**** BEGIN LOGGING AT Thu Oct 31 02:59:58 2019 Oct 31 07:13:26 Good morning. I checked out the new poky zeus (3.0) and got problems during the build of my image: Package "icu" fails at do_configure with this message: https://pastebin.com/im2277qS Oct 31 07:13:33 Has anyone else same problem? Oct 31 08:34:45 I'm trying and then wanna use it as but VAR is true in all cases. Oct 31 08:36:08 shell script and bitbake shell function have different behavior here, I'm able to perform desired operation in Linux shell script. any idea what's different for bitbake shell function e.g do_compile_append() Oct 31 09:12:05 hey guys! I need to replace git repos mentioned in SRC_URIs of a multitude of packages. SRC_URI often has patches etc, so I can't just reassign the new address, neither I want to copy-paste all details. Is there a way to just have something like SRC_URI-pn-...-replace = s//g? Oct 31 09:12:20 within localconf Oct 31 09:14:47 I guess I need to put some PREMIRRORS magic then Oct 31 09:15:56 luneff: premirrors might be a better solution Oct 31 09:16:05 or just use _remove then _append Oct 31 09:16:38 yes, I'll try PREMIRRORS. I need to build GenIVI circa 2015 and they've migrated everything to github since. Fingers crossed Oct 31 09:48:04 hmm... the cross-canadian.bbclass is the class used for toolchain packages that target the target-machine, and that run on the sdk host Oct 31 09:49:56 I don't understand what the anonymous python function at the top of the class file does. if MODIFYTOS is not "1" (the default), it will do something to TARGET_OS and TARGET_VENDOR based on TARGET_ARCH Oct 31 09:50:24 s,not "1","1", Oct 31 09:52:07 if d.getVar("MODIFYTOS") != "1": Oct 31 09:54:11 i suspect its the bit where it hardcodes linux that is bothering you right Oct 31 10:02:50 rburton, yeah, but I'm not even sure what exactly the code does after that 'if'. The stuff before the 'if' appear to modify the PACKAGE_ARCHS so that all of them have the SDKPKGSUFFIX appended. PACKAGE_ARCHS is "all any noarch ${PACKAGE_EXTRA_ARCHS} ${MACHINE_ARCH}" and nothing appears to change MACHINE_ARCH in cross-canadian Oct 31 10:03:37 so PACKAGE_ARCHS will for example afterwards contain ${MACHINE_ARCH}-${SDKPKGSUFFIX} . so the gcc-cross-canadian recipe will declare it's compatible with that package-arch. Oct 31 10:03:52 but I'm not sure what purpose the code after that MODIFYTOS-if has Oct 31 10:05:28 for example d.setVar(n, d.getVar(n)) appears to prematurely expand variable 'n' (PROVIDES and DEPENDS). but why does it do that? supposedly because afterwards it changes TARGET_OS and DEPENDS or PROVIDES might contain references to TARGET_OS that are not supposed to follow the change? Oct 31 10:07:31 all of this must have to do something with the '${LIBCEXTENSION}${ABIEXTENSION}' suffix that is set in TARGET_OS by default in bitbake.conf, because cross-canadian changes TARGET_OS to not contain those anymore Oct 31 10:08:52 regarding yocto CVE talk: Linux kernel CVE data link https://github.com/nluedtke/linux_kernel_cves/tree/master Oct 31 10:12:42 rburton, the cross-canadian tools are supposed to build their things so that the result are compatible with the image. so it appears to make no sense that cross-canadian changes anything about TARGET_OS. it makes sense that it changes HOST_OS for sure. Oct 31 10:17:54 Hi guys, I'm having trouble working with hostnames and avahi! I have 2 boards with different hostnames; if I try to ping both of them from my PC the hostname works (ping board1.local or ping board2.local)! But if I try to ping one board from the other, ping fails with: ping: bad address 'board1.local'.Do you have any suggestions to fix this issue? Oct 31 11:06:26 hey, I've question rather about u-boot, but hopefully someone here has the knowledge Oct 31 11:06:29 Hey, I just need some clarification - when no uboot.env is present on