**** BEGIN LOGGING AT Wed Sep 18 02:59:58 2019 Sep 18 05:34:18 New news from stackoverflow: BitBake: Can we use DEPENDS with package name? Sep 18 07:51:26 LetoThe2nd: thanks again for the tips about initrd instead of initramfs yesterday. Managed to boot a kernel + initrd now :) It's quite difficult to google about initrd specifically, because it seems like initrd and initramfs is used interchangeably making me very confused. Sep 18 07:52:05 iceaway: :) Sep 18 07:52:33 iceaway: using an almost standard filesystem as initrd is just so much easier. Sep 18 07:53:37 Had some issues with the kernel panicing due to "unable to handle page request" before stripping down my initrd considerably. Sounds a bit like an out-of-RAM issue, but my board has 2GB of RAM and the initrd was about 60MB unpacked. Sep 18 07:53:45 iceaway: because *technically* it is a vastly different thing. Sep 18 07:55:08 maybe not exactly an out-of-ram issue, but some other limit applying. no idea, our update initrds usually fit into 16M or so. Sep 18 07:56:30 LetoThe2nd: Managed to get below 16MB, so hopefully I can keep it there :) Not sure if I should look into the fitimage thing or not. If I do not use the fitimage, I should be able to use the same kernel image for both the "rescue" mode and normal mode, just changing the bootargs. Sep 18 07:57:13 I guess if I use a fitimage, I would have to to use one "regular" image for the normal boot, and the fitimage for rescue mode booting. Sep 18 07:57:51 iceaway: technically thats true, but if you're talking about flash storage then thats a bad idea. the update mechanism should always be capable of exchanging the productive kernel without endangering itself of being non-functional if something goes wrong. Sep 18 07:59:04 LetoThe2nd: ahh yes, of course. did not consider that aspect yet. Sep 18 07:59:33 iceaway: :) Sep 18 09:48:17 hello :-) Sep 18 09:48:57 tlwoerner: actually, I'm just realizing: are you teaching at those e-ale conferences? I was reading the slides and saw your username there :) Sep 18 09:51:35 palate: the e-ale sessions usually are presented by a couple of folks who happen to be around and available at the main conference. and yes, tlwoerner sometimes does a session. (answering for him because he's probably not around yet, tz differences) Sep 18 09:52:29 LetoThe2nd: right... this e-ale session seems super cool! :-) Sep 18 09:52:57 palate: come to ELCE in lyon, late october, and meet everybody in person :) Sep 18 09:54:33 LetoThe2nd: I've seen the dates, honestly I was tempted. But I don't really know what ELCE is about, and I thought that maybe it's overkill to take a few days off and pay for a conference that I don't know (it's usually quite expensive, right?) Sep 18 09:54:59 LetoThe2nd: but I'm also tempted, because Lyon is not so far (I'm from Switzerland :D) Sep 18 09:55:04 palate: you'll certainly need a couple of bucks, yeah. early bird is over, flights, hotels... Sep 18 09:55:17 palate: make $EMPLOYER pay Sep 18 09:55:28 LetoThe2nd: I would need to convince them :D Sep 18 09:55:52 palate: how much of your work time are you spending on embedded linux and yocto, roughly? Sep 18 09:56:08 LetoThe2nd: none, I'm spending many hours in my free time :D Sep 18 09:56:23 palate: ok, that makes it rather complicated. :P Sep 18 09:56:44 I'm thinking that yocto may be the right solution for us at work, but I need to learn more before I can argue that Sep 18 09:57:14 i see Sep 18 09:57:29 so no embedded linux job at all so far? Sep 18 09:58:01 not for me, not really Sep 18 09:58:07 Well I'll still mention the conference Sep 18 09:58:25 then its probably rather hard to get company money, yes. Sep 18 09:58:45 palate: are you using a build system already? Sep 18 09:59:04 "where developers working on embedded Linux and industrial IoT products and deployments gather for education and collaboration," Sep 18 09:59:13 How much education is there? xD Sep 18 09:59:40 palate: well, a lot of talks are given in a few days, tat's the educational part I guess Sep 18 10:00:21 qschulz: but like, talking about e.g. how to secure an IoT device, add new BSPs and stuff like this? Sep 18 10:00:22 There's also the training days Sep 18 10:00:24 collaboration is rarely done during the talk days (a few minutes after a talk, some meetings between talks, during lunch or after the day's over Sep 18 10:00:31 palate: its co-located with the open sorce summit. bascially its roughly 15 simultaneous track to choose