[01:05] <RAOF> Hm.
[01:05] <RAOF> Is the_global_event_sink really unused?L
[02:36]  * RAOF once again wishes for a pervasive signals/slots mechanism in Mir.
[06:32] <anpok_> RAOF: hm signal and slot with an explicit mentioning of the dispatcher the caller will be executed with
[06:35] <tvoss> anpok_, easiest if the respective connection maintains the dispatcher
[06:42] <RAOF> I guess you could add the latter :)
[07:38] <RAOF> duflu: Why did you push the revert in r2759?
[07:39] <duflu> RAOF: Because it removed the bit that actually detected the regression. Also I have a much smaller and simpler solution (now proposed)
[07:39] <RAOF> The test you reverted does indeed catch the regression.
[07:39] <RAOF> It tested that the regression behaviour did not occur.
[07:40] <duflu> RAOF: OK, well if you found a new way to detect it, I'm sorry. Although you did the same to me. Objectively though, a better version is now proposed
[07:40] <RAOF> I proposed my solution in your merge proposal.
[07:40] <duflu> My test was also written when the bug still existed and failed prior to it being fixed. So I assume that's what you mean too
[07:41] <RAOF> Right. I tested my test by reverting your fix, and it does indeed fail.
[07:41] <RAOF> I disagree that your test is better; doesn't your test deadlock on failure?
[07:41] <duflu> RAOF: No..?
[07:41] <duflu> Oh it might have hung
[07:42]  * RAOF will also note that my version of the test went in after 3 approvals + CI.
[07:42] <RAOF> Right.
[07:42] <RAOF> My version fails rather than hanging :)
[07:42] <duflu> RAOF: Mine also got reviewed :)
[07:42] <RAOF> Your revert?
[07:43] <duflu> RAOF: Yes, the version I reverted to was approved.
[07:43] <duflu> No need to argue. Not important. I'll redo it
[07:43] <anpok_> hm alan_g i tried you gcc-5 fix branch..
[07:44] <anpok_> demo server is working .. not sure why every attempt to start a server fails on ci
[07:44] <seb128> hey there
[07:44] <alan_g> anpok_: yes?
[07:44] <seb128> what's the status of the new mir stucked in proposed in wily?
[07:44] <anpok_> seb128: there was one obvious abi bump missing
[07:45] <anpok_> seb128: but it did not fis the issue.. so I also bumped the server graphics platform abi
[07:45] <alan_g> anpok_: nor I (yet) - I'm going to try reproducing
[07:45] <anpok_> s/fis/fix
[07:45] <seb128> so it's fix commited now? more rebuilds needed?
[07:45] <seb128> let me know if you need with uploads or such
[07:45] <anpok_> seb128: rebuilding silo 4 now.. I walso wanted to include a gcc-5 fix
[07:46] <anpok_> but there is an odd ci failure..
[07:46] <seb128> k
[07:46] <seb128> which one?
[07:46] <anpok_> the test runner fails to start mirserver - https://jenkins.qa.ubuntu.com/job/mir-mediumtests-runner-mako/5979/console
[07:47] <seb128> k, no idea about that ...
[07:49] <duflu> anpok_: If you don't need to, please keep out gcc-5 support. There's always risk
[07:49] <duflu> And this late in the game we don't need more
[07:50] <alan_g> anpok_: please don't add anything that isn't needed. There will be another release along soon
[08:00] <duflu> RAOF: Sorry, your test is restored
[08:01] <RAOF> duflu: Ta. I was mostly just surprised to see the revert without MP.
[08:01] <alan_g> anpok_: From a first look at the failure I suspect that loading (and rejecting) the 0.13 platforms in a g++-5 build drags in a second copy of boost RTTI, You shouldn't delay 0.14 on getting that resolved.
[08:01] <duflu> RAOF: It happens. Nobody yet has uncleanly reverted something. We're pretty good at it
[08:02] <duflu> Also, we have "append-only" set. So "uncommit" is disallowed, for our own protection
[08:03] <duflu> alan_g: I think the head of lp:mir/0.14 (from a few minutes ago) might fix your issue
[08:04] <anpok_> ci is using vivid+o makos or wily makos?
[08:04] <alan_g> anpok_: V+O
[08:05] <duflu> It shouldn't matter if your ABIs are right, should it?
[08:05] <anpok_> -> whats the right solution for that failure: https://ci-train.ubuntu.com/job/ubuntu-landing-004-1-build/275/console
[08:06] <duflu> anpok_: Merge lp:mir/ubuntu back to lp:mir/0.14?
