robert_ancellRAOF, are there things worth asking that we can't get apport to do?00:05
RAOFYes; run a Mir server and client, and check if *that* works.00:05
RAOFOther than that, we should be able to collect all the relevant logs.00:06
robert_ancellta, I'll ask them for that00:06
RAOFrobert_ancell: Hey, if someone starts a Mir client and attaches it to unity-system-compositor's socket, what happens?00:07
robert_ancellRAOF, they can connect, but they'll never get focussed. (Actually can't remember if Mir's default behaviour will give them focus)00:08
robert_ancellEither way, it does need to be fixed (Haven't filed a bug yet)00:08
* robert_ancell -> lunch00:24
dufluRAOF: It appears someone has started on making VBE KMS-friendly. But it never hit any kernel tree yet. Do you know anything?... http://lists.freedesktop.org/archives/dri-devel/2013-January/034058.html03:15
=== chihchun_afk is now known as chihchun
robert_ancellduflu, is bug 1192429 fixed now we have those VT changes in?03:38
ubot5bug 1192429 in Unity System Compositor "unity-system-compositor is on multiple simultaneous (and random) VTs" [High,Triaged] https://launchpad.net/bugs/119242903:38
duflurobert_ancell: I will have to update and enable it to retest...03:39
robert_ancellduflu, are you still on the PPA?03:39
duflurobert_ancell: Saucy, yes. I had planned on moving to no-PPAs soon anyway03:39
duflurobert_ancell: Is the fix in distro yet?03:40
robert_ancellduflu, yes03:40
robert_ancellnot a specific fix for that problem, but VTs seem to be working well now (aside from the input always going to XMir)03:41
duflurobert_ancell: Is there a bug for that one?03:43
robert_ancellduflu, yes, I'll just find it03:43
duflurobert_ancell: OK, please tag it: vt03:43
robert_ancellduflu, bug 119284303:43
ubot5bug 1192843 in XMir "XMir does not relinquish input on session switch" [Critical,In progress] https://launchpad.net/bugs/119284303:43
dufluThat's cool. ppa-purge results in me getting mir package upgrades03:44
robert_ancellduflu, oh, why did you bump the soname for libmirserver?03:46
duflurobert_ancell: Because the switch branch seems to have changed enough ABI things that it became necessary03:46
dufluLoading the old version by mistake was causing serious bugs03:47
robert_ancellduflu, we always rebuild dependencies so it should be impossible to have an old version03:47
duflurobert_ancell: Not so on my Nexus4, which has not been updated :)03:47
robert_ancellduflu, but after that change it did break exactly like that because debian/rules wasn't updated to fix the version03:47
duflurobert_ancell: I thought any ABI break should result in an ABI bump?03:48
dufluThat's surely why we separated ABI from version?03:48
robert_ancellduflu, only for libmirclient. libmirserver is still considered ABI unstable03:48
robert_ancellI'll send out an email clarifying03:48
duflurobert_ancell: Alright, that would be helpful03:48
MirvI keep on getting bug #1210798 (private bug) on saucy from unity-saucy-compositor, ie. not starting up. sandy bridge intel graphics. will monitor the situation as/when updates come :)04:03
ubot5Error: Launchpad bug 1210798 could not be found04:03
MirvI'll make it public, not much private there unless something ubersecret in the coredump04:03
Mirvbug #1210798 - there04:04
Mirvdamn you ubot04:04
duflurobert_ancell: Yes, VT switching works for me in USC. Bug updated.04:06
robert_ancellduflu, thanks04:06
robert_ancellRAOF, when we bump a package name (e.g. libmirserver0 -> libmirserver1) is there a manual approval for that to land in the archive?04:07
Mirvrobert_ancell: yes, I already looked at it but the deal is that I need to ack it with someone with upload rights before publishing04:07
robert_ancellbug 121079804:08
ubot5bug 1210798 in unity-system-compositor (Ubuntu) "unity-system-compositor crashed with SIGSEGV in eglGetDisplay()" [Undecided,New] https://launchpad.net/bugs/121079804:08
Mirvah, hmm, I wonder if it's about me having libhybris installed on x8604:11
robert_ancellMirv, that reminds me, we need to make u-s-c attach unity-system-compositor.log automatically04:13
* Mirv runs unity-system-compositor04:13
Mirvok, so the bug is just crashing without explanation if libhybris/etc installed on x8604:14
robert_ancellMirv, you uninstalled libhybris and that fixed it?04:15
Mirvrobert_ancell: yes. it uninstalled libhardware etc as well.04:16
tvoss_good morning :)04:16
Mirvmorning tvoss_04:17
robert_ancelltvoss_, I thought you were on holiday?04:17
tvoss_robert_ancell, nope, only Friday04:17
tvoss_robert_ancell, lool is on holiday, my alter ego :)04:18
robert_ancellah,  tag team04:18
* tvoss_ is happily running a non-slow XMir right now :)04:18
tvoss_duflu, ping04:18
tvoss_RAOF, ping04:21
robert_ancelltvoss_, ping04:23
robert_ancelltvoss_, you know you can just ask questions out loud :)04:23
MirvI think I can work with XMir, this mirror-by-default is good enough for what I need until true multi-monitor support04:23
robert_ancellMirv, nice!04:23
tvoss_Mirv, woot04:25
duflutvoss_, pong04:25
tvoss_duflu, hey there :) tried the bypass branch with xmir, seeing some occasional rendering hiccups04:25
Mirv"nice!" to you, that is :)04:25
tvoss_robert_ancell, yup :)04:27
duflutvoss_, interesting. Not sure how to gauge if it is bypass to blame, or just hiccups that were hidden by extra buffering before04:27
RAOFtvoss_: Pong04:27
duflutvoss_: What is a hiccup?04:28
tvoss_duflu, seems like a buffer arrives out-of-order04:29
duflutvoss_: Weird. I can't reproduce any such issue with plain Mir.04:29
dufluMaybe something unusual about XMir vs normal clients04:30
tvoss_duflu, ack, RAOF ^04:30
RAOFtvoss_: On intel, I presume?04:30