from, of which 3 (IIRC) are embedded linux, and probably another one or two other embedded related stuff Sep 18 10:00:55 palate: sure, many different topics: https://osseu19.sched.com/ Sep 18 10:01:05 open source summit? sounds good :) Sep 18 10:01:29 palate: the "collaboration" usually rather happens outside the tracks, in the hallways, booths, and ... pubs around! Sep 18 10:01:35 all I can say is that if you've a broad spectrum of interest and knowledge, you won't know how to chose between talks :D Sep 18 10:02:30 palate: https://events.linuxfoundation.org/events/embedded-linux-conference-europe-2019/attend/convince-your-boss/ Sep 18 10:03:05 palate: FWIW there is a hobbyist category Sep 18 10:04:27 palate: to be honest, it also can be rather overwhelming. its a great experience if you already know a couple of folks, or are involved in some project. if you just go there for the first time, theres also the chance that you are rather dazzled Sep 18 10:04:36 qschulz: awesome! Sep 18 10:04:58 New news from stackoverflow: Using OpenCV with camera on Jetson Nano running Yocto || What is difference between git:somthing.com and gitsm:something.com in .bb file in Yocto Sep 18 10:06:31 i personally remember my first ELCE as rather weird, and even the second one woudl deserve mixed reviews. but once you get the knack and know folks and your way around, its a perfect way for networking and getting updated on the latest hot s**tz in a very short time. Sep 18 10:07:30 (for the record, this year is my 9th in a row) Sep 18 10:09:39 :) Sep 18 10:09:43 LetoThe2nd: jealous! Sep 18 10:10:15 though stupid elce is basically scheduled at the same time as my son's birthday Sep 18 10:11:55 rburton: last year getting allowance was tricky. i returned from 3 months of paternal leave only like 5 weeks before ELCE. $BOSS was not exactly pleased. Sep 18 10:12:15 lol Sep 18 10:13:52 :-) Sep 18 10:32:18 haha so I read the session descriptions of ELCE. Now instead of having to learn about Yocto, I have Yocto, Deby, Isar, Buildroot xD Sep 18 10:36:35 Oh, there is something I don't understand. In Yocto, I can build for qemux86, qemuarm, etc. If I want to "emulate" my pocketbeagle, can I somehow do that with qemu? Sep 18 10:41:43 Has anyone used the kernel-fitimage class, and care to explain a bit on how to implement that in my build? It's hard to find examples of how to use it. I'm not sure where to put the required bitbake variables to make it output a fitimage. Sep 18 10:46:34 palate: yeah, there's patches in master now to configure qemu magically for beagle Sep 18 10:47:07 well the beaglebone-yocto BSP has the qemu bits setup. you can copy them into whatever bsp you're using assuming they're still right Sep 18 10:47:52 rburton: I see Sep 18 10:48:29 rburton: what is it that makes a BSP different from another? CPU arch and available devices (usb, ethernet, ...)? Sep 18 10:48:57 the question is a bit: if it works for beaglebone, does it work for pocketbeagle? Sep 18 10:51:44 iceaway: Actually you don't need to do much. I think it basically comes down to KERNEL_CLASSES = "kernel-fitimage", KERNEL_IMAGETYPES = "fitImage", KERNEL_DEVICETREE, UBOOT_LOADADDRESS, UBOOT_ENTRYPOINT. Sep 18 10:51:56 iceaway: And I have them in my machine config Sep 18 10:53:22 frsc: do you know if it is possible to create a .bbappend for a machine config? Sep 18 10:53:33 .bbappends append .bb files, as the name suggests Sep 18 10:53:48 make your own machine, and include the one you want to modify Sep 18 10:54:06 rburton: That was my guess as well. Thanks, will try that approach! Sep 18 10:54:27 I see that just after ELCE, there is a yocto summit of 2 days? Is that where the classes are? Sep 18 11:05:12 New news from stackoverflow: Using OpenCV with camera on Jetson Nano with Yocto/poky Sep 18 12:05:12 palate: yes, if you're looking to save money, i'd recommend attending the summit and skipping the conference!! lol Sep 18 12:05:58 are the presentations published online? I just realized that anyway I'm moving apartment on that week, so I won't make it :( Sep 18 12:06:16 and LetoThe2nd answered perfectly for me earlier. ironically i taught the e-ale session this past Spring at SCaLE on buildroot Sep 18 12:08:23 yes, many past presentations are available online: https://elinux.org/ELC_Presentations Sep 18 12:08:25 palate: It all goes on the YouTube'z Sep 18 12:08:59 i prefer the elinux.org site because it organizes the talks better, and provides links to