[01:33] <RAOF> In today's fun compiler-bug feature, clang-3.5 doesn't correctly destroy cross-thread capture-by-copy variables in lambdas!
[06:56] <duflu> ping mlankhorst
[07:24] <mlankhorst> pong
[07:33] <RAOF>  Would you prefer Factory? :)
[07:46] <anpok> good morning mir
[08:37] <mlankhorst> also good morning
[09:01] <mlankhorst> duflu: ?
[09:02] <duflu> mlankhorst: I've found a situation where drmModePageFlip never completes. Is that allowed?
[09:02] <mlankhorst> when, exactly?
[09:03] <duflu> mlankhorst: Schedule page flip (but don't wait), put screen to sleep
[09:03] <duflu> It might be coming on wakeup, but Mir crashes before then :P
[09:04] <duflu> Hmm, it was a while ago. Let me retest
[09:04] <mlankhorst> hm interesting, never checked that. :P
[09:04] <mlankhorst> i guess it might depend on the driver
[09:04] <duflu> mlankhorst: Yes, twas intel
[09:05] <mlankhorst> did you write a testcase? I can make one probably
[09:05] <mlankhorst> would need the arguments to drmModePageFlip though
[09:09] <duflu> mlankhorst: I need to retest to refresh my memory. I got distracted since the ping
[09:36] <mlankhorst> :/
[10:24] <duflu> alf_: You did overlapping DisplayBuffer construction/destruction to minimize visual interruption on config changes, right?
[10:28] <alf_> duflu: How do you mean overlapping?
[10:29] <duflu> alf_: When we do config changes... the old DisplayBuffers still exist when the new ones are constructed. So for a while you have 2 DisplayBuffers pointing to each CRTC
[10:29] <duflu> .. which is safe right now. But not for my future changes
[10:31] <alf_> duflu: I don't remember exactly now, but I think I had a problem when deleting the existing display buffers first, but this was in the first steps so it may not be a problem still
[10:32] <duflu> alf_: Yes deleting the old ones first solves my problem. But the "black screen pause" during config changes is much longer
[10:34]  * mlankhorst pokes duflu again
[10:34] <duflu> mlankhorst: Looking more like it might be a Mir bug. I won't bother you again till I'm more sure
[10:34] <mlankhorst> okay
[10:34] <mlankhorst> I thought as much :)
[10:40] <duflu> alf_: Whether intentional or not, overlapping the new and old DisplayBuffers seems useful. It makes mode changes appear more instantaneous with much less black screen time.
[10:40] <duflu> I will try to keep it that way
[10:41] <duflu> Although it would be nice if mesa::Display was smarter and didn't reconstruct DisplayBuffers at all
[10:41] <duflu> That would require more work
[10:45] <duflu> Awesome. After all that I achieved a two-line fix today. But at least it's fixed
[11:12] <Saviq> hey folks, do you have some machinery to generate the commit message for merging from devel to trunk?
[11:12] <Saviq> as we're looking at lp:unity8/devel for the new release process :/
[11:15] <alan_g> Saviq: duflu and kgunn are the guys to ask. I think kgunn had a script - but I'm not sure it's bug free.
[11:15] <Saviq> alan_g, thanks ;)
[11:16]  * alan_g lost interest the Nth time the process changed
[11:37] <alan_g> anpok: alf_ any opinions on how to implement parse_name() in https://code.launchpad.net/~alan-griffiths/mir/some-tidy-up-of-config-options/+merge/203384?
[11:54] <anpok> hm
[11:55] <anpok> {name.begin(), name.find_first_of(',')}
[11:55] <anpok> but then name or begin as it is called now, needs to be a string
[11:56] <alan_g> yes
[11:56] <anpok> first time I saw strcspn
[11:57] <alan_g> That's why I'm unconvinced it is clearer
[11:59] <anpok> so even with two methods and the short puzzler of substr kdbus proposal is faster to read
[11:59] <anpok> for the average mir code digester
[12:04]  * alan_g had to look up what substr(0, std::string::npos) does - the string functions are a minefield of strange indexing rules
[12:05] <alan_g> But if everyone else thinks it clearer I guess we'll do it hte consensus way
[12:11] <anpok> hm i am not everyone :)
[12:14] <alan_g> Well put your opinion in a review and I'll tally the votes
[15:04] <alf_> alan_g: Setting a new function for mir_connect_impl for a test will affect all tests that follow too, right?
[15:06] <alan_g> alf_: I seem to remember some mechanism to reset it. (But not any detail)
[15:07] <alf_> alan_g: ok, thanks, I will take a deeper look
[15:46] <ricmm> do you guys know
[15:46] <ricmm> if there is some kind of timeout
[15:46] <ricmm> for session authorization?
[15:46] <kgunn> alan_g: ^ ricmm 's query
[15:47] <alan_g> not in Mir
[15:47] <alan_g> ricmm: ^
[15:48] <ricmm> ok
[15:48] <ricmm> thanks
[19:59] <kdub_> do we have a place for code shared between the tests and the examples?