**** BEGIN LOGGING AT Mon Sep 16 03:00:52 2019 Sep 16 03:23:30 New news from stackoverflow: How should the sstate-cache directory be deleted in Yocto? Sep 16 05:51:29 what might i be missing if the crond.service is up and running, but the cron.daily, cron.hourly ... script seem to be ignored? filenames etc. have been checked, when using run-parts to test the directories everything works as expected. Sep 16 06:42:46 good morning Sep 16 07:16:49 just sent a few patches, I hope list servers are still working.. Sep 16 07:24:08 New news from stackoverflow: QA Issue: libvncserver: Files/directories were installed but not shipped in any package: Sep 16 07:56:53 for the record: cronie ships per default with a crontab that disables the /etc/cron.* directories. meh. Sep 16 08:01:30 I'm trying to build an initramfs image, and currently I bundle the kernel + initramfs together. I write that bundle to my sdcard and load it using mmc read etc, and then try to boot using booti (I'm an arm64). I get the error message "Bad Linux ARM64 Image Magic". I did some googling and now i'm wondering if I'm missing something called "uImage" container that u-boot needs. Sep 16 08:03:13 iceaway: u-boot has a variety of booting command, i think it also should support non-uImage formats. but i don't know offhand, you would have to look at the docs. Sep 16 08:06:38 iceaway: AFAIR, for arm64, you should build and boot the Image with booti, don't forget to put the DTB load addr as the third argument Sep 16 08:08:33 iceaway: can you give the full u-boot command lines you're using to boot with booti? Sep 16 08:12:17 qschulz: yup that is what I am trying. Looking at the u-boot code now and from what I can see there is no decompression done before booting, maybe that is why the magic is wrong? Should I decompress it manually first? Sep 16 08:12:30 using: booti ${loadaddr} - ${fdt_addr} Sep 16 08:15:20 have you checked that loadaddr and fdt_addr were not overlapping when loading the DTB and Image? Sep 16 08:16:52 Hello all, is there some form of tutorial out there to work with the PREEMPT RT Linux patch with Yocto Project? I just am looking to get some hands dirty with RT-Linux. Sep 16 08:17:53 iceaway: take the Image and not the Image.gz? Sep 16 08:19:09 qschulz: will look at that, it's all bundled of by bitbake so need to figure out how to get the uncompressed version. When doing my "normal" image boot the compressed image is stored in the /boot path. Not sure how u-boot does then with unpacking. Sep 16 08:19:41 I did unpack the compressed image to another RAM location from load address, and could at least get the kernel started from there. Sep 16 08:19:48 shan1: the "official" yocto kernels already come in an RT flavor Sep 16 08:23:14 LetoThe2nd does that imply that when I make a image for any board it is RT "flavour"? Sep 16 08:24:19 New news from stackoverflow: Akward line wrap in Yocto || Yocto core image for jetson nano black screen Sep 16 08:24:53 shan1: no, it means that if you explicitly select the linux-yocto-rt kernel, then they are preemp-rt enabled. for any given custom board, your mileage may vary. Sep 16 08:25:49 iceaway: KERNEL_IMAGETYPE="Image"? but AFAICT, we're using Image.gz for that one here and it boots fine Sep 16 08:26:09 iceaway: I would definitely ask those questions on #u-boot :) Sep 16 08:26:32 iceaway: I don't remember exactly where U-Boot is extracting images Sep 16 08:26:56 so the unzipping of the image might be overriding something Sep 16 08:29:06 LetoThe2nd ok thanks. Any thing I should keep in mind if I just want to get a RT-Linux on a machine like Raspberry pi or something? Sep 16 08:31:06 iceaway: might also be some Kconfig options not enabled somewhere Sep 16 08:33:20 shan1: primarily that i have no clue if meta-raspberrypi brings a preempt-rt kernel recipe :) Sep 16 08:42:56 Hello, yocto installs all files of my service correctly but the service is not started when my embedded Yocto Linux boots up. Here is the bitbake .bb file. What could be wrong?: https://pastebin.com/cmef0PZN Sep 16 08:44:18 LetoThe2nd Sure looks like it (https://github.com/agherzan/meta-raspberrypi/blob/2634621374be557f543a65d0e387ca1ec4cc4666/recipes-kernel/linux/linux-raspberrypi-rt_4.19.bb) Sep 