**** BEGIN LOGGING AT Wed Apr 21 03:01:26 2021 Apr 21 06:15:34 Hi Apr 21 06:16:37 wanted to know how to use PACKAGECONFIG for disabling python specific modules in an image Apr 21 06:16:56 like some example Apr 21 07:07:28 [10:46:37 AM] wanted to know how to use PACKAGECONFIG for disabling python specific modules in an image Apr 21 07:07:29 [10:46:56 AM] like some example Apr 21 07:39:22 yo dudX Apr 21 07:45:19 good morning Apr 21 08:03:42 LetoThe2nd can u explain using PACKEGCONFIG for removing a python-module example unittest Apr 21 08:06:58 Mahesh: what is not clear from the examples given in the video? Apr 21 08:06:59 RP: do you plan to backport bitbake-getvar to 1.50 as well after some soak time in master? (together with "tinfoil/data_smart: Allow variable history emit() to function remotely"?) I was trying it with 1.50 and without the tinfoil commit it fails with "TypeError: cannot pickle '_io.TextIOWrapper' object", but after cherry-picking both it works nicelly, LGTM Apr 21 08:07:38 JaMa: I don't know. You definitely would need both Apr 21 08:07:58 they are quite standalone so it could be possible Apr 21 09:13:06 Hello guys ! Apr 21 09:25:23 thekappe: and gals! and non-binary, and... hello people :D Apr 21 09:25:40 qschulz: "dudX" Apr 21 09:26:19 mmmm gals is not a great word I see... well.. something else then? Apr 21 09:27:10 RP: can't find the patch email to comment so doing it here: in bitbake-getvar it could be easier on the eyes to reverse the "if not args.recipe" condition Apr 21 09:27:48 LetoThe2nd: I settled on "folks" or "people", but feel free to invent your own words :D Apr 21 09:29:31 RP: also the "--unexpand" name feels strange, what about eg. "--no-expand" ? Apr 21 09:29:33 LetoThe2nd yea understood how to deploy it in my case i am not able to understand Apr 21 09:30:06 in my case i need to strip python3 so that its size gets reduced Apr 21 09:30:32 if i want to unittest module how to do i am looking Apr 21 09:30:47 sorry for asking this much questions Apr 21 09:31:42 [2:00:32 PM] if i want to remove unittest module how to do i am looking Apr 21 09:31:43 [2:00:47 PM] sorry for asking this much questions Apr 21 09:43:04 Mahesh: python3-unittest is the package you need Apr 21 09:43:16 http://cgit.openembedded.org/openembedded-core/tree/meta/recipes-devtools/python/python3/python3-manifest.json?h=master#n1151 Apr 21 09:43:31 no just an example Apr 21 09:43:54 for example if i want to remove python-unittest Apr 21 09:44:19 do not add python3-docs or python3-modules package in your image? Apr 21 09:45:37 aah yes exactly Apr 21 09:45:53 image should not have those modules Apr 21 09:46:13 end goal is reduce python size Apr 21 09:52:48 yann: those are fair comments, thanks. The patch is on the bitbake-devel list btw Apr 21 09:53:19 qschulz did u understand what i am trying to ask Apr 21 09:53:34 LetoThe2nd Apr 21 09:54:38 Mahesh: i am not suer i understand, but i am actually not that inclined to dig into the topic. packageconfigs are one thing, but i think that you are on a wrong approach anyways. Apr 21 10:03:40 tlwoerner: while preparing the kmeta bsp for a v3 I realized I forgot the .cfg was really meant as a PoC, there are so many things missing I'll have to add them before a resend Apr 21 10:05:18 LetoThe2nd for Python 3 footprint optimization is there any other methods Apr 21 10:05:48 Mahesh: https://www.yoctoproject.org/irc/%23yocto.2021-04-19.log.html#t2021-04-19T10:39:26 Apr 21 10:08:40 Mahesh: let me put it into other words. how comes you have decided that you need to do exactly that? have you looked at the package sizes? have you looked at what packages actually go into the image? how much space the take up? how the dependencies look? or have you just said "hum python is big, so i want to 'optimize' it"? so, where does that specific requirement cmoe from? Apr 21 10:24:11 LetoThe2nd i started with installing python3 basically , difference bw rootfs with python installed and not installed was about 40mb so just thought of reducing this footprint my removing unnecessary modules. Apr 21 10:25:52 Mahesh: python3-core is what you should install IIRC Apr 21 10:25:56 and not python3 Apr 21 10:27:19 QED. Apr 21 10:27:58 "i did something, i was surprised, now i want to optimize without understanding first." Apr 21 10:28:02 -> see qschulz Apr 21 10:28:27 LetoThe2nd: to be fair, the