**** BEGIN LOGGING AT Tue Aug 13 02:59:56 2019 Aug 13 08:20:46 khem: I think for speed you'd need an older bitbake :( Aug 13 08:24:06 Saur: thanks, that does help confirming that Aug 13 08:55:51 Saur: FWIW I show that commit giving around a 25% hit, then master-next is about 2.5 times slower :/ Aug 13 08:56:48 RP: Ouch. Not exactly the numbers we'd like to see... Aug 13 08:57:12 Saur: no. -next is the really really bad point Aug 13 09:04:19 Saur: I sent a summary to the list Aug 13 09:07:25 Saur: is master at least usable or are you seeing much worse than 25% degradation there? Aug 13 09:09:15 tig Aug 13 09:11:05 But honestly, why are there people relying on master being perfect 100% of the time ? It is the *development* branch... Aug 13 09:11:41 kroon: I'm not sure people are. They're reporting problems which is fair enough Aug 13 09:12:04 kroon: I'm just trying to quantify how much of a problem and from when/where Aug 13 09:17:52 RP, yeah. maybe I exaggerated a bit there, from reading that email Aug 13 09:48:37 New news from stackoverflow: apt-get failure in chroot environment Aug 13 10:16:27 sigh, both poky and meta-openembedded contain stress-ng now Aug 13 10:34:11 mcfrisk, seems like there should be a remove posted for meta-oe then? Aug 13 10:35:03 yea, testing a patch locally now Aug 13 10:35:28 also adding some stress compatibility PROVIDES Aug 13 11:10:41 "git-replacement-native" .. what is it for? Aug 13 11:14:52 There are some other "replacement-native" providers.. So what is the "replacement-native" business? Aug 13 11:22:59 armpit: I saw the note from RP earlier about the -ptest test in the autobuilder. That sounds familiar but if I remember correctly that's one of the ones I was able to get working once RP helped me get the QEMU setup working. I've since trimmed it from our internal stack of builds. Aug 13 11:26:42 armpit: that or it was related to using the "master" of the ab-helper while having to hold back all the builds on our side to thud. I had to trim some tests related to that (since they didn't exist back then). Aug 13 11:29:05 Hi all, Aug 13 11:36:51 My yocto build manifest branch is topic/yocto, so I'm getting below error with Thud do_rootfs: Error executing a python function in exec_python_func() autogenerated: Aug 13 11:36:51 The stack trace of python calls that resulted in this exception/failure was: Aug 13 11:36:52 File: 'exec_python_func() autogenerated', lineno: 2, function: Aug 13 11:36:52 0001: Aug 13 11:36:52 *** 0002:write_image_test_data(d) Aug 13 11:36:53 0003: Aug 13 11:36:55 File: '/mnt/home/topic_latest/openembedded-core/meta/classes/rootfs-postcommands.bbclass', lineno: 323, function: write_image_test_data Aug 13 11:36:58 0319: link_name = d.getVar('IMAGE_LINK_NAME') Aug 13 11:37:00 0320: testdata_name = os.path.join(deploy_dir, "%s.testdata.json" % d.getVar('IMAGE_NAME')) Aug 13 11:37:02 0321: Aug 13 11:37:06 0322: searchString = "%s/"%(d.getVar("TOPDIR")).replace("//","/") Aug 13 11:37:08 *** 0323: export2json(d, testdata_name, searchString=searchString, replaceString="") Aug 13 11:37:10 psiva87: please use a pastebin! Aug 13 11:37:10 0324: Aug 13 11:37:12 0325: if os.path.exists(testdata_name): Aug 13 11:37:14 0326: testdata_link = os.path.join(deploy_dir, "%s.testdata.json" % link_name) Aug 13 11:37:16 0327: if os.path.lexists(testdata_link): Aug 13 11:37:17 psiva87: please use a pastebin service for more than a few lines of text. Aug 13 11:37:18 File: '/mnt/home/topic_latest/openembedded-core/meta/lib/oe/data.py', lineno: 46, function: export2json Aug 13 11:37:20 0042: data2export[key] = '' Aug 13 11:37:22 0043: except AttributeError: Aug 13 11:37:24 0044: pass Aug 13 11:37:26 0045: Aug 13 11:37:28 *** 0046: with open(json_file, "w") as f: Aug 13 11:37:29 *SIGH* Aug 13 11:37:30 0047: json.dump(data2export, f, skipkeys=True, indent=4, sort_keys=True) Aug 13 11:37:32 Exception: FileNotFoundError: [Errno 2] No such file or directory: 'IMX8MQ_topic/yocto_20190813063206.testdata.json' Aug 13 11:38:25 LetoThe2nd: neverpanic: ok, will use it. Aug 13 11:40:06 jofr: if you haven't figured out the replacement-native yet, its so that ASSUME_PROVIDED can have git-native which is from hosttools. But if something needs git-native built it can depend on git-replacement-native which will build git-native and