16 08:50:09 shan1: then set PREFERRED_PROVIDER_virtual/kernel = "linux-raspberrypi-rt" in your machine, and hopefully be done. Sep 16 08:53:07 Doing "systemctl enable start-firmware" manually after Linux booted up works well, but somehow yocto fails to do a systemctl correctly. http://lists.openembedded.org/pipermail/openembedded-devel/2016-March/106629.html says systemctl of yocto is just a script but not the real tool, so I wonder if the problem is probably there, but how can I found Sep 16 08:53:07 out where it fails? Sep 16 09:04:30 is there a way to inspect the kernel options and auto-loaded modules from the image before I flash it? Sep 16 09:04:45 (i.e. in build/tmp/deploy/images/my-image)? Sep 16 09:24:29 New news from stackoverflow: Yocto change psplash image without rebuild the system Sep 16 09:30:08 LetoThe2nd Thanks it worked! The image for the Pi-4 with RT Kernel just got built! Sep 16 09:31:33 shan1: :-) Sep 16 09:33:11 shan1: on a side note: i saw your cheatsheet starter, and ... while i like the idea, unfortunately it seems to contain quite a bunch of misleading things. not sure when i get around to prepare a improvement, but at least so far i would suggest that for all the .conf options you do not use it as a reference or such, ok? Sep 16 09:34:59 shan1: it at least gave me the topic to talk about in the next live session! :) Sep 16 09:38:12 LetoThe2nd I am all for improvements. I am just an amateur with Embedded Systems with Yocto, so I would be more than happy to work with the community to improve it. This was just my __hacky__ way to understand Yocto. Sep 16 09:46:53 I also see that you guys track stackoverflow, I thought yocto was more of a Unix topic: https://unix.stackexchange.com/questions/541868/ethernet-over-usb Sep 16 09:49:34 palate: you are in fact very right, but most of the SO readers/mods don't have a grasp of the yocto concepts, and so they think it looks like code, hence it must be code :) Sep 16 09:49:55 LetoThe2nd: right :D Sep 16 09:50:13 shan1: it was not meant to discourage you, just to ... be a little careful when showing it to others. once i find time, i hope to send you a pull request. Sep 16 09:51:41 palate: you do in fact have the modules installed? Sep 16 09:52:14 LetoThe2nd (y) I will just add a note in the README.md that it is WIP and will go changes accordingly! Sep 16 09:52:59 I want to debug why SYSTEMD_AUTO_ENABLE_${PN} = "enable" does not work. Is there any logfile for this, or where can I find documentation how yocto processes that .bb file entry? Sep 16 09:53:04 hi rburton. I have recipes for qpdf 9.0.0 and cups-filters 1.25.5. I would like to push them to either openembedded-core or meta-openembedded, as meta-priting seems to be unmaintained. Where do you suggest to push those recipes? meta-openemebedded/meta-oe? Sep 16 09:53:21 shan1: thanks! Sep 16 09:53:44 shan1: the command section looks mostly fine, its really just the config stuff that needs some work Sep 16 09:54:35 New news from stackoverflow: Ethernet-over-USB with Yocto || How to reduce power button press time to turn off an imx6 board || Awkward line wrap in Yocto Sep 16 09:54:52 LetoThe2nd: I don't have any access to my pocketbeagle, yet (I will try to solder an FTDI this afternoon to get the shell) Sep 16 09:55:33 LetoThe2nd: all I know is that the kernel options mention CDCETHER, and I added that to the autoload modules. But I don't know how if I could inspect my image on my computer before flashing it on the pocketbeagle Sep 16 09:55:44 LetoThe2nd: (would be rather convenient) Sep 16 09:55:47 palate: what image fstypes do you create? Sep 16 09:56:00 palate: (and the ftdi is always handy!) Sep 16 09:56:06 LetoThe2nd: wic Sep 16 09:56:58 rburton, shame about the sdl bits, they do work properly on ubuntu 18.04.3 and opensuse. Is there a possibility to look at those logs the failure references (log.do_testimage, qemu boot log)? Sep 16 09:58:01 palate: maybe add tar.gz to it, as this can be rather easily inspected. Sep 16 09:58:48 palate: and you can always look at the image.manifest, it should contain the list of installed things. Sep 16 09:59:05 palate: so if the modules are not mentioned there, then they are not in