the slides Sep 18 12:09:06 True Sep 18 12:09:51 it's a gold mine of information! Sep 18 12:10:39 I went to a couple of ELCEs but then I decided to stop going.. I'm not extrovert enough to be "networking" with people .. and I watch all the talks online anyway.. Sep 18 12:10:53 tlwoerner: is it also a fair trade and ethically run gold mine? Sep 18 12:11:00 jofr: tlwoerner: awesome! Sep 18 12:11:23 jofr: well i don't remember you being exactly an introvert, though. Sep 18 12:11:24 jofr: it does make sense, too Sep 18 12:11:30 LetoThe2nd: xD Sep 18 12:12:54 mouaha I found one for tonight: https://www.youtube.com/watch?v=wCVYQWFIvBs&feature=youtu.be Sep 18 12:13:01 (buildroot vs yocto) Sep 18 12:13:10 (though I'm already a big fan of yocto :) ) Sep 18 12:16:59 palate: lesson #1: being a fan does not pay the bills. use the tool that fits the job best. no more, no less. :) Sep 18 12:18:44 haha, got it :) Sep 18 12:19:32 seriously, thats one of the core things i emphasize in every training, every presentaiton. its essential to know your toolbox so you can make good choices. Sep 18 12:20:58 LetoThe2nd: Haha, well I'm a complicated beast. ;) I'm not introverted when I'm already a part of the conversation. But striking up a conversation with a stranger isn't exactly what I'm known for :p Sep 18 12:21:36 LetoThe2nd: In that sense, I guess I'm just a bit "scandinavic" I guess ;) Sep 18 12:21:54 JPEW: I found a couple of tweaks needed to the patches, they're in -next Sep 18 12:22:08 JPEW: also, rburton opened a bug to do with error handling Sep 18 12:22:36 jofr: sweden? Sep 18 12:22:54 jofr: Icelandic, but I've lived in Denmark for the past 4 years. Sep 18 12:23:52 Iceland? Amazing! I went there for 10 days a few years ago, I looooved it Sep 18 12:24:25 what's funny about the conferences is: the Yocto Project will (almost always) have a booth. at some point, everyone involved with the project (and new people too) will drop by the booth. people will get chatting about what they're working on, then someone will inevitably say "oh, are you palate from IRC?" and you'll say "yes i am!" and then the other person will say "hi, i'm LetoThe2nd! we're all going to pub, why don't you join Sep 18 12:24:25 us?" Sep 18 12:25:02 :) Sep 18 12:25:08 RP: just checking in to see if there were any issues with the toolchain testing, I know there are some minor issues with the results. Wanted to make sure though that i had not missed anything major with the AB stuff Sep 18 12:25:12 tlwoerner: don't you know the story between me and jofr ? Sep 18 12:25:25 "Are you the annoying new guy asking questions on #yocto?" -> I would expect that :D Sep 18 12:25:58 palate: Well, then half the people here would be the annoying new guy .. myself included. ;) Sep 18 12:26:38 LetoThe2nd: no i don't think so, and i'm not sure who jofr is (?) Sep 18 12:26:39 nrossi: Congratulating you/us on having it mostly working was on my todo list for today! I think that one is roughly where we need it now, its great! Sep 18 12:26:50 nrossi: 1.9 million test results in that build :) Sep 18 12:27:46 RP: only 1.9 million? :) Sep 18 12:27:59 nrossi: obviously things remain but its basically good Sep 18 12:28:56 tlwoerner: when heading to prague, i told him "if you see bearded guy wearing metal clothes and cowboy boots somewhere near the yocto booth, then come say hi. its me". and when i was standing there gathering the usual troll crows for lunch, some guy steps up to me and says "Hey you told me I shall say hello. I'm jofr, and you are letothe2nd." Sep 18 12:29:29 nice! perfect ELCE story :-) Sep 18 12:30:08 thats how it works. Sep 18 12:30:15 Then we went to lunch later, with people I have no recollection of.. (sorry if you're here :p ) Sep 18 12:30:55 jofr: one or two of them might be around, but i'm not totally sure who was with us either. Sep 18 12:31:20 palate: also, a good way to get into the conference is to give a talk there, then you don't pay the ticket fee :) Sep 18 12:32:50 LetoThe2nd: so cool :D Sep 18 12:33:03 qschulz: yes, but I would need to have interesting things to say :D Sep 18 12:34:00 palate: sometimes you're doing things that you think are trivial or unworthy of other people's attnetion and it turns out it actually is very interesting and people want that kind of talks :) Sep 18 12:34:45 the perspective of people new to the project is very important to the project! Sep 18 12:35:26 New news from stackoverflow: How to use OpenCV with camera