**** BEGIN LOGGING AT Tue Jun 02 02:59:57 2020 Jun 02 06:37:22 Hi , excuse me I dont understand what is the usage of "WIC" . what is WIC in yocto ?? Jun 02 07:07:06 rburton: \m/ Jun 02 07:56:52 ndec: ping Jun 02 07:57:25 Letothe2nd: howdy. Jun 02 07:58:28 ndec: heya! just applied for the advocacy list with the mail account i use for community work. can you please approve? Jun 02 07:58:45 ndec: (I'll write up a short planning and recap soon :) ) Jun 02 07:58:56 Letothe2nd: sure! Jun 02 07:59:30 ndec: thanks! Jun 02 12:51:49 Hello, i have pretty much vanilla setup of yocto-zeus and a raspberry pi cm3 and during every boot it waits for about 20s for hwrng initialization. Similar image based on a yocto-thud does not have this issue, could you point me where to look? Jun 02 12:51:57 dmesg shows that: [ 0.282619] bcm2835-rng 3f104000.rng: hwrng registered Jun 02 12:52:24 so I suppose that hardware RNG is working, but it takes long to get to the entropy level before starting sshd Jun 02 12:52:32 is there any way to speed up this process? Jun 02 12:55:37 kpo, maybe haveged Jun 02 12:56:44 i'd try adding it to the image and see if that makes a difference Jun 02 12:57:35 thanks for the tip kroon_ ! Jun 02 13:01:47 seems like checking it out will be a little longer, because image of mine does not use systemd yet :P Jun 02 13:02:21 kpo, you shouldnt need systemd for using haveged Jun 02 13:07:36 kpo, rng-tools might be an alternative Jun 02 14:59:28 YPTM: Jan-Simon is on Jun 02 15:00:27 YPTM: Mark Van de Vyver is on Jun 02 15:00:30 YPTM: Scott Murray is on Jun 02 15:03:27 YPTM: ross in Jun 02 15:34:41 vmeson: https://www.mail-archive.com/openembedded-core@lists.openembedded.org/msg136914.html Jun 02 15:39:59 smurray: sorry, missed your previous question, who is the new platinum member? Jun 02 15:41:58 who is? Jun 02 15:42:23 M$, finally?!? Jun 02 15:43:22 Letothe2nd: you should join the call ;-) Jun 02 15:43:43 tlwoerner: you know that i can't stand talking to real people. only to webcams, when i'm drunk? Jun 02 15:43:59 (nah seriously, its a really bad timeslot for me) Jun 02 15:45:13 makes for great background "music" Jun 02 15:57:47 Thought I'd ask again today, running into a predicament. Python3 availabel from `oe-core` is at 3.8.2, but `psycopg2` provided by `meta-openstack` is at 2.7.5. Python 3.8 isn't supported until `psycopg2@2.8.4`. There's some patch work being done in each instance to bypass the need for `PG_CONFIG` to be set, obviously needs updated for the newer Jun 02 15:57:47 version. Could anyone point me into the right direction for possibly getting `psycopg2` included in the image I'm trying to build? Jun 02 17:14:17 bsmerbeck: zeddii is the maintainer of meta-openstack, suspect he'd be willing to take a patch to upgrade psycopg2 Jun 02 17:15:32 yup. it’s not version frozen for any reason in this case. Jun 02 17:18:14 smurray: thanks for the heads up. I'll have to work on that today then, never done anything like it before Jun 02 17:20:45 bsmerbeck: the recipe looks simple other than that patch, devtool upgrade may work, see: https://www.yoctoproject.org/docs/3.1/dev-manual/dev-manual.html#gs-using-devtool-upgrade Jun 02 17:32:51 Is there any info on how to interpret bitbake-diffsigs output? I'm trying to understand why sstate cache isn't being reused Jun 02 17:33:59 paulbarker: The few times I used it I found it informative, what's going on with yours? Jun 02 17:34:11 I see "basehash changed", I also see that the only differences in runtaskdeps are the paths to the recipes Jun 02 17:34:52 I also see that some dependencies changed hash. Do I have to manually walk down that dependency tree to find out what changed hash at the bottom? Jun 02 17:35:42 Moving the recipes to a different path shouldn't be enough to prevent sstate reuse Jun 02 17:39:43 paulbarker: my intuition tells me you're right. It should walk down the dependency tree if there are two sigdata (previous and current) available for dependency recipes I think Jun 02 17:40:45 