the image. Sep 16 10:04:55 LetoThe2nd: aha... I don't find any "ether" mentioned in the image manifest Sep 16 10:05:41 palate: nah, not exactly "ether". kernel module packages are by defintion called "kernel-module-***" in OE builds. Sep 16 10:05:51 palate: do they make it into the image? Sep 16 10:06:08 LetoThe2nd: nope, none of them Sep 16 10:06:28 LetoThe2nd: (as far as I can tell, i.e. there is no mention of "kernel-module" in the manifest) Sep 16 10:07:01 palate: then i'd start out by creating an image recipe basing off whatever you want, and add "kernel-modules" to IMAGE_INSTALL Sep 16 10:07:11 LetoThe2nd https://imgur.com/a/txqEdgj Added a Note for the configuration section. Sep 16 10:07:11 kanavin_: hopefully Sep 16 10:07:33 kanavin_: https://autobuilder.yoctoproject.org/typhoon/#/builders/72/builds/1056/steps/8/logs/step5c is qa-extras2 on the new tumbleweed machine failing in testimage Sep 16 10:07:37 related? Sep 16 10:08:10 shan1: looks good, thanks! Sep 16 10:08:51 LetoThe2nd: ooh, "kernel-module" Oo. I was reading the features here and did not see that this was an option: https://www.yoctoproject.org/docs/current/ref-manual/ref-manual.html#ref-features-image Sep 16 10:09:26 palate: :) Sep 16 10:10:48 LetoThe2nd: and apparently it's not part of the core-image-minimal. Trying that \o/ Sep 16 10:11:12 minimal means minima Sep 16 10:11:19 there's a reason there's also a core-image-base Sep 16 10:11:24 or core-image-full-cmdline Sep 16 10:12:27 LetoThe2nd: btw I think it's "kernel-modules" (with an "s"), as the other one was not found ;-) Sep 16 10:12:46 palate: i think so too. read again what i wrote :) Sep 16 10:13:29 LetoThe2nd: haha you're right xD. Annnnnd... it works! At least modules (including cdc-ether) appear in the manifest! \o/\o/ Sep 16 10:13:51 palate: badum-tsh! Sep 16 10:14:47 LetoThe2nd: I'll check what I can get with a tar.gz, now. It's important to get basics of how to debug an image :D Sep 16 10:15:09 LetoThe2nd: thanks a lot for the help! Sep 16 10:15:40 shan1: and for some nitpicking nomenclature (you don't have to change the note!): i am officially a member of the OE project, but not of yocto :) Sep 16 10:17:06 Anybody here with previous experience of using Plymouth in a initramfs? Sep 16 10:17:42 rburton, ahven't seen the tumbleweed one before Sep 16 10:29:58 kanavin_: its a new worker so all new error cases :) Sep 16 10:32:20 step5 is just with systemd Sep 16 10:34:09 rburton, the xorg test does pass however - maybe the 'error' is actually not a real one? it seems to say something about udev, which is systemd Sep 16 10:34:17 right Sep 16 10:35:32 LetoThe2nd: the pocketbeagle is still not doing anything when I plug it in my computer (`dmesg` says nothing), but that's one step forward anyway. Next step is probably the FTDI :-) Sep 16 10:39:18 qschulz: thanks for the tips, will investigate more and join #u-boot if required :) Sep 16 11:24:51 New news from stackoverflow: SYSTEMD_SERVICE_swupdate value swupdate.service does not exist Sep 16 12:38:42 JPEW: 2019-09-16 13:35:12,476 - oe-selftest - INFO - Reproducibility summary for deb: same=4707 different=0 missing=0 total=4707 Sep 16 12:38:42 2019-09-16 13:35:14,221 - oe-selftest - INFO - Reproducibility summary for ipk: same=4707 different=0 missing=0 total=4707 Sep 16 12:39:48 Hi all, I would have some advice about version management inside Yocto framework. I have several recipes under version control (1.0.bb, 1.2.bb ...) I have set preferred_version for each of them. I'm not sure but I can't put the preferred version inside our distro.conf because OS and applications are not in the same meta. I let conf/local.conf for Sep 16 12:39:49 users stuff. So where is the best place for that? Thx. Sep 16 12:40:12 distro.conf is exactly where that should go Sep 16 12:40:20 kanavin_: did you get anywhere with strace? Sep 16 12:41:12 RP: rburton: i already decided the next live session will be on DISTRO vs. MACHINE vs. IMAGE Sep 16 12:41:25 rburton Thx, but there is no trouble to set preferred_version if any recipe exists ? Because there are in our "application layer" Sep 16 12:42:02 its just a variable Sep 16 12:42:05 no problem at all Sep 16 12:42:27 LetoThe2nd: sounds good Sep 16 