on Jetson Nano with Yocto/poky Sep 18 12:36:45 to most of us, how to do various things is "obvious", but sometimes having someone say "i wanted this, did that, but something else happened" is interesting. to us it might be obvious why you got something else :-) Sep 18 12:37:22 tlwoerner: "i wanted to be a rock star, so i bought a guitar and booze, but i only got drunk" Sep 18 12:40:50 xD Sep 18 12:45:02 LetoThe2nd: haha, i actually lol'ed Sep 18 12:45:33 tlwoerner: :) Sep 18 12:47:04 so busybox comes with udhcpd already. Meaning that I can quite easily setup a dhcp server, and thanks to the dropbear feature, now I can SSH into my pocketbeagle from my computer over usb \o/ Sep 18 12:49:09 next step is to automate all that, and then to add that to my yocto setup so the image works out of the box :) Sep 18 12:49:15 palate: https://www.youtube.com/watch?v=roRQ2mNwMMQ Sep 18 12:50:08 :) Sep 18 13:09:24 LetoThe2nd: metal clothes as in black tees with skulls and fire on, or as in medieval armour? Sep 18 13:10:24 rburton: the former. the latter usually just gives me headaches at the airport. Sep 18 13:18:14 rburton: you didn't think tinfoil hat? :) Sep 18 13:19:18 RP: I think you might want to chdir back to the starting directory after starting the hash server also? Sep 18 13:19:44 RP: Also, I noticed a lot of AB failures due to do_locked_sigs? Sep 18 13:20:21 JPEW: is that call site serialised? I worked on the assumption it was in its own process so doesn't matter in reality Sep 18 13:20:34 JPEW: patch in -next for it Sep 18 13:20:48 * alessioigor waves all Sep 18 13:20:51 Is there a way to make images of different size of the same (image) recipe? Sep 18 13:21:04 RP: Ah I see that patch now. Sep 18 13:21:21 JPEW: I need to clean up the commit message. Its testing atm Sep 18 13:21:33 JPEW: I am worried though as sstate reuse isn't happening when it should Sep 18 13:21:38 JPEW: I don't know why Sep 18 13:22:12 JPEW: compare and contrast https://autobuilder.yoctoproject.org/typhoon/#/builders/59/builds/1046 and https://autobuilder.yoctoproject.org/typhoon/#/builders/59/builds/1048 Sep 18 13:22:29 difference is that locked sigs task change and the bitbake hashserv tweaks Sep 18 13:22:37 so it shouldn't be rebuilding Sep 18 13:27:48 JPEW: selftest is throwing errors with hashequiv set :/ Sep 18 13:28:09 RP: In the hashequiv tests, or elsewhere? Sep 18 13:29:38 JPEW: elsewhere Sep 18 13:41:38 JPEW: good news is no performance issue that I can spot so I think we're good on that front Sep 18 13:41:59 JPEW: on that basis I think I merge the patches, build M3 and we move on to bug fixing Sep 18 13:50:31 Where is the recipe that builds python in yocto? Sep 18 13:50:48 I'm looking for the recipe that provides the file /usr/bin/python3 Sep 18 13:54:52 the recipe is python3 in oe-core Sep 18 13:59:54 rburton, how useful is sysprof? Sep 18 14:00:15 rburton, the new version makes polkit non-optional, so I am deliberating between moving sysprof to meta-oe vs moving polkit to oe-core Sep 18 14:09:18 JPEW: something is really wrong. Its not reusing sstate on a build I'm experimenting with on the autobuilder Sep 18 14:10:50 thanks RP Sep 18 14:10:54 thanks rburton Sep 18 14:11:55 ecdhe: oe-pkgdata-util (or whatsitcalledexactly) lets you directly search for the package and therefore recipe that provides a given path file in your build Sep 18 14:14:17 Thanks LetoThe2nd, really appreciate it! Sep 18 14:15:42 ecdhe: have fun! Sep 18 14:29:59 RP: I've been trying to do some fixing on the strace issue myself. I can see the same sort of runtime in 5.3 as well. so before I go to the list, I'm making sure I've tried a few things. Sep 18 14:32:21 zeddii: ok. We're going to need to do something for M4 but I can use the timeout workaround to get M3 built Sep 18 14:36:30 kanavin: haven't tried it for a long time Sep 18 14:41:17 khem: meta-oe down to two failures: https://autobuilder.yoctoproject.org/typhoon/#/builders/88/builds/5 Sep 18 14:48:56 rburton, yeah, I wonder if there are users at all Sep 18 14:52:59 RP: Hmm, weird. Do you think that is because of the server? I wouldn't expect it to be since the information it is reporting hasn't changed. Sep 18 14:54:38 Architecture question: I have a single board for which I'm writing two different applications, each of which may include some additional custom utilities. I thought about making a layer for each application, but then I realized there is some common configuration between