but considering the time of the day, I'd say it's too late for my brain to not spew non-sense so I'd rather not misguide you. Jun 02 17:41:56 80 char lines are so out of fashion - http://lkml.iu.edu/hypermail/linux/kernel/2005.3/08168.html Jun 02 17:41:59 But IIRC, when I had all sigdata, (no removal of tmpdir) it would walk the tree and find the original issue like 5 dependencies deep. Jun 02 17:42:17 qschulz: Ok. I have a guess of the culprit so I'll go looking a different way Jun 02 17:43:58 Looks like `bitbake-diffsigs` is useful when you know which recipe is the root of the changes not if you just know "bash shouldn't have been rebuilt" Jun 02 17:45:16 paulbarker: maybe something to report? If it effectively does not work, might need to be fixed. Otherwise, if it does not work but is intended to work like that, maybe some doc is needed? Jun 02 17:45:23 paulbarker: good luck. let us know Jun 02 17:47:55 khem: it was already a big issue within v4l2 and drm where the function names are terribly long and they stopped following the 80char limit a while back. 80char is a nice default limit though, it forces you to not have more than 2 or 3 indented code parts which is usually great for readability IMO. Anyway, Linus has spoken :D Jun 02 17:49:54 I stopped doing it when I started using GUI based systemd Jun 02 17:49:59 err systems Jun 02 17:50:23 but I am glad that this disucssion is happening Jun 02 17:55:02 perhaps next would be html emails Jun 02 17:58:09 Looks like my last CI build was before https://git.openembedded.org/openembedded-core/commit/?h=dunfell&id=5a2bc3bfa9e1a4f37b6e26a5c40a4a9c025d03f1 and that change is causing a rebuild of everything Jun 02 17:58:43 I'll run a new CI build and re-test sstate cache usage later Jun 02 17:59:09 paulbarker: walking the sigs turns out to be hard for bitbake to do accurately and you often have to do it manually Jun 02 17:59:20 hi Jun 02 17:59:48 paulbarker: often you can get an idea of that what changed and see the issue though and you can walk the trees quite quickly if you have an idea of the earliest things that are rebuilding Jun 02 18:00:05 smurray: gave the devtool upgrade a try, but it seems the patch it what's most important. It's somehow allowing `psycopg2` to build without needing the variable `pg_config`, which is configured after postgresql in installed. Knew this would be a fun one! Jun 02 18:00:32 i have a repo which contains the code for 2 MCU which are required in a project. How do i write single recipe to build hex files for both MCU ? Jun 02 18:01:13 build tool is cmake Jun 02 18:01:17 RP: It was fairly easy for me as I have some intuition on what -native task is likely to be low in the dependency tree and I can easily guess that if `populate_sysroot` shows a difference in the hash of `install` I can probably jump back to `prepare_recipe_sysroot` without needing to look at configure/compile/etc Jun 02 18:01:42 khem, rofl Jun 02 18:02:12 RP: May be a challenge for less experienced users though. It's well outside my area of experience with the bitbake internals though. Jun 02 18:02:47 khem, we introduced checkpatch for better control and suddendly it changes Jun 02 18:03:25 My real issue is that my CI system was using old checkouts of poky instead of pulling updates but I'm submitting patches to kas to fix that Jun 02 18:03:28 paulbarker: the desire to do better is there, nobody has managed to make it work like we'd like though Jun 02 18:03:45 * RP lacks the time to spend on it Jun 02 18:03:57 khem, the one with html-mails is a lost war Jun 02 18:04:54 RP: One thing that threw me off and may be low-hanging fruit to fix is `runtaskdeps changed` highlighting things where only the recipe path has changed Jun 02 18:04:59 saurabhj: we support multiconfig for mcu+cpu combos but it seems you are looking for a standalone build env, you perhaps can build SDK and have youe own logic then to build it or try to look into multiconfig Jun 02 18:05:35 RP: https://pastebin.com/HxBNXptz Jun 02 