12:42:33 why not make the distro layer depend on your other layers? :) Sep 16 12:42:36 LetoThe2nd should be good idea, in fact there are any info about Yocto and how to use it but not so much about create and/or configure project :):) Sep 16 12:42:49 RP: i kinda realized that it is needed. Sep 16 12:43:18 LetoThe2nd: YES PLEASE. Sep 16 12:43:23 rburton sounds a good advice :) thx Sep 16 12:48:43 LetoThe2nd: awesome, sounds like an great topic for us newcomers. Sep 16 12:49:19 * Copyright (c) 2003 Intracom S.A. Sep 16 12:51:39 RP: fired an enhanced reprod test at the ab Sep 16 12:51:47 ipkg and sato :) Sep 16 13:08:48 rburton: took me a minute to understand that :) Sep 16 13:09:21 reproducible :) Sep 16 13:15:16 rburton: Awesome! Sep 16 13:16:51 JPEW: adelcast cleaned up my opkg-utils patch so that's opkg sorted. fired a build on the AB testing opkg with sato to see what the damage is Sep 16 13:25:16 New news from stackoverflow: How to fix error 501 while using yocto build tool? Sep 16 13:42:16 hmm so I soldered my FTDI, connected it to the pocketbeagle and to my computer, and it is not recognized by my computer :/ Sep 16 13:42:43 It detects Product: TTL232R, Manufacturer: FTDI, which is my FTDI chip Sep 16 13:43:27 on the other side (pocketbeagle), I connect as shown here: https://github.com/beagleboard/pocketbeagle/wiki/System-Reference-Manual#56_Serial_Debug_Port Sep 16 13:43:54 (TX to RX, RX to TX, GND to GND) Sep 16 13:44:09 I power my pocketbeagle over USB, and then I connect the FTDI Sep 16 13:44:26 my computer doesn't see anything. I would expect that dmesg shows a serial device connected, right? Sep 16 13:44:59 Am I doing something wrong, or could it be that my yocto image just doesn't boot? It boots when I build for qemux86... Sep 16 13:46:17 palate you should try first with one of this images : https://beagleboard.org/latest-images Sep 16 13:46:23 just to check HW ;) Sep 16 13:46:34 palate: do you have the FTDI drivers installed on your computer? Sep 16 13:47:42 PinkSnake: yes that works, I use that pocketbeagle with other images. I'm just trying to learn a bit of Yocto :) Sep 16 13:48:00 qschulz: not that I remember... that's probably a "no"... Sep 16 13:48:33 palate: I think what PinkSnake meant is to check your UART with the official image? Sep 16 13:48:46 qschulz: oooh, got it, sorry PinkSnake :) Sep 16 13:48:58 qschulz: I have libftdi installed, is that it? Sep 16 13:49:39 palate: noidea.gif :) but in dmesg you should have a new serial device enumerated right after detection of the FTDI chip IIRC Sep 16 13:49:41 palate qschulz Apologize my English guys ;( Sep 16 13:49:48 I see libftd2xx that I haven't installed: "Library that allows a direct access to a USB FTDI2XX chip based device" Sep 16 13:50:39 palate: no you certainly don't need that one Sep 16 13:50:54 palate: its a totally standards compliant usb-serial device Sep 16 13:50:59 no lib neede Sep 16 13:51:05 :/ Sep 16 13:51:17 what OS is your dev host on? Sep 16 13:51:35 I guess the next step is to do what PinkSnake suggests, then: official image, and try to see if the serial is recognized? Sep 16 13:51:40 LetoThe2nd: Arch Linux xD Sep 16 13:51:52 palate: reboot Sep 16 13:52:02 Is that an arch linux thing? :D Sep 16 13:52:06 ooooh... update Sep 16 13:52:27 palate: archlinux has the speciality that it doesn't recognize new usb devices if there was a kernel update since last reboot Sep 16 13:52:43 palate: and i am exactly 100% sure that the ftdi works OOTB on arch Sep 16 13:52:59 LetoThe2nd: is it only an arch specialty? Sep 16 13:53:09 * palate reboots... Sep 16 13:53:16 RP: I did, check the bugzilla :) Sep 16 13:53:29 RP: https://bugzilla.yoctoproject.org/show_bug.cgi?id=13506 Sep 16 13:53:31 Bug 13506: normal, Medium+, 2.8 M3, alex.kanavin, IN PROGRESS DESIGN , [master-next] strace ptest timed out Sep 16 13:57:03 LetoThe2nd: FTDI USB Serial Device converter detected -> looking better Sep 16 13:57:29 palate: :) Sep 16 13:57:40 palate: and yes, it *is* an arch speciality Sep 16 13:57:58 LetoThe2nd: haha! I'm in \o/ \o/ \o/ Sep 16 13:58:10 Thanks a lot guys :-) :-) Sep 16 13:59:00 So now I see that cdc_ether is there in /lib/modules, that's something Sep 16 14:00:01 If I do "lsmod", it