**** BEGIN LOGGING AT Tue Oct 27 02:59:57 2020 **** BEGIN LOGGING AT Tue Oct 27 06:45:16 2020 **** BEGIN LOGGING AT Tue Oct 27 06:48:07 2020 Oct 27 07:37:57 good morning Oct 27 07:42:07 yo dudX Oct 27 08:38:00 Hmm, question about latest version of psplash? Oct 27 08:38:26 If I want to bbappend a brand change (e.g. change colors, logo) what would be the best way to go about doing this? Oct 27 08:38:41 I just keep getting the openembedded graph... :/ Oct 27 08:41:10 wertigon: You could take a look how this is done in meta-raspberrypi Oct 27 08:41:14 https://github.com/agherzan/meta-raspberrypi/blob/master/recipes-core/psplash/psplash_%25.bbappend Oct 27 08:41:33 https://github.com/agherzan/meta-raspberrypi/blob/master/conf/machine/include/rpi-base.inc#L93 Oct 27 08:54:39 wertigon: and meta-poky overrides the image too Oct 27 08:55:50 rburton: what was said before that? I seem to have had a slight connection glitch >_< Oct 27 08:56:27 paulbarker said wertigon: You could take a look how this is done in meta-raspberrypi Oct 27 08:56:27 https://github.com/agherzan/meta-raspberrypi/blob/master/recipes-core/psplash/psplash_%25.bbappend Oct 27 08:56:27 https://github.com/agherzan/meta-raspberrypi/blob/master/conf/machine/include/rpi-base.inc#L93 Oct 27 08:56:30 wertigon: You can always check https://www.yoctoproject.org/irc/ Oct 27 08:56:48 paulbarker: ah, thanks :) Oct 27 09:00:38 Ok, so if I understand this correctly, I should just be able to add our own color scheme by replacing/patching psplash-colors, yes? Oct 27 09:02:52 And then I need to create a new psplash image as a header file Oct 27 09:03:19 What confuses me is that the source indicates I should be able to just add a png file in the proper place now Oct 27 09:05:17 rburton: question, does anything openstack in meta-cloud-services work or build ? its a bit confusing. Oct 27 09:05:28 no idea, never tried it Oct 27 09:16:55 rburton: yupp, more approrpiate for the viertualization list i guess, thanks though Oct 27 09:26:55 hi, it seems patches (at least meta-java ones on oe-devel) are no longer showing up in patchwork. Anybody knows whats going on there? Or am I missing something? Oct 27 09:27:04 OutBackDingo: Did you send this email? https://lists.yoctoproject.org/g/meta-virtualization/message/5972 Oct 27 09:27:16 Using "Yocto" as the author on the email is a really good way to confuse people Oct 27 09:28:11 <__ad> hi, have a group of recipes not compatible with my custom machine, as imx-dpu-g2d imx-dpu-g2d xf86-video-imx-vivante. Is it the right way to go, to create a .bbappend for all, addinc COMPATIBLE_MACHINE ? Oct 27 09:57:26 __ad: assuming you do not want NXP components in your BSP, you can set IMX_DEFAULT_BSP = "mainline" in your local.conf, or search where IMX_DEFAULT_BSP is set to "nxp" and change it to "mainline". Oct 27 09:58:28 __ad: you forgot to mention which BSP you based your custom BSP on, and which NXP machine your "custom machine" is based on. Oct 27 10:21:40 paulbarker: yes, i changed the name to Y L reddy Oct 27 10:22:15 OutBackDingo: If you look in the archive the name still just shows as "Yocto" - https://lists.yoctoproject.org/g/meta-virtualization/message/5972 Oct 27 10:22:16 no idea why its still says Yocto, anyway, i guess ill have to unsub all yocto lists and resub as a different user Oct 27 10:22:28 PITA Oct 27 10:22:32 :( Oct 27 10:22:36 That is confusing! Oct 27 10:22:52 paulbarker: i already unsubbed Oct 27 10:22:57 OutBackDingo: Are you posting via the web interface or by email? Oct 27 10:23:00 unsubbing them all Oct 27 10:23:08 thunderbord Oct 27 10:23:20 *bird Oct 27 10:23:31 Could you send me a direct email, not via the list? Let's check how that looks Oct 27 10:23:52 To pbarker@konsulko.com Oct 27 10:24:18 That should let us isolate whether the problem is with the list or with your email client Oct 27 10:25:05 paulbarker: sent Oct 27 10:25:22 Thanks, I got that. It looks correct, the name is Y L Reddy Oct 27 10:25:31 soooo..... Oct 27 10:25:32 It could definitely be a groups.io issue Oct 27 10:25:37 found a bug !!! Oct 27 10:25:41 LOL Oct 27 10:26:18 paulbarker: no worries mate, ill unsub all lists and resub as a diff user Oct 27 10:26:23 If you log in to lists.yoctoproject.org then visit https://lists.yoctoproject.org/account?page=viewprofile you should see your identities Oct 27 10:27:09 If you edit the main account profile (the top entry), hopefully it will update the rest Oct 27 10:29:01 paulbarker: yeah it