robert_ancellRAOF, bug 1211063 - does XMir load color profiles on startup and just not handle changes which would be via XRANDR yet?04:31
ubot5bug 1211063 in Unity System Compositor "Color profile change is not instant when unity-system-compositor is running" [Undecided,New] https://launchpad.net/bugs/121106304:31
RAOFrobert_ancell: We don't handle colour profiles *at all* at the moment.04:31
robert_ancellRAOF, I thought so too - Is Jussi confused04:32
RAOFI think so?04:34
tvoss_RAOF, ack04:34
tvoss_RAOF, is that a known issue?04:35
RAOFtvoss_: No, but it changes the plausible causes.04:35
tvoss_RAOF, true :) anything else I can help to track down/triage the bug?04:35
RAOFeg: We have in the past failed to flush correctly on radeon, so the buffer we submit to Mir didn't necessarily have fresh content.04:35
RAOFtvoss_: What would be really good is a vsync'd app, displaying an increasing counter, counting up one per vblank, and a 60+Hz film of said client ;)04:36
tvoss_RAOF, hah, nexuiz has got such a display, let me see if I can trick it into vblank ;)04:37
RAOFXMir *is* the client we have that's most likely to show out-of-order artefacts, as it's the only client we have that doesn't submit a new frame each vblank.04:37
tvoss_RAOF, true :)04:41
* tvoss_ needs more coffee04:42
RAOFActually, the requirement for vsync is spurious; all we need is something that displays a monotone counter that increases at least once per vsync.04:42
duflutvoss_: I will modify some demo client(s) to increase my visual test coverage... after lunch04:42
tvoss_duflu, yeah, I will move over to the new place, back in 3004:43
* RAOF heads off to the doctors with Zoë. Taking my laptop, so I should be around.04:51
robert_ancellthomi, do you remember what the trigger was to make Jenkins do a test on a MP? I would like Jenkins reviews of https://code.launchpad.net/~a7x/lightdm/multiseat/+merge/178511 ideally before I approve the branches05:19
thomirobert_ancell: I take it a7x is not a member of ~canonical?05:20
thomirobert_ancell: I believe in that case you need to vote to approve the MP (you don't need to top-approve it thought)05:20
robert_ancellthomi, actually, he is a member of ~contributor-agreement-canonical - perhaps we should allow that group to trigger Jenkins05:21
thomirobert_ancell: you can ask, but I doubt it'll go through. Signing the contributor agreement doesn't mean Canonical has any kind of leverage over them, in case something bad happens05:21
thomibut I agree, it's a real PITA05:22
robert_ancellthomi, well, you could kick them out05:22
robert_ancellbut yeah, I know what you mean05:22
thomiif we had more time we should set something up on prodstack as a firewalled builder05:22
thomishouldn't be too hard to achieve05:22
robert_ancellthomi, is there a way to manually trigger a build, e.g. like the form we use for rebuilding?05:24
thomiyeah, you don't want to add an approve vote?05:24
smartboyhwrobert_ancell, why does unity-system-compositor only work in lightdm05:25
robert_ancellsmartboyhw, because no-one has used it in anything else. There's no particular reason why it couldn't be used elsewhere05:26
smartboyhwrobert_ancell, oh05:26
thomirobert_ancell: I think I found the right job, just trying it now05:27
thomihmmm, nope!05:28
thomirobert_ancell: I'm not sure how to do that, sorry. I can email fginther, and get back to you tomorrow, if you like05:29
robert_ancellthomi, np05:29
robert_ancellit's not that important05:29
duflutvoss_: I still can't reproduce any "hiccups" using bypass on mir_demo_server_shell with the example clients06:37
tvoss_duflu, hmmm, might be XMir specific then06:37
dufluI'm using Intel BTW06:37
dufluRAOF: Is XMir now a well-behaved normal client? (no GBM-specifics)06:38
RAOFduflu: No.06:38
dufluRAOF: How so? Please forgive my forgetfulness...06:39
RAOFBecause it doesn't use GL to render; it uses the DDX's methods to render.06:39
RAOFAnd it's not going to use GL to render for the forseeable future.06:40
dholbachgood morning06:42
dufluRAOF: Hmm, OK. I don't really need to know what that means, other than to ensure that XMir only uses the buffer given to it by the client library, and for never longer than till the next swap. That's still true right?06:43
dufludholbach: Morning06:43
RAOFduflu: That is true.06:43
RAOFI have a plausible hypothesis.06:43
dholbachhi duflu06:43
RAOFHow many times to you allocate buffers?06:43
RAOFOr: how many times will you send a buffer that the client hasn't seen before?06:44
dufluRAOF: Max 3 unique ones. Allocated lazily (when they're first required)06:44
RAOFI think the current code fails to handle the case where buffer->age == 0, but that should only happen once per buffer allocation.06:48
RAOFtvoss_: How many times do the hiccups occur?06:49
tvoss_RAOF, hmmm ... difficult to tell, but regularly06:49
RAOFSo, unlikely to be that you're hitting the age==0 case.06:49
dufluI can also see some damage/age-related bugs when toggling CCSM>OpenGL options06:50
dufluBut that's not really related06:50
duflutvoss_: What are you running to see hiccups in XMir? I suspect this might also be an issue - https://bugs.launchpad.net/xmir/+bug/121118607:31
ubot5Launchpad bug 1211186 in mir (Ubuntu) "[regression] GL(X) apps don't have vsync when running under XMir" [Undecided,New]07:31
tvoss_duflu, gnome terminal07:32
duflutvoss_: Hmm, which branch or Mir/bypass and what revision? Keep in mind the input lag fix only just landed07:32
duflu-or +of07:32
tvoss_duflu, only landed now?07:32
tvoss_duflu, as of Friday iirc07:33
duflutvoss_: Only landed in r95507:33
dufluSo looks like Friday night/Saturday your time07:34
tvoss_duflu, let me recheck07:34
duflutvoss_: Would have only landed in lp:mir midnight Friday night your time07:36