boot partition, it creates it's default based on what happened in uboot.scr at boot? and second question - boot.env *always* is 16K and *mustn't* be modified by hand (question arised because of non-literal characters in file)? Oct 31 11:07:05 kpo: its not even exactly u-boot specific, but u-boot-on-specific-hardware-specific Oct 31 11:07:42 kpo: so i would actually suggest you ask over at #u-boot, and name your specific board in use Oct 31 11:07:58 ok, thanks :) Oct 31 11:09:24 kpo: good luck! Oct 31 11:16:24 New news from stackoverflow: How to override a .conf file in Yocto || Yocto: remove packageconfig item Oct 31 12:15:55 hi folks, im writing a blog post on how to set the gcc version in yocto, does anyone fancy taking a look to check if i got it correctly? https://gist.github.com/milloni/0018e810c61694f80c4421689b30dbb2 Oct 31 13:04:44 rburton: Is that a new QA test? Oct 31 14:07:17 why are there two Variable glossaries in the mega manual? Oct 31 14:07:42 litb: because more is more Oct 31 14:08:01 one in Chapter Chapter 45. and one in Chapter 33. Oct 31 14:08:07 Oct 31 14:08:20 litb: creation failure probably Oct 31 14:08:21 BB_GENERATE_MIRROR_TARBALLS is listed in two ones with different documentation Oct 31 14:08:25 litb: can you file a bug with links Oct 31 14:08:34 JPEW: one i hacked up locally but yes Oct 31 14:09:09 JPEW: the idea of a pod2man wrapper isn't super trivial, as you can get it to write to stdout. Oct 31 14:09:23 JPEW: so a wrapper would have to emulate pod2man arg parsing, intercept the output, etc. Oct 31 14:09:34 doable but complex Oct 31 14:09:48 rburton, yeah I will file Oct 31 14:09:59 JPEW: alternatively, we fixup the manpages in do_package. complicated by the fact that man pages could be compressed. Oct 31 14:10:24 JPEW: the easy solution is to depend on perl-native everywhere but that's horrid. Oct 31 14:10:36 rburton: Right Oct 31 14:11:05 rburton: I'd prefer to fix it at the source (pod2man) rather than try to munge the pages post-mortem Oct 31 14:12:08 presumably upstream won't take a patch to remove it from the output considering the reproducible builds people don't care for host environment changing yet Oct 31 14:14:27 rburton, I sent a patch that tests ability to create gpl3-free images - suggestions for what other images could be tweaked and tested? Oct 31 14:14:36 awesome will look Oct 31 14:14:58 I picked minimal and full-cmdline, because that is what meta-gpl2 is tested with Oct 31 14:15:13 JPEW: the easiest thing would be to rip out the pod2man arg parsing code directly but i hate perl so violently i don't think i could manage that Oct 31 14:15:58 maybe i'll write a dumb wrapper in py Oct 31 14:16:06 port the arg parsing to argparse Oct 31 14:16:28 and switch out the output to a temporary file so we can munge it then either write to the right place, or output on stdout Oct 31 14:18:30 rburton: Nothing fullfills a hatred for perl more than replacing it with python. Cackling while you do so doesn't hurt either :) Oct 31 15:02:31 JPEW: well https://pastebin.com/ucGVx03Z isn't *too* horrible and nearly complete Oct 31 15:03:19 hm could use an argparse action to write the string as it is parsed Oct 31 15:04:07 rburton: Can use use parser.parse_known_args() instead? Oct 31 15:05:30 rburton: https://docs.python.org/3/library/argparse.html#argparse.ArgumentParser.parse_known_args Oct 31 15:05:41 ah that might work Oct 31 15:05:45 didn't know of that Oct 31 15:06:23 rburton: It's *really* helpful for these types of wrapper scripts Oct 31 15:06:58 *so close* Oct 31 15:07:08 eg opkg-utils does -r '' Oct 31 15:07:14 but this gets eaten and turned into just -r Oct 31 15:07:42 yeah because it doesn't know -r takes an option Oct 31 15:08:12 pod2man has an mandatory input and optional output argument Oct 31 15:08:24 so i'm trying having just those setup and then the parser Oct 