**** BEGIN LOGGING AT Wed Nov 23 03:00:00 2016 Nov 23 12:47:26 khem: I know you have somehow been playing with compilers, did you try to get oe to generate a baremetal aarch64 toolchain? Nov 23 12:47:51 it seems that insane.bbclass has to be patched Nov 23 20:12:03 Hello. I see on http://www.openembedded.org/wiki/Features that you have reproducible builds. How are you achieving that? Nov 23 20:14:58 lamby: they aren't reproducible in the "exact binaries" sense Nov 23 20:15:39 reproducible as in using the same input we get the same output within a reasonable way.. date stamps, embedded (build) paths, etc may change.. Nov 23 20:15:53 but you can do binary diffs on product executables and they are the same (otherwise) Nov 23 20:16:00 but since we are building the cross compiler & toolchain, and a good portion of the native tools, there's less chance of anything external leaking in Nov 23 20:16:15 you can compare build logs, etc and show (other then those items) they are the same from one build to the next (with the same config/input) Nov 23 20:16:17 (that would change the output in a material way) Nov 23 20:16:41 lamby: check https://bugzilla.yoctoproject.org/show_bug.cgi?id=5866 Nov 23 20:17:51 it would be great to fix at least the issues which cause binary size differences Nov 23 20:18:16 then it will be easier to spot the bigger differences in e.g. files-in-image.txt reports Nov 23 20:18:24 Thanks :) Nov 23 20:18:36 (I work on https://reproducible-builds.org/ in general) Nov 23 20:19:13 (Where we use reproducible in the exact binary sense. Silly terminology…) Nov 23 20:26:08 lamby: there are efforts to be proper reproducible, and we have magic to strip exact source paths from eg debug info to reduce the delta Nov 23 20:26:43 -fdebug-prefix-map etc Nov 23 20:27:26 in theory OE would be a good way to experiment with reproducible, as we always build our own gcc and forcibly control every compiler flag Nov 23 20:27:37 just need someone with motivation Nov 23 20:32:18 re. -fdebug-prefix-map, have you seen https://gcc.gnu.org/ml/gcc-patches/2016-11/msg00182.html ? Nov 23 20:33:17 curl in being a bit stupid shocker Nov 23 20:33:41 It's nowhere near just curl :) Nov 23 20:34:19 well, we're mainly using it to stop embedding build paths in binaries so blatently Nov 23 20:34:42 if the paths are relative to the target prefix you can relocate them trivially to a local sysroot Nov 23 20:39:08 gotta go, if you want to talk about this more then both me and RP are interested (and in the UK) Nov 23 20:45:11 :) Nov 23 20:47:12 lamby: I'm the RP who just asked about the Yocto Project getting involved in reproducible builds Nov 23 20:48:21 lamby: we've been working at this in our own way for a while but happy to work with others. We actually want to go one step further than some of what you want to do as we want the binaries to be identical regardless of the path we build them in for example Nov 23 20:58:09 RP: hey :) Nov 23 20:58:31 RP: Just to correct you there (:p), we want the binaries to be identical regardless of the path… Nov 23 21:29:16 lamby: hmm, that wasn't what I inferred from your website Nov 23 21:29:44 lamby: I struggle for time as I juggle many things but I do want to find a way to work together on some of this Nov 23 21:40:35 RP: If so, it's likely out of date. In the Debian Reproducible Builds team we temporarily fixed the build path, but we now don't do that. This could be source of the confusion. Nov 23 21:40:56 If you are interested, please join #debian-reproducible and/or #reproducible-builds on OFTC. **** ENDING LOGGING AT Thu Nov 24 03:00:00 2016