put it into the recipe-sysroot-native Aug 13 11:46:53 nrossi: Yeah, I had kindofsortof figured that out. But thanks! Aug 13 11:55:39 nrossi: I had made a recipe a long time ago that depended on git-replacement-native (and I can't remember why I did the "replacement" bit), but it kept breaking. Exec format error on the git it built. Changing it to git-native instead solved my problem. Aug 13 11:56:24 jofr: you probably needed some feature that was not in your host git at the time ;) Aug 13 12:02:31 nrossi: Perhaps. Kind of doubt it, though.. All it's doing are a couple of git-clones for dependencies for this homegrown package. What puzzles me even more is why I'm getting the exec-format-error in the git-replacement-native binary. "file" says it's an x86-64 as it should be. I'm thinking, perhaps the interpreter (loader) is somehow broken within that sysroot. But all other executables execute just fine.. Aug 13 12:40:19 Hi #yocto, I have a question about initramfs. I working in a project which has the requirement to use two kernels. One is a standard kernel without initramfs which gets normaly booted with a root-kernel parameter to a appropiate partition on emmc. Aug 13 12:40:20 The second kernel should include an initramfs with some tooling stuff for maintainance work. Aug 13 12:40:21 How can I make this with yocto? Aug 13 12:46:35 silviof: are you asking about creating the initramfs? or how to construct a project steup with two kernel settings? i don' completely understand what the main question is, sorry. Aug 13 12:48:26 LetoThe2nd: The second one. I need a project setup with two kernels. First **without** initramfs and the second one **with** initramfs. Aug 13 12:49:07 Maybe you have a project example on github or so?! ๐Ÿ™‚ Aug 13 12:49:55 silviof: ok. i personally would go the route of two separate distributions. they can include a common .conf that carries all the non-changing stuff, and only select different kernel configurations then. Aug 13 12:50:38 silviof: i don't think you can do all in one build run, at least not without massive tricks Aug 13 12:52:27 LetoThe2nd: Just to ensure I have understood you fully. Its not possible to build two kernels within one project with yocto? I have to need to build 2 different distributions. One with initramfs and one without initramfs? Aug 13 12:52:29 ouu Aug 13 12:52:56 silviof: i am not awre of a simple way to build two kernels, yes. Aug 13 12:54:42 silviof: well, you can certainly try to build an image that explicitly depends and therefore builds to specific kernels, but you'll have fun working around the virtual/kernel provider thing. and on top, the initramfs has to come from somewhere too, it usually is an image build of its own. Aug 13 12:55:39 silviof: so in a nutshell, any setup like this i have seen so far was a "tasteful" combination of several builds :) Aug 13 12:58:03 tihi okay "tasteful" ๐Ÿ˜ I understood. I will dig a bit the web what tells me about "more than one kernel builds". Could not think that this wasn't a usecase in all the years for yocto. Aug 13 12:58:04 Thanks so far LetoThe2nd . Aug 13 13:01:02 it is often a use case, we hear it regularly. but in my (very personal) opinion, its a question of the mind set. yocto is not a "complete project all in one building mechanism", but a "distribution building mechanism". so if your complete project requires more than one distribution - which seems to be the case - then the approach should probably be to use a building mechanism around yocto, that triggers Aug 13 13:01:08 the two biulds and then combines them ... Aug 13 13:01:10 ... into whatever artifact you need. Aug 13 13:01:31 the multiconfig stuff would work well for this use case, I think Aug 13 13:02:34 and I think there may have been some discussion of building two kernels on the oe-core mailing list a while back, worth searching for that Aug 13 13:02:53 smurray: care to elaborate? happy to be proven wrong respectively corrected. Aug 13 13:06:01 LetoThe2nd: since multiconfig allows building two different configs with the same bitbake invocation, it seems like it'd be somewhat straightforward to have the one config set up initramfs. Getting it all packaged into a single output image is perhaps tricky, though Aug 13 13:10:33 ouu, smurray this looks intresting, have search about