* didrocks again sees issues with ATI07:37
didrockssil2100: doesn't seem around, let me try to match with a commit :/07:37
sil2100I'm here, what's up?07:37
didrockssil2100: didn't you see my ping?07:37
sil2100Did I miss a ping?07:37
didrocks09:19:30 didrocks | sil2100: the -ati machine seems wracky again, it was ok on Friday/Saturday07:37
didrocks09:19:38 didrocks | sil2100: I think it's mir being the issue with that07:37
didrocks09:19:48 didrocks | (the tests are ran in mirslave)07:37
didrocks09:20:10 didrocks | can you see if anything was committed between latest good mirslave run and the new bad one in mir itself?07:37
didrocks09:20:24 didrocks | (so look at when exactly mirslave check started to fail)07:37
didrocks09:20:37 didrocks | and potentially match that with a mir rebuild which happened just before?07:38
didrocks09:21:20        * | didrocks reboot ati to save it first07:38
didrocksmultiple ones even ^07:38
sil2100Ah! Ok, just saw the ping ones, didn't manage to read the rest of the context ;) Looking at mirslave07:38
sil210011th was fine it seems07:39
RAOFGrargh ATI07:40
mlankhorstisn't memory corruption fun? :P07:41
didrockssil2100: no, look closer, sometimes the autopilot job is green, but just have one test running07:42
didrocks(on ati)07:42
didrocksand the job went stuck07:42
sil2100I know, since job 72 failed since it only ran 1 job07:42
sil2100But 71 seems ok, so checking changes in mir etc07:43
* didrocks hopes they were some07:43
didrockssil2100: also package list (both post-setup) maybe?07:43
RAOFmlankhorst: I personally prefer texture corruption across suspend/resume cycles. That's more fun.07:45
mlankhorstpfft :p07:46
mlankhorstbut I guess I'll try to isolate the issue at least07:48
mlankhorstand figure out why ati is kernel corrupting on me07:48
sil2100didrocks: still looking, but hm, doesn't seem like there were any intrusive changes - the build started failing when running Mir version from 12.1, but the only changes made since the time it was ok are the packaging change (shlib in rules) and one CMake change ;/ And unity-system-compositor had no new changes since 10th07:49
sil2100Looking at the postsetup list for anything else07:50
didrocksyeah, let's hope we can find something07:51
sil2100But why is it always ati affected ;/?!07:51
sil2100didrocks: how exactly did the ati machine hang up?07:52
didrockssil2100: so, what we saw last week, is that (it was 100% of the time at first):07:53
didrocksu-s-c couldn't start or starts and hangs07:53
didrocksunity couldn't start (stuck on opengl plugin, when duflu is running unity_system_compositor)07:53
didrocksso the first opengl app hangs07:53
didrocksthen, the machine times out of 2h07:54
didrocksif you then restart with regular xorg07:54
didrocksnothing happens07:54
dufludidrocks: duflu?07:54
didrocksthe machines hangs07:54
didrocksduflu: sorry, I meant you run unity_support_test from compiz07:55
didrockssil2100: so, it's really annoying, everytime, that's what screwing up the ati machine07:55
* duflu barely remembers what that means07:55
didrocksduflu: you do run it if the unity plugin is setup to know if unity can be ran in that session07:56
didrocksor should fallback on llvmpipe mode07:56
sil2100didrocks: since diffing the postsetup pkg lists show that only mir packages, unity-system-compositor and unity-autopilot are different between runs07:57
didrockssil2100: ok, no kernel, no driver?07:57
didrocksweird, we had the issue 100% of the time07:57
didrocksand then, not anymore07:57
didrocksat all, in mutiple runs07:57
didrockssil2100: when did you see that starting again?07:57
dufludidrocks: We've seen that problem before... https://code.launchpad.net/~vanvugt/ubuntu/quantal/nux/fix-1039155/+merge/128422/comments/28115207:58
sil2100didrocks: I compared the last ok run (which ran more than 1 test), which was from yesterday07:58
sil2100With the one today, which started failing07:58
didrocksduflu: right, here the issue is that with Mir/u-s-c on ATI, the program hangs07:59
dufludidrocks: Yes there have been problems on various machines where unity_support_test causes bugs if executed from within Compiz. I think I was advocating not running it from within Compiz (which solved the issues at the time)08:00
didrocksduflu: but this only happen when u-s-c is running08:01
didrockswith regular Xorg, no issue08:01
didrocksand not everytime :p08:01
dufludidrocks: The issue is much older than that and has been happening for an unlucky few for a year08:01
sil2100didrocks: can I try re-running the mirslave stack? Since I missed last week and want to make sure the failure now is reproducible08:01
didrocksduflu: I mean, we run compiz hundreds of times everyday08:01
sil2100(or does it crash ati for good?(08:01
didrocksif not thousands08:01
didrocksfrom the past 8 month on this particular machine08:02
didrocksduflu: only running xmir shows that issue08:02
didrockssil2100: it's running every 4h automatically now08:02
didrockssil2100: so let's wait for next run08:02
dufludidrocks: I know and I could never reproduce it either. However those who did have the problem before, found the solution was to unity_support_test externally, as in https://code.launchpad.net/~vanvugt/ubuntu/quantal/nux/fix-1039155/+merge/12842208:04
didrocksduflu: but it would be good to know why Xmir is showing the issue more08:05
didrocksthat doesn't explain the reason08:05
dufludidrocks: True :(08:05
dufluRAOF: Fun fact: DRM will successfully scan out non-scanout GBM buffers (!?)08:08
didrocksduflu: do you think we should try it now?08:08
didrocksduflu: and that will be compatible with u-s-c?08:08
dufludidrocks: I can't remember all the details and how things fit together, but I do think it would be good to avoid any fork/exec from compiz08:09