says "cdc_ether 16384 0 - Live 0xbf01f000" -> so I presume the cdc_ether module is running Oo Sep 16 14:00:03 palate: lsmod? Sep 16 14:00:12 ah damn Sep 16 14:00:15 too quick for me :) Sep 16 14:00:32 haha :) Sep 16 14:00:37 ip address assigned? Sep 16 14:00:40 on the correct subnet? Sep 16 14:01:04 so the issue I have is that my computer doesn't see the pocketbeagle as a network interface. Like "ip link" doesn't show it, and "dmesg" doesn't say anything (except for the FTDI) Sep 16 14:01:21 Which I guess should come before the IP setup, right? Sep 16 14:01:59 I'm new to usb gadgets, so my current naive idea is that "I load cdc_ether, plug the pocketbeagle into my computer, and my computer sees a network interface over USB" Sep 16 14:03:37 if dmesg does not say anything, that does not smell so good Sep 16 14:04:24 ¯\_(ツ)_/¯ Sep 16 14:05:00 that's the reason why I went for FTDI actually, hoping to see something from the pocketbeagle Sep 16 14:05:47 but now I'm completely lost. I just know that the official image does ethernet over usb (I don't know how) Sep 16 14:07:44 RP: Do we really need both BB_HASHSERVE and SSTATE_HASHEQUIV_SERVER variables? Can I remove SSTATE_HASHEQUIV_SERVER as redundant? Sep 16 14:08:22 palate: run the official image and look at its loading scripts. thats what i would do. Sep 16 14:09:06 JPEW: probably Sep 16 14:11:06 LetoThe2nd: can you give me a keyword about the loading scripts I'm interested in? I guess I would check the modules (is it cdc_ether or g_ether?), but I don't really know what else Sep 16 14:11:21 LetoThe2nd: as I said, USB gadgets are still magic to me xD Sep 16 14:11:37 palate: i think it should be g_ether Sep 16 14:12:04 palate: just boot it up, look at lsmod for starters. and join #beagle :) Sep 16 14:12:47 oh, #beagle :) Sep 16 14:13:14 LetoThe2nd: ok I'll do that! Thanks a lot again for the help Sep 16 14:13:38 I was going for `cdc_ether` because I believe that `g_ether` is not recognized on Android, and I was curious about that Sep 16 14:13:55 LetoThe2nd: why would you think it should be g_ether instead of cdc_ether? Sep 16 14:14:16 palate: because g_ stands for "gadget" Sep 16 14:15:50 oh :) Sep 16 14:16:28 palate: and if i'm not super mistaken, then cdc_ether should be the hostside equivalent to it Sep 16 14:19:05 damnit, I got it wrong again xD Sep 16 14:20:03 but then if Android doesn't detect g_ether, I'm screwed I guess Sep 16 14:20:36 palate: why would anyone care about andoird anyways? Sep 16 14:24:49 LetoThe2nd: xD. I'd like my pocketbeagle to be a network interface to a robot. So you connect your Android phone to my pocketbeagle, it creates a new network interface on your phone, and then you can just communicate with the robot with networking Sep 16 14:25:19 palate: my personal suggestion is: do one ting at a time. Sep 16 14:25:51 LetoThe2nd: it works e.g. if you plug a router into your Android with the right adapters. So I'm trying to understand how my pocketbeagle could imitate that Sep 16 14:26:14 and as long as you don't have a solid grasp on the PB handling itself, then android will only complicate things. Sep 16 14:26:32 right, that's a fair point Sep 16 14:38:52 rburton: just saw the results, that's awesome Sep 16 14:52:15 zeddii: Is there something like https://lkml.org/lkml/2019/5/13/703 we should pull in? Sep 16 14:57:34 possibly, but it requires research to why 5.2-stable wouldn't have already gotten the fix if it went mainline. I'll poke around a bit. Sep 16 15:00:56 it's this commit: git show 05b289263772b Sep 16 15:02:10 RP: interesting. that's already in 5.2, are you seeing this with linux-yocto-5.2 ? you shouldn't Sep 16 15:07:54 zeddii: sorry, I'm misreading. https://bugzilla.yoctoproject.org/show_bug.cgi?id=13506 Sep 16 15:07:55 Bug 13506: normal, Medium+, 2.8 M3, alex.kanavin, IN PROGRESS DESIGN , [master-next] strace ptest timed out Sep 16 15:08:04 zeddii: looks like there is another issue with that change :/ Sep 16 15:16:58 * zeddii reads and ponders Sep 16 15:23:48 i have built a (non-extensible) SDK using "bitbake image -c populate_sdk". do i need to run the "xyz.sh" script to install it on the build system? this step isn't mentioned explicitly in the doc Sep 16 15:23:58 