python3 recipe is not really like any others we have, so I can understand the confusion Apr 21 10:28:34 other* Apr 21 10:28:59 the recipe, granted. not looking at the package sizes? nope. thats really the 101. Apr 21 10:29:16 A customer is asking for CPEs information about packages... I know we have a mechanism for CVEs since at least Zeus but... is there anything about CPEs? Honestly, clueless what they need it for but asking here first, to be sure i'm not missing out on something? Apr 21 10:30:17 LetoThe2nd: there are many 101 things before looking at package sizes :) Optimizing a rootfs size is not a beginner task to me. Anyway :) Apr 21 10:30:36 qschulz: mabye. Apr 21 10:34:16 I have a main recipe that install some files in ${D}/boot/ Apr 21 10:35:54 since I don't want some of them I've added in my .bbappend, in the do_install_append() task, something like: Apr 21 10:35:55  rm -rf ${D}/boot/i-don-t-want-you Apr 21 10:36:41 bitbaking the recipe leads to Apr 21 10:36:58 do_package: QA Issue: bitstream-extraction: Files/directories were installed but not shipped in any package: /boot Apr 21 10:37:56 Do I have to remove ${D}/boot and not only ${D}/boot/I-don-t-want-you ? Apr 21 10:38:41 thekappe: sure, you could also just move those files to a different package if you wanted and not install this package Apr 21 10:39:33 @qschulz, thanks ! (again XD) Apr 21 10:50:48 Hello, have anyone successfully built rust 1.51 and cargo 1.51 on master branch? or any recent branch. I'm trying to figure out if this error is on my end or the repo https://pastebin.com/JBg7zHxg Apr 21 12:04:04 How can I add a new package for a specific recipe so that I can add some files to it ? something like FILES_${PN}-hoho = "hohoho.txt" Apr 21 12:12:03 thekappe: PACKAGE_BEFORE_PN, or PACKAGES variables Apr 21 12:12:34 remember that a file can only be in one package, so the first to have a matching pattern will get it Apr 21 12:12:42 "first" is leftmost to rightmost in PACKAGES Apr 21 12:13:24 yeah thanks, I've noticed that Apr 21 12:14:06 I was using "dbg" to put the files I don't want to be isntalled but I don't know wy they are still present in rfs Apr 21 12:21:40 Hi, do you know what the default size is for the RAM disk in /tmp? And how can I configure a limitation for that size? Apr 21 12:23:15 I don't explicitely configure that RAM disk in /tmp (AFAIK), and that's why I think it's somewhere in the yocto framework Apr 21 12:47:27 what is IIRC Apr 21 12:53:50 IIRC, IIRC means "If I remember correctly" Apr 21 13:40:50 qschulz: the CVE_PRODUCT variable in recipes is a list of known CPEs for that recipe. list because the cve database is terrible. Apr 21 13:47:30 rburton: thank you :) Apr 21 14:30:50 yann: sounds good Apr 21 14:34:37 tlwoerner: btw if you have any comments on the approach/contents, it's not a bad timing since I'm on it right now :) Apr 21 14:35:04 (note I have cleaned up a couple of dirty things already) Apr 21 15:25:27 RP: ndec: LetoThe2nd: (others?) just to make sure that company branded slides are ok for presentations at YP summit 2021? Apr 21 15:28:49 tlwoerner: ^^^ Apr 21 15:28:55 qschulz: such was no problem for YPS2020 (https://www.youtube.com/watch?v=PLl8jwZOpS0) so i don't think it would be an issue this time. Apr 21 15:32:00 LetoThe2nd: yeah I got kind of confused by the "one pdf with many slides all with Linux Foundation template" thing we had at YP Summit 2019 in Lyon Apr 21 15:34:48 to me branding includes the template too, let me know what is the decision of YP/LF committee about branding with logo and branding with logo+template :) Apr 21 15:52:32 Running containers built with VOLATILE_LOG_DIR = "yes" (the default) fails with "Failed to create /var/log/journal: No such file or directory". Setting it to "no" fixes this. What are the pros for VOLATILE_LOG_DIR = "yes"? Apr 21 15:53:01 RP: Morning! with the help of halstead, I was able to reproduce the centos/qmp issues in a standalone AB launched build, but when I ran the same command directly it passed, so the AB is introducing some change I don't understand. Apr 21 15:53:03 anyone having issues to fetch linux-imx from coudeaurora.org? Apr 21 15:53:34 RP: I did diff the 2 saved -st conf directories and they are identical. Apr 21 15:55:59 RP: thanks :-) Apr 21 15:56:19 qschulz: we will be making a slide template available to presenters that we're hoping