muticonfig. Thanks for this hint. (). Aug 13 13:12:43 silviof: note that it works better in the last couple of releases (thud and warrior) than it did back in rocko Aug 13 13:15:23 JPEW: sadly whilst I'd say the performance is better, we still have timeouts :( Aug 13 13:19:19 RP: Is that with the NO_LINGER option set? Aug 13 13:23:46 RP: I also may have found a source of slowness in the hash server: If there are a lot of entries for the same taskhash (e.g. with different outhashes when builds aren't reproducible), it has to create a temporary b-tree to sort them. Aug 13 13:24:27 This is easy to test; temporarily rename the hash server database file and see if it gets faster with an empty database. Aug 13 13:32:25 JPEW: yes Aug 13 13:45:49 Hi there ! I am trying to use devtool on Yocto Warrior. It works quite well, but I have a weird (I think) issue Aug 13 13:46:25 I want to use the "autorev" functionnality on a git repository, but the devtool finish step does not work Aug 13 13:46:42 I basically get : Aug 13 13:46:44 oe.patch.CmdError: Command Error: 'sh -c 'git format-patch --no-signature --no-numbered AUTOINC -o /tmp/oepatchxthoybo4 -- .'' exited with 0 Output: Aug 13 13:46:45 fatal: bad revision 'AUTOINC' Aug 13 14:23:53 JPEW: hash equivalence appears to break the layer check script: https://autobuilder.yoctoproject.org/typhoon/#/builders/39/builds/927 :/ Aug 13 14:27:18 RP: Hmm, INTERNAL SERVER ERROR.... I think that usually means there was an exception in the server Aug 13 14:29:40 Anyway.... Does it even make sense to use the hash equivalence server when running the check layers script? I thought it was using the taskhashes to figure out if things are changing when they should not.... hash equivalence seems like it would confuse that Aug 13 14:48:45 JPEW: no, doesn't make sense. We'll need to guard against it though Aug 13 14:53:23 * LetoThe2nd announces stream starting at 17:00 Europe/Berlin -> https://www.twitch.tv/yocto_project Aug 13 14:53:29 that is, in 7 Minutes :) Aug 13 14:54:37 LetoThe2nd: hmm, I've had the wrong dates in the weekly status report :/ Aug 13 14:55:05 i guess its the vacation session anyways. but i want to keep my self-imposed schedule Aug 13 14:55:34 LetoThe2nd: sadly I have the weekly engineering meeting, as ever... Aug 13 14:57:18 RP: yes, i'm already thinking about how to deal with this, maybe change my schedule. i wasn't aware of it when planning, as i only knew of the public tech call every first tuesday. Aug 13 14:59:11 but as far as i can tell we're seeing good view counts of the recordings on youtube :) Aug 13 15:01:02 Engineering meeting at https://zoom.us/j/990892712 Aug 13 15:02:03 halstead: waiting for the host to start? Aug 13 15:02:11 JPEW: yes, different audiences Aug 13 15:02:15 was I on the wrong channel. I got booted Aug 13 15:02:16 RP, I was in it a moment ago. Aug 13 15:03:07 halstead: got in second try Aug 13 15:30:33 is it possible to post comment on twitch stream without account ? Aug 13 15:38:08 cheers :) Aug 13 15:42:19 baahahah. I finally got on the meeting, just to hear everyone hang up :D Aug 13 15:42:39 zeddii: its ok, we gave you the action items Aug 13 15:43:29 :D I'd expect no less! Aug 13 15:43:44 but I'll also be around next week (no ELC), so I'll follow up with them all completed! ;) Aug 13 15:44:06 zeddii: I'm not at ELC either... Aug 13 15:44:25 after my two weeks of vacation. I have a lot of changes to move along, traveling was not an option. Aug 13 15:44:52 * RP has no travel budget Aug 13 15:45:29 so no ELCE for you?!? Aug 13 15:45:59 i'm constantly messing up today. Aug 13 15:49:09 LetoThe2nd: have budget for ELC-E, just not ELC Aug 13 15:50:19 question on twitch: do we have an out-of-tree module example somewhere, a proper one? Aug 13 15:51:59 LetoThe2nd: we have a skeleton one, not sure about a proper one Aug 13 15:52:05 zeddii: ^^ Aug 13 15:57:27 RP: Ok, there are 50,000 duplicate task hashes (of 250,000 total) in the AB database, so the sqlite performance issue might be relevent Aug 13 15:58:01 lol strace pulling mesa, cairo, bluez, ... :D Aug 13 15:58:12 why is that ? Aug 13 16:01:58 we just use the skeleton one as the example. there are others in some packages. but they all follow the same format. Aug 13 16:04:35 most awful session i've ever done for sure. Aug 