/srv/irclogs.ubuntu.com/2013/11/26/#ubuntu-mir.txt

kdubhey duflu, I wanted to sync with you on adding overlay support to the compositor01:04
kdubpretty much, my plan is to have the platform provide a 'filter' for the scene that will skip GL composition on a surface if the hardware can handle composing that surface01:05
kdubso, if the platform can bypass or overlay-composite any given buffer, it simply tells the compositor to skip rendering that surface with GL01:07
duflukdub: Sounds about right. Although all the "filter" code will be replaced by Q*01:07
kdubright, i pinged greyback about it, he seemed optimistic that it can be worked into the qt scenegraph01:07
duflukdub: Could be a generalization of BypassFilter, in a way. We just have to remember that if/when Q* take over the scene graph to ensure there's a working Qt replacement01:09
kdubright, just like a 'hardware optimization filter'01:11
kdubgreyback said it wouldn't be too hard to integrate a 'skip rendering these layers' into his work, so hopefully the bypass filter and this overlay work will mesh nicely with that01:12
duflukdub: Before we add yet more filters, I had planned on tidying up and getting back to only having a single surface class passed around filters/renderers.01:16
kdubduflu, yeah, part of the work that will need to be done is cleaning up the filter/operators on the SurfaceStack01:16
kdubi don't know that the model we have makes sense anymore, with the way we're using it these days01:16
duflukdub: I'm not committing to a particular class or name. Just that it became obvious to me there's so much coupling on the filter/operators that it would make more sense to refer to just something::Surfacelike instead. A single interface01:17
kdubperhaps, i haven't really felt out what I like best yet... just that it is pretty coupled01:19
duflukdub: I think it could be dangerous to be specific about the interface. It needs to be a general Surface-like interface to encompass CompositingCriteria, Renderable (buffer acquisition), flagging for overlay/bypass/etc01:23
dufluThat all fits into some kind of Surface interface I think. If we're lucky, an existing one01:23
dufluHmm, maybe it would be the reintroduction of a "Renderable"01:24
kdubgot a bit lost01:25
kdubwe'll probably hash all that out next week though :)01:27
dufluYes. Oh look; final confirmation.01:30
kdubyeah, should be a fun long flight01:41
dufluActually, significantly shorter than Boston :)01:50
truebattleaxei've lost my volume control and ability to vol + and - with the keyboard. is there a fix for this01:53
truebattleaxethis happened when i installed mir01:54
duflu_truebattleaxe: Not sure. That's odd because XMir does not affect input or sound. Those are unchanged to regular X01:58
duflu_truebattleaxe: Try uninstalling Mir and make sure the problem goes away. Otherwise it's somewhere else :)01:58
truebattleaxehmm maybe its just how i installed kde01:59
truebattleaxeis there any way to update mir? or see if there are updates01:59
duflu_truebattleaxe: it's just in normal Ubuntu updates02:00
=== duflu_ is now known as duflu
truebattleaxegotcha. i did uninstall ubuntu software manager and center. i have noticed its probably just muon but for some reason it doesn't get root access02:01
duflutruebattleaxe: sudo apt-get update ; sudo apt-get upgrade02:03
dufluWill do a regular update (including Mir)02:04
truebattleaxedoing that now. just wondering if thats normal with muon. to get the error of no root access02:04
truebattleaxehow often is mir getting updated right now?02:05
duflutruebattleaxe: On 14.04 every couple of weeks. On 13.10, not at all. (https://launchpad.net/ubuntu/+source/mir)02:06
truebattleaxegotcha. im on 14.04 so that is definitely good to hear02:06
duflutruebattleaxe: But XMir does not use Mir input. And Mir has nothing to do with sound. So you should probably search for other causes/solutions02:07
truebattleaxei think i may have uninstalled the sound manager some how.02:10
truebattleaxethanks duflu02:11
truebattleaxeso far i've had no problems with mir. It actually seems like my system is faster02:11
duflutruebattleaxe: Probably not faster, but very likely smoother. There's extra buffering and page flipping is enforced. So it could really make a visual difference if the existing X driver is below-par02:13
dufluUmm, "above" par??02:14
truebattleaxeyes i would say above par :P02:15
truebattleaxeway smoother02:15
truebattleaxethanks duflu for the info. definitely makes me a bit easier knowing there are updates every few weeks02:18