them. Sep 18 14:54:54 JPEW: I did some digging, I'm wondering if its its "ripple" effect from non reproducibility :/ Sep 18 14:55:18 JPEW: -Hash for dependent task /home/pokybuild/yocto-worker/qemux86/build/meta/recipes-devtools/libtool/libtool-cross_2.4.6.bb:do_populate_sysroot is 1ea317b018c3e4809c46581de8044a13a607ae87fe1ee6ea972a4e731a9ec7f8 Sep 18 14:55:18 +Hash for dependent task /home/pokybuild/yocto-worker/qemux86/build/meta/recipes-devtools/libtool/libtool-cross_2.4.6.bb:do_populate_sysroot is bb5eeab5a12213c63d056b449a8d1dfef455b94ef16f03181572118936da07c5 Sep 18 14:55:18 -Hash for dependent task /home/pokybuild/yocto-worker/qemux86/build/meta/recipes-devtools/python/python3_3.7.4.bb:do_populate_sysroot is c07a38e24cb582a6976cbe3d6d9b96ffd91bca5e311b19b9df903eb74ec99f15 Sep 18 14:55:18 +Hash for dependent task /home/pokybuild/yocto-worker/qemux86/build/meta/recipes-devtools/python/python3_3.7.4.bb:do_populate_sysroot is 49fcc8535e00073cfe2a10b0cb9e281857ece956b6e17ab09f098cd8495ee608 Sep 18 14:55:49 RP: Quite possibly. If you can get it, the build output the signature files that is hashed to get the output hash in ${T} Sep 18 14:56:00 JPEW: but even if it wasn't reproducible, it still should have generated valid sstate :/ Sep 18 14:58:14 JPEW: I don't understand :( Sep 18 14:59:25 RP: Agreed. Whats that diff from? Sep 18 14:59:50 So then I thought perhaps I should make a single layer, meta-myapps and expose the different collections of packages as seperate recipes... Sep 18 15:01:39 JPEW: its from the output of dumpsigs on btrfs-tools:do_prepare_sysroot in two different builds that should be the same (same sstate, same hashequiv server) Sep 18 15:02:03 JPEW: its rebuilding btrfs-tools since its saying python3 and libtool cross changed Sep 18 15:03:53 JPEW: more alarmingly, ls tmp/stamps/core2-32-poky-linux/libtool-cross/ Sep 18 15:03:53 2.4.6-r0.do_populate_sysroot_setscene.1ea317b018c3e4809c46581de8044a13a607ae87fe1ee6ea972a4e731a9ec7f8 Sep 18 15:03:54 2.4.6-r0.do_populate_sysroot.sigdata.bb5eeab5a12213c63d056b449a8d1dfef455b94ef16f03181572118936da07c5 Sep 18 15:04:13 JPEW: correct hash would be bb5eeab for both Sep 18 15:04:54 JPEW: that is a smoking gun of some kind Sep 18 15:07:33 JPEW: just gets worse as I look. The build shows do_populate_sysroot (56090): log.do_populate_sysroot.56090 Sep 18 15:07:33 then later do_populate_sysroot_setscene (35314): log.do_populate_sysroot_setscene.35314 Sep 18 15:07:42 definitely shouldn't do that Sep 18 15:08:05 JPEW: runqueue problem though, not the server Sep 18 15:11:09 * armpit hmm wonder who is setting patches to suspended with out tell me Sep 18 15:11:45 armpit: i noticed something/someone is setting oe-core patch status to master-next etc Sep 18 15:14:27 I tend to set the ones for stable to "under review" and my name Sep 18 15:57:55 RP: I filed this https://bugzilla.yoctoproject.org/show_bug.cgi?id=13536 which should fix dfu-util-native Sep 18 15:57:56 Bug 13536: normal, Undecided, ---, richard.purdie, NEW , Add glibc-devel-static to prerequisites on OpenSUSE based host distributions Sep 18 15:58:18 IMO its a gap in prerequisites for SuSE based distros Sep 18 15:58:34 that will get us down to 1 Sep 18 16:02:49 and the last failure is kernel-selftest, kernel-selftest needs clang to be installed on build host or alternatively use meta-clang, its usually easier to install clang as prerequisite and then add clang llc to HOSTTOOLS Sep 18 16:02:59 khem: I do not want to add this, I'd much rather we fix that recipe Sep 18 16:03:18 khem: and no, we're not installing clang Sep 18 16:03:31 khem: worst case we'll have to blacklist these in poky I guess Sep 18 16:03:37 RP: problem is not recipe I hope you go tthat Sep 18 16:03:53 khem: the problem is that patch which enabled static linking? Sep 18 16:04:09 yes its not problem Sep 18 16:04:48 problem is that prerequisites that we have for redhat and debian like distros ( glibc-devel ) already install libc.a Sep 18 16:04:52 but suse does not Sep 18 16:05:08 khem: and nothing else in the system seems to need this Sep 18 16:05:49 sure, for the subset that is being built Sep 18 16:06:09 I'm open to a case for adding things but we don't just do it arbitrarily. One recipe with a dubious patch isn't a case for adding a dependency Sep 18 16:06:16 what I am saying is this is just a symptom that can happen with other packages too Sep 18 16:06:36 so either we dont require libc.a on everyhost then its fine too Sep 