**** BEGIN LOGGING AT Wed Jan 06 02:59:58 2021 Jan 06 03:06:52 asenm: neither of these are build-related issues. i'd suggested contacting the mtd developers regarding the first issue, and your board manufacturer about the watch-dog issue Jan 06 03:07:35 tlwoerner both of them are same company and they refuse to help us Jan 06 03:09:11 asenm: i don't think the mtd developers are board manufacturers. get in touch with the people on OFTC:#mtd about the mtd issue Jan 06 03:09:50 asenm: you can look through the kernel source code to figure out where the watchdog message is being generated, it might help figure out the problem. does the board actually end up rebooting? Jan 06 03:11:11 tlwoerner no its not rebooting but i think this is for the gps/4g modem that is rebooting Jan 06 03:12:41 tlwoerner there are 2 companies together (same shareholders (wahatever) ) one of them is producing the hardware the other one is producing the software and both of them refuse to help us we have 15 000 devices that have tons of problems Jan 06 03:13:03 so any advice apart from talk to the people who did that would be great Jan 06 03:14:16 asenm: that sounds terrible Jan 06 03:14:39 asenm: this channel (#yocto) is on the freenode server. there are many other IRC servers, one of them is OFTC Jan 06 03:14:54 asenm: on OFTC you'll find the #mtd channel, which is where the linux kernel mtd developers hangout Jan 06 03:15:10 asenm: you'll have better luck asking about your mtd issue on that channel, than this one Jan 06 03:15:26 thanks mate will try to figure it out Jan 06 03:16:01 asenm: btw, using a FAT fs on flash doesn't sound like a good strategy Jan 06 03:16:30 i'm not a flash expert, but (if i remember correctly) FAT insists on using the same sectors over and over again, leading to premature failure Jan 06 03:16:50 the fact that this runs on yocto project doesn't help me at all however its kernel version 3.14 and having no internet connection doesn't help my debugging at all :) Jan 06 03:17:00 for things like the boot sector and file directory Jan 06 03:17:45 asenm: better choices are UBI/UBIFS and FSFS, but again this is just a guess Jan 06 03:18:15 yeah however we can't really change that since I can't even figure out what is on this chip ... Jan 06 03:19:57 if my understanding is correct, using a non-flash-friendly fs on a flash device is a severe design decision failure, and that's bad news for your 15,000 devices Jan 06 03:20:42 asenm: but i realize that's not your fault Jan 06 03:23:00 yet i'm the one who gets the blame when devices fail :( Jan 06 03:24:02 can you help me with other issue? I have to understand why the 4g modem stops working randomly ? and starts working randomly at some point? Jan 06 03:36:09 tlwoerner btw i have mmc card as well and this mtd error are not related right? Jan 06 03:51:56 asenm: mmc cards are managed, so they don't use mtd Jan 06 03:53:02 asenm: i don't have much experience with 4g modems. presumably it's an off-the-shelf part that's been added to your design? say it's a hardware issue and you're "just" a software person (lol) Jan 06 03:54:41 tlwoerner so any idea how to check whats on that MTD and other mtds? Jan 06 03:55:03 The script that cause this error doesn't have anything that writes to mtd? how even this error is triggered? Jan 06 03:57:17 asenm: your device has a (probably) small "disk" somewhere for storing things like configurations and such? maybe NAND or SPI-NOR? Jan 06 03:57:39 i guess Jan 06 03:58:29 according to lsblk there are few mtd devices Jan 06 03:58:30 https://pastebin.com/gQiJ9P73 Jan 06 04:01:06 obu? you work for kapsch or using their devivces? Jan 06 04:03:25 nah not really obu is called onboard unit so plenty of vehicles have it Jan 06 04:04:51 * tlwoerner worked for markiv/kaspch a while back Jan 06 04:13:21 tlwoerner never heard of them :) Jan 06 06:28:59 been a long time since i've used irc Jan 06 06:31:43 hey everyone! I'm brand