31 15:08:47 but it feeds input as '' and output as what i put as input, then the actual output becomes a unknown argument Oct 31 15:08:49 so, no. Oct 31 15:16:01 Is it just '-r'? Oct 31 15:16:10 that takes an argument? Oct 31 15:35:30 no Oct 31 15:35:33 most do Oct 31 15:35:37 and can be '' Oct 31 15:40:55 all these years and i'm not sure i even knew we had a meta-skeleton Oct 31 15:43:02 spooky Oct 31 18:01:31 JPEW: a working thing in ross/mut now Oct 31 18:01:38 about to do a world build to see if it breaks Oct 31 18:01:45 cool Oct 31 18:01:48 it was a fun forkbomb with the wrapper calling itself Oct 31 18:01:57 lol Oct 31 18:02:36 hosttools will happily link to scripts, so even though i was removing scripts from PATH before searching, it found itself still Oct 31 20:17:51 New news from stackoverflow: how to get the manifest of an image before building the image Oct 31 21:07:17 nice, after a couple small tweaks and one workaround, and adding the coreutils+gnu-sed gnubin to the PATH, pyrex mel builds work on my macbook Oct 31 21:08:18 of course, had to use a case sensitive disk image too, but that's to be expected Oct 31 21:14:42 mel is mentor linux? Oct 31 21:14:48 what is pyrex? Oct 31 21:18:37 yep. pyrex is one of those wrapper tools for running bitbake in docker Oct 31 21:18:43 https://asciinema.org/a/9n4XXJwMzDX2e02bOJMkF54ef Oct 31 21:18:49 er, wrong link, though related Oct 31 21:18:52 https://github.com/garmin/pyrex Oct 31 21:18:53 there Oct 31 21:19:17 ah, so the builds happen in a linux container Oct 31 21:19:43 yeah Oct 31 21:20:28 unlike crops, it's transparent, doesn't muck with your workflow. i was using custom scripts for it, but pyrex has a .ini to configure what paths fromt he host get mounted in the container and the like, which is convenient Oct 31 21:20:37 nice Oct 31 21:20:45 https://github.com/kergoth/meta-mentor/blob/pyrex/conf/pyrex.ini.sample Oct 31 21:29:02 yay pod2man reproducibility wiped out Oct 31 21:41:35 adelcast: nice one about verification. Oct 31 21:43:35 maybe show a warning (guard behind -v 2+) if the checksum file is missing Oct 31 21:45:20 why not use sha* as well? (i found file_sha256sum_alloc() at least) Oct 31 21:58:18 ^Piraty: I try to model this after what debian does. Currently they only support md5 as the feature is meant to be used as a way to determine if you messed up with installed files. Is not meant to be a cryptographically secure way to detect tampering. Oct 31 21:58:51 debian tools for auditing, etc, now should work with opkg Oct 31 21:59:47 regarding warnings, currently debian doesn't enforce that all packages have checksums. Conffiles are an example were you explicitly don't want to have md5sums. Oct 31 22:00:17 basically, its up to the package builder to decide, but there are tools to snapshot the system by creating md5sums of all installed packages Oct 31 22:00:39 why not surpass dpkg in this ? i mean debian clearly isn't i wise yardstick to choose Oct 31 22:00:52 then with some bash you can see if you are off the golden standard and reinstall the packages that have files that changed Oct 31 22:01:41 so I debated regarding sha256....I was kind of the fence to see if I should just support .sha256sums files too Oct 31 22:01:47 i know, i use that alot with the packagemanager in my distro Oct 31 22:02:02 I don't think I am opposed to it, and it should be simple to implement I think Oct 31 22:04:05 still think that tripwire and similar tools should fill the tampering detection hole, but should be simple to have sha256 too Oct 31 22:08:25 is /dev/shm still a thing? Oct 31 22:25:16 rburton: it is on my system ? Oct 31 22:25:33 same here but wondering if thats historical Oct 31 22:25:41 appears to be used by glibc still for some apis Nov 01 00:18:29 New news from stackoverflow: Unable to start bitbake server **** ENDING LOGGING AT Fri Nov 01 03:00:13 2019