Hmm, strange... but I'll give it a try (y) Oct 29 11:15:57 kanavin_home: Looks like master-next failed in glib, have you looked at that? Oct 29 11:17:30 RP: I can take a look. Oct 29 11:18:10 RP: which build? Oct 29 11:18:21 a-full 1519? Oct 29 11:18:26 kanavin_home: https://autobuilder.yoctoproject.org/typhoon/#/builders/83/builds/1519 Oct 29 11:18:30 kanavin_home: yes Oct 29 11:18:47 kanavin_home: I suspect its the patch added in master interacting badly? Oct 29 11:19:06 kanavin_home: I spotted additional issues with gstreamer in no-x11 and reported those on the list Oct 29 11:19:33 hello, I've got an old broken version of u-boot that gets confused about libfdt headers from the dtc-dev package. Can I blacklist a package from getting included in the recipe-sysroot-native direcory? Oct 29 11:20:03 beacuse I do not need /recipe-sysroot-native/usr/include/fdt.h Oct 29 11:20:13 to build u-boot Oct 29 11:20:40 RP: yes, I think you need to revert Ross's commit http://git.yoctoproject.org/cgit/cgit.cgi/poky/commit/?id=5689a70b23df001d5502bf756f115a84b29b0159 Oct 29 11:21:16 kanavin_home: ok, I'll try that Oct 29 11:23:21 kanavin_home: updated and build fired, thanks for confirming Oct 29 11:24:19 RP: I am also doing extra confirmation that the patch is indeed already included in the new version Oct 29 11:31:54 RP: it is Oct 29 11:35:26 kanavin_home: thanks. I've squashed that into the upgrade Oct 29 11:36:52 RP: thanks, I will send out additional timezone fixes (that also need a dunfell backport) and a 2.66.2 update that includes them - apparently 2.66.1 doesn't quite fix all of the issues https://gitlab.gnome.org/GNOME/glib/-/releases#2.66.2 Oct 29 11:41:00 kanavin_home: sounds good. I know Steve is looking at how to fix glib in dunfell so we can take th tz updates Oct 29 11:48:51 RP: nevermind, the 2.66.2 fixes apply only to 2.66.1, and not to 2.64.x Oct 29 11:49:03 however, 2.66.2 update still needed, I will send that Oct 29 12:05:52 is there any documentation about the artifacts in the deploy directory? I get a core-image-minimal-initramfs-geodelx.cpio.gz and bin file there and everything and not really sure why... Oct 29 12:15:47 carlsb3rg: I think that would come from cpio.gz being in INITRAMFS_FSTYPES Oct 29 12:16:12 actually that is the default value Oct 29 12:22:22 * fbre likes initramfs Oct 29 12:28:32 yeah...I figured that out eventually with bitbake -e Oct 29 13:07:11 Hello guys !! I want to build a kernel module out of the tree. to do that I've generated the SDK (adding TOOLCHAIN_TARGET_TASK_append = " kernel-devsrc" to my conf), I've installed it and I've sourced the variables with Oct 29 13:07:34 source /environment-setup-aarch64-xilinx-linux Oct 29 13:08:03 then I've run Oct 29 13:08:31 $ cd /sysroots/aarch64-xilinx-linux/usr/src/kernel Oct 29 13:08:40 $ make scripts Oct 29 13:08:59 after that the compilation process exits with: Oct 29 13:09:05 HOSTCC scripts/extract-cert Oct 29 13:09:13 but if I run Oct 29 13:09:25 $ cd /sysroots/aarch64-xilinx-linux/ Oct 29 13:09:33 $ find . -name bio.h Oct 29 13:09:47 ./lib/modules/4.19.0-xilinx-v2019.1/build/include/linux/bio.h Oct 29 13:10:09 I think that I am missing some flags somewhere Oct 29 13:32:02 hmm, buildhistory failures with master-next :( Oct 29 14:10:53 BitBake devs, how do you interactively debug the bitbake daemon subprocess? Do you have any wiki/readme that explain a certain setup that works? I could probably try gdb... Oct 29 14:12:26 hi guys Oct 29 14:12:43 I have a really dumb issue Oct 29 14:13:20 after changing the kernel configuration, my image still loads, and I can see that there is no kernel panic or something Oct 29 14:13:53 but bootup procedure does not ends up with a command prompt Oct 29 14:15:16 linums: well what was the kernel configuration change? :D Oct 29 14:15:46 