18 16:06:51 khem: sure. We can fail because of all kinds of missing dependencies Sep 18 16:07:01 khem: right now we don't rely on libc.a as far as I'm concerned Sep 18 16:07:20 khem: has anyone asked JaMa if that patch has specific history? Sep 18 16:08:15 yes hee mentioned that he does not use it, but it was tool that was shipped to users so statically linking it made it host independent Sep 18 16:09:26 I am actually inclimed to fix it since it was a revert of upstream patch and probably pass -static via CFLAGS/LDFLAGS but still static binary made sense Sep 18 16:10:48 khem: we now have uninative and sdk relocation so presumably those obsolete this then Sep 18 16:20:28 with uninative it might be since the tool is installed in deploydir Sep 18 16:24:48 RP: kernel-selftest I think can be blacklisted, although it will be a good addition for test matrix eventually Sep 18 16:25:19 khem: right, I don't really want to blacklist anything :( Sep 18 16:26:32 kernel-selftest uses bpf and you need clang for that backend Sep 18 16:28:46 either add clang to prereqs or add meta-clang to layermix are only two other options besides blacklisting, pick whatever you wish Sep 18 16:28:48 khem: right, we don't want to install clang on the workers and add that as a dependency Sep 18 16:29:18 khem: the recipe should probably disable itself it clang isn't present somehow Sep 18 16:29:29 hmm Sep 18 16:29:47 do we have some helper functions in metadata to detect host tools ? Sep 18 16:29:57 I would certainly do that Sep 18 16:30:26 khem: I'm not aware of any. Just talking to rbuton about meta-intel's need for git-lfs :/ Sep 18 16:30:38 we do have an open feature request for recipe specific host tools Sep 18 16:30:46 that in itself shouldn't be hard Sep 18 16:42:18 RP: yikes, as a user of meta-intel, that sounds unappealing Sep 18 16:42:30 RP: I forwarded you email, dfu-util is shipped for end-users so they can flash devices, and it need libusb1 so I guess easiest is/was to built a static binary and ship it for minimal host deps Sep 18 16:44:38 khem: right, I understand the history and why. I'm not adding extra build dependencies to the autobuilders without good reason though and this doesn't make the bar IMO Sep 18 16:47:12 probably its best to add these both recipes to BLACKLIST for poky then Sep 18 16:47:29 this makes the best for everyone I guess Sep 18 16:51:36 khem: its that or we make it a PACKAGECONFIG Sep 18 16:51:50 khem: poky can then config it off Sep 18 16:52:32 yeah packageconfig is better I guess for dfu-utils Sep 18 16:52:44 infact I would like to keep static disabled by default Sep 18 16:53:41 not sure for kernel-selftest Sep 18 16:54:21 this recipe has been a pain and submitter has dropped off chart after sending the first patch to add it Sep 18 16:55:43 khem: I remember the discussion about it for core. I'd really like it there but not with large dependencies Sep 18 16:56:01 JPEW: I fixed the chdir thing and merged it Sep 18 16:58:20 RP: OK, looks good Sep 18 17:00:23 RP: When I original did the hash equiv work, I used persist_data as the storage mechanism. IIRC the reason for this was that there were places where the unihash was needed in both the parent bitbake runqueue parent process and the child workers; you replaced that with the new message to set the unihash in the parent process, but maybe it's racing? Sep 18 17:03:36 JPEW: right, I wondered. Its actually a problem I've already kind of worked on in that horrible WIP patch I had in -next Sep 18 17:03:47 JPEW: I need to get back to that and figure out how to test it Sep 18 17:04:01 JPEW: at least that is what I suspect Sep 18 17:04:27 RP: Is that patch still sitting around somewhere? Sep 18 17:05:06 JPEW: I've kind of misplaced it :/ Sep 18 17:07:01 RP: Remember any part of the title I can grep for? It might still be dangling in my git tree Sep 18 17:08:41 JPEW: I think its the one I just managed back into -next Sep 18 17:08:51 JPEW: need to go afk but I've thrown it at the autobuilder Sep 18 17:09:17 just *mangled* was what I was trying to say Sep 18 17:09:53 RP: OK, sounds good Sep 18 17:10:17 the bitbake selftest may be bust as that part of the patch failed to apply Sep 18 17:10:28 * RP shrugs Sep 18 17:11:00 Is this correct? PREFERRED_VERSION_systemd = “243” http://layers.openembedded.org/layerindex/recipe/5979/ Sep 18 17:12:31 bentech `PREFERRED_VERSION_pn-systemd = "243"` I think Sep 18 17:13:09 Ah thanks Sep 18 17:13:22 I wouldn't have