new-ish to Yocto and I'm working on building a recipe for openauto for a go kart I'm building, here's my repo https://github.com/eromanowski/clickityos Jan 06 06:32:37 I got open auto running on a PI, but I thought this would be a cool project to learn Yocto Jan 06 06:45:13 EricWalter: good for your, thank you for sharing :) Jan 06 06:57:05 seeing this error when trying to run my build Jan 06 06:57:07 fatal error: QBluetoothServiceInfo: No such file or directory Jan 06 06:57:07 | 21 | #include Jan 06 06:57:32 doesn't look like the headers path is getting passed in, is there a way to hack that in? Jan 06 06:57:40 or what would be the proper way to do that? Jan 06 07:07:54 should I patch the make file to include CXXFLAGS and add CXXFLAGS_prepend = "-I${S} " to my bb file? Jan 06 07:34:49 EricWalter: make sure SW component build scripts use the cross compile build flags from bitbake, and only append to them if needed. Otherwise compilation will fail or binaries will not work on target. So yes, CXXFLAGS etc from bitbake must be used. Jan 06 09:46:03 kanavin: https://autobuilder.yoctoproject.org/typhoon/#/builders/83/builds/1728 so far gio postinst failure and mingw packaging issue Jan 06 10:08:58 hi Jan 06 10:10:05 when building with yocto, does it automatically add all the device drivers available or do we have to create our own layer and then create a recipe? Jan 06 10:53:48 is there any iptables integration in yocto, or do I have to provide my own systemd service to set it etc? Jan 06 11:22:34 kanavin: and ERROR: Cannot use 'hosttools/python3', Python >= 3.6 is required Jan 06 11:22:46 so the new qemu has a python 3.6 min version requirement Jan 06 11:22:58 * RP knows what rburton will say Jan 06 11:23:13 lol Jan 06 11:24:03 * RP remembers the days py 2.4 was controversial Jan 06 13:21:24 RP: we can insert python3-native there? Jan 06 13:29:07 kanavin: probably a bad idea and not really needed :/ Jan 06 13:32:37 kanavin: builds are mostly finishing, it looks better than I expected to be honest Jan 06 13:49:37 RP: do you have enough pain points to raise a ticket with qemu upstream? Jan 06 13:54:44 Hi, quick question, I'm trying to make a recipe that simply copies a textfile into the /home/root directory, but I'm getting a "No such file or directory error" on "cp ${WORKDIR}/hello-world.txt ${D}/home/root/hello-world.txt"... Jan 06 13:54:54 Could you tell me what's wrong with the command? Jan 06 13:58:25 ${D}/home/root doesn't exist yet Jan 06 13:58:26 mkdir it first Jan 06 13:58:37 ${D} is initially empty in do_install Jan 06 14:02:42 rburton: Isn't ${D} the image destination? Shouldn't it be populated at the time of installing packages? Jan 06 14:03:01 but do_install is when you generate the package, not build the image Jan 06 14:03:18 images are just extracting packages, packages are created in isolation Jan 06 14:04:34 ahh, I guess I still don't understand basic terminology a week from starting with yocto Jan 06 14:05:08 recipes generate packages, packages are unpacked into images Jan 06 14:06:43 So it probably isn't good practice to manipulate /home/root from inside recipes, correct? Jan 06 14:07:37 since that gets created when the image is made Jan 06 14:09:16 you can put a file in a package to drop files into there,that's fine Jan 06 14:12:47 get it, thank you. Jan 06 14:30:23 Ok, one more quick one; where do I specify "FILES_${PN} += ..." It doesn't work, when I put it in the hello-world recipe.. Jan 06 14:31:01 I'm trying to fix "hello-world: Files/directories were installed Jan 06 14:31:03 but not shipped in any package Jan 06 14:32:52 don't put ${D} in the value Jan 06 14:37:57 rburton: 2worked! thank you! Jan 06 14:38:14 kanavin: I suspect I've more work to do before I can do that Jan 06 14:38:52 Hello, I'm trying to build a userspace application that uses libnl3. However, during compilation I get "fatal error: netlink-private/object-api.h: No such file or directory". I've