didrocksduflu: mind proposing an updated branch?08:09
dufludidrocks: I'm on the critical path to deliver bypass support, so want to finish that first08:09
didrocksduflu: sure, do you have any idea when you can tackle that one? (roughly?)08:10
dufludidrocks: 1 week ? :/08:10
didrockswe have to find something/someone to do it beforehand08:10
didrocksas otherwise it will mean no more Mir delivering in a week08:11
sil2100Unacceptable ;/08:11
sil2100didrocks: so hm, the workaround/solution would be to remove running unity_support_test from within compiz and just running it like in the branch duflu mentioned, i.e. separately (maybe upstart?)08:14
sil2100I wonder if this would indeed help08:15
didrockssil2100: yeah, we need someone to try that with xmir though08:15
didrockssil2100: /etc/X11/Xsession.d/ is fine for now08:15
didrocks(see #ubuntu-desktop)08:15
duflusil2100: You need to do it in a parent process of the X session, because if unity_support_test fails then export LIBGL_ALWAYS_SOFTWARE=1 has to be inherited by the compiz process08:19
sil2100duflu: right, thanks!08:19
RAOFduflu: How did you discover that?08:19
dufluRAOF: I was just curious. Not sure why. It's weord08:20
sil2100didrocks: I understand that you're hacking on it right now? Or should I try? But I won't be able to test it on xmir though, so will have to find someone running it08:20
didrockssil2100: please do, I'm in a ETOOMUCH already :/08:20
didrockssil2100: why can't you try on xmir?08:20
RAOFduflu: We might accidentally allocate scanout-capable gbm buffers anywoy08:20
dufluRAOF: Makes me think GBM's idea of scanout is different to DRM's, which would be important to know08:20
sil2100didrocks: I am afraid it will make my system boom ;) Is it safe?08:20
didrockssil2100: it's easy to revert anyway08:21
didrocksinstall unity-system-compositor08:21
didrocksand uninstall it08:21
didrocksfor removing it08:21
didrockssil2100: what driver are you using?08:21
dufluRAOF: All I have to decide right now is (width >= 800 && height >= 600)    ... :/08:21
sil2100Ok, sounds easy08:21
sil2100didrocks: the open radeon driver08:21
RAOFduflu: It's also possible that there's *no* buffer which is incapable of scanout, depending on how funky intel's scanout hardware is.08:22
didrockssil2100: excellent, that's what we need :)08:22
dufluRAOF: Yeah I figured drivers would be different on this. I will probably have to play with video cards08:22
didrockssil2100: I would advise you to first try with regular Xorg08:22
didrockssil2100: and ensuring that the script is ran08:22
sil2100\o/ will do08:22
dufluRAOF: Tho if you try to DRM flip a buffer with wrong dimensions that's pretty fatal08:29
dufluAt least, Mir makes it fatal08:30
mlankhorstsuperrr deadly08:30
sil2100didrocks: I'm quickly rebuilding compiz to remove that patch for running it from compiz, once this is done I test the scripts08:32
didrocksthanks sil2100!08:34
dufluBack in 30ish08:37
dufluRAOF: XMir is buffer usage hardware, right? :)09:12
tsdgeosguys, using mir_demo_server + mir_demo_client_accelerated on the Nexus 4 i get a black screen09:15
tsdgeosany clue of what may be wrong?09:15
sil2100brb, reboot09:15
duflutsdgeos: Probably surfaceflinger is still running?09:15
tsdgeosduflu: nope09:16
tsdgeosphablet@ubuntu-phablet:~$ ps -A | grep surf09:16
duflutsdgeos: I have seen similar issues... Mir lacks explicit support for turning the screen on if it is off. But that's not usually an issue if you have fully disabled surfaceflinger (and reboot): https://wiki.ubuntu.com/Touch/Testing/Mir#Switch_from_SurfaceFlinger_to_Mir09:17
duflutsdgeos: Also, that issue I have seen is obvious because Mir will throw exceptions09:17
duflumir_demo_server* will throw exceptions09:18
tsdgeosnot here09:18
tsdgeosstuff is here09:18
tsdgeoson the shell09:18
tsdgeoswithout complaining09:18
duflutsdgeos: Any hint that the backlight is *on* ?09:18
tsdgeoshow could i know?09:18
tsdgeosit seems to be09:18
ogra_use powerd-cli  to force it on09:18
dufluIt's usually obvious (not properly black)09:18
tsdgeosi.e. if i reboot i get the "google" logo09:18
tsdgeosuntil i start mir_server_demo09:19
duflutsdgeos: Yeah09:19
tsdgeosthat turns gray09:19
tsdgeoslooks "ok" to me09:19
duflutsdgeos: OK, so your server is running but not showing clients then09:19
duflutsdgeos: Because it sounds like the glClear from mir_demo_server* is working09:19
tsdgeosshould it show some FPS readings?09:20
tsdgeosbecause the unaccelerated one does09:20
tsdgeosbut i still can't see anything on screen09:20
duflutsdgeos: Yes it will always be 60 FPS on Nexus409:20
tsdgeossee no output here09:20
duflutsdgeos: Does the client start/hang?09:20
tsdgeosi do see 60fps with the unaccerelated09:20
tsdgeosbut the accelerated only gives09:21
tsdgeosphablet@ubuntu-phablet:~$ mir_demo_client_accelerated09:21
tsdgeos__pthread_gettid -209:21
tsdgeosSurface created09:21
tsdgeosand that's it09:21
duflutsdgeos: OK then you have a GL-specific issue. Probably related to hybris09:21
tsdgeoswho do i nag ?09:22
duflutsdgeos: https://bugs.launchpad.net/mir/+filebug   ;)09:22
tsdgeosso it's going to be another day without any work done09:22
tsdgeosawesome :D09:22
duflutsdgeos: Did you get all your binaries from the phablet image or hand built?09:23
tsdgeosall image + apt-get upgrade09:23
ogra_did you check that there are no bits in the apt-get upgrade that reach into the android side ?09:24
duflutsdgeos: Same result for mir_demo_client_egl* ?09:24
ogra_(are yoou sure surfaceflinger works at all after you upgraded)09:25
tsdgeosduflu: the egl ones print stuff to the shell09:25
ogra_(if not, Mir wont either ... apt-get has to be used with *lots* of care)09:25