duflutruebattleaxe: No problem. In fact we're trying to get 0.1.2 into trusty this week02:18
truebattleaxea stable release?02:18
duflutruebattleaxe: They're all *stable* but it's a snapshot of the latest development02:20
truebattleaxegotcha02:20
truebattleaxeonce that update comes out i'll compile my desktop setup so i can pass it to my dad. He wants to use linux but he's iffy switching from windows because of ease to use.  So i made something that has all the apps he would use and ease and taking out that "edgyness" of linux to give him the experience he needs02:22
=== chihchun is now known as chihchun_afk
=== philipballew is now known as philip
=== chihchun_afk is now known as chihchun
didrockshey duflu, how are you?07:05
dufludidrocks: Hello. Good. You?07:06
didrocksduflu: I'm fine, even if latest image we promoted isn't that nice, so we'll have to race again today :)07:06
dufludidrocks: Well, I didn't land it this time :)07:06
didrocksduflu: I was coming to news, I think you heard about Mir making unity8 crashing?07:06
dufluBut happy to help07:06
didrocksduflu: it's the previous Mir release07:06
dufludidrocks: I haven't seen any evidence of Mir making Unity8 crash... ?07:07
didrocksurgh, the info wasn't sent?07:07
didrocksI guess it was https://bugs.launchpad.net/ubuntu/+source/unity8/+bug/125368507:07
ubot5Ubuntu bug 1253685 in unity8 (Ubuntu) "unity8 crashed with SIGSEGV in QMetaObject::activate()" [Critical,Confirmed]07:07
dufludidrocks: Ah, right. See comment #1107:07
didrocksso it's unity-mir (still associated with Mir in my head)07:07
didrocksyeah, just reading :)07:07
=== chihchun is now known as chihchun_afk
dufluIt's very confusing that unity-mir has classes that sound like they're in Mir07:08
didrocksyep07:09
didrocksso, I guess I have to wait for greyback07:09
dufluRAOF: How evil is it to limit width/height of a surface to 65535?07:28
RAOFNot very evil?07:29
RAOFI guess it depends on what the units are.07:29
dufluRAOF: Pixels... so that's 4-gigapixels07:30
RAOFWell, not necessarily. It could be a thin strip of a window.07:30
RAOFBut that's not at all evil.07:30
dufluRAOF: Yeah I know. But is it foreseeable for Mir to support multi-gigabyte surfaces?07:30
RAOFIt's forseeable for GPUs to act on multi-gigabyte *buffers*, but I don't think they're likely to be displayed.07:31
dufluI'm conflicted because the most elegant solution to resizing seems to be to make size a MirSurfaceAttribute. Therefore int value = (width << 16) | height;07:32
dufluAlthough if it was beyond 65535 you could just use zero and ask the client to query the buffer properties, as they already do07:33
RAOFThat (width << 16) | height would be entirely hidden in mir_client_library, right?07:34
Mirvfiled bug #1254986 and bug #1254987 against unity-mir and unity-system-compositor respectively07:34
ubot5bug 1254986 in unity-mir "unity-mir FTBFS against libmirserver11" [Critical,New] https://launchpad.net/bugs/125498607:34
ubot5bug 1254987 in Unity System Compositor "unity-system-compositor FTBFS against libmirserver11" [Critical,New] https://launchpad.net/bugs/125498707:34
dufluRAOF: Yes, behind mir_surface_get_size07:34
dufluAlthough a client could observe events and see the odd value if it wants07:35
RAOFIt shouldn't be able to do that.07:35
RAOFIf it's monitoring for size events it should receive a synthesised event, not the raw protocol value.07:35
* duflu discovers yet another redundant Surface class and is now much more sad08:18
=== chihchun_afk is now known as chihchun
* alan_g enjoyed their Java and Python IDEs: http://www.jetbrains.com/objc/features/cpp.html09:39
alf_alan_g: looks interesting, but mac os x only :/09:43
alan_galf_: that's just the current xcode plugin. They're working on a cross platform C++ IDE09:44
* alf_ is pretty happy with vim + YouCompleteMe + Syntastic09:46
* duflu thinks IDE == heresy. Pure code should be pure, accessible, readable and buildable with a text editor and shell :)09:47
dufluAlthough I do now accept syntax highlighting. And feel slightly dirty.09:49
* alan_g thinks e.g. selecting a block of text and saying "make that a member function" is less disruptive to thinking about the code than doing the steps by hand.09:50
duflualan_g: Pick the odd one out:09:54
duflu    std::shared_ptr<graphics::Buffer> snapshot_buffer() const;09:54
duflu    void swap_buffers(std::shared_ptr<graphics::Buffer>& buffer);09:54
dufluI think we need some kind of consistency, somehow09:55