checked the working directory and in fact that appears to be missing. I can only see "recipe-sysroot/usr/include/libnl3/netlink/object-api.h", in fact the whole netlink-private folder is missing. Could you steer me in the right direction to look for the problem? Jan 06 14:40:16 justas: I know nothing about nl3 but the "private" sounds like internal headers it shouldn't be pulling in if used correctly externally Jan 06 14:47:04 I need to do some work with netlink sockets, and for example struct nl_sock is defined in "netlink-private/types.h", the non-private version is very different. Jan 06 14:53:33 Well it looks like you're right. Checked the source code of libnl and all private headers are set to noinst_HEADERS. Thank you :) Jan 06 15:13:13 hi, how to avoid that custom python libs executed in "python do_install" create caches to path of the python library, e.g. meta-bla/libs/foo.py creates mata-bla/libs/__pycache__/ directory into the git tree after running the tasks? "inherit python3" in the bbclass isn't enough. Jan 06 15:15:29 oh, poky tree is also full of __pycache__ directories after build. this is ugly. Jan 06 15:17:17 you can't really stop python doing that, as it runs bytecode not the scripts Jan 06 15:47:38 Hello to everyone, could someone suggest me the EASIEST way to get a graphical build?? vnc, X11, framebuffer?? Jan 06 15:54:10 Did somebody get experience on graphical build with yocto? Jan 06 15:54:38 warlock: it depends on your hardware Jan 06 15:56:47 Jpew you are right. My build is just for test and the platform is qemux86 Jan 06 15:59:39 Jpew, I tried with core-minimal-xvnc, but the problem is the virtual network device, it doesn't function, probably I should use KVM but the build is already inside a LXC Jan 06 16:00:11 warlock: Ah.... not sure I know then. The only qemu graphical builds I've ever done are KVM + weston Jan 06 16:01:34 Jpew, probably the best answer, but I cannot use KVM from the LXC inside Jan 06 16:01:39 :-( Jan 06 16:02:01 warlock: Ya.... I would recommend using a docker container instead of LXC FWIW Jan 06 16:02:25 framebuffer culd be a solution, but I didn't meet someone who has experience on it Jan 06 16:02:52 warlock: Actually.... LXC is just a container, not a VM, so you might be able to do it still.... Jan 06 16:03:08 Can you access the build workspace from the host? Jan 06 16:03:17 Jpew Docker is a over-structure on LXC Jan 06 16:03:25 (e.g. outside the container) Jan 06 16:03:42 the psplash program is a graphical application that runs on the framebuffer Jan 06 16:03:55 (i.e. /dev/fb0) Jan 06 16:04:02 tlwoerner, wooow please tell me more Jan 06 16:04:28 http://layers.openembedded.org/layerindex/recipe/587/ Jan 06 16:04:40 tlwoerner, does an image exist? Jan 06 16:04:49 core-image-x11 Jan 06 16:05:11 but you have to do: IMAGE_FEATURES += "psplash" Jan 06 16:05:26 tlwoerner, great!! have you tried it? Jan 06 16:05:32 actually, all you need is core-image-minimal Jan 06 16:05:40 yes, and with qemu Jan 06 16:05:57 just do core-image-minimal + IMAGE_FEATURES += "psplash" and you'll see it on bootup in qemu Jan 06 16:06:24 https://twoerner.blogspot.com/2020/12/psplash-improvements-for.html Jan 06 16:06:35 I aready built core-image-minimal, but it didn't function without nographic flag Jan 06 16:07:16 using the most up-to-date (i.e. as of yesterday) master? Jan 06 16:08:52 JPEW: would it be safe to assume you'll be using resulttool in the near future? if so, do you know which MACHINEs you're thinking of using? Jan 06 16:08:54 tlwoerner, do you think the problem can be the environment? I mean I build the image inside a LXC Jan 06 16:09:35 warlock: not sure. i build on my build machine natively, and i use the sdl backend for qemu Jan 06 16:09:48 (or runqemu, as it were) Jan 06 16:10:33 warlock: hahah, oops. i point this out explicitly in my blog then made the mistake anyway Jan 06 16:10:50 it's: IMAGE_FEATURES += "splash" Jan 06 16:10:54 (not psplash) Jan 06 16:11:52 ok thank you all for your help and support Jan 06 16:12:20 thank you so much. See you Jan 06 16:12:52 good luck Jan 06 16:14:50 JPEW: ps, looks like rock-pi-4 kernel builds fail with the new linux-5.10. they split the DTS file into 3 separate ones, one for each board revision (A, B, and C) Jan 06 16:15:11 so i'll have to add separate MACHINEs for each Jan 06 16:16:11 tlwoerner: Yep, my CI *just* caught that :) Jan 06 16:17:00 I looked into it briefly, but it doesn't look like they have any way of identifying the machines from u-boot (or at least none that they have published), so I think separate MACHINES will be required :( Jan 06 16:17:07 but master supports the building of (at least) 3 different kernel versions (5.4, 5.8, and 5.10 + dev (which is linux-stable, i think?)) Jan 06 16:17:40 so the fix has to be kernel-version specific… somehow? Jan 06 16:19:08 not that it helps, but I'll be deleting 5.8 shortly. Jan 06 16:19:17 that still leaves you one-before that change, and one after Jan 06 16:19:31 plus -dev, which is after (?) Jan 06 16:19:45 tlwoerner: Ya, -dev should be after Jan 06 16:19:50 yah. -dev is on 5.11 -rc2 (but I was asked to not push it until rc3) Jan 06 16:20:18 tlwoerner: We can adjust COMPATIBLE_MACHINE so it only works with the newer kernels for those machines Jan 06 16:20:47 tlwoerner: I suppose we could even keep the old machine if we wanted too.... but that might just rot Jan 06 16:21:45 hmm okay, i'll look into it this afternoon then (if that works for you) Jan 06 16:21:57 Yep Jan 06 16:22:03 ok Jan 06 16:22:11 CC me if you make a change and I'll look at it Jan 06 16:22:22 sounds good Jan 06 16:22:44 JPEW: do you plan to use resulttool on any rockchip MACHINEs? Jan 06 16:23:07 tlwoerner: Ideally, it'd use it on all my test boards... but I'm a long way from that Jan 06 16:23:35 tlwoerner: I still have to figure out how to tie labgrid into selftest Jan 06 16:23:51 okay, i'm hoping to look at resulttool sooner rather than later Jan 06 16:24:16 tlwoerner: Cool... I'll be curious to see what you find Jan 06 16:25:33 watch this space for updates: → https://twoerner.blogspot.com/ (lol) Jan 06 16:27:15 put it on your geocities page Jan 06 16:30:18 zeddii: hahah :-P Jan 06 16:31:51 * zeddii bets that flew over the heads of some young 'uns Jan 06 16:32:00 * tlwoerner recently found out a non-technical family relation of mine still pays ~$20CAD/month for AOL in order to have an email account Jan 06 16:32:35 wow. do they pay the CRA in google gift cards as well ? ;) Jan 06 16:33:02 no, silly, the CRA only accepts Apple gift cards, everyone knows that! Jan 06 16:33:15 ahahaha!! that's why they sent mine back. Jan 06 16:33:21 I thought it was visa gift cards! Jan 06 16:34:46 I'm still sad that all my 1 and 2 dollar bills are no longer legal tender here in the great white north! Jan 06 16:35:55 except, it isn't so white here. I can see my grass again. Jan 06 16:42:12 you're the 3rd person i know from the ottawa area who has complained to me about not having snow (lol) Jan 06 16:42:41 we are used to a bit of it!! Jan 06 16:42:46 ironically we don't usually have snow late december, but this year it snowed christmas eve and we've had it ever since :-) Jan 06 16:43:49 * tlwoerner waits for TrevorG and Randy to chime in… Jan 06 16:47:45 zeddii: now you have something to add to your collection of $3 bills Jan 06 16:48:14 :D Jan 06 17:05:12 good afternoon from the UK Jan 06 17:08:21 got a question as am a newbie on yocto, I am trying to create a layer to include a project of mine in the rootfs, now this project depends on a number of librairies (193 of them to be precise) and only a few of them exist in the image I am building (core-image-sato) the image builds fine and boots and runs ok Jan 06 17:09:25 question is how do I include these librairies in the filesystem, nearly all of them have rpms and adding them shouldn't be complicated Jan 06 