tsdgeosbut can't see stuff on screen either09:25
duflutsdgeos: How about mir_demo_client_fingerpaint?09:25
dufluThat should work (software rendering)09:25
tsdgeosand it ended (returned to shell) after a while09:26
tsdgeosseems the demo server quite09:26
RAOFduflu: yes indeed it is buffer_usage_hardware :P09:26
duflutsdgeos: It sounds a lot like your client and server are mismatched. :/09:27
tsdgeosmay be09:27
tsdgeossadly s-jenkins:8080/job/ubuntu-touch-phablet-image-saucy-mir/ hasn't had a correct build for a long time09:27
dufluRAOF: Good. Because bypass cripples software buffer performance slightly. I have disabled bypass of those09:28
dufluNot surprising. I would expect pushing pixels across the bus to GPU-land to be a bit slower09:29
RAOFBut you need to do that bypass or not bypass?09:29
tsdgeosogra_: what's your suggestion then to get something "up to date" given the last successful image is from 6 days ago?09:30
mlankhorstduflu: well just do staging transfers from GART to VRAM :-)09:30
ogra_tsdgeos, poke rickmm ... once he is up ... and beyond that use the latest normal image, diable SF there and install the Mir stuff you need (i dont think the Mir image does much different)09:31
duflumlankhorst: I don't know what you mean09:31
tsdgeosogra_: then why is it red :'(09:32
mlankhorstperform a gpu assisted copy from system memory to VRAM09:32
* tsdgeos tries to do something else until ricmm or grayback are online09:32
dufluRAOF: In buffer_usage_hardware we're just sending (smaller) instructions across the bus and the filling/blitting originates from the GPU. It's quite different09:32
ogra_tsdgeos, no idea, i'm not affiliated with that image09:32
duflumlankhorst: Sounds interesting. Mir doesn't seem to with Intel at least.09:33
tsdgeosogra_: sure, not blaming you, just crying out loud09:33
mlankhorstduflu: intel has no dedicated VRAM :P09:34
ogra_tsdgeos, well, grab the latest normal image, disable SF and install Mir ... shouldnt be different :)09:34
duflumlankhorst: I know. Which makes me surprised there's any performance difference. But I would expect it with real cards09:34
tsdgeoscan try that i guess09:35
tsdgeosogra_: tx09:35
sil2100How can I see if xmir is currently used?09:48
didrockssil2100: unity-system-compositor should be running09:49
sil2100Ok, don't see it running here - is there anything else I need to do besides installing unity-system-compositor ? Any configuration switch?09:51
didrocksduflu: RAOF: ^09:52
didrockssil2100: cat /var/log/lightdm/unit*09:52
* duflu checks09:52
duflusil2100: You should have unity-system-compositor running, and also "ps auxw | grep mir" reporting mir options to X09:53
sil2100This is from /var/log/lightdm/unity-system-compositor.log09:53
RAOFThat's *very* informative :)09:53
duflusil2100: Wrong URL?09:54
sil2100hm, good url but it seems the log has some non-ascii characters in it09:54
sil2100Let me paste just the relevant parts09:54
sil2100inkerlinker.c:1095| ERROR: Library '/system/lib/libGLESv2.so' not found09:55
sil2100Segmentation fault (core dumped)09:55
didrockssil2100: hybris installed on your machine?09:55
RAOFYou win the "I've got hybris installed" award.09:55
sil2100I wonder what's hybris doing on my system09:55
seb128we need to stop diverting libgles from there...09:55
tvossseb128, in the works as I understand it09:58
seb128tvoss, great ;-)09:58
tvossseb128, I have had that issue multiple times in the recent past :)09:59
seb128tvoss, you are not the only one ;-)09:59
sil2100Ok, xmir works now with all things hybris uninstalled10:04
sil2100Thanks guys!10:04
sil2100didrocks: the Xsession.d detection seems to work fine10:04
didrockssil2100: you do have the file in /tmp/?10:05
sil2100I also did the test that duflu recommended in that merge10:05
dufluWhy do we even have libhybris on desktop? https://launchpad.net/ubuntu/+source/libhybris10:06
sil2100i.e. made unity_support_test -> /bin/false and had a fallback to llvmpipe10:06
sil2100Works both for xmir and standard x10:06
sil2100didrocks: now the steps to proceed... I will fill in a merge to lp:nux with the xsession.d file, but what about compiz? Since we don't daily release that yet10:07
sil2100didrocks: (we should start daily-releasing this week btw.)10:07
sil2100didrocks: should I modify the packaging and queue it up for sponsoring?10:07
didrockssil2100: juste prepare for a manual upload10:09
sil2100Ok, doing one more check and then preparing everything10:09
didrocksthx sil210010:10
duflusil2100: Can you link it to bug 1066764 too?10:10
didrocksmirslave just passed FYI this time10:10
ubot5bug 1066764 in Nux "Unity fails to load on old hardware (compiz enabling LLVMpipe has no effect and Mesa tries to use hardware still)" [High,In progress] https://launchpad.net/bugs/106676410:10
sil2100didrocks: ;/ Still, if this is the root cause, I guess it's better to have it resolved10:11
sil2100duflu: ok!10:11
sil2100didrocks: btw. is DESKTOP_SESSION and GDMSESSION ubuntu-2d still used? Since my desktop didn't want to start when I was setting those when doing a fallback to llvmpipe10:13
didrockssil2100: I don't think they do work10:14
sil2100didrocks: is setting those still required for the fallback mode or can I remove those and just use LIBGL_ALWAYS_SOFTWARE10:14
sil2100didrocks: ok10:14
didrockssil2100: yeah LIBGL_ALWAYS_SOFTWARE10:14
sil2100So they're leftovers then ;)10:14
didrocksprobably :)10:14
RAOFBah. One-surface-per-output is annoying.10:16
dufluRAOF: How?10:19
RAOFBecause it means sharding the single X root window.10:22
RAOFWhich is awkward.10:25
sil2100duflu: btw. in the old nux quantal branch I see you also removed the 01_blacklist_llvmpipe.patch patch as well, is that necessary?10:28
