**** BEGIN LOGGING AT Mon Sep 14 02:59:57 2020 **** BEGIN LOGGING AT Mon Sep 14 03:09:52 2020 **** BEGIN LOGGING AT Mon Sep 14 05:36:02 2020 **** BEGIN LOGGING AT Mon Sep 14 05:38:52 2020 Sep 14 06:28:30 hello guys, I'm struggling with kernel patches. I've to patch the defconfig and a Kconfig file inside a kernel tree. So, I've downlaaded the kernel with bitbake, modified the mentioned files inside the git/ folder. So that if I run "git status" I can see the previous files as modified. The I create the patch running "git diff > mypatch.patch". Then Sep 14 06:28:31 I copied the patch insid the "files" folder in the recipe folder. By the way the do_patch tasks seems to do nothing as the final .config files is not modified at all. ANy suggestion ? Sep 14 06:36:57 ilkappe: check in recipe which file its using for creating .config Sep 14 06:39:53 sometimes defconfig is added via file:// and used Sep 14 06:42:18 good morning Sep 14 06:47:52 khem, yep, no defconfig is added via file:// it's the one in the kernel's config folder. The strange thing by the way is that the patches seems to have no effect on the repo since if I run git status in the kernel folder I can't see any new commit on the tree Sep 14 07:23:40 ilkappe: In addition to adding the patch to files/, do you also add it to SRC_URI? Sep 14 07:27:33 erbo, yes I have ! The strange thin is also that the patches are not in the WORKDIR dir Sep 14 07:49:23 Hi! I need some custom bootloader arguments in my bake and realize I can use the canned directdisk-bootloader-config from poky/scripts/lib/wic/canned-wks/, but I want the directdisk-bootloader-config.cfg to be a part of my bsp layer and not change the original directdisk-bootloader-config.cfg in the poky directory. How do I go about that? Sep 14 07:53:21 I was thinking of .bbappending the recipe that provides "WKS_FILE" and the rest of the WKS stuff to change the directory where it searches for wks files, but that might not be the right way to go. The directdisk-bootloader-config.wks file includes common.wks.inc which is also in the poky directory. I guess what I'm saying is that I can't find any Sep 14 07:53:22 way of changing the bootloader parameters without modifying files in the poky directory. Sep 14 08:00:19 ilkappe: That is strange. If you do e.g. "bitbake -e virtual/kernel | grep ^SRC_URI", do you see that the final SRC_URI really contains your patch? Sep 14 08:07:10 erbo, thanks, I've missed a typo in SRC_URI Sep 14 09:41:42 hi. is there a way to stop wic from writing to /etc/fstab ? Sep 14 09:42:21 for reasons, I need a manually created /etc/fstab, and I don't want wic to touch it **** BEGIN LOGGING AT Mon Sep 14 09:54:18 2020 **** BEGIN LOGGING AT Mon Sep 14 09:59:59 2020 Sep 14 10:05:49 xantoz: `--no-fstab-update` Sep 14 10:06:04 xantoz: As an argument to the wic program Sep 14 10:06:21 right. where does it get called, normally? Sep 14 10:06:50 xantoz: It's executed from image_types_wic.bbclass Sep 14 10:07:09 ok, so I should override that somehow? Sep 14 10:07:13 I think `WIC_CREATE_EXTRA_ARGS` is the variable to add command line args for wic Sep 14 10:07:21 ah, that might be easier Sep 14 10:07:32 It's not terribly well documented, you may have to look at https://git.openembedded.org/openembedded-core/tree/meta/classes/image_types_wic.bbclass Sep 14 10:07:37 right now I tried just removing the "--fsoptions" in the .wks file Sep 14 10:07:46 And https://git.openembedded.org/openembedded-core/tree/scripts/wic Sep 14 10:09:18 thanks for the tips Sep 14 11:02:37 Hi all -- a question regarding multiconfig: What are the advantages of using a "multiconfig" arrangement over just having multiple yocto build directories pointing to the same source layers? Sep 14 11:14:24 jtrimble: multiconfig lets you build a single artifact that has multiple configurations in Sep 14 11:30:09 think an image that ships a hypervisor, a linux filesystem, and a rtos app Sep 14 11:30:33 or firmware, or whatever Sep 14 11:55:22 kinda stupid question but why do we not disable configure and compile tasks for allarch inheriting recipes? Sep 14 11:55:47 because you can still configure and so on Sep 14 11:55:54 think an icon theme Sep 14 11:55:58 like adwaita Sep 14 11:56:17 that configures and "compiles" (converts SVG to