https://www.yoctoproject.org/docs/2.1/sdk-manual/sdk-manual.html Sep 16 15:24:17 the tmp/deploy/sdk/xyz.sh script, that is Sep 16 15:26:13 yes Sep 16 15:26:20 yes, captain obvious... Sep 16 15:30:15 zeddii: not sure if kanavin_ is planning to take it further but we should get it reported. This issue is blocking M3 :/ Sep 16 15:35:09 zeddii, you ok if i send your 4.14 update patch to the list? Sep 16 15:35:30 no worries. fire away. Sep 16 15:35:37 k Sep 16 15:39:27 RP: I am sadly out of ideas at this point how to drive this further. We need to ask the author of that commit. Sep 16 15:41:16 RP: we can apply a short timeout to the tests, but it won't resolve the issue, just will make the tests complete within reasonable time Sep 16 15:54:47 kanavin_: in the short term lets do that as having X failures rather than timeout would be better Sep 16 15:55:00 kanavin_: are you going to report to upstream or should someone else? Sep 16 16:01:21 I tried to install myfile.py file into /usr/local/bin/ from a do_install(), but yocto copied to /usr/lib/python3.6/myfile.py instead... are there are any example layers that show how to deploy a python source app in a custom layer? Sep 16 16:01:53 Most of the examples I see all compile helloworld.c and put it in /usr/bin (${bindir}) Sep 16 16:10:35 RP: if you or Bruce already follow lkml, I'd rather one of you do the report Sep 16 16:10:44 ecdhe: your do_install can put files wherever it wants Sep 16 16:10:56 kanavin_: thats fine, just want to ensure someone is doing it Sep 16 16:11:00 zeddii: you or me? Sep 16 16:28:40 Anyone ever used git send-email behind a proxy? I've never used git send-email before and can't seem to get it working yet Sep 16 16:42:51 I'm getting this error: ERROR: sqlite3-3_3.23.1-r0 do_package_qa: QA Issue: /usr/bin/sqlite3 contained in package sqlite3 requires libpthread.so.0, but no providers found in RDEPENDS_sqlite3? [file-rdeps] Sep 16 16:43:00 Do I need to add glibc as an RDEPENDS? Sep 16 16:43:07 That doesn't seem right to me Sep 16 16:43:26 I can see that libpthread.so.0 is in the glibc package Sep 16 16:43:36 tmp/work/armv7vet2hf-neon-poky-linux-gnueabi/glibc-locale/linaro-2.25-r0/locale-tree/lib/libpthread.so.0 Sep 16 16:43:53 sorry wrong file Sep 16 16:43:54 tmp/work/armv7vet2hf-neon-poky-linux-gnueabi/glibc/linaro-2.25-r0/packages-split/glibc/lib/libpthread.so.0 Sep 16 16:46:28 Has someone ever incurred into "/lib/ld.so.1: No such file or directory" error? Sep 16 16:48:45 alessioigor: when? Sep 16 16:49:38 rburton: Building gobject-introspection-data on powerpc (ppc7400) and powerpcspe (ppce500v2). Sep 16 16:49:50 On master Sep 16 16:51:40 alessioigor: probably trying to find ld.so in a path where it isn't Sep 16 16:51:55 alessioigor: possibly due to a host/target libdir confusion? Sep 16 16:52:26 rburton: I'll try to give a look. Sep 16 16:55:41 alessioigor: ppc host or x86 Sep 16 16:55:43 ? Sep 16 16:56:08 rburton: x86 Sep 16 16:56:13 hm Sep 16 16:56:58 the AB should be exercising those paths for poky at least Sep 16 16:59:28 yeah and they work Sep 16 17:10:03 does anyone know how "glibc" gets implicitly added to RDEPENDS during do_package_qa? Sep 16 17:10:43 because a binary links to libc.so.6, and bitbake knows that libc.so.6 is in the glibc package Sep 16 17:10:52 its during do_package, not do_package_qa Sep 16 17:10:58 it's not getting added in sqlite3 Sep 16 17:11:18 something broke in sqlite then Sep 16 17:11:24 it's happening in multiple recipes Sep 16 17:11:27 not just sqlite3 Sep 16 17:11:37 ok, something broke in your toolchain Sep 16 17:11:40 right Sep 16 17:11:49 that's why I'm asking where that RDEPEND gets implicitly added Sep 16 17:11:50 can you replicate with nodistro/poky? Sep 16 17:11:54 it's not getting added in my environment Sep 16 17:12:03 I'm using poky distro Sep 16 17:12:06 maybe check that glibc contains the library you expect Sep 16 17:12:10 it does Sep 16 17:12:29 However, I can see in do_package_qa with/without my toolchain there's a difference Sep 16 17:12:46 in my toolchain, glibc is not getting added to RDEPENDS, with the default tool chain it is Sep 16 17:12:50 who is adding glibc to RDEPENDS? Sep 16 17:12:53 ah, so your toolchain is bust Sep 16 