RAOFSo much so that it's time to investigate the CrtcSetScanoutPixmap route.10:28
duflusil2100: I have no idea. It was a long time ago10:28
sil2100It wasn't mentioned in the change and it works here without removing it, so I guess I'll skip that10:29
davmor2didrocks: I did a fresh install of saucy from fridays image over the weekend and am using mir from main that seems to of fixed the issues I was having thanks for the advice to drop the ppa :)10:35
didrocksdavmor2: great! :)10:36
sil2100didrocks: btw. since I see we already had a conffile /etc/X11/Xsession.d/50_check_unity_support which we rm_conffile'd - you think I should add it with the same name or maybe choose a different one?10:55
didrockssil2100: same name is fine10:55
didrockssil2100: of course, remove the rm_conffile :)10:58
sil2100Yep ;)11:00
sil2100didrocks: for the compiz manual upload... should I create a bug for that version, or just poke you with the source package etc?11:17
didrockssil2100: no need for a bug11:18
sil2100didrocks, duflu: https://code.launchpad.net/~sil2100/nux/unity_support_test/+merge/179666 <- it's the same change as duflu did in the past, just without removing the whitelist patch11:29
sil2100didrocks: I sent the compiz source pkg to you by e-mail11:29
didrockssil2100: approved and sponsored compiz, thanks!11:38
duflusil2100: Thanks. But I must run away.11:42
=== tvoss is now known as tvoss|quick_lunc
=== chihchun is now known as chihchun_afk
=== alan_g is now known as alan_g|lunch
mlankhorsti think my radeon hd 6570 is just busted, if I disable UVD the memory corruption appears to be gone12:20
mlankhorstbut it still goes into lockup recovery every 10 seconds, until after 5 minutes of that it gives up and shuts itself down12:20
=== alan_g|lunch is now known as alan_g
=== alan_g is now known as alan_g|tea
=== alan_g|tea is now known as alan_g
ricmmhey guys14:44
ricmmmsh::SessionListener::stopping() gets called when the socket's client end disappears?14:44
ricmmor when the client exits _explicitly_14:44
ricmmgreyback_: ping15:05
greyback_ricmm: pong15:05
greyback_sorry, dropped out there15:06
ricmmI asked this when you dropped15:06
ricmmmsh::SessionListener::stopping() gets called when the socket's client end disappears?15:06
ricmmor when the client exits _explicitly_15:06
ricmmas-in, would we get that when the process is SIGKILL'd15:06
ricmmor only SIGTERM / exit()15:06
greyback_yes, I just tested it now (you missed my reply) I "kill -9" an app, SessionListener::stopping() was called my Mir15:07
ricmmgreyback_: awesome \o/15:10
xjuniorhey guys15:25
xjuniorI know, I know... I only get here with problems15:25
xjuniorNow what I'm seeing here, using a intel graphics card is "intel(0): [drm] failed to set drm interface version."15:26
xjuniorusing the sleep trick doesn't work for me15:26
xjunioranother workaround I've seen was to use GDM instead of LightDM, but I'm trying to avoid that15:30
tsdgeosogra_: this morning you suggested using powerd-cli to make sure the screen was "on"15:42
tsdgeosogra_: http://paste.ubuntu.com/5977723/15:42
tsdgeosthe output feels not totally right15:42
ogra_sudo -u pahblet -i15:43
ogra_make sure you are in the users session ... phablet owns the session dbus15:43
tsdgeosogra_: but then it says i need to be root15:44
tsdgeosphablet@ubuntu-phablet:~$ powerd-cli display on15:44
tsdgeosYou must be root to run powerd-cli15:44
tsdgeosdo i sudo?15:44
tsdgeossame warning if i sudo15:44
ogra_tsdgeos, hmm, ask sfoshee ... it should work :(15:50
SamuRayalguien habla español?15:57
xjuniorSamuRay: portugues15:58
SamuRayxjunior, al menos entiendes lo que digo?15:58
SamuRayxjunior, puedes ayudarme con un problema que tengo con mir?15:58
xjuniorSamuRay: depende... qual o problema?16:00
SamuRayyo he venido usando mir desde hace algunos meses, el dia viernes hubo una actualizacion y me da un error16:01
SamuRayE: /var/cache/apt/archives/libmirserver1_0.0.8+13.10.20130808.2bzr948saucy0_amd64.deb: intentando sobreescribir `/usr/lib/x86_64-linux-gnu/libmirplatformgraphics.so', que está también en el paquete libmirserver0:amd64 0.0.8+13.10.20130808.1bzr945saucy016:01
SamuRaymorning racarr16:01
xjuniorSamuRay: tente16:01
xjuniorSamuRay: sudo dpkg --install --force-overwrite /var/cache/apt/archives/libmirserver1*.deb16:02
xjuniorracarr: morning!16:02
xjuniorracarr: would you try to help me with one intel issue? it seem to be failing to becomr DRM master or something16:02
SamuRayxjunior, esto ha ocasionado que el ligthdm no inicie automaticamente16:03
xjuniorSamuRay: con usted?16:03
SamuRayxjunior, no entiendo16:04
xjuniorSamuRay: o que esta ocasionando que el lightdm não inicie automaticamnte?16:04
racarrxjunior: I can try a little but I don't have an intel card or any ideass16:05
racarrso nothing so far :p16:05
xjuniorhave you seen people complaining about it?16:06
ricmmkdub: hey kevin16:19
ricmmtsdgeos is unable to bring up Mir on his nexus4 since a while back16:19
ricmmseems to only be working on maguro16:19
racarrxjunior: Not yet.16:21
racarrricmm: I think kdub is on vacation? (t/f?)16:21
racarrit hasnt been that long since I tried on nexus 4 but I can try again today16:21
ricmmah right, you have one too16:21
ricmmif you could that'd be great16:21
ricmmtheres no obvious error, just nothing on screen16:22
racarrok will do soon16:23
tsdgeosracarr: yeah it'd be great if you can try, just to make sure i'm not doing something really stupid (which tbh i don't see what can be)16:24
tsdgeosricmm: racarr: ok, i'm eod'ing now, will be back in a few hours in non-working-mode in case you need me16:27
greyback_racarr: hey, I'm looking into dandrader's work on the OSK. He was trying to add an extra surface type to Mir, but got resistance to that idea.16:34