well, I've changed pretty much everything :D Oct 29 14:16:00 I just tried the current 5.8 kernel config from ubuntu Oct 29 14:16:40 but actually this machine works with that kernel from ubuntu Oct 29 14:18:07 was it a bad idea? :D Oct 29 14:19:09 i have no clue what you are talking about. because the condensed question sounds like "hey. i've got a machine that works with the ubuntu kernel. and works with a kernel i have. now i turned the config upside down and it doesn't work anymore. why?" Oct 29 14:23:48 well, I am trying to figure out why the video card is not working with my kernel, so I just had this idea, that if I use the same configuration as the ubuntu kernel was built, maybe I can find what I need for that video card Oct 29 14:26:29 in my mind I thought that it can not be considered as an upside down turn for the configuration Oct 29 14:26:33 maybe it's not true Oct 29 14:31:00 video often involves a lot more than just the kernel config, sadly. especially on x86. i really can't comment there, being an arm guy Oct 29 14:31:56 yeah, this I've faced in the previous weeks :( Oct 29 14:33:49 well, I've added the kernel module what this video card needs, but than I had to realize, that some kernel parameters should be added to use it Oct 29 14:34:35 so after just enabling the amdgpu related stuff I received a kernel panic Oct 29 14:36:37 so I just decided to try the whole ubuntu kernel config, to check if the kernel config is the issue, but actually I was not expecting this issue Oct 29 14:37:24 at least a kernel panic would give me a hint why it does not boot up fully :D Oct 29 14:41:07 I've only worked with arm devices too until now, and it was just better :D Oct 29 15:36:31 roussinm: I'm afraid I'm oldschool, printk style debugging. You probably could make it start pdb on a port you could connect to? Oct 29 15:37:19 pdb? Never heard of it probably? Oct 29 15:37:22 yeah pdb works Oct 29 15:37:37 https://github.com/rossburton/meta-ross/blob/master/classes/pdb.bbclass Oct 29 15:38:13 call pdb and it will start it on a socket Oct 29 15:38:18 so you can connect and step Oct 29 15:38:35 Let's see if I can integrate that inside VSCode. Thanks! Oct 29 15:42:10 * rburton didnt consider that vscode has pdb integration and gets a bit excited Oct 29 15:44:13 Hello all, i have a problem with mutiple version of openssl , and used the PREFERRED_VERSION Oct 29 15:44:26 rburton: go go go https://youtu.be/hsjfAO870bk Oct 29 15:44:27 roussinm: vscode uses https://github.com/microsoft/debugpy should be trivial to hook that up in a bitbake class Oct 29 15:44:55 LetoThe2nd: have you considered writing a bot to auto-reply with metal covers of words that people write? Oct 29 15:45:00 in the local.config : PREFERRED_VERSION_openssl_forcevariable = "1.1.1%" Oct 29 15:45:18 ERROR: Multiple versions of openssl are due to be built Oct 29 15:45:18 rburton: actually, nope. Oct 29 15:45:23 rburton: in a bitbake class, sorry I'm really a newbie, I just started to look at the bitbake repository to start contributing! So I guess there is some kind of learning curve. Oct 29 15:45:33 Only one version of a given PN should be built in any given build. You likely need to set PREFERRED_VERSION_openssl to select the correct version or don't depend on multiple versions. Oct 29 15:46:31 Sorry i have just signed in and didn't know there is a discussion Oct 29 15:46:44 heh no problem GeGo Oct 29 15:47:17 roussinm: ok debugpy can literally hijack existing python binaries so just give it the bitbake pid and you're sorted, in theory Oct 29 15:48:18 rburton: I would have to start bitbake first though... will it follow the daemonized thread? Oct 29 15:48:37 depends what you're debugging. bitbake spawns a lot of processes Oct 29 15:49:13 GeGo: maybe you have something that has a dependency