**** BEGIN LOGGING AT Thu May 14 02:59:57 2020 May 14 03:03:52 jonmason, JPEW: while our RTOS team hasn't moved to gcc-arm-none-eabi 9 yet, it's in the works, so fine with me. May 14 04:16:42 * prpplague waves at kergoth May 14 04:17:11 prpplague: howdy! how are you? May 14 04:17:21 denix: dandy May 14 04:17:24 denix: all good May 14 04:17:29 denix: you? May 14 04:17:52 prpplague: great to hear! May 14 04:18:05 prpplague: fine here as well May 14 04:19:23 prpplague: balls deep in work - pardon my french :) you? May 14 04:19:32 denix: same May 14 04:21:21 denix: focused on mostly hardware, but i needed to do a build using yocto this evening and had a bit of nostalgia May 14 04:21:27 prpplague: haven't heard from you in a long time - how's new order treating you? :) May 14 04:21:40 just curious if old kergoth was still hanging out on IRC May 14 04:22:07 denix: yep all good, just moved away from active engagement in FLOSS May 14 04:23:06 prpplague: yeah, kergoth is around - I'm sure he'd be able to help you May 14 04:23:46 prpplague: yeah, don't forget about us FLOSS folks here... :) May 14 04:24:07 denix: hehe, well it was more of a "howdy" to kergoth, since i still have IRC logs of him when he was at TI and first came up with the initial concept of openembedded May 14 04:26:11 prpplague: those were the days!!! May 14 04:26:25 indeed May 14 04:27:51 prpplague: it seems everyone had to be at TI at some point of their career... good or bad, nevertheless May 14 04:29:16 denix: hehe right of passage May 14 04:31:25 prpplague: regardless of employment, it is great to hear from you and nice talking to you again, my friend! ping me if you need anything! May 14 04:31:39 denix: will do! thanks May 14 04:58:13 jonmason: I agree with denix , I think baby steps are fine, but lets also keep the end goal in mind and dont forget to set it up properly with TCLIBC=baremetal/newlib May 14 08:37:15 Hello there, I am building core-image-minimal for beaglebone using meta, meta-poky, meta-yocto-bsp, meta-openembedded/meta-oe layers. It builds successfully, but when I try to login, just after I wrote username it gives me an error that says: getty: tty1: can't execute '/bin/login'. I checked my root partition and all executables are broken May 14 08:37:16 symbolic link to /bin/busybox.nosuid. May 14 08:37:48 *all executables under /bin May 14 08:38:33 ilkmc2rd: on which version? also using meta-ti? May 14 08:39:20 tried with version zeus and warrior. same result. May 14 08:39:37 No not using meta-ti May 14 08:40:40 just using 4 layers that i mentioned above, and not doing any changes on local conf. just changing machine to beaglebone-yocto May 14 08:41:19 ilkmc2rd: while that sounds like a bug worth investigating, for real world applications i'd suggest adding meta-ti, as thats the proper board support for the BBB. the yocto-internal one is more like, lets call it a "test vehicle" May 14 08:42:43 I will try with meta-ti, but will it fix that broken symbolic link bug ? May 14 08:43:29 ilkmc2rd: last time i build meta-ti for the BBB it happily logged in. if not, please file a bug. May 14 08:43:38 Is there any way to not use busybox for that image? May 14 08:44:17 ilkmc2rd: getting rid of busybox completely is somewhat involved, as far as i know May 14 08:45:45 ilkmc2rd: see: https://youtu.be/MKpLQ2bGsAk May 14 08:46:26 Letothe2nd: Yes tried something and didn't succeed. Will try build using meta-ti and will check the youtube link May 14 08:46:43 Letothe2nd: thank you May 14 08:47:37 Letothe2nd: which layer should I replace with meta-ti ? May 14 08:47:49 ilkmc2rd: none, just add it. May 14 08:48:59 Letothe2nd: thank you, trying now. May 14 08:53:53 Letothe2nd: Should I change MACHINE? It is beaglebone-yocto now. May 14 08:56:11 Making it am335x-evm May 14 08:58:24 ilkmc2rd: why not making it "beaglebone"? May 14 08:58:33 ilkmc2rd: i think the video shows that pretty nicely. May 14 08:59:01 Huh sorry, ok making it beaglebone. May 14 08:59:33 ilkmc2rd: and: https://git.yoctoproject.org/cgit/cgit.cgi/meta-ti/tree/conf/machine/beaglebone.conf May 14 12:04:03 ok... is there a better target for x86_64 bare metal then MACHINE ?