thought you'd need to set it? Sep 18 17:14:08 I need this feature https://github.com/systemd/systemd/pull/12429 Sep 18 17:14:49 I gather you've copied the new recipe back to an older branch? Sep 18 17:15:29 bentech: Ah, that might not work either. The PV has the git SHA1 in it, which you might also need (I don't recall ATM). Sep 18 17:16:44 I think just PREFERRED_VERSION_systemd = “243%" would work Sep 18 17:17:50 Ah, ya. I always forget about the wildcards Sep 18 17:31:09 hmm, i’m not having much luck.. There are no warnings it just hasn’t changed Sep 18 17:31:33 i’m using bitbake -s mybuild to check Sep 18 17:33:08 maybe i need the +X in the version Sep 18 17:38:25 bentech: The suggestion from smurry to use the wildcard is probably better Sep 18 17:41:55 bentech: And I was wrong on needing the "pn-" in the override. Sorry. Sep 18 17:42:45 It seems to work with it Sep 18 17:43:03 My issue was that *cough* i hadn’t updated the poky folder Sep 18 17:44:19 so the 243 reciepe wasn’t there. It was commited today apparently Sep 18 17:45:15 bentech: Ah, that would do it. Sep 18 17:45:46 Can i use depends instead of preferred so it might give a warning? Sep 18 17:47:44 Strange coincidence that the code i needed was commit 6 hours ago Sep 18 17:47:50 bentech: Not sure I follow Sep 18 17:48:25 If i add PREFERRED_VERSION_pn-systemd = “500000” i don’t get a build warning Sep 18 17:48:48 PREFERRED_VERSION doesn't take a pn- Sep 18 17:49:37 bentech: Ya, my mistake. I was confusing it with e.g. PACKAGECONFIG. Sorry :( Sep 18 17:53:00 bentech: if you just pulled in poky master and got the updated recipe, you shouldn't need to set PREFERRED_VERSION Sep 18 17:54:27 I am currently using Thud, i’ve just downloaded those files from master Sep 18 17:54:36 It should all get sorted when i upgrade to Zeus Sep 18 17:56:59 bentech: okay Sep 18 18:07:08 That was optimistic :( Sep 18 18:13:18 How do you require a glibc-utils for use during the build? I’m trying to use getent which is on the host system Sep 18 18:24:26 hi. does anyone have experience with ocaml in yocto? Sep 18 18:28:17 beneth: needing getent is a bit unusual. $PATH is sanitised heavily, either use an absolute path (hack) or HOSTTOOLS+=getent in local.conf Sep 18 18:36:33 beneth: if its the new kernel then current yocto exports KBUILD_BUILD_USER and _HOST to remove host details leaking into the kernel Sep 18 18:40:39 I'm struggling with a second bug which is again some files that appear and shouldn't or don't appear and should in sigdata for a task. Sep 18 18:40:55 I'm trying to understand how is this list created so I can better debug it Sep 18 18:41:04 http://git.yoctoproject.org/cgit.cgi/poky/tree/bitbake/lib/bb/siggen.py#n893 Sep 18 18:41:34 I've dug up for a bit but it leads me to the cache and then it's the big pandora box Sep 18 18:42:43 I'd like to learn how to debug this part as I'm very new in this area of Yocto Sep 18 18:42:55 so any pointer is appreciated :) Sep 18 18:52:54 hey all, I have a custom distro conf with a few variables defined that are used in various tasks in multiple recipes, I'd like to invalidate/trigger the build step on all these recipes whenever the var changes...any thoughs Sep 18 18:53:49 checksums are done on a per-task, not per-recipe, basis. you could just add them to vardeps of do_compile or something Sep 18 18:54:55 vardeps...forgot about that...thanks Sep 18 19:08:30 Is the archiver.bbclass code tested on the AB? Sep 18 19:26:47 JPEW, there is a archiver selftest so I would think so Sep 18 19:27:13 see https://git.openembedded.org/openembedded-core/tree/meta/lib/oeqa/selftest/cases/archiver.py Sep 18 19:34:35 RP: are we building master-next of meta-openembedded on AB Sep 18 19:36:32 RP: if so then I have queued a fix for dfu-utils there Sep 18 19:55:09 khem: whoa, poky is now building with meta-oe on AB? nice! Sep 18 20:17:35 Khem we can make it build any branch we when from any repo.. ie contrib Sep 18 20:18:02 * armpit smacks himself.. Sep 18 20:18:29 but its poky repo format then meta-oe Sep 18 20:19:54 khem, do you think the kernel-selftest should be a dynamic-layer thing for clang? Sep 18 20:20:18 that is the other failure Sep 18 20:20:59 khem the default it master btw Sep 18 20:31:45 khem, there are two a-full build currently running. do you want to kick off a meta-oe build? Sep 18 20:31:53 later? Sep 18 21:32:52 hm, are things from DEPENDS supposed to appear in the $PATH in devshell? Sep 18 21:37:38 mischief: yes