was emtpy so i put in Y L Reddy and now they all show Oct 27 10:30:50 paulbarker: nope, makes 0 difference Oct 27 10:30:58 still shows as Yocto Oct 27 10:31:14 must be grabbing the sender addie Oct 27 10:32:00 OutBackDingo: Hopefully new emails you send to the list will use the new display name though Oct 27 10:32:34 Ah I see your email now, still the same. I think we have a groups.io issue Oct 27 10:32:34 paulbarker: nope, they dont Oct 27 10:34:20 That is really frustrating from groups.io. Thanks for trying at least! Oct 27 10:41:32 <__ad> zandrey, sry, i am on imx6dual based stuff, and upgraded to dunfell. Seems the issue is that i want to use vivante blob stuff, while dunfell is set to use etnaviv way. Oct 27 10:42:10 <__ad> (this at least from a first look) Oct 27 10:46:51 paulbarker: ok all resubbed as dingo Oct 27 10:48:30 * OutBackDingo feels like a toruble maker now :) Oct 27 11:09:07 __ad: ok, now it is more clear. iirc, packages you listed are not supported for imx6dual machine, and adding it to COMPATIBLE_MACHINE is not guaranteed you would have a working runtime. there are few issues reported in meta-freescale github regarding vivante, i think you should check if some of them applies in your case. Oct 27 11:11:08 __ad: it would also be good if you can give an example of what you're trying to achieve with those packages included. Oct 27 11:16:03 <__ad> zandrey, thanks. Well, i am working on a custom board upgrading from poky to dunfell, where it is mandatory (seems for now) to use nxp/vivante approach Oct 27 11:18:16 __ad: probably you are migrating from a version previous to dunfell, not poky though https://wiki.yoctoproject.org/wiki/Releases Oct 27 11:19:31 Hello everybody! I was bitbakeing a recipe and got a warning in red letters saying that there are multiple providers for a shared library. I understand which recipes lead to that confusion. What is not clear to me is how to explicitly say which shared library to use under which circumstances. Any hint where I might find the answer to that is highly welcome :) Oct 27 11:21:48 emrius: in newer YP versions in the red letters there is a hint to solve the issue Oct 27 11:22:18 emrius: however a pastebin sould help us to help you better ;-) Oct 27 11:22:28 *should Oct 27 11:23:34 mckoan: ok. I mean I'm on dunfell thus I'd guess I'm using a rather recent version. But I probably overlooked the hint. Hang on... Oct 27 11:24:58 https://pastebin.com/Kdgmk3gK Oct 27 11:25:30 That's all. Thus, it's not really throwing an exception but looks rather like an intense warning... Oct 27 11:31:11 kanavin_home: rburton: (or others), what are the implications of http://git.yoctoproject.org/cgit/cgit.cgi/poky/commit/?id=1ec91031165f8801eeb5662aa98c73dc6be88cc1 that we can document for the migration guide? what would nvidia binary users need to do (not use this functionality ?) Oct 27 11:37:03 bluelightning: not use it, use it at their own risk (it is almost certainly not going to work at all), or install nouveau driver. Oct 27 11:56:31 Good afternoon. My task is to configure the Yocto Project* - based BSP for the Gen 3 Intel ® Movidius ™ VPU (Keem Bay) Platform. But the problem is to build bitbake core-image-ese-initramfs-fit Oct 27 11:57:04 Can I describe the problem here? Oct 27 12:02:20 vadim61: yep, please jsut go ahead. Oct 27 12:06:45 I asked a question on https://stackoverflow.com/questions/64554007/bitbake-core-image-ese-initramfs-fit-problem Oct 27 12:07:02 please take a look Oct 27 12:22:54 did you checked the openvino version in your project? bitbake-layers show-recipes openvino Oct 27 12:25:57 moto-timo: sakoman really later for the conversation but I will eventually start growing coffee, any chance your place in CR is an airbnb? haha Oct 27 12:29:32 Hi, I have a wks-file which states that the rootfs partition has a fixed size of 120 MB. Yet, the generated rootfs is 147 MB. When I increase the fixed size, the size of the generated image gets bigger too. So the generated rootfs is always too big for the partition. What determines the size of the generated image? Oct 27 12:29:54 armpit: tlwoerner and I'm not just starting to dislike it, I did before and still do haha, I'd pay for the rapid test if it allowed us to get back to in person conferences Oct 27 12:33:49 caiortp Summary: There were 3 WARNING messages shown. Oct 27 12:33:56 sorry , Oct 27 12:34:02 late for the party Oct 27 12:34:49 vadim61 Error: Oct 27 12:34:49 Problem: conflicting requests