[03:37] <tjaalton> 4/win 21
[03:37] <tjaalton> bah
[03:59] <tjaalton> hmm, either mesa 9.1 has no bugs, or noone tested it
[04:02] <bryce> heh
[04:02] <ScottK> Not testing is the more sure way to have no bugs.
[04:03] <tjaalton> yep :/
[04:04] <tjaalton> ScottK: do you have intel gfx btw, and if so mind giving it a go with kwin?
[04:04] <ScottK> I do, but the relevant machine is dead ATM.  Need to do a reinstall first.
[04:04] <tjaalton> ah, ok
[04:04] <ScottK> (not happening tonight as it's almost time for bed)
[04:05] <tjaalton> sure thing
[04:11] <RAOF> I've been using the Mir mesa for some time - that's based on 9.1-rc2 and so suffers from everyone's favourite terrible dash performance, but aside from that everything works.
[06:12] <tjaalton> RAOF: yeah, and that's fixed on the ppa one
[06:14] <mlankhorst> morning
[06:15] <mlankhorst> tjaalton: hey I run mesa 9.1 here, though not the official packaged version, just the cherry picked git snapshot
[06:27] <mlankhorst> bryce: did you upload ubuntu5 xserver?
[06:30] <bryce> no
[06:31] <mlankhorst> ok good :)
[06:36] <tjaalton> hmm yeah I think we could just update mesa to the 9.1 snapshot
[06:38] <tjaalton> and 9.1.2 coming next friday-ish
[06:38] <mlankhorst> just keep what's there now and get it in ubuntu first..
[06:38] <tjaalton> ok :)
[06:40] <tjaalton> I'll just add e6616948b
[06:48] <mlankhorst> looks good to me
[06:48] <tjaalton> pushed
[07:38] <mlankhorst> I'm going to test the drm_device_keep_trying on panda first before I push it, seemed to have worked on my macbook now, tegra too, so just omap remaining
[09:23] <mlankhorst> more annoying than I thought, blob is fragile, easiest way seems to be unsquashing squashfs, and trying from there
[09:31] <mlankhorst> welp, it worked
[09:38] <mlankhorst> [    20.218] (II) config/udev: Adding drm device (/dev/dri/card0) card0 /sys/devices/pci0000:00/0000:00:01.0/0000:01:00.0/drm/card0
[09:38] <mlankhorst> [    20.218] (II) config/udev: Ignoring already known drm device (/dev/dri/card0)
[09:38] <mlankhorst> cute, race triggers on nfs too
[10:17] <mlankhorst> hm maybe not a race, triggers second time too
[10:55] <tjaalton> so i915 gen5 still has slow blur
[11:01] <tjaalton> well, didn't even know it had regressed until now
[11:24] <mlankhorst> just push and sru later?
[11:29] <tjaalton> maybe
[11:44] <mlankhorst> I think I found my suspend failure
[12:32] <ppisati> guys, i heard there was a fix for x-input in R bt that was hard to backport to Q
[12:32] <ppisati> something about touchscreen input that impacts our nexus7 img
[12:32] <ppisati> can anyone confirm it?
[12:32] <ppisati> (and sorry for the lame description :) )
[12:33] <tjaalton> ther is no fix in raring
[12:34] <tjaalton> what I have crashes the server after opening the dash and touching somewhere
[12:36] <ppisati> tjaalton: ah
[12:36] <ppisati> tjaalton: i was told different
[13:40] <mlankhorst> tjaalton: hm shall I try to figure out the touch crashing issue in 1.14 branch?
[13:41] <tjaalton> mlankhorst: it's not the same crash
[13:41] <tjaalton> aiui
[13:42] <mlankhorst> tre
[13:42] <mlankhorst> true*
[13:43] <mlankhorst> but I want to get 1.14 with patches working stably first
[13:56] <tjaalton> ok
[16:00] <mlankhorst> wow, reading through the input stack is weird
[16:11] <mlankhorst> it's like an onion, layers, layers, layers..
[16:19] <ogra_> and smelly ... 
[16:48] <mlankhorst> hm fun, guess it's a magic union
[16:50] <mlankhorst> any bets on the magic thing where it crashes on being pointer barriers?