Sep 18 21:44:36 i see, i forgot a ${D} :) Sep 18 21:51:08 armpit: dynamic-layers might be a good idea, although its not a hard dep Sep 18 21:51:24 so infact someone can install clang on build host and use that too Sep 18 21:51:38 or maybe remove it from world build? Sep 18 21:51:39 but its okay to mark it dynamic layers Sep 18 21:51:51 that way we remove dep on clang being on host Sep 18 21:52:18 yes we can add it BLACKLIST dynamically on AB Sep 18 21:52:39 yep Sep 18 21:53:32 armpit: can you help adding that Sep 18 21:53:51 I think so. let me check Sep 18 21:53:54 and can we pull master-next branch from openembeeded-core repo Sep 18 21:54:14 yes, we can Sep 18 21:54:36 even from any repo.. its what i do for the stable branches Sep 18 21:54:45 https://git.openembedded.org/meta-openembedded/log/?h=master-next Sep 18 21:54:50 is what we need to test Sep 18 21:56:22 armpit:see https://github.com/kraj/jenkins-jobs/blob/master/jenkins-job.sh#L376 Sep 18 21:56:30 k Sep 18 21:56:38 we can do something similar for kernel-selftests Sep 18 21:56:51 the issue is I don't have write perms for the AB sources Sep 18 21:56:58 hmm Sep 18 21:57:13 who does Sep 18 21:57:24 RP Sep 18 21:57:43 maybe MH Sep 18 21:58:22 let me think about.. there may be another way Sep 18 21:58:43 Or I can add poky override to recipe e.g. COMPATIBLE_HOSTS_poky = "null" Sep 18 21:59:02 ideally meta-poky should do it via a bbappend Sep 18 21:59:45 but this means the folks who use poky+meta-clang will have to do something to unblacklist it Sep 18 22:01:50 best is to blacklist it on AB since that solves the problem in a good way Sep 18 22:10:05 armpit:can you change branch ? Sep 18 22:10:29 yes Sep 18 22:10:42 ok then can you change one job to do master-next ? Sep 18 22:11:00 yes Sep 18 22:11:51 want me to use poky master with meta-oe master-next ? Sep 18 22:18:35 armpit: OK I Pushed a potential change into master-next, actually use master-next of poky too Sep 18 22:18:47 build started Sep 18 22:18:48 master-next of everything Sep 18 22:18:58 oh cool so you seem to be sitting on crank :) Sep 18 22:19:12 link ? Sep 18 22:19:47 ah build #6 Sep 18 22:21:59 you can also track it via build log https://wiki.yoctoproject.org/wiki/BuildLog#meta-oe_57928_-_master_YP_CHASH_on_Typhoon-Cluster Sep 18 22:22:13 its what swat looks at Sep 18 22:52:06 thanks Sep 18 22:58:00 new build Sep 18 22:58:04 https://wiki.yoctoproject.org/wiki/BuildLog#meta-oe_57928_-_master_YP_CHASH_on_Typhoon-Cluster Sep 18 22:58:40 armpit: “Typhoon-Cluster” sounds ominous ;) Sep 18 23:11:13 hm, how can i add a custom 'host tool'? Sep 18 23:20:50 mischief: add it to HOSTTOOLS, I believe Sep 18 23:22:36 neat, thanks Sep 18 23:22:44 i'm cheating a bit since its too hard to build Sep 18 23:27:01 armpit, khem: back now, still need me? Sep 18 23:29:07 JPEW: still a load of odd task hash mismatch failures using hash equiv, there is some kind of regression as this was working :/ Sep 18 23:45:30 RP: we sorted it out Sep 18 23:46:34 RP: armpit helped me with starting meta-oe job see https://autobuilder.yoctoproject.org/typhoon/#/builders/88 Sep 18 23:46:55 RP: my user id seems to be not able to have the perms to start jobs Sep 18 23:47:06 I sent this to halfhalo Sep 18 23:47:15 ups halstead I meant Sep 18 23:48:31 khem. Let me see what's up. Sep 18 23:49:03 halstead: I did restart the controller btw so that is done now Sep 18 23:49:18 halstead: I also added the ability to select the worker a job runs on Sep 18 23:49:40 its slightly broken at the moment as '*' isn't a valid value, will tweak tomorrow Sep 18 23:49:48 RP very cool. Sep 18 23:50:07 halstead: long requested, everything aligned to make it work Sep 18 23:50:34 RP, oh I watched that build until I fell asleep and forgot to follow up. Thank you for restarting. Sep 18 23:51:07 halstead: it was still going this morning for me :/ Sep 18 23:52:51 * RP -> Zzzz Sep 18 23:56:08 Goodnight. Sep 19 00:02:19 khem, I've added your account in the right spot. Just started a build with your creds https://autobuilder.yoctoproject.org/typhoon/#/builders/15/builds/1295 Sep 19 01:02:38 halstead cool thx Sep 19 01:02:46 :) Sep 19 02:17:08 halstead: I submitted https://autobuilder.yoctoproject.org/typhoon/#/buildrequests/58474 but it seems its incomplete ? did I miss anything ? Sep 19 02:18:09 it seems the builder I requested is busy np Sep 19 02:18:58 It takes a moment. Especially with builds in progress **** ENDING LOGGING AT Thu Sep 19 03:01:42 2019