Oct 27 12:34:50 - nothing provides openvino >= 2020.3.0_3362_dc895cd9dcf_releases_2020_kmb_pv1 needed by vpual-libs-vpual_4b8d399_WW37.4_Pre_Int_nightly-r0.aarch64 Oct 27 12:35:51 vadim61: do a bitbake -g and see what is requiring openvino, and then figure out where openvino is supposed to come from, if you got that far supposedly openvino should have been built but it looks like for whatever reason the naming or the package version doesnt match Oct 27 12:40:12 I'm not sure what I'm doing Oct 27 12:40:39 polaris: you can set IMAGE_ROOTFS_EXTRA_SPACE in a .conf or --extra-space in the WKS file to add more space to the partition Oct 27 12:40:39 Summary: There were 2 ERROR messages shown, returning a non-zero exit code. Oct 27 12:40:40 |#####################################################################################################################################################################################################################| Time: 0:00:00 Oct 27 12:40:40 Typo?#################################################################################################################################################################| ETA: 0:00:00 Oct 27 12:40:41 |###################################################################################################################################################################################################################| Time: 0:00:00 Oct 27 12:40:42 recipes available for: Oct 27 12:40:45 Resolving any missing task queue dependencies Oct 27 12:40:47 are available for nativesdk-zlib (nativesdk-zlib, nativesdk-zlib-intel) Oct 27 12:41:16 alejandrohs: :-) i gave up reading backscroll a long time ago ;-) Oct 27 12:43:51 carlsb3rg: --extra-space sounds interesting. I will give it a try. But do you know why the generated rootfs size is bigger than the partition? Usually, I get an error message when I use up more space. For example, when I add a big package that exceeds the partition size. Oct 27 12:46:32 is it possible to mount wic images? Oct 27 12:48:01 hpsy: not by definition. Oct 27 12:49:07 hpsy: if you happen to have a specific wic image that has a known partition type, size and offset, and it is a fs type that can be loop mounted, then you can do that theoretically. a wic image per se, no. Oct 27 12:50:14 you can bmap tool the wic to a blob Oct 27 12:50:17 and mount it Oct 27 12:50:58 carlsb3rg: nope. not by definition. only if your specific wic file happens to be in a form that actually is supported by that workflow. Oct 27 12:51:39 oki...then mine happened to be in that form since a colleague of mine managed to do it Oct 27 12:52:33 the key point is to understand that the suffix "wic" does not imply a specific file structure. it could literally be a couple of megs of random output. it only denotes that it has been created by wic. Oct 27 12:53:01 or event partition structure... Oct 27 12:53:02 I just thought if I can dd it to an SD card then I should be able to mount it, this is for a intel-corei7-64 MACHINE, until zeus I just used the hddimg to mount and wic to flash, but from dunfell it looks like creating a hddimg is no longer possible. Oct 27 12:53:23 hpsy: its also not given that a wic file relates to a sd card. Oct 27 12:53:38 polaris: leto and the smart guys here probably know...but my only guess is that some recipe is bypassing the "yocto" way of doing things, like adding something large IMAGE_PREPROCESS_COMMAND or IMAGE_POST_PROCESS command Oct 27 12:53:44 hddimg has a hard 4 Gb limit. That was a problem Oct 27 12:55:42 alejandrohs: it has a very rustic bodega. VERY RUSTIC. Oct 27 12:56:52 alejandrohs: but there is a nice budget backpacker/traveller style bed and breakfast near by Oct 27 12:57:34 alejandrohs: http://www.turrialtico.com/ Oct 27 13:09:07 The size is less the 4GB for sure, it is just that the parsing errors out in core-image-minimal-initramfs.bb (even though I am building the core-image-minimal) Oct 27 13:13:14 I have the same issue...it does "something" with initramfs even when I'm using core-image-minimal Oct 27 13:56:05 Is anyone of you building a yocto system in CI? Oct 27 14:01:03 many Oct 27 14:02:45 That was a stupid question :) Oct 27 14:02:45 I've got no idea how to set up my CI system due to a typical yocto project being distributed over so many repositories. I'm using Teamcity (from Jetbrains, I think). Oct 27 14:03:24 literally just finished was a talk at ELC-E that covered repo vs submodules vs kas for