people will use Apr 21 15:57:06 qschulz: we don't have a hard and fast rule about it, we're hoping people will use the template Apr 21 15:57:32 no idea about how to configure the RAM disk /tmp with yocto? Apr 21 15:58:26 qschulz: presenters are more than welcome to include a "who am i" slide and indicate their employer Apr 21 16:02:28 sgw1: I'd suspect it is something being written into the configuration by the test Apr 21 16:06:46 RP: If I understand oe-selftest correct, it creates a new build-st- dir for each test it runs, the ab-helper setup uses -K to keep that directory after, and I ran the same os-selftest inside the build-renamed dir from the failed build (hope that all made sense). Then I compared the conf directory from the 2 builds , they were identical. Apr 21 16:08:06 RP: hi -- since {DISTRO,MACHINE}_EXTRA_RDEPENDS only affects packagegroup-base, I'd suggest to rename the later to packagegroup-extra to make it obvious. "base" is confusing and unclear IMHO. Apr 21 16:08:36 sgw1: the test can write out extra config to the local.conf as part of the test though. Remind me which test is failing Apr 21 16:10:09 vdl: hi, I'm not sure there was ever an intent to match the names like that but I agree things are rather confused. I'd have to look at the differences to remind myself of them to see if that would make sense Apr 21 16:13:59 btw if I want to have simple host and container image recipes without bothering with multiconfig, having the kernel-image and kernel-devicetree in MACHINE_EXTRA_RDEPENDS seems like a good practice, so that a minimal container image not installing packagegroup-base won't pull in the boot artifacts. Correct? Apr 21 16:23:29 Hmm, it'd be nice if the layer index could also search for python libraries, like selftests, wic subcommands, recipetool/devtool commands Apr 21 16:23:43 RP: test_testimage_virgl_gtk_sdl, we have a AB-helper branch halstead/contrib that just runs that one test. I have failed and passed builds on centos8-ty-2:/home/pokybuild/yocto-worker/oe-selftest/build Apr 21 16:23:43 -1331210 failed (by AB) / -92132 passed (by Hand) / but I just tried again and by hand failed -2856961 , I am going to try again with the by hand failed configuration Apr 21 19:09:06 how can I update do_image_wic[depends] dynamically from a list of file? e.g. do_image_wic[depends] += "foo-image-*:do_deploy" Apr 21 19:10:58 your only option would be python, but it doesn't sound like that choice would lead to very reproducible builds, so doesn't sound like a great one imo Apr 21 19:12:28 kergoth: it's for development anyway. Do you have an example of such python call? Apr 21 19:12:55 plenty of recipes and classes either use inline python (${@}) or anonymous python to set flags.. Apr 21 20:59:12 sgw1: I assume you've read the test code for that test? Apr 21 21:00:43 sgw1: Note how it is dynamically configuring PACKAGECONFIG for qemu-system-native and setting IMAGE_FEATURES/IMAGE_INSTALL - are you doing that? Apr 21 21:06:33 RP: you mean this: PACKAGECONFIG_append_pn-qemu-system-native = " gtk+", yes I have it included in my local.conf along with the kmscube append Apr 21 21:07:14 sgw1: and you're running two different testimage calls with the different configs? Apr 21 21:07:48 I can get the failure by hand now with the "gtk gl" qemu params Apr 21 21:08:10 more consistently on the AB then my home nuc hardware Apr 21 21:08:18 sgw1: ok, so you've at least solved the reproduction issue :) Apr 21 21:08:36 it's been a long strange road! Apr 21 21:09:15 sgw1: one step further along it :) Apr 21 21:11:03 Yeah, would be great if I could recreate it locally on my nuc, but your correct, now I can try more debugging on the AB Apr 21 21:52:01 sgw1: I dream of reproducible bugs! Apr 21 23:28:43 Do I need a specific distro feature in order to connect via ssh with root? Apr 22 00:13:30 Hi everyone, is there a public project similar to home-assistant (so OTA updates, overlayfs etc) but based on yocto? Apr 22 00:18:36 vdl: I think that is still part of debug-tweaks, it really should move to one of it's own Apr 22 01:14:19 RP: are you telling me that I cannot connect via ssh with root unless I use debug-tweaks? Apr 22 01:34:01 vdl, not being able to connect via ssh as root is usually a good idea, but you can probably just allow that by changing the sshd config **** ENDING LOGGING AT Thu Apr 22 02:59:57 2021