17:12:57 correct Sep 16 17:13:03 that's why I'm asking how glibc gets added to RDEPENDS Sep 16 17:13:05 so I can fix it Sep 16 17:13:12 was that one you built out of tree, or a vendor provided it? Sep 16 17:13:21 it's the linaro toolchain Sep 16 17:13:26 provided by meta-linaro Sep 16 17:13:28 i'd complain to linaro then Sep 16 17:13:33 sure Sep 16 17:13:39 but the question still stands Sep 16 17:13:46 how does glibc get added to RDEPENDS? Sep 16 17:13:57 [18:10:43] because a binary links to libc.so.6, and bitbake knows that libc.so.6 is in the glibc package Sep 16 17:13:58 I'm not complaining, I'm looking for help Sep 16 17:14:09 so for you've asked me a bunch of questions and haven't provided any help yet Sep 16 17:14:27 so check that the library is in glibc Sep 16 17:14:32 that's not an answer to the question Sep 16 17:14:47 when and how in does glibc get added to the RDEPENDS variable? Sep 16 17:14:57 it is in glibc, I already answered that question Sep 16 17:15:11 do_package knows what libraries exist and where they come from Sep 16 17:15:18 note that if I run bitbake -e, it's not in RDEPENDS whether or not I'm using either toolchain Sep 16 17:15:34 so check that glibc has built that library correctly Sep 16 17:15:34 but it gets added somewhere inside the code before do_package_qa Sep 16 17:15:38 it is in there Sep 16 17:15:40 [18:10:52] its during do_package, not do_package_qa Sep 16 17:15:51 so presumably the shlibs are broken then Sep 16 17:15:52 you're not listening to me and you keep asking the same questions I've already answered Sep 16 17:16:04 I just have one question Sep 16 17:16:09 who is adding glibc to RDEPENDS? Sep 16 17:16:11 where is the code? Sep 16 17:16:16 the do_package Sep 16 17:16:20 task Sep 16 17:16:45 who is adding to it when do_package_qa runs? Sep 16 17:16:50 I'm taking about the python data structures Sep 16 17:16:54 they get reset at every task Sep 16 17:17:07 so if do_package sets the RDEPENDS variable, it doesn't affect do_package_qa Sep 16 17:18:28 yocto finds that /usr/bin/sqlite3 depends on libc, but libc is not added to RDEPENDS_sqlite3 Sep 16 17:18:48 I could make a bbapend to added glibc (which works) but that's obviously wrong Sep 16 17:18:56 how is glibc supposed to be added to RDPEENDS_sqlite3? Sep 16 17:19:27 I can tell you that with or without linaro, it does not appear when you but "bitbake -e sqlite3" Sep 16 17:19:41 correct, do_package adds it for you Sep 16 17:19:43 it gets added somewhere in the task code, but I don't know where, and I don't know why it's not getting added when I use lianro Sep 16 17:20:02 Is there a way to get the bitbake -e output within a task? Sep 16 17:20:10 have a look in pkgdata/shlibs for glibc Sep 16 17:20:14 A way to print a variable and all the places it has been set within a task? Sep 16 17:20:15 compare with both toolchains Sep 16 17:20:19 it's not a problem with shlibs Sep 16 17:20:27 shlib is detecting the dependency Sep 16 17:20:38 and you've verified that the shlibs are correct? Sep 16 17:20:43 the problem is that it detects the dependency, but sees that the dependency wasn't declared in RDEPENDS Sep 16 17:20:54 so it then asserts an error that RDEPENDS is missing the dependency Sep 16 17:20:58 YES! Sep 16 17:21:21 I've asked 2 questions that I haven't gotten an answer to, you've asked me a dozen questions Sep 16 17:21:28 please can you answer one of my 2 questions? Sep 16 17:21:31 ffs Sep 16 17:21:45 no you can't do 'bitbake -e' on a task Sep 16 17:21:53 that's not what I asked Sep 16 17:22:01 I asked if you could print the same output as bitbake -e within a task Sep 16 17:22:02 [18:20:02] Is there a way to get the bitbake -e output within a task? Sep 16 17:22:03 yes you did Sep 16 17:22:09 what was the other question? Sep 16 17:22:10 that's not the same Sep 16 17:22:37 out is the output of bitbake -e printed? Maybe there is a library or a function I can call to get the output within a task Sep 16 17:22:46 *how is the output... Sep 16 17:23:33 i'm not sure how that is relevant. the problem is that the binaries are obviously different. run ldd on the working sqlite binary and the broken sqlite binary, maybe the linkage is bust Sep 16 17:23:45 you're still not understanding