on a specific version of openssl, and therefore your PREFERRED_VERSION doesn't work as expected? just guessing Oct 29 15:50:55 rburton: but seriously, what would you think of me if i would leave the tedious work of annoying you with carefully selected hilarious metal videos to something as boring as a bot? Oct 29 15:51:18 you'd actually get more respect Oct 29 15:52:10 rburton: i don't need yours, i already have https://youtu.be/SQnrbdChHrg Oct 29 15:52:18 veni, vidi, vici! Oct 29 15:53:03 rburton: I'm trying to explore the startup of bitbake, before it starts building stuff. Oct 29 15:54:31 LetoThe2nd: nicely done Oct 29 15:59:31 Hmm, how come when I bitbake and image (e.g. core-image-minimal) and everything is in sstate, it still re-runs do_rootfs? Oct 29 16:01:05 At least, on a new sandbox.... if I do `bitbake core-image-minimal; bitbake core-image-minimal` the second does nothing Oct 29 16:01:46 RP: as in -Os? Oct 29 16:02:05 alejandrohs: yes Oct 29 16:02:29 JPEW: we don't store images in sstate Oct 29 16:03:00 JPEW: we do use the sstate code for images to allow their cleanup, there is no sstate object written Oct 29 16:03:13 RP: I'll run a test, and check the improvement Oct 29 16:03:14 RP: Ah, OK Oct 29 16:03:31 JPEW: we probably could, its just slow as the things are so large Oct 29 16:04:54 Sure. I'm testing a CI setup with NFS sstate and was confused as to why it still was running that task when I retriggered the build Oct 29 16:05:13 zeddii: I forgot to mention, can you add qemuarm64/tiny? it blows up as soon as you try to build it Oct 29 16:05:51 yah. no problem. I'm just updating to 5.10, and can test it against that. Oct 29 16:07:27 alejandrohs: the arm people need to get a grip ;-) (hi jonmason, rburton :) Oct 29 16:08:02 wait what? Oct 29 16:08:29 for the arm people (jonmason, rburton) here you go: https://youtu.be/9yH6-CIt7DE Oct 29 16:08:45 ok you're so fast now i think you are a bot Oct 29 16:08:48 * LetoThe2nd strikes, ducks and runs... time to call it a day until open mic Oct 29 16:09:12 rburton: i just happen to *know* so much already that its no searching anymore, just finding the link. Oct 29 16:11:05 zeddii: thanks Oct 29 16:16:21 open a bug, we'll get to it Oct 29 16:31:09 any idea why qtbase and qtwayland would produce empty ${PN}-src packages? qtdeclarative-src works correctly Oct 29 16:35:37 RP: 15% size improvement on core-image-minimal Oct 29 16:35:55 x86 Oct 29 16:36:24 seems to work properly as well Oct 29 16:36:32 to me thats enough for me to send a patch Oct 29 16:39:53 alejandrohs: go for it! Oct 29 16:40:08 alejandrohs: I'd like to see what some of the other things in khem's presentation do too Oct 29 16:40:30 I'm literally checking loop unrolling rightnow Oct 29 16:40:38 *disabling Oct 29 16:40:42 :) Oct 29 16:41:29 theres a patch for poky-tiny pending to update and reduce busybox size on the ML btw, no rush, just in case you hadn't noticed Oct 29 17:04:34 alejandrohs: I saw it, I'm struggling a little with getting everything queued Oct 29 17:14:24 * zeddii is saving his patch bombs for another week. Oct 29 17:27:54 zeddii: the thing is yours are probably tested so less of an issue Oct 29 17:31:08 JPEW: we're still seeing that netbase reproducible issue on the autobuilder. I guess that my patch isn't enough to invalidate the bad equivalence entries? :/ Oct 29 17:41:08 RP: a couple of kilobytes with specific compilation flags IMO not enough to enable by default Oct 29 17:41:14 RP: patch set Oct 29 17:44:12 sent* Oct 29 17:44:41 alejandrohs: queued in -next, thanks Oct 29 17:52:41 alejandrohs: are you using LTO Oct 29 17:53:02 that should help a bit with size too Oct 29 17:55:17 openvswitch shrunk by 30% with lto Oct 29 18:46:20 ugh, wow, meta-imx is .. not good Oct 29 18:46:36 bbappending busybox in a bsp