build management Oct 27 14:03:37 personally, I like kas Oct 27 14:04:55 My current preference is repo as the underlying system, but something else as a front end to build and control the manifest.. I don't like the who repo 'lets store a manifest, and now it's immutible' Oct 27 14:06:30 whereas my current preference is that repo is nuked from orbit Oct 27 14:13:32 for me it allows me to use git commands on individual repositories, or global commands on the whole set of repositories.. Oct 27 14:13:56 problem I've seen is a lot of the manifests hardcode certain things.. but just setting branch names keeps things really flexible, but yet 'connected' Oct 27 14:14:16 then I can iterator of each repository -- or run repo sync etc to update.. Oct 27 14:14:28 (I'm usually working with > 10 different repositories) Oct 27 14:14:37 but I think most of us can agree submodules are terrible Oct 27 14:16:07 (submodules are only good when you NEED to lock two repositories in lock step, where one is dependent on another, but the other changes in an uncontrolled way) Oct 27 14:21:47 I dunno, we have 50 submodules, and it works fine Oct 27 14:26:56 at lest now they support tracking branches Oct 27 14:27:37 last time I had to deal with it (6 months ago) it was still a pain in the ass... but if they have generic branch tracking, that gets them closer to what the advantage of repo is/was Oct 27 14:29:52 Assuming you have set the remote branches, I believe that `git submodules update --remote` will do that Oct 27 14:31:21 We don't use it because we have a policy to mirror everything locally, so I have a custom script to add the proper upstreams as a remote, pull in the latest, merge it into our local mirror and push it Oct 27 14:46:47 manuel1985: we do use TC, welcome to the club. :) do you have a specific question? Oct 27 14:54:01 zandrey: I don't even know how to phrase my question right. ;) We're using repo right now. Hence, the manifest.xml lives in a seperate repository. So I guess I have to have TeamCity watch the repository in which my metalayer lives in, but not clone it. Because cloning it is done through repo. Oct 27 14:54:01 So if a change at my metalayer occurs, TeamCity should start from a clean directory, call `repo init` and `repo sync` there, wait for it to complete, and then issue the build. Oct 27 14:54:01 Is this how it's usually done? Oct 27 14:55:25 :D this is indeed the same problem that we do have! Looks like it is a common issue with TC and it's implementation of "VCS Root" concept. Oct 27 14:57:06 Ah, good. :) Haha Oct 27 14:57:33 iirc, you can attach another VCS Root to your build configuration, and setup custom checkout rule to "move it away" into a separate folder which build would not take. then setup your VCS trigger to watch it, and upon arrival - kick the repo on another VSO root Oct 27 14:57:51 s/VSO/VCS/g Oct 27 14:58:15 just as i said: welcome to the Club! :D Oct 27 14:59:16 Yeah having several VCS roots is what I also thought about Oct 27 14:59:39 Thanks ;) Oct 27 14:59:59 That's some club I'd prefer not to be part of tbh ;0 Oct 27 15:00:54 if you want to always execute clean build - you can tick the "Clean build:" in "Version Control Settings" -> "Additional Options". in that case your Agent would be forced to start from "square 1" Oct 27 15:01:30 "That's some club I'd prefer not to be part of tbh ;0" <- same here, but "we have what we have"... ;) Oct 27 15:03:30 Yeah same here Oct 27 15:04:15 It seems every company acts the same. At first they choose their toolstack on a more or less random basis, and then lather it gets hard if not impossible to switch it out Oct 27 15:58:00 Hello All, is it possible to run yocto on ubuntu 20.04. I am having lots of errors with building m4. Oct 27 15:58:42 funk123: depends on the version of yocto you are trying to build Oct 27 15:58:50 Daisy Oct 27 15:59:06 funk123: please go and slap whomever told you to build it. Oct 27 15:59:19 i'm having a few issues building dunfell on u20 myself. need to dig through to check for needed cherry picks Oct 27 15:59:25 but daisy, wow Oct 27 15:59:35 i'd use docker to build in an older distro Oct 27 16:00:03 funk123: daisy is so massively outdated you'll have a hard time on building it on anything other than a container/vm from that time. Oct 27 16:00:53 containers are great for that, can continue to use your normal current development tools on a recent host and just build in an old container Oct 27 16:01:03 trying to