greyback_racarr: I'm considering alternative ideas, would you have a few mins to discuss it some time today?16:35
ricmmtvoss: ping16:37
tvossricmm, pong16:37
ricmmaccording to the MR above, which seems to be stalled for a month now16:38
ricmmwould it make more sense for OSKs to use the overlay type?16:38
ricmmhow do we stack overlays and make sure the OSK has priority16:38
ricmmwe need to unblock that work today16:38
xjuniorfilled this bug, racarr: https://bugs.launchpad.net/ubuntu/+source/xserver-xorg-video-intel/+bug/121140316:40
ubot5Launchpad bug 1211403 in xserver-xorg-video-intel (Ubuntu) "intel failed to become DRM master" [Undecided,New]16:40
tvossricmm, looking16:41
racarrricmm: greyback_: I thought we decided to use16:41
racarrthe overlay surface type16:41
racarrAm updating my phone now to test 416:41
ricmmracarr: well I think the problem here is how would the shell know which of the overlay surfaces is the OSK?16:42
tvossracarr, that contradicts what we have been discussing with design16:42
greyback_racarr: I don't recall the decision being made anyway. I worry about needing the OSK surface to always be on top of everything else, so I need a unique way to identify it. I could identify the surface by its session16:42
racarrgreyback_: You still sort of need to identify it by session because16:43
racarrit should be the only surface allowed to take the16:43
racarrtype we are using16:43
greyback_racarr: indeed. But instead of just authorising it to take that session type, I can mark that surface in shell as special somehow, instead of relying on the surface type.16:44
greyback_I don't particularly like that solution though16:44
racarryes, it feels like the surface type is a nice way to make that mark sort of though16:44
racarrI would be happy to land that branch now :)16:44
ricmmnot so much the sort but the original identification is the issue I think16:44
ricmmisnt it?16:44
racarrtvoss: ricmm:?16:44
racarrricmm: No, because, its the same as clients16:45
racarrwith the PID based authorization16:45
racarrthen if the client is called OSK, you can trust it really is the OSK16:45
ricmmright, but maliit-server isnt launched by the shell right now16:45
ricmmgreyback_: the plan is to launch it from the shell?16:45
ricmmcurrently its a separate service, it needs to be up by the time the shell comes up16:45
racarrat least the shell needs the PID right16:45
racarrbecause the default policy is to dissaprove connections16:45
greyback_ricmm: I figured shell would launch it.16:45
ricmmiirc if you launch maliit-server after a shell is up, Qt fails to find it as an input method16:46
ricmmnot entirely sure how that works, but im sure it can be sorted16:46
racarrright I mean the worst case is the shell is passed16:46
ricmmand you are right, we need to authorize all sessions16:46
ricmmso it has to happen after anyways16:46
racarr --with-maliit-pid=716:46
greyback_oh that's good to know, I wasn't aware of that concern16:46
ricmmracarr: now thats ugly16:46
ricmmthe shell could launch the upstart job, thats fine16:47
racarrOf course XD, that's the literal worst case I can think of16:47
ricmmits a session job anyways16:47
ricmmwe just need to make sure that maliit-server wont block in that respect16:47
racarrand if there are issues with Qt detection or something16:47
ricmmI'll take that up with tmoenicke16:47
racarrthat can either be resolved or it can do it before initializing Qt16:47
ricmmwe can figure it out16:48
ricmmlets agree on the surface type thing16:48
ricmmwho were the ones against it? daniel and alan?16:48
racarrAlan had no strong opinion I think daniel thinks we should use the overlay type16:49
racarrtvoss is right the design documents call for16:50
racarran input method surface type16:50
racarrbecause input methods are still stacked under hints (i.e. tooltips), and the final overlay layer, which so far is only for things like16:50
racarrmagnifier or screenshot area selection16:51
tvossracarr, ricmm I set the MP back to needs review16:51
racarrI am happy to approve it there is one thing though16:51
racarrdoes this really have to bump16:51
racarrmirclient so version?16:51
racarrI am16:52
racarrsuper super sure no one is using mir_surface_type_arraysize_16:52
alan_gracarr: if they are then they are the sort of idiot that cannot be protected from themselves.16:53
racarralan_g: Later this afternoon we find that someone was using it and it was me :p16:54
ricmmwe can live with that16:54
ricmmalan_g: are you ok with approving this branch?16:54
alan_gricmm: sorry? I don't see the context16:55
ricmmthe input method surface type branch discussed above16:56
racarrTop approved17:05
=== alan_g is now known as alan_g|EOD
racarrBye alan :)17:07
racarrripping the cloggable event sink out of client-focus :D17:07
ricmmracarr: \o/17:12
racarrmerging ~kdub/mir/connect-display-request17:29
racarrif anyone wants to jump in and stop it before jenkins has its way :p17:29
racarrconnect-display-info has some of the worst merge conflicts I have ever seen :(18:13
racarreven with weave merge, its basically still total file ocnflict on all the tests18:13
racarrunit-tests got really slow18:20
racarrover the last week18:20
racarrprobably takes 10x as long to run18:20
racarrWhen did we get all these tests with18:22
racarrsleeps, etc18:22
racarrI think it's a bad idea. sometimes things that take <100ms locally, I have seen take > 10,000 ms on valgrind in qemu18:22
racarrso you can never choose a reasonable sleep value18:22
tsdgeossleeps in tests are not cool18:25
racarrand now already the unit tests take 20 seconds, but if we have to bump the values because of failures18:25
racarrbler :p ill look in to it after lunch18:26