layer? really? Oct 29 18:46:38 * kergoth bangs head on desk Oct 29 18:46:56 still?? Oct 29 18:47:23 and the stuff in their layer.conf is an abomination Oct 29 18:51:49 kergoth: I don't dare look :( Oct 29 19:16:32 Hi I'm wondering if someone can point me in the right direction here.. I'm working on building chromium-ozone-wayland from meta-browser's master branch (claims dunfell compatibility). I'm targeting a boundary devices imx8 SBC, using their official bsp support for dunfell. The issue I've run into is specifically during do_compile - I get an error Oct 29 19:16:33 saying "ninja: build stopped: subcommand failed". When I look at the compile log I see an error regarding "ld.lld: error: unable to find library -latomic". I've tried adding "libatomic-ops libatomic-ops-native libatomics-ops-native" to the DEPENDS via a bbappend, and have added libatomic to the RDEPENDS_${PN} via bbappend but am still seeing this Oct 29 19:16:33 issue. Can anyone point me in the right direction here? Oct 29 19:30:29 kernelsandals: libatomic should come with gcc-runtime Oct 29 19:31:16 Agreed, when I tried to add libatomic to the DEPENDS though I got an error that gcc-runtime only RPROVIDES it Oct 29 19:31:36 The chromium-ozone-wayland recipe is built using clang as well Oct 29 19:31:55 check recipe-sysroot/ inside the chromium build do you see this .so file ? Oct 29 19:32:17 yes I am aware of it being using clang, but runtime is still gnu Oct 29 19:33:33 kernelsandals: I'd start building it for a similar architecture that you know it works qemuarm64 for example, if it does work on that one start checking what the bsp for imx8 is overriding, cant say fo sure but I think libatomic DEPENDS should be just libatomic since it comes from gcc-runtime Oct 29 19:33:35 e.g. I see recipe-sysroot] Oct 29 19:33:38 % find . -name "libatomic*" Oct 29 19:33:40 ./usr/lib/libatomic.so.1.2.0 Oct 29 19:33:42 ./usr/lib/libatomic.so.1 Oct 29 19:33:44 ./usr/lib/libatomic.so Oct 29 19:33:46 ./usr/lib/libatomic.a Oct 29 19:34:46 kernelsandals: what khem said Oct 29 19:35:24 Hmm, so recipe-sysroot in build/tmp/work/aarch64-mx8mm-fslc-linux/chromium-ozone-wayland/85.0.4183.102-r0/recipe-sysroot/usr/lib64/ does have libatomic.so.1.2.0, libatomic.so.1, libatomic.so, libatomic.a Oct 29 19:35:49 ok so is it failing in some native part of build then ? Oct 29 19:36:19 Yeah, its failing on specifically yocto_native/brotli Oct 29 19:36:46 right, it perhaps is them expecting this on your build host Oct 29 19:37:27 I have installed libatomic via yum on my build host - packages for libatomic, libatomic-static and libatomic_ops-devel Oct 29 19:37:36 eg. on archlinux gcc-libs provides it Oct 29 19:39:02 kernelsandals: can you pastebin your failing complete linker cmd somewhere ? Oct 29 19:42:36 @khem This is the log I have from do_compile -- https://pastebin.com/B7xASDME Oct 29 19:43:25 @khem This is the log I have from do_compile -- https://pastebin.com/B7xASDME Oct 29 19:51:11 kernelsandals: yeah i see, so its using clang-native with lld, so one option would be to try add -fuse-ld=gold to HOST_LDFLAGS Oct 29 19:51:36 another option might be to add -L/usr/lib Oct 29 19:52:01 but clang++ should have been able to find it if its installed in /usr/lib Oct 29 19:54:45 can you try a simple testcase with clang++ from recipe-sysroot-native with latomic something like clang++ hello.cpp -latomic Oct 29 19:54:50 and see if that works ? Oct 29 19:59:24 Tried adding -L/usr/lib and -fuse-ld=gold to HOST_LDFLAGS but got the same error. I'll have to see if I can put something together for that second testcase.. Oct 29 20:01:34 kernelsandals: you could add verbose flags to the linker to get more info on whats happening Oct 29 22:43:51 \o/ 10th anniversary, thank you all **** ENDING LOGGING AT Fri Oct 30 02:59:57 2020