actually develop in an old vm isn't fun :) Oct 27 16:01:48 https://github.com/Xilinx/meta-xilinx Oct 27 16:01:59 that is the reason I am picking older yocto Oct 27 16:02:34 Hi I was wondering where I can find the python api to extend a bbclass, I need to look in openembbed repo's I guess? Oct 27 16:02:41 ah no bitbake probably Oct 27 16:02:47 that doesn't really explain it, meta-xilinx supports many versions. unless it doesn't support your bsp with a more recent one? Oct 27 16:02:56 funk123: um. just because something is listed at the top, its not the most recent one. xilinx happily supports zeus for example. Oct 27 16:03:26 funk123: https://wiki.yoctoproject.org/wiki/Releases Oct 27 16:03:38 funk123: daisy is from 2014.. Oct 27 16:03:58 LetoThe2nd: thanks a lot Oct 27 16:04:06 I was just looking at the dates too Oct 27 16:04:28 Maybe I should ask my question more directly, but is there a way to extend the bbclass programmatically from python? Lets take as an example add echo's to do_compile (with a key value pair) Oct 27 16:05:27 fenrig: do_compile is normally a shell function, you don't need python to use bitbake variables from there. do_compile_append () { bbwarn "${FOO}"; } Oct 27 16:05:48 if you really want to use python for that, you'd have to use prefuncs or postfuncs flags to run a separate python function before or after the task, since the task isn't python. Oct 27 16:06:07 bitbake's _append/_prepend are just concatenation, so concatenating shell to python or vice versa will cause issues Oct 27 16:06:09 and I can't use python to extend the shell function Oct 27 16:06:24 i wanna use python to add shell commands in fact :) Oct 27 16:06:32 do_compile[postfuncs] = "some_function"; python some_function (){ bb.warn(d.getVar('foo')) } Oct 27 16:06:48 ah, do you want the python to run at task execution time or parse time? Oct 27 16:06:58 parse :) Oct 27 16:07:05 in that case, inline python is your best bet Oct 27 16:07:29 do_compile_append () { ${@my_function(d)} }; def my_function(d): return 'echo some random shell command' Oct 27 16:07:38 the python function returns the text appended to do_compile in this case Oct 27 16:07:41 would that do? Oct 27 16:08:04 then you can use bitbake -e to examine the final do_compile command with the returned commands Oct 27 16:08:20 and then in the end result: the do compile only gets the "echo some random shell command" appended? Oct 27 16:08:20 though I dread to think why you'd want to do that Oct 27 16:08:23 When I switch to zeus, I get this error
poky > source oe-init-build-env Error: TEMPLATECONF value points to nonexistent directory 'meta-yocto/conf'
Oct 27 16:08:39 fenrig: yep Oct 27 16:08:44 Is that as simple as editing: poky/build/conf/templateconf.cfg ? Oct 27 16:08:45 funk123: that would be because in zeus, meta-yocto doesn't exist and its called meta-poky Oct 27 16:10:15 fenrig: https://github.com/MentorEmbedded/meta-mentor/blob/master/meta-mentor-common/classes/sdk_extra_vars.bbclass is sort of similar, you can see it produces the bits to be cat'd to the script using inline python which returns those Oct 27 16:10:36 funk123: might i point you to https://www.youtube.com/playlist?list=PLD4M5FoHz-TxMfBFrDKfIS_GLY25Qsfyj for some introductory content :) Oct 27 16:10:37 rburton: well I have a bbclass in python here from my company that parses config options and hands them to the application, but I noticed that when I change the config option (from a bbappend) then yocto doesnt properly recompile the recipe. So I was thinking of echo'ing these config options with their values and add them to do compile so that when Oct 27 16:10:37 the value changes, the do compile hash is changed. and it would trigger a recompile Oct 27 16:11:04 seems reasonable Oct 27 16:11:07 LetoThe2nd : thanks again Oct 27 16:11:23 can't you just mark the variables as dependents? Oct 27 16:11:32 how do I do that? Oct 27 16:11:42 Just searching YT for yocto was throwing up too much Oct 27 16:12:09 fenrig: [vardeps], oe-core has loads of examples Oct 27 16:12:26 if you were writing shell bitbake would notice Oct 27 16:12:31 but you're not, so you need to tell it Oct 27 16:12:44 or something Oct 27 16:12:51 the magic variable tracker has edge cases Oct 27 16:13:39 do_compile[vardeps] += "CONFIG_OPTION" Oct 27 16:13:43 just like this? Oct 27 16:13:50 yeah Oct 27 16:13:51 can I call it from python? Oct 27 16:13:58 call what? Oct 27 16:14:01 to add it to the bbclass Oct 27 16:14:10 just add