PNG), but the output is arch-independent Sep 14 11:56:38 allarch does break configure so if it looks for a proper compiler, it fails Sep 14 11:59:44 ok, outreachy project submitted Sep 14 11:59:46 we may have a corresponding influx of potential contributors soon (hopefully !) Sep 14 12:00:06 ndec ^^ FYI Sep 14 12:04:06 rburton: makes sense, thx! Sep 14 12:10:12 bluelightning: That looked like a really useful project, I hope we have plenty of applications :) Sep 14 12:10:23 bluelightning: thanks for doing it! Sep 14 12:34:22 bluelightning: thanks! Sep 14 12:45:26 RP: Well, I think the bitbaker cooker is a lot more stable now. After rebasing, I ran 4000+ iterations of the torture test this weekend and I stopped it without it deadlocking. Before that it was only getting about 100 before something would break Sep 14 12:46:08 JPEW: that is good news :) Sep 14 12:46:41 JPEW: I also now know where its blocking causing the timeout problems on the autobuilder, its the SyncThread. Any ideas on why that could be so slow? Simply IO bound? Sep 14 12:47:20 Possibly. Could be waiting on semaphore or something also Sep 14 12:47:36 JPEW: a sempahore from where? :/ Sep 14 12:47:53 bitbake is otherwise "gone" at that point Sep 14 12:48:15 RP: Ya, I don't know. I/O seems more likely. I don't think I ever saw that one in the torture test, so I/O seems most likely Sep 14 12:48:18 JPEW: ideas for better debug to help diagnose welcome, I've not gotten to thinking about it yet Sep 14 12:48:41 JPEW: does your torture test include invalid metadata to parse? Sep 14 12:48:46 It doesn't Sep 14 12:49:04 That's a good idea though **** BEGIN LOGGING AT Mon Sep 14 12:51:41 2020 Sep 14 12:56:57 RP: On the bad news, parsing seems to have gotten slower. A simple clean `bitbake -p` takes ~35s for me when it used to take < 20s (sorry, I don't have the exact numbers from before). Not sure if this is just my local setup or if it did indeed get slower... Sep 14 13:00:49 RP: Nevermind. Must just be me. I went back and it's still slower. I must have changed something in local.conf w/o realizing **** BEGIN LOGGING AT Mon Sep 14 13:05:47 2020 Sep 14 13:33:59 Is anyone around here building yocto on AWS? Which kind of machine instance do you use? That is, which combination of CPU/memory is ideal? I don't want to pay for ressources I don't use. Sep 14 13:34:53 spin the machine down when its not building, just ensure you have persistent storage for sstate Sep 14 13:59:03 manuel1985: When I was experimenting with it, I chose C5aXlarge nodes. Seemed to have the best size/cost tradeoff (at the time). Sep 14 13:59:33 the graviton2 arm nodes are cheap and in my benchmarking of comparable speed Sep 14 14:00:51 JPEW, rburton: Thanks! Sep 14 14:04:04 manuel1985: I also just setup my nodes to mount a EFS volume for the sstate cache. Maybe not the most cost effective, but it was really easy and works really well Sep 14 14:04:49 Fully disclosure, I'm not doing anywhere near production volume builds, just playing around :) Sep 14 14:12:49 JPEW: Is there a reason you chose EFS over EBS or instance store? Is it cheaper? Or because you can access EFS simultanously from different hosts? I'm very new to AWS... Made my account just yesterday... Still trying to make sense of all these options there are... Sep 14 14:15:47 Right. You could do EBS I think, but you'd have to stand up an NFS server instance in front of it so that your nodes can access it. This seemed more complicated than I wanted to do for my expirement. Sep 14 14:28:23 Facing an issue compiling binutils. The error says:fatal: repository '/opt/grays-poky/build-grays/downloads/git2/sourceware.org.git.binutils-gdb.git/' does not existWhereas I see this directory over there. Can also do a git clone, so it is not a network issue. Sep 14 14:29:06 Facing this issue with binutils Sep 14 14:43:01 JPEW: I've wondered about performance. Sadly its not something I've had any time on recently Sep 14 14:43:18 JPEW: that sync threads thing was a performance measure, wondering if we should just get rid of it Sep 14 15:01:08 RP: I'm trying to figure out what the sync thread is doing Sep 14 15:08:31 JPEW: its writing out bb_caches to disk Sep 14 15:08:49 JPEW: so iterating over the caches calling .sync() Sep 14 15:09:13 