17:10:59 **** Time to select the CVE raffle winner **** Jan 06 17:11:17 for 2020, I suspect SolarWinds won Jan 06 17:11:38 Looks like there are 50 entries: Jan 06 17:11:42 steve@octo:~/Desktop$ wc -l cve-raffle-sorted-nodups Jan 06 17:11:43 50 cve-raffle-sorted-nodups Jan 06 17:12:04 And the winner by random selection is: Jan 06 17:12:21 steve@octo:~/Desktop$ shuf -n 1 cve-raffle-sorted-nodups Jan 06 17:12:22 CVE-2020-1971: Robert Joslyn Jan 06 17:18:44 FRAUD Jan 06 17:18:49 I DEMAND A RECOUNT Jan 06 17:21:05 sakoman: nice :) Jan 06 17:23:19 zeddii: :-) recount demands seem to be quite popular these days Jan 06 17:24:50 zeddii: OK, I recounted. There are still 50 entries ;-) Jan 06 17:25:05 the contents of that file were switched at 3am! Jan 06 17:25:12 we didn't have any auditors present! Jan 06 17:25:24 #altfacts Jan 06 17:25:27 :D Jan 06 17:58:50 seebs: does http://git.yoctoproject.org/cgit.cgi/poky/commit/?h=master-next&id=84249a2b477f3a2ada7641bbb1a10511b4f838e6 look right? Jan 06 18:37:38 Yeah. I was going to ask why you were adding SYMLINK_NOFOLLOW support to the NO_REAL_AT_FUNCTIONS case, but the previous thing explains it. Jan 06 18:37:50 i gotta say i am not hyped about lchmod existing Jan 06 18:39:08 I might have some peach pie this week. Jan 06 18:39:23 :D Jan 06 18:39:54 with brown sugar on top Jan 06 18:39:57 not orange meringue ? Jan 06 18:40:14 orange peeps are only for microwave experiments Jan 06 19:51:33 Hello Everyone. Please I need some help. Jan 06 19:51:39 https://www.irccloud.com/pastebin/YqdnDG5T/ Jan 06 20:08:58 ERROR: Running idle function########################################################## | ETA: 0:00:01 Jan 06 20:09:16 KeyError: 'exc' Jan 06 20:09:26 while running bitbake for yocto Jan 06 20:47:15 idadel: how does it come to deduce the license to be GPL-2.0 WITH Linux-syscall-note ? Jan 06 21:01:13 seebs: it doesn't seem like a great idea to me either but it seems it does now :/ Jan 06 21:01:30 seebs: thanks for the sanity check :) Jan 06 21:22:20 sanity. hah. Jan 06 21:22:27 this is not a great day for sanity checks i think Jan 06 21:22:35 indeed it is not Jan 06 21:23:20 #2020.2.0 Jan 06 21:27:58 seebs: I was just looking at the news :( Jan 06 21:31:04 i am just so happy for all the lawtwitter people Jan 06 21:31:15 who have spent YEARS angrily saying "no, that's NOT sedition" Jan 06 21:31:24 and finally get to say "okay sure that's sedition" Jan 06 22:01:05 rburton: fwiw I think there is only debian9 using py 3.5 Jan 06 22:01:24 i.e. not a lot of reason to keep 3.5 as the min Jan 06 22:02:02 kanavin: deleting patches from qemu seems to be fixing failures :) Jan 06 22:02:36 RP: that's how I do those restoring-recipe-sanity version updates :) Jan 06 22:03:05 RP: start off by deleting *all* patches, then bring them back on a strict needs-to-be-present basis Jan 06 22:03:54 no 'someone may have an issue if this patch is absent' nonsense :) Jan 06 22:05:51 kanavin: I do know the history of most of the qemu ones but the failures definitely pointed at the patches as now being problematic Jan 06 22:06:26 kanavin: I'm forcing a full rebuild for other changes so will be interesting to see how this build looks Jan 06 22:07:49 JPEW: would you be able to help with https://autobuilder.yoctoproject.org/typhoon/#/builders/89/builds/2903/steps/12/logs/stdio ? (new qemu on mingw) Jan 06 22:08:53 JPEW: I suspect the switch to meson upset something Jan 06 22:09:04 RP: Ok Jan 06 22:10:11 JPEW: its building things so hopefully just a path configuration issue, thanks! Jan 06 22:12:31 RP: Ugh 32-bit MinGW :/ Jan 06 22:12:58 RP: hi there, while it builds can you maybe comment about using ASSUME_SHLIBS for custom libfoo.so without soname? Jan 06 22:13:24 seems rather an hack but googling I found worse... Jan 06 22:14:22 ant__: it does seem a bit of a hack. Not sure we do that anywhere in core? Jan 06 22:14:27 well, these should maybe be PRIVATE_LIBS so the question