[04:08] <RAOF> Bah! Test the new code, not a copy of the old code somewhere else!
[04:14]  * duflu takes notes
[04:19] <racarr_> namespace mb = mir::benchmarks
[04:19] <racarr_> WHYA NOT
[04:30] <duflu> racarr_: Because it's 9:30pm?
[04:34] <racarr_> duflu: Oh im migrating to burning man time.
[04:34] <racarr_> which is at least 3 time zones east of west coast
[05:33] <RAOF> Hm. RPATH!!!!!
[06:17] <RAOF> Alright. So, stripping out the rpath from our binaries fixes the doubly-included protobuf issue.
[06:19] <duflu> RAOF: Propose away
[06:20] <RAOF> I'm currently testing it manually :(
[06:21] <duflu> RAOF: Oh we have an rpath in released binaries? To /usr ?
[06:21] <duflu> Or rather non-released binaries
[06:21] <RAOF> CMake is all about the rpath.
[06:21] <RAOF> It embeds a lot of them.
[06:21] <duflu> I know :P
[06:21] <duflu> make install strips them
[06:21] <duflu> But that doesn't help CI much
[06:22] <RAOF> For example: lib/libmirclient.so: RPATH=/home/chris/Canonical/Mir/mir/private-library-loading/partial-armhf-chroot/lib:/home/chris/Canonical/Mir/mir/private-library-loading/partial-armhf-chroot/lib/arm-linux-gnueabihf:/home/chris/Canonical/Mir/mir/private-library-loading/partial-armhf-chroot/usr/lib:/home/chris/Canonical/Mir/mir/private-library-loading/partial-armhf-chroot/usr/lib/arm-linux-gnueabihf
[06:22] <RAOF> And rpath conveniently preempts basically all other ld behaviour.
[06:23] <duflu> RAOF: Except LD_LIBRARY_PATH I think
[06:23] <RAOF> Nope, preempts that too.
[06:23] <RAOF> At least empirically. I've got LD_LIBRARY_PATH exported, and the tests fail until I strip the rpath.
[06:26] <RAOF> Let's try to unrpath the CMake!
[06:27] <duflu> RAOF: CMake uses it for basic stuff like "make test". It might be hard to avoid completely
[06:27] <RAOF> Nope, that'll be easy to handle.
[06:27] <duflu> Kay
[06:27] <RAOF> We've already got the infrastructure required to set LD_LIBRARY_PATH
[08:17] <RAOF> Alright. And with that, EOW!
[08:23] <alan_g> Have a good weekend