JPEW: we do this so bitbake can get on with doing whatever it wants whilst the cache is saved out in parallel Sep 14 15:22:30 RP: Oh, I misunderstood the "get rid of it" part. You mean get rid of the thread and sync on the main thread, not get rid of the sync() completely? Sep 14 15:23:22 JPEW: right Sep 14 15:23:29 JPEW: just block there Sep 14 15:23:46 RP: It seems like final_cleanup() is always right after shutdown(), so the thread likely isn't really gaining anything Sep 14 15:24:18 JPEW: under many workloads it would be long done by then though Sep 14 15:24:42 JPEW: the cleanup is just to block if its still there Sep 14 15:25:01 Well, I guess the thread parallelizes the parse_cache_savemerge(), fetcher_parser_done(), and profile saving? Sep 14 15:25:25 JPEW: right Sep 14 15:36:01 JPEW: FWIW there is an open bug about parsing failures and unclean bitbake shutdowns causing locks so it may be worth a second torture run like that Sep 14 15:39:05 K, I'll modify the test to do that and give it a go Sep 14 15:43:00 JPEW: thanks. Reporter said the issues weren't fixed :/ Sep 14 15:45:52 Hey hey guys! Had anybody experience in making graphics driver autodetection for yocto? Sep 14 15:47:17 i'm not sure this is specific in any way to Yocto? Sep 14 15:48:49 Yes and no (I'm not sure). I was trying to to it with alpine and totally failed. That's why I am asking before I switch to something different :) Sep 14 15:49:25 I am trying to build a distribution with sole purpose of giving presentations. That's the reason for graphics driver Sep 14 15:51:22 sxiii: why do you need to know about the graphic driver? Sep 14 15:53:55 Let's say we have nvidia and two screens (one full-hd laptop screen and a 4K projector). We need to be sure that resolutions are okay. Or we can have AMD graphics and 1 screen. Or intel graphics... Etc. Ideally, I need automatic videodriver detection and support. Or I might try to stick with some generic driver, too, but I'm not sure that this will give some reasonable results Sep 14 15:54:49 I don't actually need to know, I just want the full-resolution of the device to be automatically utilized. Sep 14 16:44:20 Hey, so when I used boot2qt. It generated a configure-qtcreator.sh script for setting up the kit in qt creator for new machines. Is there such a thing for eclipse? Or is there any good instruction to setup eclipse? All seem to use the yocto plugin which is not supported for newer eclipse releases. **** BEGIN LOGGING AT Mon Sep 14 16:56:54 2020 Sep 14 17:36:17 RP: I know there is a bug with the config fragment to reproduce that issue, but I can't find it ATM. Do you happen to know what it is? Sep 14 17:42:42 https://bugzilla.yoctoproject.org/show_bug.cgi?id=14034 Maybe? **** BEGIN LOGGING AT Mon Sep 14 17:45:38 2020 Sep 14 18:04:45 Hey all, running into a weird issue with Git repos and bitbake. It's failing to clone a specific repo as a submodule (permission denied (publickey) error), but it doesn't have the same issue cloning other repos on the same git server using the same keypair. I am also able to clone this repo with no incident manually (outside of bitbake) Sep 14 18:05:05 Any ideas what may be going on here? Sep 14 18:07:59 submodules, are each cloned one at a time in OE. So if you are xpecting the original SRC_URI to point to the credentials it won't work right.. Sep 14 18:08:27 the other thing you can do is manually specify the URIs for all of the submodules within your SRC_URI of the recipe, that might give you more info why it's failing.. Sep 14 18:08:37 (basically use "git:..." for each URL not gitsm) Sep 14 18:11:50 So the username is encoded in the request URI (git@[submodule URI]) and the credentials should be handled by ssh (as configured in .ssh/config). What I'm seeing is that gitsm had issues with user and password included in the URI, but that is not the case with my setup Sep 14 18:16:08 In fact it looks like bitbake isn't cloning the submodule manually (via a URI in a recipe) but as a consequence of a recursive clone of the parent repository. Sep 14 18:16:50 Or, sorry, it configures a number of submodules and then does a recursive submodule update --init then sync Sep 14 18:21:54 username is not transfered to the submodules Sep 14 18:22:08 the submodule uri's