[03:52] <RAOF> Man, I _really_ wish that libprotobuf didn't have an __init section.
[04:34] <duflu> RAOF: Yeah I noticed too... https://code.launchpad.net/~vanvugt/mir/fully-defined/+merge/234261
[04:35] <RAOF> Incidentally, isn't https://code.launchpad.net/~vanvugt/mir/version-drivers/+merge/234278 fully subsumed by https://code.launchpad.net/~raof/mir/versioned-library-loader/+merge/234062 (at least, once the latter lands!)
[04:36] <duflu> RAOF: No, you need versioned package names as well as library names to resolve bug 1293944 (which is blocking any further bumping of the server ABI)
[04:36] <RAOF> Oh, bah, sorry, wrong MP: https://code.launchpad.net/~raof/mir/privatise-all-the-things/+merge/234063 is what I meant.
[04:37] <RAOF> Man I wish CI didn't take so long.
[04:37] <RAOF> Or, alternatively, that I could easily reproduce the CI environment locally.
[04:37] <duflu> RAOF: Don't know. I'm ignoring it on size alone right now. That and I got proof last night my smaller branch sufficiently unblocks CI \o/
[04:38] <RAOF> Oh, yeah, it will.
[04:38] <RAOF> It's a bit wrong, though :)
[04:38] <RAOF> The binding between the driver and the platform isn't libmirplatform's SOVER; it's just the ABI defined by what gets dlsym'd by libmirplatform.
[04:39] <duflu> Madness madness everywhere. We just have to stop adding features for a year or two. Clean up what we have, then carry on :)
[04:41] <RAOF> We actually _have_ a defined ABI for that, though :)
[04:41] <RAOF> And it hasn't changed over the last 6 months.
[04:42] <duflu> Except the server ABI, that changes every week or two :)
[04:43] <RAOF> Right, but the platform drivers don't care about that.
[04:44] <duflu> OK, platform-api is now on my list. It won't ever get forgotten in SDK testing again
[04:44] <duflu> I really didn't expect platform-api to be full or Mir server header usage
[04:48] <RAOF> Oh.
[04:48] <RAOF> We've overprivatised  things.
[04:50]  * RAOF should do an out-of-tree X11 platform module. It'd make greyback happy.
[04:58] <RAOF> In fact...
[04:58] <RAOF> I'm tired and annoyed. Let's call that a Friday Labs project and have a play.
[04:59]  * duflu -> lunchtime sunshine
[05:17] <RAOF> HURRAY!
[05:17] <RAOF> CI finally passes!
[07:30] <RAOF> Ok, so why do we have PendingCallCache::force_completion(), and why do we think it might do something useful?
[07:32]  * duflu shrugs. It's undergone at least one rewrite since I last looked there
[07:35] <RAOF> I'm not entirely sure how “let's call the function that handles the reply, except handing it a non-reply” is supposed to be a good idea :)
[07:39] <RAOF> _Particularly_ since it's not clear that it's possible for the reply handler to determine whether or not the reply object is a reply or not.
[07:41]  * duflu shrugs again. Trying to not get caught up in new topics on a Friday afternoon