tvossracarr, ping19:46
racarrtvoss: Pong19:48
tvossracarr, hey there19:49
tvossracarr, quick one: Do we hand over the appid to unity when a surface is requested?19:49
racarrthe app id19:49
racarrlike the session identifier19:49
racarror do you mean like the PID or something19:49
racarrwe do hand over the session identifier19:49
racarrit merged not so long ago19:50
tvossracarr, ack. So whenever an app requests a surface, the session information is handed to unity?19:57
racarrtvoss: Ok well not entirel when worded that way19:58
racarrthe surface now contains the session, so in all the19:59
racarrsurface created hooks etc the session is passed through19:59
racarrhowever there are some componenets19:59
racarrmainly the_shell_placement_strategy19:59
racarrwhich deal in terms of msh::SurfaceCreationParameters19:59
racarrhere, we don't pass the session19:59
racarrit seems like a good thing to pass19:59
tvosswhy not?19:59
tvossyeah, I think so, too19:59
tvossotherwise the implementation always needs to query the state from some tracking entity19:59
racarrtvoss: I guess just omission :)20:10
racarrbecause the only placement we have design for, can be decided entirely based on20:11
tvossracarr, can you fix that please?20:11
racarrsurface parameters20:11
racarrSure, what is it for?20:11
racarrhmm I guess you would probably want it for OSK20:12
tvossracarr, well, session and surface are two things that belong together from my pov20:13
tvossracarr, that should be reflected whenever requesting a policy decision20:14
racarrtvoss: mm20:16
racarrtvoss: Ok ill propose it in an hour or so just finishing up some stuff from pre lunch now20:17
tvossracarr, thx20:18
=== tvoss is now known as tvoss|eod
tsdgeosracarr: any luck with my nexus4+mir problem?22:07
racarrtsdgeos: Not yet. my nexus 4 seems to be having battery issues, I upgraded, then it died in abttery halfway through and wouldn't come back on for 20 minutes, finished upgrade, started getting mir22:08
racarrrinse repeat22:08
racarrstarted building mir22:08
racarrrinse repeat...22:08
racarrI will get to it before EOD definitely though22:09
racarrMan I have never had so much fun on a monday22:39
racarrIn particular considering I've spent almost the entire day with strange merge conflicts and the like22:39
RAOFracarr: Re: focus notification: we should move away from having the enum guard value be named foo_arraysize_, possibly to foo_enum_max_?22:47
racarrRAOF: That seems more reasonable22:48
racarrI'm not really sure what it's purpose is in some cases but I didn't want to argue it in thisMP XD22:48
RAOFActually, now that I think of it - I don't suppose the MirSurfaceFocusState enum could go away entirely?22:49
RAOFEnumerating boolean values is odd :)22:50
RAOFAnd you've already chosen to have it implicitly convertible to bool (deliberately, I presume?)22:51
racarrRAOF: Hmm22:53
racarrI dunno it's a little unclear because the state API still deals in integer values so22:53
racarryou don't necessarily know as a client library consumer that mir_surface_state_focus can only take on two values22:54
racarrI mean, that would be a really good guess :p but22:54
RAOFWe could document that mir_surface_get_state(mir_surface_attrib_focus); returns bool ☺22:54
RAOFLeaving the internal implementation as-is.22:55
racarrMm, I wonder if an enum is the preferred way to do that22:55
racarrover using a comment though22:55
robert_ancellRAOF, have you seen http://pad.ubuntu.com/xubuntu-mir?22:55
RAOFrobert_ancell: I suspect I've seen all the bugs on there, but I hadn't seen the pad itself; checking.22:56
racarrRAOF: Also strange requirements from 3d headtracking displays may force us to add mir_surface_quasifocused in 201822:56
robert_ancellRAOF, yeah, can you just double check we have bugs against everything. It would be good to add bug links so we catch them all22:56
RAOFracarr: Actually, that's a reasonable point; it's entirely possible that we'll want to do gaze-tracking focusish in the not-too-distant future (I, for one, would love it); an enum it is.22:57
RAOFracarr: Or, probably, a bitfield, but we can transparently upgrade the unfocused/focused enum to a bitfield without changing API.22:58
racarrRAOF: Mm ok sounds good23:01
RAOFracarr: For explicitlness' sake, can you add an “= 0” to mir_surface_unfocused's definition?23:02
racarrRAOF: Ok yeah23:03
racarrI think I need to remove IPC semaphore too23:03
robert_ancellRAOF, can you triage bug 121031623:14
ubot5bug 1210316 in xserver-xorg-video-ati (Ubuntu) "x crashes with radeon hd7850 video card with error /usr/bin/X: symbol lookup error: /usr/lib/xorg/modules/drivers/radeon_drv.so: undefined symbol: exaGetPixmapDriverPrivate" [Critical,Triaged] https://launchpad.net/bugs/121031623:14
RAOFOooh, that's probably why.23:22
RAOFrobert_ancell: Do you think anyone would be interested in read-only XRANDR support? If so, I can push it to staging.23:23
robert_ancellRAOF, in XMir?23:23
robert_ancellRAOF, yeah, staging should be safe for testing since no-one should be running it. I suspect XRANDR clients might get confused if they can't apply changes?23:24
RAOFXRandR calls are allowed to fail for pretty much no reason at all.23:25
robert_ancellRAOF, so clients just listen for events and attempt to apply changes and don't really care if they fail?23:26
robert_ancellRAOF, also, currently clients just get a fixed config back, right?23:26
RAOFWell, clients care if they fail, but they need to handle the case that they fail.23:28
RAOFYeah. Clients get whatever Mir set up.23:28
racarrRAOF: pushed the changes you suggested (plus removed ipc_semaphore.h)23:52

Generated by irclog2html.py 2.7 by Marius Gedminas - find it at mg.pov.lt!