**** BEGIN LOGGING AT Tue Sep 15 02:59:57 2020 **** BEGIN LOGGING AT Tue Sep 15 03:25:22 2020 **** BEGIN LOGGING AT Tue Sep 15 05:48:42 2020 **** BEGIN LOGGING AT Tue Sep 15 06:39:35 2020 **** BEGIN LOGGING AT Tue Sep 15 06:55:18 2020 **** BEGIN LOGGING AT Tue Sep 15 07:54:01 2020 **** BEGIN LOGGING AT Tue Sep 15 08:43:12 2020 Sep 15 08:46:06 rburton: does https://autobuilder.yoctoproject.org/typhoon/#/builders/60/builds/2460 look familiar? :( Sep 15 08:46:22 armpit: this is why we don't give qemumips 512 memory :( Sep 15 08:49:55 hmm, qemu there is still half alive Sep 15 08:50:06 won't let me get to an ssh login though Sep 15 08:52:21 it will let me see dmesg, nothing there Sep 15 08:52:44 * RP wonders what sensible debug of this would look like Sep 15 09:13:55 RP: someone to understands mips needs to dive into the qemu memory implementation, i've still got a hunch something is not right Sep 15 09:14:41 rburton: I agree Sep 15 09:15:09 i wish i made notes from when i tried researching it a few years ago Sep 15 09:15:10 rburton: do we just disable core-image-sato-sdk and core-image-full-cmdline for now for systemd I wonder? Sep 15 09:15:33 there are odd memory layout restrictions in mips hardware Sep 15 09:15:48 rburton: it was the HIGHMEM stuff iirc Sep 15 09:15:55 yes Sep 15 09:16:11 hardware dates from the "who'd need 512MB memory?" **** BEGIN LOGGING AT Tue Sep 15 10:02:27 2020 **** BEGIN LOGGING AT Tue Sep 15 10:16:04 2020 **** BEGIN LOGGING AT Tue Sep 15 10:43:03 2020 Sep 15 11:59:33 I'm working on branch dunfell, and am having curious behavior with bind in recipes-communication, where configure fails due to two flags not being being supported: --disable-silent-rules and --disable-dependency-tracking. The configure script doesn't support them, and from what I can tell the recipe isn't adding them, so I am wondering how they Sep 15 11:59:34 could be magically showing up in the call to configure? Any ideas would be helpful. Sep 15 12:00:51 I have also witnessed this behavior on master and zeus. I don't want to infer this is experienced only on dunfell. Sep 15 12:08:00 sorry, that's ^^ recipes-connectivity, not enough coffee yet :p Sep 15 12:09:48 rpcme: bitbake -e bind and look for the "history" of the variable (the few lines above CFLAGS= (grep -e "^CFLAGS=")) Sep 15 12:12:58 rpcme: seems like those are set by default in autotools.bbclass Sep 15 12:14:12 Does it really fail though? I have this in my log.do_configure for bind on dunfell: "configure: WARNING: unrecognized options: --disable-silent-rules, --disable-dependency-tracking" Sep 15 12:15:06 yes I just saw that :) this is good stuff. I wonder what is going on then ... everyone uses this. maybe this is just a symptom of a larger underlying problem in my environment since I also get this: ERROR: bind-9.11.13-r0 do_configure: QA Issue: This autoconf log indicates errors, it looked at host include and/or library paths while determining Sep 15 12:15:07 system capabilities. Sep 15 12:15:15 you can add those to UNKNOWN_CONFIGURE_WHITELIST but if configure doesn't support them then its most likely not actually autoconf Sep 15 12:20:03 Thank you. I do a lot of yocto builds across architectures and whenever a vendor's distribution definition includes bind I run into this problem. It is the only package that I've had fail this way. I will need to keep looking. Sep 15 12:21:08 erbo thank you for confirming that you see it on your machine as well and it's effectively benign Sep 15 12:24:30 hi, is RDEPENDS able to pick up runtime dependencies based on DEPENDS or do you have to explicitely state every runtime dependency using RDEPENDS Sep 15 12:24:50 RDEPENDS will automatically include anything that your binaries link to Sep 15 12:25:13 rburton < thanks **** BEGIN LOGGING AT Tue Sep 15 12:27:53 2020 Sep 15 12:49:54 hello, anyone here using libmodbus with RS-485 on embedded Linux? would like to PM and ask questions about it (since perhaps this topic is not appropriate here) Sep 15 12:50:20 have not really found a channel where discussing libmodbus would be appropriate, though Sep 15 12:54:00 Hello guys. I-m looking to> https://github.com/Xilinx/meta-xilinx-tools/blob/master/recipes-bsp/fpga-manager-util/fpga-manager-util_1.0.bb Sep 15 12:54:35 what-s the purpose of the python (function/task?) Sep 15 12:55:14 when is it called ? before every task ? After all of the tasks ? In the middle of something ? Sep 15 12:56:09 ilkappe: on parse Sep 15 12:56:54 rburton, so before everything if I am right.. Is it correct ? Sep 15 12:57:06 it runs before a task is executed, yes Sep 15 12:57:38 rburton, thanks dude Sep 15 13:06:56 halstead: auto upgrade helper was unable to sent mail: https://autobuilder.yoctoproject.org/typhoon/#/builders/112/builds/9/steps/8/logs/step1d Sep 15 13:07:02 ERROR:Could not send email: [Errno 110] Connection timed out Sep 15 13:07:13 the mail server in use is smtp1.yoctoproject.org:25 Sep 15 13:07:18 RP: ^^^ Sep 15 13:09:02 kanavin_home can you update to mail.yoctoproject.org and try again? Sep 15 13:13:34 halstead: can you please disable oe-commits ML? Sep 15 13:15:49 For those who answered my questions about an hour ago - I changed base directories and it's configuring and compiling fine. On my workstation, I have 1TB mounted at /src, which is where I put all my source and build trees. Somehow, the bind recipe is detecting this as a host directory, even though /src is not in the FHS. Sep 15 13:17:39 halstead: yes, I'll try - thanks Sep 15 13:22:45 halstead: mail.yoctoproject.org seems to be unresponsive as well Sep 15 13:22:45 akanavin@ubuntu1804-ty-3:~$ nc -v mail.yoctoproject.org 25 Sep 15 13:22:45 ^C Sep 15 13:57:14 has anyone here managed to get meta-zephyr working? Sep 15 14:28:32 moto-timo: is naveen on irc? **** BEGIN LOGGING AT Tue Sep 15 14:48:44 2020 Sep 15 14:58:14 is there a technical call this week? Sep 15 14:58:43 tlwoerner: yes Sep 15 16:57:25 After further inspection, the issue with bind has nothing to do with the base checkout directory. bind 9.11.x recipe is def busted for me, where configure is finding datadir on my local machine and not in the basedir. on master, it works fine - the recipe is 9.16.x, changed a lot. Unfortunately jumping jellyfish (AGL) is using the busted one Sep 15 16:57:26 (for me) so I must find the root cause Sep 15 16:58:41 If anyone has deep experience in this recipe and/or is the person who changed from 9.11.x to 9.16.x, I would love to hear from you. I am basically starting from zilch to narrow exactly what needs to change Sep 15 17:27:31 dumb question. How can I set NATIVELSBSTRING? Sep 15 17:33:29 or, perhaps a better question, did the sstate cache directory structure change at all between 2.7 and 3.x? Sep 15 17:56:17 I believe the format of the sstate-cache directory has changed Sep 15 18:08:47 Is there a way to rebuild the sstate directory without cleaning everything? Sep 15 18:32:45 fullstop: that question doesn't make sense. Either it finds what it needs in sstate and things are fine and it doesn't need to rebuild, or it rebuilds it? Sep 15 18:33:01 RP: I'm in a strange situation, then. Sep 15 18:33:56 I have built an image using one machine, which migrated from 2.7 to 3.1.2.. Sep 15 18:34:40 and building from a second machine, using the first machine's sstate only finds about 35% of the objects and proceeds to build a bunch of stuff that shouldn't be needed. Sep 15 18:35:03 At first I thought that it was because of a NATIVELSBSTRING mismatch, but that is not the case. Both are "universal" Sep 15 18:35:24 I guess I can rebuild everything at night and see where things are tomorrow. Sep 15 18:41:33 fullstop: it sounds like the machines don't have as much overlapping state as you think. You probably need to compare the signatures between the two machines Sep 15 18:42:11 good news is you don't actually have to run the build to do it, "bitbake XXX -S none" for each machine and them look in tmp/stamps Sep 15 18:42:42 Will do. The git hashes for the layers are identical and there are no pending changes. Sep 15 18:42:56 Same versions of everything.. guess I'll find a difference somewhere. Sep 15 18:45:35 fullstop: you can also use openembedded-core/scripts/sstate-diff-machines.sh which calls bitbake -S none and helps a bit to compare the results Sep 15 18:53:47 odd.. Sep 15 18:54:24 https://pastebin.com/raw/4aN5tfFA Sep 15 18:54:40 Signatures are different, but the versions are the same Sep 15 18:56:54 use bitbake-diffsigs on them, but do_build will just point to some other task being different Sep 15 18:59:52 I don't know if this makes a difference or not, but this sstate setup was working fine before I pulled 3.1.2 on both sides.