that line Oct 27 16:14:21 you can't "add to a bbclass". just put the line in the bbappend for the recipes that use it Oct 27 16:14:33 next to your do_compile() that uses CONFIG_OPTION, add that vardeps line Oct 27 16:14:52 oh ok so all the recipes need this? Oct 27 16:15:03 that use those kind of options Oct 27 16:15:06 you said you have a class, so set it in there Oct 27 16:15:15 unless you can directly adjust the bbclass, or globally inherit a class that conditionally operates in anonymous python based on whether the class was inherirted Oct 27 16:15:17 yes but the config options are component specific Oct 27 16:15:17 if you copy/paste the same code then yes youll need to copy paste this bit too Oct 27 16:15:38 if you want to add to vardeps programmatically from anonymous python, you can use d.appendVarFlag() Oct 27 16:15:52 from what you described you have a class that has a do_compile() in that uses CONFIG_OPTION to do stuff Oct 27 16:16:01 rburton: yes Oct 27 16:16:02 so put the vardeps line in the class Oct 27 16:16:24 funk123: you should be able to build it with zeus just fine, but I think you set it up with an older release and not switching properly, if youre having trouble, I suggest you do thesetup agani but start with zeus from the beginning Oct 27 16:16:32 I think kergoth's idea is far superior Oct 27 16:16:35 without an example showing your use case, this is mostly speculation based on what we're guessing you need :) Oct 27 16:16:42 adding them programmatically would be ok Oct 27 16:17:03 imho, kergoth is simply telling you what *can* be done Oct 27 16:17:24 right, there are options, you'll have to decide what to use based on your requirements Oct 27 16:17:26 I am dashed new to all this. So I appreciate the advice Oct 27 16:17:35 Even if it is don't work with daisy Oct 27 16:17:39 yeah well the bbclass already parses the config options, so the loop is already there Oct 27 16:18:49 I just did not know how to pick a branch from https://github.com/Xilinx/meta-xilinx that would work on my native ubuntu 20.04. Oct 27 16:19:10 funk123: when in doubt, the latest Oct 27 16:19:17 Trying Zeus now. Oct 27 16:19:46 I just deleted the build directory and started all over again after switching to Zeus. Oct 27 16:20:01 Is there a better way to clean the build directory? Oct 27 16:20:34 funk123: well, master works with dunfell and gatesgarth, so mayb try oe-core dunfell + meta-xlinix master Oct 27 16:21:29 rburton : how did you figure out which branch is compatible with meta-xlinix master? Oct 27 16:21:41 https://github.com/Xilinx/meta-xilinx/blob/master/meta-xilinx-bsp/conf/layer.conf#L14 Oct 27 16:22:16 Making notes. Thanks a lot. Oct 27 16:22:36 hmm, wonder if ithe layer index should be highlighting what a given branch supports, yocto release wise Oct 27 16:22:48 http://layers.openembedded.org/layerindex/branch/master/layer/meta-xilinx/ doesn't mention it, but folks should probably know ift he layer is out of date Oct 27 16:22:52 (not the case for this one, but..) Oct 27 16:23:11 We (Xilinx) never made a Dunfell branch.. a few people have said master worked for them -- but Zeus was the last, and gatesgarth (master) is the current development Oct 27 16:23:18 Just to summariise: oe-core dunfell + meta-xlinix master + yocto master ? Oct 27 16:23:30 funk123: what is yocto master? Oct 27 16:23:43 yocto master + meta-xilinx master works -- except for Microblaze.. Oct 27 16:23:57 This is for the Zynq Oct 27 16:23:57 (but to be clear, it's still very much in development, so it's "rough") Oct 27 16:24:31 i'm assuming funk123 means meta-yocto, since there's no such thing as 'yocto master' :) Oct 27 16:24:32 Zynq has microblaze for the embedded firmware as well as if you enable softcore in your bitstream.. but if you aren't using any of that, you can ignore it Oct 27 16:24:55 yocto master, I assume that means the 'poky' repository, which included oe-core and meta-yocto Oct 27 16:25:19 d.appendVarFlag('do_compile', 'vardeps', key) Oct 27 16:25:22 should do it right? Oct 27 16:25:28 fenrig: yep, but don't forget the leading space Oct 27 16:25:38 the python indents? Oct 27 16:25:41 it's direct string concatenation, and the flag is a space separated list of words Oct 27 16:25:46 if it has a value already.. Oct 27 16:26:04 i.e. vardeps = 'foo', you d.appendVarFlag('do_compile', 'vardeps', 'bar') now it's 'foobar', not 'foo bar' Oct 27 16:26:10 so d.appendVarFlag('do_compile', 'vardeps', key) in a loop would overwrite