[08:06] <anpok_> but there is no changelog entry yet
[08:06] <duflu> Ah, that old problem
[08:06] <duflu> anpok_: Then download it :)
[08:06] <anpok_> and add it to mir/0.14 and add another one?
[08:07] <duflu> anpok_: Yep.... https://launchpad.net/ubuntu/+source/mir
[08:07] <duflu> Distro needs a continuous timeline. You shouldn't pretend the first attempt didn't happen
[08:13] <duflu> Debs are messy. Can we have something new now?
[08:13] <duflu> :)
[08:15] <RAOF> Urgh! What the hell is happening here: http://s-jenkins.ubuntu-ci:8080/job/mir-clang-ts-wily-amd64-build/69/console ?
[08:16] <duflu> VPN says no
[08:16] <RAOF> I can't reproduce that locally; *my* threadsanitiser make test works.
[08:17] <duflu> tsan with clang presumably?
[08:17] <RAOF> Indeed.
[08:17] <duflu> Or is the "clang" a misnomer in the task?
[08:18] <duflu> Also.. for some value of "wily"
[08:18] <RAOF> Well, my wily is up to date as of today; the builder should be too (as it's pulling from the archive)
[08:18] <duflu> For some value of "today"
[08:19] <RAOF> :)
[08:20] <RAOF> I suspect that it's got to do with the fork() involved in the death-tests.
[08:21] <RAOF> Because at least one of the failures is talking about the main thread of ClientLibraryErrorsDeathTest_createing_surface_on_garbage_connection_is_fatal racing with the RPC thread created by ApplicationNotRespondingDetection_responding_client_is_not_marked_as_unresponsive
[08:22] <RAOF> (And now that typo has crossed my threshold of annoyance ☺)
[09:01] <RAOF> anpok_: You're familiar with the vagaries of ThreadSanitizer, right?
[09:02] <anpok_> not really
[09:03] <anpok_> I even fail to look up vagaries in the dictionary
[09:04] <RAOF> :)
[09:16] <anpok_> RAOF: but you can try treating me like a rubber duck - whats wrong?
[09:18] <RAOF> ThreadSanitiser complains about an RPC thread started from ApplicationNotResponding's acceptance test racing with mir_connect_sync from ClientLibraryErrorsDeathTest.
[09:20] <RAOF> From a heap block allocated from the ANRDetection test.
[09:20] <RAOF> And the ANR detection tests start after the ClientLibraryErrorsDeathTests have already finished.
[09:21] <RAOF> So I'm suspicious that the problem is actually that the fork() in the death tests is confusing ThreadSanitizer.
[09:37] <alan_g> anpok_: alf_ hangout?
[09:38] <alf_> alan_g: coming
[09:39] <anpok_> now
[12:53] <mterry> RAOF, poke about subscribing mir-team to abi-compliance-checker
[13:03] <RAOF> mterry: Subscribed as of a couple of hours ago.
[13:04]  * mterry hugs RAOF
[13:04] <RAOF> Sorry about the delay.
[13:04]  * RAOF was on holiday last week, too.
[14:45] <alan_g> camako: with mir-on-x I ought to be able to see clients in a desktop window. Right?
[14:47] <anpok_> alan_g: the server in a x-window
[14:48] <alan_g> I see an X window (and titlebars painted in the server) but nothing from the client
[14:48] <camako> alan_g, is it an intel gpu?
[14:49] <alan_g> camako: yes
[14:49] <camako> alan_g, try a non egl client
[14:49] <camako> intel's gpu has support problems for the extension I am using
[14:49] <camako> for egl clients only though
[14:49] <alan_g> camako: Ah! multiwin works
[14:50] <camako> \o/
[14:50] <camako> alan_g, egl clients work but do not render.. you can try eglplasma and hit 'q'.. it'll quit fine.
[14:51] <alan_g> Confirmed
[14:52] <alan_g> Is mouse input supposed to work?
[14:55] <ogra_> nah, it is unity8 ... its all gestures (you need to know the secret handshake to log in) :)
[14:57] <alf_> alan_g: no, mouse is not supported yet
[15:29] <racarr> Morning!
[16:10] <kgunn> racarr: o/
[16:13] <racarr> kgunn: Howdy :)
[17:39] <racarr> kdub: Can you rereview client-specifie-dinput-regions when you get a chance I had to repropose it with a dependency
[17:39] <racarr> (symbols.map stuff)
[17:39] <racarr> no code changes
[17:39] <kdub> sure
[17:43] <racarr> kdub: Same thing for relative-pointer-events...had to resubmit without prereq :)
[17:43] <racarr> thanks