are used -exactly- as speified in the submodule Sep 14 18:22:27 so if you need to provide credentials for the submodules, you will need to break it out manually and specify each URI within the recipe itself.. Sep 14 18:22:59 The behavior I am describing started sometime last year if I remember correctly (2019) Sep 14 18:24:00 Well, my submodule URIs do specify the username, and credentials should be handled by my ssh agent (using keypairs specified in ~/.ssh/config). Is there something about the bitbake configure process that would interfere with the ssh agent being able to provide those credentials? Sep 14 18:24:57 This is in .gitmodules - the recipe does not specify separate URIs for submodules, just adds shell commands to update the submodules we need for a given Yocto build Sep 14 19:14:05 hi, i have a recipe which pulls a private git repo. in order to connect i have my credentials in .gitconfig using git credential helper Sep 14 19:14:22 however, bitbake does not seem to pick this up. is this by design? Sep 14 19:26:55 elcfd: not really an answer to your question, but I usually go for git over ssh, and ssh key authorization for private repos Sep 14 19:28:01 erbo < normally i would but work have their git server setup without ssh access. it seems im going to have to go into meetings with IT to get this changed! Sep 14 19:28:20 was hoping to avoid having to do this because it wont be a quick process Sep 14 19:28:22 Ah, the dreaded IT meeting :) Sep 14 19:28:40 indeed - you want what port opened! Sep 14 19:29:08 i nmapped the server they are filtering the ssh port :( Sep 14 19:32:05 elcfd: Hmm, this (old) SO post seems to indicate that it should work to have it in .gitconfig Sep 14 19:39:15 erbo < do you have the link for the post? Sep 14 19:40:34 elcfd: right, I intended to paste that at the end of the last message :) https://stackoverflow.com/questions/53656233/yocto-fetch-private-repo-using-https Sep 14 19:50:26 erbo < thanks - i think it may actually be the work proxy causing the authentication error message i am seeing. Sep 14 20:21:05 RP: Ok, that was pretty easy to reproduce. Looks like another multiprocessing termination bug. FWIW, the re-write I've been working on seems to fix it Sep 14 20:21:20 will run more tests overnight **** BEGIN LOGGING AT Mon Sep 14 20:30:32 2020 **** BEGIN LOGGING AT Mon Sep 14 20:57:22 2020 Sep 14 20:57:23 JPEW: hmm. So its definitely not fixed by existing patches? :( Sep 14 20:57:55 JPEW: you found a fragment I guess? I was afk Sep 14 21:02:48 RP: Ya, I did. It's the exact one from the bug report. And it's pretty easy to reproduce, just CTRL+C after the stacktrace appears from the invalid expansion Sep 14 21:03:20 JPEW: ah, I wasn't doing the Ctrl+C Sep 14 21:03:34 RP: The torture test finds it in less than 10 iterations for me, and most of those are where it completely misses the window because it's too early or too late Sep 14 21:03:36 JPEW: I wonder if blocking the SIGINT from the children would help Sep 14 21:04:10 RP: Hmm, maybe. I can try it Sep 14 21:04:46 JPEW: or maybe blocking it in the parent Sep 14 21:04:57 JPEW: worth a try since you can easily reproduce! Sep 14 21:15:21 RP: `signal.signal(signal.SIGINT, signal.SIG_IGN)` seems to fix the problem. It's done over 100 iterations now Sep 14 21:15:24 hm there is no 3.2 schedule page on the wiki? https://wiki.yoctoproject.org/wiki/Yocto_3.1_Schedule exists, but https://wiki.yoctoproject.org/wiki/Yocto_3.2_Schedule is 404 Sep 14 21:15:44 given that gatesgarth (3.2) is planned for october, that is a little weird ... ? Sep 14 21:17:48 RP: Patch sent Sep 14 21:18:24 JPEW: great, thanks! Sep 14 21:18:47 dv_: its a sign that we don't have enough people to do everything :( Sep 14 21:38:45 sakoman: did you make any progress with https://autobuilder.yoctoproject.org/typhoon/#/builders/83/builds/1389 - the python version issue? Sep 14 21:41:48 sakoman: its a bit invasive but I think we split utils.setup_buildtools_tarball into two and then call the second half from the qa-send script Sep 14 22:18:11 RP: oh.. Sep 14 22:18:26 RP: could this lead to gatesgarth being postponed? Sep 14 22:18:56 dv_: the release is going ahead as planned, we just don't have all the planning documents and wiki updated :/ Sep 14 22:19:03 