Sep 16 17:23:48 the binaries are fine Sep 16 17:23:55 the problem is that glibc is missing from RDEPENDS Sep 16 17:24:37 so I have to see who is adding glibc to RDEPENDS with the default toolchain Sep 16 17:24:42 and why it's not getting added with the linaro toolchain Sep 16 17:25:17 the piece of code that adds that dependency is do_package Sep 16 17:25:29 ok...where in do_package? Sep 16 17:25:31 based on extracting the linkage from the binary and using the shlibs files Sep 16 17:25:46 b*sh Sep 16 17:26:15 wrong window :) that was an emacs command Sep 16 17:26:41 let me double-check the shlibs Sep 16 17:27:13 https://www.growingwiththeweb.com/2017/03/mastering-vscodes-terminal.html Sep 16 17:27:14 right dinner Sep 16 17:27:36 kergoth: if you can write a vscode plugin to make bbclass work in the outline view i'd give you actual money Sep 16 17:27:51 shlibs look identican with and without linaro Sep 16 17:27:56 *identical Sep 16 17:28:10 that'd be nice. sadly not a js guy Sep 16 17:28:27 cat pkgdata/shlibs2/libsqlite3.list Sep 16 17:28:29 damnit, i need to fix meta-external-toolchain due to the libnsl split Sep 16 17:28:40 cat libsqlite3.list Sep 16 17:29:06 so where is the dependency on glibc added? Sep 16 17:37:31 marler89976: http://git.yoctoproject.org/cgit/cgit.cgi/poky/tree/meta/classes/package.bbclass#n1858 Sep 16 17:38:08 shlibs is not the problem Sep 16 17:38:48 the problem is that yocto is detecting that sqlite3 depends on libc.so.6, but glibc is not in the RDEPENDS_sqlite3 list Sep 16 17:39:03 RDEPENDS_sqlite3 is completely empty Sep 16 17:39:07 Yes, it is. The dependency from your sqlite package to the package providing libpthread.so.0 would normally be automatically added by this code, satisfying the QA check. For some reason that is not happening. Sep 16 17:39:23 RDEPENDS_sqlite3 should be empty when the recipe runs, the packaging step fills it appropriately Sep 16 17:39:23 oh? Sep 16 17:39:34 I'm printing it inside the do_package_qa code Sep 16 17:40:07 Without linaro I can see that glibc gets added to RDEPENDS_sqlite3, with linaro it isn't Sep 16 17:40:24 And in both cases, bitbake -e shows RDEPENDS_sqlite3 as empty Sep 16 17:40:41 That may be, for example, because the glibc package built by linaro doesn't have libpthread.so.0 in its shlibs providers Sep 16 17:40:50 it has the file Sep 16 17:40:56 let me check Sep 16 17:41:40 cat pkgdata/shlibs2/glibc.list Sep 16 17:41:44 librt.so.1:/lib:linaro-2.25 Sep 16 17:42:02 it's in there Sep 16 17:42:47 That's librt.so, not libpthread.so? Sep 16 17:43:23 that's glibc.list Sep 16 17:43:33 it provides both librt and libpthread Sep 16 17:44:00 Also, what makes you think the file-deps QA check uses RDEPENDS_sqlite3? It looks at the packaging input data written by do_package, see insane.bbclass around line 715 Sep 16 17:44:19 Well, if it provides both, but your glibc.list only contains one of them, that's your problem. Sep 16 17:44:28 right it's both Sep 16 17:44:48 glibc doesn't get added explicitly to RDEPENDS in any configuration Sep 16 17:44:55 it's getting added using setVar at some point Sep 16 17:45:00 so bitbake -e won't show it Sep 16 17:45:10 but I'm just not sure where Sep 16 17:45:13 IIRC it never gets added to RDEPENDS at any point. Sep 16 17:45:18 line 687 Sep 16 17:45:23 I print it and it's in there Sep 16 17:45:34 without linaro it gets added to RDEPENDS Sep 16 17:45:47 rdepends = bb.utils.explode_deps(localdata.getVar('RDEPENDS') or "") Sep 16 17:46:06 note that these lines are right after adding the pkg override Sep 16 17:46:16 so it's really getting RDEPENDS_sqlite3 Sep 16 17:46:42 I'm on thud right now so line number might be off Sep 16 17:47:05 Obviously something is adding glibc to RDEPENDS here, I just don't know where Sep 16 17:47:40 Well, the function name should tell you, as should the comment in the header of the file: "read_shlibdeps - Reads the stored shlibs information into the metadata" Sep 16 17:47:56 but glibc isn't in there Sep 16 17:48:01 no matter what package Sep 16 17:48:06 So as I have been trying to tell you, and other have been trying to tell you, something's broken with your shlibdeps detection Sep 16 17:48:06 I'm guessing it's special Sep 16 17:48:13