13 16:04:47 * LetoThe2nd needs some metal to recuperate. Aug 13 16:04:54 s/recuperate/recover/ Aug 13 16:05:34 thanks for your time Aug 13 16:05:42 thanks for joining us! Aug 13 16:06:05 I did not known this kind of format it is interesting Aug 13 16:06:20 amaury_d: every second tuesday a month Aug 13 16:06:42 on my agenda now :) Aug 13 16:06:55 feel free to drop questions, remarks, everything in here, i'm constantly monitoring this place :) Aug 13 16:07:07 but now, dinner! Aug 13 16:07:09 * LetoThe2nd AFKs Aug 13 16:35:40 JPEW: hmm, I wonder if we can tell somehow. I think we're going to need stats Aug 13 16:36:02 RP: Yep. I'll work on that Aug 13 16:36:28 JPEW: thanks. I'm going to restart the autobuilder build and try a load of my fixes. Anything I should try on the server? Aug 13 16:36:51 Not ATM Aug 13 16:38:28 ok. I found three issues with runqueue so I'm hoping that is a decent set of the issues Aug 13 16:40:11 RP: Good. On the plus side, it looks like we are going to have a pretty good regression test suite for the runqueue when you are done. It looks like it's been needing one for a while Aug 13 16:49:56 New news from stackoverflow: Yocto - Can't build library for host SDK Aug 13 17:09:51 hey, i'm new to yocto and i'm using it for work. we'd started on sumo and just moved to warrior.. is there an easy way to delete old stuff built from sumo without just deleting tmp/? Aug 13 17:10:49 JPEW: I'm struggling to make test cases for all of the issues :/ Aug 13 17:11:03 JPEW: but yes, its better to have what is now there than nothing Aug 13 17:16:57 LetoThe2nd, question from twitter Aug 13 17:16:59 Dan CaJacob Aug 13 17:16:59 @astrojockey Aug 13 17:16:59 ยท Aug 13 17:16:59 30s Aug 13 17:16:59 Replying to Aug 13 17:17:00 @yoctoproject Aug 13 17:17:02 and Aug 13 17:17:04 @balister Aug 13 17:17:09 What happens on the Yocto Project Twitch channel? Live building? Aug 13 17:46:07 hey is it ok to ask for help with yocto in here? Aug 13 17:58:44 xteixeira, just ask Aug 13 17:59:09 it may take a while for someone to see the question and answer Aug 13 17:59:17 oh ok Aug 13 18:00:48 so ive been developing a custom distro in yocto, but i cant get the kernel to load the initramfs on boot if i use the INITRAMFS_IMAGE_BUNDLE=1 option Aug 13 18:02:17 if I set INITRAMFS_IMAGE_BUNDLE=0 and include the initramfs image in .cpio.gz format separately and configure grub to point to it using the "initrd" option in the boot entry, it works. Aug 13 18:02:20 which machine ? Aug 13 18:02:42 im booting in a regular x86_64 computer Aug 13 18:02:52 using the linux-intel kernel from the meta-intel layer Aug 13 18:03:46 building everything for the intel-corei7-64 machine Aug 13 18:07:13 and btw, i've been using these instructions to install my distro to a hard drive: https://wiki.yoctoproject.org/wiki/How_do_I#Q:_How_do_I_put_Yocto_on_a_hard_drive.3F Aug 13 18:08:19 im not sure (and couldnt find anything on it) if i need a kernel parameter for the kernel to try to load a bundled initramfs image? Aug 13 18:08:59 quick question, I've set the PREFERRED_PROVIDER for one of my recipes in the distro config, however, it's still picking the newest version. I've verified that the bitbake -e output has the PREFERRED_PROVIDER set. How would I go about finding out why yocto is selecting the new version? Aug 13 18:09:19 omg, it's supposed to be PREFERRED_VERSION Aug 13 18:10:18 yes you need to use PREFERRED_VERSION_${PN} Aug 13 18:43:49 <__angelo> how can i find the path of the toolchain used by yocto ? Aug 13 18:55:36 __angelo: do you mean used *by* yocto to build its own tooling, or the one that yocto builds for itself and uses to build stuff? Aug 13 18:56:09 <__angelo> LetoThe2nd, sry. The one used to cross-build the kernel. Aug 13 18:56:32 __angelo: the former is your standard toolchain of your system. the latter is actually prepared per-recipe as far as i know Aug 13 18:56:37 __angelo: why do you ask? Aug 13 18:57:02 <__angelo> well i found it. Just to build the kernel separately Aug 13 18:58:00 __angelo: reacing directly into the build is very error prone. if you need a cross toolchain thats compatible with a specific image, its better to go the sdk route. Aug 13 18:58:45 Crofton|work: hum, what happens on twitch? you can see me mess up things, talking about random weird stuff