**** BEGIN LOGGING AT Mon Oct 31 02:59:56 2022
Oct 31 04:02:38 <zwelch> i'm trying to compile `jim` in kirkstone, so i can use it instead of the builtin verison to openocd.  interestingly, the cross version builds fine, but the native version fails because the jim bits do not make it into the sysroot for openocd. They needed files are installed in ${WORKDIR}/image for jim-native, but they do not appear to make it past that spot. I feel like I'm missing something simple and obvious
Oct 31 09:23:51 <kroon> RP, bluelightning, can I ask what your thoughts are about splitting up the buildhistory FILELIST entries into one file per line ? I'm asking because I think it makes reviewing the git history diff easier
Oct 31 09:27:47 <kroon> for instance the glibc-src package, that line is ~190000 characters long
Oct 31 09:58:18 <rburton> zwelch: share the recipe?
Oct 31 14:59:28 <zwelch> rburton: https://github.com/openembedded/meta-openembedded/compare/kirkstone...zach-welch-aquabyte:meta-openembedded:kirkstone
Oct 31 15:05:48 <rburton> zwelch: inherit autotools then delete your EXTRA_OECONF lines, do_compile, do_install, and change do_configure to be a prepend which just does the bootstrap line.  you're basically badly reimplementing autotools.bbclass.
Oct 31 15:06:15 <rburton> also LIC_FILES_CHKSUM is relative to S so you can just do file://LICENSE
Oct 31 15:09:57 <zwelch> i tried interhiting autotools, but it gave me other trouble
Oct 31 15:10:03 <zwelch> i'll give it another stab
Oct 31 15:10:24 <rburton> you can tweak how autotools works, much easier than reimplementing it
Oct 31 15:13:48 <rburton> zwelch: also no need to depends on binutils
Oct 31 15:30:40 <rburton> zwelch: if the problem is that jim doesn't work with out-of-tree builds then inherit autotools-brokensep (and file a bug with jim)
Oct 31 15:31:38 <rburton> zwelch: so the problem is that you're hardcoding a prefix of /usr but that isnt the right prefix for native or nativesdk
Oct 31 15:31:51 <rburton> as i said, don't reimplement stuff, you'll do it wrong :)
Oct 31 15:32:45 <zwelch> rburton: well, i like the look of the simpler recipe, but it fails in exactly the same way. (Pushed for your consideration.)  However, now, the sysroot-destdir for the jim-native build is bonkers; a single file gets installed therein: `$workdir/sysroot-destdir/$workdir/sysroot-recipe-native/sysroot-providers/jim-native`    yes, the full path to workdir gets installed
Oct 31 15:33:48 <zwelch> and, yes, autotools-brokensep is required
Oct 31 15:34:00 <zwelch> (or at least seems to be)
Oct 31 15:36:55 <zwelch> one detail that probably is confounding things: the configure script is decidedly not autotools; jim has its own custom "autosetup" scripts, written in TCL.
Oct 31 15:37:11 <rburton> oh if its not actually autotools then good luck with that
Oct 31 15:37:18 <zwelch> heh, indeed
Oct 31 15:37:27 <zwelch> thus my ditching the autotools class
Oct 31 15:37:54 <rburton> i saw configure.ac and assumed it was autotools
Oct 31 15:38:10 * rburton puts this on his list of software to hard glare at
Oct 31 15:38:42 <zwelch> well, i went down this rabbithole because openocd-native does not build for me in stock kirkstone
Oct 31 15:39:40 <zwelch> i'd be curious to hear if it builds for others (and thus my failure is an artifact of my environment) or whether it is broken for everyone
Oct 31 20:17:43 <bluelightning> kroon: good idea - I think it would be sensible to split it out to a separate file alongside. We sort of have that in files-in-package.txt, but I believe the latter doesn't get written out when restoring from sstate (which probably should be filed as a bug if it isn't already)
Oct 31 20:18:57 <bluelightning> not to muddy the waters, we can consider that as a separate issue and still have FILELIST split out to a separate file e.g. files-in-package-names.txt with one per line
Oct 31 20:27:55 <rburton> buildhistory doing what pkgdata has done and moving to a machine-readable format might be nice too
Oct 31 20:31:27 <rburton> (pkgdata can be emitted as json)
Oct 31 21:01:23 <kroon> bluelightning, rburton, ok. I was worried people cared about some sort of backwards-compatibility in the buildhistory output, but maybe that is not such big of a concern
Oct 31 21:15:33 <zwelch> rburton: i figured it out after digging into the bowels of oe. I've pushed my latest patches to github for your consideration, but i'll submit them to the oe-core list at some point.
Oct 31 21:18:30 <bluelightning> rburton: the current format is reasonably machine readable ;)
**** BEGIN LOGGING AT Mon Oct 31 21:41:32 2022
**** ENDING LOGGING AT Tue Nov 01 02:59:56 2022