= "genericx86-64" in ycoto May 14 12:09:10 OutBackDingo: core-x64-i7 or whatsitcalled in meta-intel? May 14 12:10:31 no idea how good that works if you really go baremetal, though. May 14 12:34:45 Letothe2nd: welp its a start right May 14 12:51:47 Hi. I'm trying to figure out why a package gets included in my image, and wanted to run bitbake -g -u taskexp sp-image, but I get an python AttributeError: 'TreeModelFilter' object has no attribute 'sort_new_with_model'. Working on Dunfell. May 14 12:53:24 sveinse: isn't it just bitbake -g? respectively looking at the buldhistory? May 14 12:55:53 Hmm, this crash caused the previous run to not clean up after it, so bitbake refuses to do anything, it's just stuck at retrying server connection. How do I reset it? May 14 12:56:15 kill it and clean up the build dir. May 14 12:56:18 * Letothe2nd guesses. May 14 12:56:24 some socket file somewhere, IIRC May 14 12:57:15 * sveinse attempt deleting bitbake.lock and cross his fingers May 14 13:02:14 I never really haven't been able to show any of the yocto generated dot files with xdot because of their sheer size May 14 13:03:18 use less :) May 14 13:03:36 sveinse: don't do that May 14 13:03:37 d'oh :D May 14 13:03:55 sveinse: try lsof on bitbake.lock May 14 13:04:15 RP: hmm, I already did :( May 14 13:04:34 remove it I mean May 14 13:04:42 sveinse: then very carefully go through the output of ps ax and find the thing which is still in memory May 14 13:04:56 there will be something still in memory May 14 13:05:12 deleting that lock file is a really really really bad idea May 14 13:05:39 RP: Can I suggest a command-line feature for doing proper recovery in bitbake? May 14 13:06:52 Hi guys, my BSP has a line with WKS_FILE = "${@bb.utils.contains_any("EFI_PROVIDER", "systemd-boot rmc-boot", "systemd-bootdisk-initrd.wks", "grub-bootdisk-initrd.wks", d)}" May 14 13:07:24 sveinse: sure, patches would be awesome! May 14 13:07:47 I'm trying mender and it doesn't work out-of-the-box (No kickstart files from WKS_FILES were foun│ PID USER PRI NI VIRT RES SHR S CPU% MEM% TIME+ Command May 14 13:08:01 Sorry tmux paste lol May 14 13:10:08 I'm trying mender, and it errors out with No kickstart files from WKS_FILES were found: grub-bootdisk-initrd.wks. Even if I specify in local.conf EFI_PROVIDER="systemd-boot" it does not work. I had to manually override WKS_FILE in machine.conf. May 14 13:10:46 kriive: bitbake -e image May 14 13:11:27 kriive: oopsie, typed enter too quickly. From there, have a look at the lines before the one starting with EFI_PROVIDER. It'll tell you who sets it May 14 13:12:00 that way, you know where EFI_PROVIDER is set and maybe it's overriding the variable instead of loosely assign it or I don't know May 14 13:18:56 qschultz: thank you so much! I found out that for some reason EFI_PROVIDER is set to "grub-efi" May 14 13:19:28 kriive: well, with -e it gives you the "some reason" part :) May 14 13:20:42 kriive: also, there is usually autocompletion for nicks in IRC clients, so you don't mistype my nick (i'll just miss the highlight, no biggie) May 14 13:23:14 Yeah, I found the "some reason". The question now shifts on why it doesn't find the correct WKS_FILE (the error is pretty clear: EFI_PROVIDER is grub-efi, I'll check-out if I find a WKS file named grub-bootdisk-initrd.wks and for some reason I cannot find it) May 14 13:23:51 qschulz: woooow hahah TAB autocompletion even on IRC, good tip, thx May 14 13:24:08 and sorry for mistyping (: May 14 13:25:00 Do I need to rebuild everything if I change EFI_PROVIDER? May 14 14:32:49 Have you noticed that IRC logs are stuck to yesterday or is it my cache? May 14 14:39:00 kriive: I've noticed that also May 14 16:28:53 Hi all -- I've got a few config tweaks that I've made to my linux kernel in a " May 14 16:30:19 ... in a "fragment.cfg" file which I've included in my SRC_URI in a .bbappend file. I can see it show up in the SRC_URI when use bitbake -e virtual/kernel, but the tweaks don't seem to get applied to the final kernel .config. Any good way to get some insight into why it's not being applied? May 14 16:31:20 jeremytrimble: which kernel recipe is it? May 14 16:31:31 i really should write to kernel tutorial.... May 14 16:31:57 linux-ti-staging from meta-ti May 14 16:32:19 does that even use the config fragment stuff? May 14 16:32:27 it doesn't May 14 16:32:32 just use a defconfig May 14 16:32:49 Ah ok will do. May 14 16:32:56 jeremytrimble: kernel fragments are only supported by recipes inheriting kernel-yocto which isn't the case of many May 14 16:32:58 That's only a linux-yocto thing? May 14 16:32:58 maybe the kernel class should emit a huge warning if it looks like you're trying to use a fragment in a recipe that uses a defconfig May 14 16:33:02 zeddii: & May 14 16:33:04 zeddii: ^ even May 14 16:33:12 I’d send my patch to move parts of that into kernel.bbclass, but the bikeshedding would send me over the edge at the moment. May 14 16:33:28 rburton: why? May 14 16:33:38 so people wouldn’t ask the question. May 14 16:33:51 zeddii: but that is a valid use case May 14 16:34:04 I could want one defconfig and then add fragments on top of it May 14 16:34:08 not if you don’t use linux-yocto it isn’t. that’s the point. May 14 16:34:15 that’s what rburton is saying May 14 16:34:31 you can perfectly implement it yourself without linux-yocto May 14 16:34:34 qschulz: and yes, I’m aware of the use cases ;) May 14 16:34:43 ok. I’m out of here. bikeshedding begins May 14 16:34:52 zeddii: didn't mean to :) May 14 16:35:10 it was about the warning message only May 14 16:35:40 jeremytrimble: KERNEL_CONFIG_FRAGMENTS += "fragment.cfg" May 14 16:36:38 oh, so it is supported. denix is your man for the proper word! May 14 16:37:03 @denix: Oh -- didn't see that, but I do see that being referenced in meta-ti/.../setup-defconfig.inc. let me give that a shot May 14 16:38:52 zeddii: yeah, rburton just hates me - he keeps on perpetuating the lie... :) May 14 16:39:06 outed! :D May 14 16:39:17 haha! May 14 16:39:49 I blame the lockdowns May 14 16:41:27 khem: Hello! I wanted to know if you tried doing a full clang, thin/full-lto build on your distro? Do I have to do INHERIT += 'lto' in my local.conf? If so, python and cairo doesn't like the lto options. May 14 16:41:57 denix: fine, fragment in SRC_URI but that not set :) May 14 16:47:26 OK, I ended up having to do: May 14 16:47:43 SRC_URI += " file://myfragment.cfg " May 14 16:47:49 roussinm: I have not done full distro builds but targetted apps thus far May 14 16:48:02 it will be disirable May 14 16:48:20 KERNEL_CONFIG_FRAGMENTS += "${WORKDIR}/myfragment.cfg" May 14 16:49:12 For Python, I run into this issue: https://bugs.python.org/issue28015 May 14 16:49:45 Kind of a simple question, but was wondering. I'm trying to build an image based on meta-tegra for the Jetson Nano Developer Kit. I've read that using IMAGE_INSTALL_append isn't the best practice for managing the image. Would the best option be adding to meta-tegra? I'm using core-image-minimal as a base right now but I don't know which way I May 14 16:49:45 should go. For instance, I had to add a .bbappend to core-image-minimal to modify /etc/sudoers. Guess I'm just needing some guidance on what direction I should go... May 14 16:49:50 It should be fixed upstream, so I don't why I hit this particular issue. May 14 16:50:49 khem: the exact issue is the ax_cv_c_float_words_bigendian=no error. not the first message. May 14 16:59:54 ax_cv_c_float_words_bigendian that seems like a config check is failing look into config.log May 14 17:00:57 I noticed that there is no PACKAGECONFIG for lto either in the python recipe. May 14 17:03:24 is it possible to add the PACKAGECONFIG that depends on the TOOLCHAIN= clang? May 14 18:12:30 zeddii: the