and having a drink. Aug 13 18:58:57 Aug 13 19:05:38 Well, contra to what I said earlier, we can fix the performance problem with runqueue as its not where I thought it was Aug 13 19:06:38 * RP was quite surprised when glancing at the profile Aug 13 19:08:08 RP: pray tell Aug 13 19:08:33 LetoThe2nd: its in the scheduler which I didn't really change. Seems the list comprehensions there are really slow Aug 13 19:08:59 RP: fun! Aug 13 19:10:08 and 50 million calls to taskname_from_tid which is just bizarre Aug 13 19:10:54 hey i'm sure you can reach 100 million! you can do it! beleive in yourself! Aug 13 19:12:10 I don't even understand how the code did that Aug 13 19:16:31 hrhr Aug 13 19:19:19 http://git.yoctoproject.org/cgit.cgi/poky-contrib/patch/?id=5c79999b61e1232ff97e1f07667a98a04a58d4fb for anyone interested Aug 13 19:19:43 Could have been the generator never expanded to "empty" Aug 13 19:20:31 RP: Heh, that was a pretty hairy generator expression Aug 13 19:21:03 JPEW: it had grown a bit :) Aug 13 19:22:55 RP: The new code doesn't update self.buildable like the old one did Aug 13 19:23:11 JPEW: right, its ok though as it doesn't need to Aug 13 19:23:15 Ok Aug 13 19:23:23 JPEW: I did check! :) Aug 13 19:24:18 JPEW: old versions did poke around buildable but it was error prone so we have holdoff_tasks for that now Aug 13 19:24:30 RP: Makes sense Aug 13 20:00:49 RP: I was trying to put downloads/ folder into a nfs share, and fetcher gets stuck "NOTE: Fetching uninative binary shim from http://downloads.yoctoproject.org/releases/uninative/2.6/x86_64-nativesdk-libc.tar.xz;sha256sum=133387753a9acf3e1b788103c59fac91e968e2ee331d7a4b9498e926ada7be57" Aug 13 20:01:02 RP: I was trying to put downloads/ folder into a nfs share, and fetcher gets stuck "NOTE: Fetching uninative binary shim from http://downloads.yoctoproject.org/releases/uninative/2.6/x86_64-nativesdk-libc.tar.xz;sha256sum=133387753a9acf3e1b788103c59fac91e968e2ee331d7a4b9498e926ada7be57" Aug 13 20:01:42 If I wget this manually ( under same username ) in same location it works fine Aug 13 20:02:06 any hints where to look at, Aug 13 20:02:32 I see that it created a corresponding .lock file and then it spins Aug 13 20:08:49 I see that uninative is calling a normal fetcher which iin this case would be wget I suppose Aug 13 20:09:14 If I use local disk for downloads it works ok then Aug 13 20:43:57 khem: I'd have to guess at locking not working on the NFS mount? Aug 13 20:44:19 khem: the main other thing the fetcher would do is take out a lock using bb.utils.lockfile() Aug 13 20:44:28 khem: easy to test. Could try flock too Aug 13 20:49:27 RP: is there a way to enable locking on NFS drive Aug 13 20:52:28 khem: missing portmap/rpcbbind or lockd? Aug 13 20:52:44 khem: I don't remember :/ Aug 13 20:59:37 RP:its creating .lock file so does that mean its creating the lock fine Aug 13 21:16:33 khem: does it actually get a lock? I'm wondering if that is what is hanging? Aug 13 21:16:41 khem: where does strace or gdb show it? Aug 13 21:31:17 RP:now its working, I needed to mount it with nolock,local_lock=all Aug 13 21:31:44 the server is running v3, I wished it ran v4 Aug 13 21:41:09 khem: restricting the locks to local might be a bit risky? Aug 13 22:50:35 i'm making some recipies for kernel modules but i'm confused about where the module files end up. the package files (.deb) are empty. where do they go? Aug 14 00:22:44 On one of my minimal builds i am trying to remove a package ( ofono). when i type apt-get remove --auto-remove ofono i get the following: E: flAbsPath on /var/lib/dpkg/status failed - realpath (2: No such file or directory)E: Could not open file - open (2: No such file or directory)E: Problem opening E: The package lists or status file could no Aug 14 00:22:45 t be parsed or opened. Aug 14 00:23:11 I've added package management and opkg to my local conf Aug 14 00:23:32 is there anything else i could be missing? Thank you in advance! Aug 14 00:24:15 there was a post on stack exchange but the two suggested fixes did not work Aug 14 00:26:18 and even after adding a status folder under dpkg i get: Reading package lists... Error!E: Read error - read (21: Is a directory)E: The package lists or status file could not be parsed or opened. **** ENDING LOGGING AT Wed Aug 14 02:59:56 2019