[16:58] <tjaalton> the guy on the upstream bug says the crasher was there with 1.13 too
[16:59] <tjaalton> so not that
[17:08] <mlankhorst> tjaalton: the specific function hasn't been touched since before 1.12 or so
[17:09] <mlankhorst> anyway it looks like I may have a fix
[17:11] <tjaalton> mlankhorst: this was the memory corruption crasher you were able to repro with synaptics?
[17:20] <mlankhorst> yeah I think so
[17:20] <mlankhorst> if you look at ProcessTouchEvent, there are 2 destinct structs, ev->device_event and ev->touch_ownership_event, in a union with different sizes
[17:21] <mlankhorst> and first thing that function does is int emulate_pointer = ! !(ev->device_event.flags & TOUCH_POINTER_EMULATED);
[17:21] <mlankhorst> device_event.flags is at a different offset than touch_ownership_event.flags
[17:24] <mlankhorst> ev->device_event.corestate
[17:24] <mlankhorst> ends up saving at some random offset
[17:41] <mlankhorst> seems to be a lot better at surviving now..
[17:42] <mlankhorst> yeah, that fixed the valgrind error for me
[17:43] <mlankhorst> \o/
[17:43] <mlankhorst> no more random memory corruptions
[17:59] <tjaalton> cool, i'll try if it fixes my crasher too
[18:30] <mlankhorst> considering each time you pressed you ended up corrupting some random memory, most likely
[18:31] <mlankhorst> seems I didn't fix my nouveau suspend bug when I leave my pc off for longer periods though, boo :((
[18:31] <tjaalton> well the one with 1.13 + backports happened almost instantly
[18:31] <tjaalton> look at the backtrace on the upstream bug
[18:31] <tjaalton> open dash - touch the indicators - boom
[18:32] <mlankhorst> random memory corruption is random :P
[18:37] <tjaalton> well the way to reproduce was different than on 1.14, and very easy :)
[18:37] <tjaalton> but anyway, I'll try it
[18:58] <tjaalton> mlankhorst: doesn't apply on top of the backports ;)
[18:58] <mlankhorst> I'm sure you can find a way to do it manually :p
[18:58] <tjaalton> yeah, tomorrow
[19:02] <tjaalton> ah, it was easy
[19:05] <tjaalton> building
[20:09] <tjaalton> rebooting the nexus
[20:11] <tjaalton> crashed
[20:11] <tjaalton> so it wasn't that :)
[20:11] <mlankhorst> boo different bug then, I could reproduce the corruption on 1.13.3 too though iirc
[21:12] <tjaalton> hope it isn't fixed by the barrier work in 1.14
[21:13] <tjaalton> testing one other commit now
[21:28] <tjaalton> nope, not that..
[21:39] <llstarks> bryce, re your changes to the hybrid spec. shall i assume this work is postponed until amd and nvidia publish egl drivers?
[21:41] <tjaalton> the diff is just changing the assignee?
[21:41] <bryce> yeah just an assignee change
[21:41] <llstarks> to in-house?
[21:41] <tjaalton> raof is in-house :)
[21:41] <bryce> raof's busy with mir now, so freeing the WI's for anyone else that might want them
[21:42] <bryce> community owners would definitely be fine; I probably should have set them to ubuntu-x-swat actually
[21:42] <llstarks> when i see canonical-x, i interpret that as "secret project for the next few months" :)
[21:44] <bryce> llstarks, fair enough; updated
[21:45] <llstarks> that wasn't the problem, i was referring to the changes a few days ago. i'm just wondering where this work fits in when slutty snail is going to probably focus on mir and the converged platform
[21:46] <bryce> llstarks, I think you answered your own question there...
[21:48] <llstarks> thanks for the clarification
[21:49] <bryce> llstarks, basically it's backburnered in favor of phone stuff, but as low hanging fruit becomes available we can certainly look into it.  And help on anything in this space is welcomed.
[21:50] <llstarks> i plan to do some experimental mir+optimus stuff over the next few weeks
[21:50] <llstarks> uncharted territory
[22:18] <tjaalton> mlankhorst: so it's the same trace after all what I'm seeing, both with 1.13+backport & 1.14
[22:18] <tjaalton> xi2mask_isset returns 0
[22:19] <tjaalton> enough for today..
[23:06] <RAOF> llstarks: Cool. You should be able to have a bit more fun with that soon, after use-dma-buf has landed.
[23:09] <llstarks> fun, but not much in the way of code though. maybe a script or two if i have time. modprobe.d hybrid stuff needs an overhaul, so maybe i'll look at that too.