kernel/reproducibility patches are causing parse issues with meta-ti May 14 18:14:24 http://sprunge.us/roq8BI May 14 19:33:29 khem: RP already reported that. I suggested he toss it in the bin. May 14 19:34:35 zeddii: Whats the reason for needing the fallback SDE extraction? May 14 19:34:45 it’s how it was. May 14 19:34:49 I didn’t write the logicv May 14 19:36:53 zeddii: ok tossing now May 14 19:37:13 the handling of the exception in that git call is fixed here: http://git.yoctoproject.org/cgit.cgi/poky-contrib/commit/?h=zedd/kernel-next&id=dfffdf63c57ad949e4ff44755251ea46cfafbc91 May 14 19:37:39 but someone that cares about the factoring of the code can run with it. the cut and paste shell block works fine, and is why I did it in two separate patches May 14 19:37:39 zeddii: Ah, Ok. I understand... May 14 19:37:45 JPEW: do you use rk 3399 ? May 14 19:37:57 I have a Rock-pi-4 on my desk May 14 19:38:32 there is aarch64 tune overhaul I have send its in master-next give it a shot May 14 19:38:37 zeddii: The kernel needs a human readable value, not a unix timestamp... make sense May 14 19:38:46 Ok, will do May 14 19:39:42 yah. I only got tangled up in it since I changed a config and the modules got tangled up in it :D May 14 19:39:48 that’s a lot of tangling May 14 19:39:53 I need a thesaurus May 14 19:41:29 jonmason: Doh! I think GCC 10 broke op-tee :( May 14 19:41:48 hahahaha May 14 19:42:02 zeddii: entanglement ! perhaps you need quantum computers May 14 19:42:12 jonmason: Well, it worked for a day *brushes hands off* ;) May 14 19:42:14 JPEW: in what way May 14 19:42:23 we can put a dep on arm gcc, since we don't have gcc 10 yet ;) May 14 19:42:50 make: *** No rule to make target '/projects/poky/build/tmp/work/qemuarm64_secureboot-poky-linux/optee-os/3.8.0+gitAUTOINC+023e33656e-r0/recipe-sysroot-native/usr/lib/aarch64-poky-linux/gcc/aarch64-poky-linux/9.3.0/include/stdbool.h' May 14 19:43:10 denix: were you going to do a v2 of your tf-a patches (to remove bool)? May 14 19:43:32 why would you access 9.3 headers when you are using 10.1 May 14 19:43:41 JPEW: so it assumes gcc 9.x? May 14 19:43:47 khem: Hmm.... weird May 14 19:44:07 I don't know. maybe my sandbox is confused for some reason May 14 19:44:09 so is it really using 10.1 May 14 19:44:18 zeddii: thesaurus is my fave dinosaur May 14 19:44:20 I would say delete your tmp/ May 14 19:44:33 git clean -dffx May 14 19:44:44 that fixes everything May 14 19:44:58 or sudo rm -rf / May 14 19:45:07 khem: hahaha May 14 19:45:16 jonmason: thesaurus is my preferred dinosaur. May 14 19:45:18 khem: my favorite solution! how did you know! May 14 19:45:19 see what I did there. May 14 19:45:22 :P May 14 19:45:45 khem: I'm building for rock-pi-4 now, so I'll look in a bit ;) May 14 19:48:32 JPEW: If you do some benchmarking on rock-pi4 It would be interesting to see numbers on master and master-next May 14 19:49:13 khem: You know, I acually have some numbers already from a few weeks ago.... May 14 19:49:32 zeddii: nice May 14 19:52:17 JPEW: fantastic, I would be interested to know new numbers May 14 19:52:21 and difference May 14 19:54:47 So I just observed gcc10 in my archlinux update today, given that OE got gcc10 yesterday with all wierd architectures to support is cool May 14 19:56:56 jonmason: will do May 14 19:57:10 nice May 14 19:57:39 Any volunteers to be the maintainer of python3-cryptdome(x) & python3-elftools in oe-core? May 14 19:58:52 zeddii: Out of curosity, are you actually seeing it use the fallback to calling git in the kernel recipe? I wouldn't expect it to go there since SOURCE_DATE_EPOCH is set by reproducible.bbclass May 14 19:59:31 I don’t have it on in any of my builds (reproducibility), but I forced all of the conditions when testing it May 14 20:11:25 zeddii: nodistro doesn't enable reproducibile builds iirc May 14 20:16:39 jonmason: meta-arm-toolchains is benin if you don't do anything with it correct? It's going to be a depenency for BSP layers that needs the