everytime? Oct 27 16:26:16 you'd need to appendVarFlag ' bar' or ' ' + key Oct 27 16:26:32 ah okay Oct 27 16:26:33 no, it's appending, not overwriting. if you want to blow away an existing value, use setVarFlag Oct 27 16:26:45 i forgot the leading space Oct 27 16:26:47 it seems Oct 27 16:26:47 just it's space separated, so if you want to add, include a leading space with your appended value Oct 27 16:26:50 * kergoth nods Oct 27 16:27:02 fray: that is right. Just building for the PS part Oct 27 16:27:47 Hi ! I am trying to build a disk image for x86-64, using WIC to generate the ESP and bootloader configuration. The thing is I have multiple rootfs and I just can't seem to figure out how to have WIC generate multiple bootloader entries for each rootfs. By looking at how WIC is actually implemented, it does not seem to support this. Do I necessarily have to create a WIC plugin to achieve that, or is Oct 27 16:27:49 there an other way to generate the bootloader configuration? Oct 27 16:28:01 another question, i dont need to pass the value of the key right. It's being checked dynamically right? Oct 27 16:30:03 it's just python, assuming key is a variable holding the string you want to add, it's fine Oct 27 16:30:32 is there a way to check the vardeps parsed result at the end? like in a tmp/do_compile.log Oct 27 16:30:33 frey: what advice do you have for someone starting out given that you said both gatesgarth and zeus are the latest? Which should I pick? Oct 27 16:30:46 key is the string of the variable name in the yocto recipe Oct 27 16:31:20 Yumasi: IIRC creatig your own plugin should allow you to do that, you just have to make wic uderstand and process the several bootloader entries Oct 27 16:31:44 .bb contains CONFIG_SOMETHING ??= "y" . Then key is "CONFIG_SOMETHING" Oct 27 16:31:59 fenrig: yeah, should be fine. ' ' + key Oct 27 16:32:09 GREAT <3 Oct 27 16:32:32 yocto is awesome, too bad it's a pretty steep knowledge hill Oct 27 16:33:00 fenrig: I have the same feeling :) Oct 27 16:33:16 Yumasi: specifically the bootimg-partition.py file would need some mods Oct 27 16:34:10 funk123: it depends of what you want, if you want the latest then pick gatesgarth Oct 27 16:37:46 alejandrohs: That's what I thought. Thanks for your answer! (I am using EFI but I see how bootimg-efi.py would need to be modified for that). I suppose I can't have multiple partitions with the "/" mountpoint (I guess it would have trouble generating the fstab, amongst other things)? Oct 27 17:01:02 khem: have you tried building with host ld being clang? Oct 27 17:02:44 rburton: not sure I want to know ;-) Oct 27 17:03:57 no, i don't think you do Oct 27 17:09:48 hah Oct 27 17:45:00 rburton: not as ld but as host compiler yes Oct 27 17:46:09 gentoo and mageia already use clang to bootstrap so its much better to build native packages now a days Oct 27 17:47:25 ok Oct 27 17:47:28 i blame user error Oct 27 17:47:43 give it a shot we can fix things up nothing is impossible Oct 27 18:15:53 khem: you're not supposed to tell people that, they may hold you to it ;) Oct 27 18:22:11 if the do_compile changes then all programs that recursively depend on that changed recipe should be recompiled right? Oct 27 18:41:37 yes Oct 27 19:02:27 smurray: that was addressed to rburton specifically not universal statement :) Oct 27 19:11:53 khem: heh Oct 27 19:42:48 JPEW: I hacked up and sent out a patch for that hashequiv issue, tweaked the one I had previously. Will see how testing looks with it Oct 27 20:22:15 RP: seeing WARNING: busybox-1.32.0-r0 do_populate_lic: busybox: No generic license file exists for: bzip2-1.0.4 in any provider Oct 27 20:23:53 I guess we also need to add bzip2-1.0.4 license text in meta/files/common-licenses Oct 27 20:24:07 along with Yann's change Oct 27 20:50:04 khem: will be the busybox license patch :/ Oct 27 20:54:56 is there a way to completely disable/delete/overwrite a task defined in a recipe by appending it? Oct 27 21:03:35 mbulut: anonymous python and deltask it? or just specify your own task to overwrite it? Oct 27 21:04:39 yeah thx, I was going for deltask but it seems not available in rocko Oct 27 21:04:49 will overwrite it Oct 27 21:13:47 https://www.yoctoproject.org/community/events/ Oct 27 21:13:57 not a lot to see there Oct 27 21:43:47 * RP discovers the scatter gather offload on a network card was breaking NAT Oct 27 22:18:25 noooo Oct 27 23:34:19 Thanks to advice I received earlier, I was able to compile (without