18:05:39 paulbarker: yes, could you file a bug please? or patch really welcome too! :) Jun 02 18:06:14 ant_home: yeah, but I think adapting to new world is a good thing, I have sat in many freebsd meetings where they debated endlessly about such things without any change to effect Jun 02 18:06:59 khem: ha ha Jun 02 18:07:08 right now in #openbsd Jun 02 18:07:20 eval `resize` can fix terminal size issues Jun 02 18:07:29 Even today I think BSD code is far better but very few projects use it Jun 02 18:08:15 yes but you are a bit on your own, like linux in the '90 Jun 02 18:08:51 khem:it is not standalone my cpu recipe is dependent upon mcu recipe.Nonetheless i will look in multiconfig. Jun 02 18:21:58 RP: My grep skills are letting me down. Do you have a rough idea where those task names (like `bzip2/bzip2_1.0.8.bb:do_populate_sysroot:virtual:native:/w/layers/poky/meta/recipes-extended/bzip2/bzip2_1.0.8.bb`) get generated in bitbake? Jun 02 18:35:31 a bit more general question. any shot I could get python 3.5 or 3.7 from a previous release into my dunfell project? Jun 02 18:35:45 ugh Jun 02 18:35:54 or should I consider just going down to the previous release entirely Jun 02 18:37:40 are you hard stuck at python version particularly Jun 02 18:38:32 I think using dunfell would be preferred with python3 version it provides for best results Jun 02 18:38:57 I am not aware of any incopatibilities between 3.7 and 3.5 Jun 02 18:39:26 khem: Pretty much. There's a psql driver package called `psycopg2` that's incompatible at its current rev with python 3.8. Getting it to install in general is tricky business in yocto. So my options are to upgrade psycopg2 and figure out a new patch to get it working or downgrade from 3.8 to at 3.6 or earlier Jun 02 18:40:09 I would think getting psycopg2 onto py3.8 would be progressive solution Jun 02 18:40:44 python3 is quite core to yocto and a given version is well tested for that release Jun 02 18:41:13 and I dont think we have a case of parallel installing different versions of python3 Jun 02 18:41:24 it will be messier but certainly can be done Jun 02 18:42:09 I would agree, it'll probably take me a while to understand how the patch worked at all - that's the biggest hurdle. The python package requires postgresql installed and a pathway for its pg_config binary...which it doesn't have during the buildtime. The previous patch was four or so lines, but this one I'm thinking will be larger as the setup.py Jun 02 18:42:09 and setup.cfg files for psycopg2 have changed quite a bit. Jun 02 18:43:11 Thought I'd be able to squeeze out a quick solution, but it's probably for the best I give the upgrade a shot and bump the psycopg2 recipe Jun 02 19:29:32 paulbarker: lib/bb/siggen.py ? Jun 02 21:20:59 hello people Jun 02 21:21:41 So I have made distribution using yocto project and qt5 and I'm having problems with text not being displayed when i run my qt application Jun 02 21:22:02 Am I missing fonts or something, does anyone idea whats going on. Jun 02 21:23:12 So, I have successfully booted, logged in, ran my eglfs application and everything is displayed except for the Labels. Jun 02 21:30:23 I think just fonts are missing, I will try installing them Jun 02 21:44:59 That was it! Jun 02 21:45:07 everything iw roking now Jun 02 22:18:40 Hey there, How can I disable a task if a certain override is set? Jun 02 22:19:55 Or to put it into other words: How can I negate on override? Jun 02 22:22:52 Basically, I have a `do_install_append()` in one recipe that I want to disable when I certain `OVERRIDE` is set. Anyone? Jun 02 23:26:24 JPEW: any ideas on https://autobuilder.yoctoproject.org/typhoon/#/builders/89/builds/1994 ? Jun 02 23:28:18 kanavin_home, sno: https://autobuilder.yoctoproject.org/typhoon/#/builders/87/builds/1002 - perl make install race in manpages :( Jun 03 00:22:02 nobody reads manpages anymore. Just delete the target rule from the Makefile Jun 03 00:22:05 * paulg runs **** ENDING LOGGING AT Wed Jun 03 02:59:57 2020