is a bit pointless Jan 06 22:14:36 no, no, just a layer uses it Jan 06 22:15:38 ant__: PRIVATE_LIBS is definitely nicer Jan 06 22:16:06 ah, ok, then adding the RDEPENDS Jan 06 22:18:24 JPEW: sorry. If we want to jettison that in favour of 64 only... ;-) Jan 06 22:19:13 RP: Maybe... it looks like perhaps the actual problem is that the qemu build is eliding all the TLDs ("bin", "share", etc.) when installing Jan 06 22:19:30 Which, to be fair *is* how it gets installed on Windows.... Jan 06 22:23:26 JPEW: there was a small change to the "suffix" option which I had to adapt the recipe too, wonder if that is related? :/ Jan 06 22:26:20 RP: Maybe... let me try a build Jan 06 22:41:43 hi, i have a crusty old build environment that works and new docker one that doesn't to generate my image ... the difference appears to be in the rpm spec files where some libs aren't being put in the Requires and Provides for my (unnecessarily complex) package and i can't quite figure out why :( Jan 06 22:43:16 ahead of the spec file i found the pkgdata, which shows some ordering differences and some content differences that i've not quite exhaustively correlated yet -- but i'm at a loss to understand how the same 'code' would produce such wildly different results (one includes the sstate-cache that is preserved and shared across many builds, which may be "the key difference") Jan 06 22:44:38 my best guess is to instrument the part of the system that discovers dependencies by looking at binaries, so any direction as to where to find that would be helpful :) Jan 06 22:46:18 opello: the code is in package.bbclass and lib/oe/package.py Jan 06 22:47:28 opello: the data is written out to pkgdata so if that doesn't look right, that is somewhere to look and follow up Jan 06 22:47:59 RP: thanks! yeah in the pkgdata i see the FILERDEPENDS missing for the missing spec file elements, so that's pretty compelling ... Jan 06 22:48:45 adding debug prints to package.bbclass ends up re-running quite a few tasks, i imagine i can just override them for my recipe too? Jan 06 22:48:47 opello: that is generated by rpmdeps iirc Jan 06 22:49:25 opello: writing a locked-sigs.inc and including that to force the hashes Jan 06 22:49:47 (will stop things rebuilding) Jan 06 22:50:41 ah, i've seen that documented but haven't used it -- will try to make use of that too Jan 06 23:03:06 RP: the other/worse hack to handle no soname https://www.yoctoproject.org/pipermail/yocto/2015-February/023715.html Jan 06 23:03:14 RP: Ah, found it: https://github.com/qemu/qemu/blob/master/configure#L1573 :/ Jan 06 23:04:29 JPEW: oh :/ Jan 06 23:05:21 RP: Supper time. I'll give it a think for a bit Jan 06 23:06:08 JPEW: fair enough. I'm going to run more builds/tests of qemu overnight and I've my hands full with other issues (like the missing pseudo lchmod wrapper breaking reproducibility) Jan 06 23:06:13 JPEW: thanks! Jan 06 23:06:45 ant__: That does look worse/horrible Jan 06 23:07:31 meta-rockchip shoud have the ASSUME_SHLIBS thingy, which I understend as coupled with ASSUME_PROVIDED so not for target Jan 06 23:08:52 https://wiki.yoctoproject.org/wiki/TipsAndTricks/Packaging_Prebuilt_Libraries Jan 06 23:09:08 ^ add the no-no examples :) Jan 06 23:09:30 ant__: docs patches weclome! Jan 06 23:09:37 well, that is a wiki :) Jan 06 23:10:13 he, I am still newb against unversioned libs w/out soname..never happened before... Jan 06 23:23:53 heh - an old co-worker pointed me at this 6-7y old patch and asked me why it wasn't merged... Jan 06 23:23:58 https://www.yoctoproject.org/pipermail/poky/2014-July/009821.html Jan 06 23:24:55 I have no idea. I'd still say using "/home/root" sucks swamp water even today. Jan 06 23:42:12 paulg that one certainly got lost! Jan 06 23:47:02 sgw, maybe kernel people aren't allowed to muck with userspace... ;-) Jan 06 23:48:29 need somebody like RP to give it the holy blessing. **** ENDING LOGGING AT Thu Jan 07 03:00:55 2021