errors) yacto with meta-xilunx (zeus branch). I have added some information here: https://solardata.s3-ap-southeast-1.amazonaws.com/zynqboot.pdf Oct 27 23:35:09 Can anyone provide some advice on what I did wrong? Oct 27 23:36:32 That pdf is some text and screenshots Oct 27 23:43:37 you might want to check where the switch to reading vfat should happen. Oct 27 23:44:31 it seems your bootenv is stored on spi flash Oct 27 23:44:59 and that should perhaps have info about where to boot from Oct 27 23:45:07 dump it and read it Oct 27 23:46:25 Sorry, I lost the message since my webclient dies. Oct 27 23:46:37 Can you repeat your comments again, khem? Oct 27 23:46:50 (I was user 'row' ) Oct 27 23:47:55 And I posted this problem about yocto not booting: https://solardata.s3-ap-southeast-1.amazonaws.com/zynqboot.pdf Oct 27 23:48:53 it seems bootloader env is being read from spi flash Oct 27 23:49:10 amd that perhaps has logic to choose where to boot from Oct 27 23:49:27 Strangely I can boot if the ethernet cable is attached. Oct 27 23:49:28 if you are able to break the boot at bootloader shell then you can see Oct 27 23:49:44 I copied the exact files to my tftp server Oct 27 23:50:11 yes thats becuase perhaps network boot is ahead of emmc boot in order Oct 27 23:51:01 sometimes uboot waits for a second or two before loading kernel you can press return key and see if it breaks in boot shell Oct 27 23:51:53 Load device from filesystem (FAT by default) (Xilinx only) Oct 27 23:51:53 loadfs [dev] [address] [image size] [blocksize] Oct 27 23:51:53 [] Oct 27 23:51:53 loadmk [dev] [address] Load device generated with mkimage Oct 27 23:51:53 For loadmk operating on FIT format uImage address must include Oct 27 23:51:53 subimage unit name in the form of addr: Oct 27 23:51:54 Hit any key to stop autoboot: 0 Oct 27 23:53:49 khem: Ah, I see what you mean Oct 27 23:55:12 https://solardata.s3-ap-southeast-1.amazonaws.com/Screenshot_2020-10-27_18-54-28.png Oct 27 23:55:23 That shows it is trying to mmc boot Oct 27 23:55:41 so how do I change it to SD card? Oct 28 00:03:59 break the build at bootload prompt by pressing ENTER key few times Oct 28 00:04:19 then type printenv and perhaps it should have option to boot from SD Oct 28 00:04:37 which you can run then Oct 28 00:07:53 https://i.imgur.com/fK9e0Wh.png Oct 28 00:12:43 zybouser: sdboot is the command youre looking for Oct 28 00:13:10 zynqmp issued that automatically, I cant remember if zynq does it as well Oct 28 00:14:05 zybouser: you could issue those commands (the ones expanded from sdboot) one by one in case one fails you know what happened Oct 28 00:14:08 change "loadkernel=fatload mmc 0 ${kernel_load_address} ${kernel_image}" to "loadkernel=fatload sdboot 0 ${kernel_load_address} ${kernel_image}" in uEnv.txt? Oct 28 00:14:49 zybouser: no so from the prompt Oct 28 00:15:02 as khem said you interrumpt booting and you get a prompt Oct 28 00:15:13 you can type sdboot there Oct 28 00:15:18 I am at the "Zynq>" prompt now Oct 28 00:15:30 Zynq> sdboot Oct 28 00:15:30 Unknown command 'sdboot' - try 'help' Oct 28 00:16:06 if you look at the screenshot you sent it says sdboot= echo Copying ... Oct 28 00:16:42 that should be in your environment already Oct 28 00:18:51 ah got your point Oct 28 00:18:55 trying that now Oct 28 00:19:19 so sdboot = load mmc 0 ${load_addr} ${fit_image} && bootm ${load_addr}? Oct 28 00:20:03 right Oct 28 00:20:21 run that cmd Oct 28 00:20:48 https://imgur.com/pEeF5C9.png Oct 28 00:24:22 no, something's off Oct 28 00:24:23 "run bootkernel" Oct 28 00:24:31 are you doing the spl flow to boot? Oct 28 00:24:32 got it up and running Oct 28 00:24:37 sorry I have to go Oct 28 00:24:42 thanks a lot Oct 28 00:24:42 oh well Oct 28 00:24:45 im glad Oct 28 00:24:52 thanks again for the ponter Oct 28 00:26:12 I think you're not using the rootfs from the sdcard Oct 28 00:26:27 if thats what you want then thats fine Oct 28 00:26:41 I'd give this a read anyway: Oct 28 00:26:43 quick question: how do I edit uEnv to use rootfs from sd card Oct 28 00:26:45 https://github.com/Xilinx/meta-xilinx/blob/master/meta-xilinx-bsp/README.booting.md#loading-via-sd Oct 28 00:27:54 zybouser: I dont believe you would need to and it should work out of the box, so I think you're missing a step somewhere, but if you really want to, the u-boot-zynq-uenv recipe on meta-xilinx-bsp is what populates that Oct 28 00:28:07 * alejandrohs goes afk **** ENDING LOGGING AT Wed Oct 28 02:59:57 2020