ah I see Sep 14 22:19:25 well, assuming we actually build M3 (feature freeze). That is two weeks behind right now but I'm not overly concerned about that Sep 14 22:19:33 I originally wanted to look up in what milestone we are in Sep 14 22:19:57 and I assume that gstreamer 1.18 (which got released ~7 days ago) won't make it in .. too late for that Sep 14 22:20:20 dv_: how big a change is it and are there good reasons we'd want it? Sep 14 22:20:40 its late and past feature freeze so we'd need a compelling reasons Sep 14 22:20:40 well it is always good to stick to the latest gstreamer version. support is much easier then Sep 14 22:20:47 oh thats what you mean Sep 14 22:20:57 no, I dont think there are pressing reasons for an exception Sep 14 22:21:27 I understand latest is "best" but we do have to draw a line somewhere! Sep 14 22:21:56 of course, of course. thats why I wanted to know the current milestone. if it were an early one, I'd try to get 1.18 in Sep 14 22:22:01 but M3 is a late one Sep 14 22:22:18 anyway, thanks for the info Sep 14 22:22:56 dv_: np, thanks for checking! Sep 14 22:27:49 you're welcome. Sep 14 22:28:40 oh, one last thing (different topic), then I anyway have to go: why does the SDK's environment file add optimization and debug symbol flags to CFLAGS, CXXFLAGS, LDFLAGS, and also to build system specific scripts like meson.cross? Sep 14 22:29:21 this gets very annoying when you try to build a debug build, wanna step through it with gdb, only to find that stuff has been optimized out Sep 14 22:29:42 optimization and debug symbol flags (-g etc.) shouldn't really be part of a BSP setup Sep 14 22:30:13 err, part of an SDK setup Sep 14 22:30:15 dv_: I'd imagine people do want the optimization flags there in general Sep 14 22:30:40 I'd assume we're thinking people will split the debug symbols out like we do with in tree builds Sep 14 22:30:46 RP: I'm sending some -stable bumps and a fixed up lttng-modules patch in the next few minutes. just one more build to see pass. Sep 14 22:30:48 for yocto builds, yeah. Sep 14 22:30:58 but if you use the SDK, the situation is different Sep 14 22:31:31 dv_: It depends. If we remove that info its hard to someone to get it back but easier for them to remove it if needed Sep 14 22:31:52 zeddii: ok, I'm about to start a build and head to bed. I wonder whether sleep will win :) Sep 14 22:31:56 hm. maybe some flag that can enables/disables it? Sep 14 22:32:35 I'll discuss this with my BSP colleaguees Sep 14 22:32:36 -e Sep 14 22:32:43 dv_: I was just wondering about two environment files. That has big impact on the docs though Sep 14 22:32:57 dv_: its worth thinking about Sep 14 22:32:59 true Sep 14 22:33:02 alright Sep 14 22:33:11 anyway, gotta go, l8r! Sep 14 22:33:58 it built. sending the queue. feel free to let sleep win. Sep 14 22:35:30 zeddii: series doesn't apply? :/ Sep 14 22:35:41 some kind of strange encoding on one patch. Sep 14 22:35:51 git send email complained on the way out. Sep 14 22:37:07 zeddii: patch just before the ones you sent in the branch? Sep 14 22:37:13 linux-yocto/config: netfilter: Enable nat for ipv4 and ipv6 Sep 14 22:37:50 bugger. Sep 14 22:37:52 I missed one. Sep 14 22:38:17 I can resend, or you can grab it off the branch. apparently I can't count. Sep 14 22:38:24 rushing never pays off. Sep 14 22:38:46 zeddii: I have it :) Sep 14 22:39:09 I'll follow up to my 0/N in case someone out there is trying to grab the patches and apply them. Sep 14 22:39:24 my AB run is green now with those changes, so in theory, it shouldn't cause you any issues. Sep 14 22:39:31 zeddii: I was rushing to get to sleep to which obviously meant it all was doomed ;-) Sep 14 22:39:53 zeddii: I'm off to bed now, will see what happened tomorrow Sep 14 22:39:54 my first AB runs .. didn't go well, hence why I ended up with lttng changes :D Sep 14 22:39:57 later! Sep 14 22:44:27 2020 and ltt is still a thing. Sep 14 22:44:44 ltt-ng-ng-ng Sep 14 22:46:13 wouldn't have predicted that back in 2005 or whatever. **** BEGIN LOGGING AT Mon Sep 14 23:12:46 2020 Sep 15 01:19:54 heh, but year of linux desktop is still elusive **** BEGIN LOGGING AT Tue Sep 15 01:43:13 2020 **** BEGIN LOGGING AT Tue Sep 15 02:01:12 2020 **** ENDING LOGGING AT Tue Sep 15 02:59:57 2020