arm-none-eabi-gcc toolchain I don't want including it to magically switch me to arm-gcc May 14 20:39:15 is there any way to prune symlink to /bin and /sbin ? The image is setup to run in r/o so it make no sens to keep link to let's say /bin/touch.coreutils . I know I could modify the recipe in bbappend but I would rather use something alreay in place (if there is any) May 14 20:39:33 khem: Looks like I'm seeing about a 2% performance improvement with dhrystone.... that's against my old numbers, I'd need to go back to just master to be 100% sure May 14 20:41:16 khem: That's also on the A72s big.LITTLE gets real confusing for benchmarking ;) May 14 20:42:52 khem: About the same (2%) on the A53's May 14 20:58:39 aleblanc: that is like an image optimisation, you could add a rootfs postprocess command? May 14 20:58:57 JPEW: meta-arm-toolchains isn't going anywhere. It's just fun to joke about May 14 21:04:48 RP: that's what I'm planning so far May 14 21:05:19 RP, for git sources, instead of backporting patches, wouldn't it be possible to just specify the git sha commit of the patch to backport in the SRC_URI instead ? May 14 21:08:52 kroon: not as things stand today May 14 21:09:02 kroon: and that would only work if the patch applies cleanly May 14 21:10:04 RP, right. well, maybe its not worth the trouble, backporting patches is pretty easy as it is May 14 21:10:39 jsut figured it would mean a lot less git churn.. May 14 21:12:06 JPEW: I see your patches moving python modules - have you looked into replacing old pycrypto with pycryptodome in optee? May 14 21:14:11 kroon: it is a neat idea May 14 21:14:40 kroon: maybe worth writing a feature request bug in bugzilla? May 14 21:14:42 denix: Yep thats the plan May 14 21:15:08 denix: It seems to work just to s/python3-pycrypto-native/python3-pycrytodome-native/ May 14 21:15:18 RP, sure, I'll do that May 14 21:16:08 JPEW: ah, interesting. last time I tried it didn't work, but I haven't looked closer... May 14 21:17:00 denix: It's possible my qemuarm64 isn't acutally exercising the codepaths where it matters.... but I'm not sure how to if thats the case May 14 21:18:02 jonmason: I'm not sure I get the joke May 14 21:54:27 khem: Ah, sorry, bad news. I retried with master and got the same numbers as master-next, so no real improvement May 14 22:03:58 khem: Ah, I remember, my older numbers are from dunfell, so I think we can say that it's gotten 2% faster since then, but not perhaps that you change makes it faster :) May 14 22:06:12 * paulg somehow reads "skyfell" every time seeing "dunfell" May 14 22:36:40 JPEW: 2% is not bad May 14 22:36:51 its perhaps gcc10 May 14 22:39:25 JPEW: dhrystone is very limited, perhaps something like phronix testsuite might reveal morre May 15 00:26:42 JPEW: hmm, it seems to work fine for me too. I wonder if in the past I tried replacing pycrypto with pycryptodomeX and that has own namespace, not a drop-in replacement. of course, if would be best if upstream can update their code and only use cryptodomeX exclusively,,, May 15 01:40:52 Hi there, is there an easy way to determine what recipes/layers are setting a particular kernel config? I have a fragment that sets CONFIG_PROC_KCORE=y as well as CONFIG_PROC_VMCORE=y but for some reason in my final kernel KCORE is not set. It seems like something overriding it. May 15 02:25:58 anyone around that can grant me openembedded-core-contrib access ? Crofton|cloud ? khem ? May 15 02:32:58 zeddii, meh - just fork the project. May 15 02:33:13 * paulg runs May 15 02:33:32 I’m rewriting it in rust May 15 02:35:36 what - no clang? May 15 02:35:56 zeddii: pfft so mainstream, haskell is where the cool cats are B-| May 15 02:36:17 elisp May 15 02:36:28 cobol May 15 02:36:37 paulg: clang is behind rust. so you get to levels of goodness! May 15 02:37:25 * paulg was just trying to bait vmeson May 15 02:50:59 * zeddii calls it a night May 15 02:54:22 * paulg calls a 1-900 number **** ENDING LOGGING AT Fri May 15 02:59:57 2020