alan_gduflu: they are different use-cases (note the const) but I do agree09:56
alan_gWe'll get there09:56
mlankhorstok, mesa 10 test build in ppa:canonical-x/x-staging soon10:24
mlankhorstmight need some mir testing :P10:24
ogra_did anything with the input layer change recently ? i cant wake up my maguro anymore without tapping the screen10:28
=== chihchun is now known as chihchun_afk
didrocksogra_: seems you are not popular here :)12:45
didrockskgunn: so, we confirmed that ogra's bug started in the previous Mir release: https://bugs.launchpad.net/mir/+bug/125504512:45
ubot5Ubuntu bug 1255045 in mir (Ubuntu) "screen does not turn on on maguro when pressing the power button" [High,Confirmed]12:45
ogra_heh12:45
didrockskgunn: so, putting the current Mir release on hold until this one is fixed12:46
didrocksmind having some people from your team looking at it? (and answering on IRC in the future ;))12:46
ogra_well, i didnt ping anyone specifically and didnt expect an answer :)12:49
didrocksogra_: still, would be nice if upstream in working hours was reading IRC :)12:49
alan_gdidrocks: I only answer when I have something to say12:50
didrocksalan_g: would be still nice if downstream doesn't have to play catching up to proove/search for which component regressed (this happens a lot) and that can be done with a little bit more communication12:51
alan_gI don't have a Galaxy Nexus and don't see the problem12:51
didrocksstill, I guess you have an idea on what enters Mir trunk12:51
didrocksmore than us anyway12:51
alan_gdidrocks: I rarely touch the input stuff. I think dandrader was making changes there a few weeks ago.12:54
alan_gThose changes might be (finally) on trunk12:55
didrocksalan_g: yeah, so at least, sharing this information would have been a start :)12:55
alan_gogra_: which version of Mir are you looking at?12:57
ogra_alan_g, well, the changeset with which it started is http://people.canonical.com/~ogra/touch-image-stats/20131120.2.changes12:58
ogra_so somewhere between misetver 9 and 1012:58
ogra_*mirserver12:58
alan_gogra_: Ok, that would be from "a few weeks ago". :(13:02
ogra_well, it entered th distro on the 20th13:03
alan_gogra_: there's a long and painful story about how stuff gets to Mir trunk. I don't like to think about it.13:05
didrocksthis is due to a long a painful story to be able to release Mir as well, but we already discussed it13:06
alan_gI can't see anything immediately suspicious. The stuff dandrader did in -c 1150 lp:~mir-team/mir/development-branch/ is only test code.13:10
dandraderalan_g, there's 1160.1.33 "android-input - Assign more unique touch ids"13:11
dandraderalan_g, that needs a recent qtubuntu which finally got a release only yesterday. but it's easy to tell if there's a mismatch there as there would be a crash after you tap on the screen 16 times :)13:12
dandraderbut that's all about touch events. nothing to do with buttons/keys.13:13
* alan_g has to go. (If someone can lend him a Galaxy...)13:14
=== alan_g is now known as alan_g|lunch
=== ara is now known as Guest1754
=== ara_ is now known as ara
=== ara is now known as Guest55288
=== dandrader is now known as dandrader|afk
=== Guest55288 is now known as ara_
Kaleoshouldn't https://bugs.launchpad.net/mir/+bug/1255045 be marked as Critical?13:53
ubot5Ubuntu bug 1255045 in mir (Ubuntu) "screen does not turn on on maguro when pressing the power button" [High,Confirmed]13:53
=== dandrader|afk is now known as dandrader
=== alan_g|lunch is now known as alan_g
=== ara_ is now known as ara
=== dandrader is now known as dandrader|lunch
greybackdidrocks: there's no kgunn for the rest of this week I believe15:55
kgunngreyback: whats up ? (on a hangout....)15:55
greybackkgunn: oh, I thought you were on hols. Sorry!15:55
didrocksgreyback: ok, ah no :)15:55
didrockskgunn: you probably missed my pings15:55
kgunnvery soon ....like minutes...not hours15:55
* greyback goes back under his rock15:55
kgunndidrocks: something wrong ?15:58
didrockskgunn: yeah, basically we have https://bugs.launchpad.net/mir/+bug/125504515:58
ubot5Ubuntu bug 1255045 in mir (Ubuntu) "screen does not turn on on maguro when pressing the power button" [Critical,Confirmed]15:58
didrockskgunn: this is on previous mir release15:58
didrocks(and it's promoted in the current image now)15:59
didrockskgunn: I think I'll have to hold any Mir release on it15:59
didrockskgunn: not, it can be on qtubuntu (which had to be upgraded due to the ABI breakage), you have the package version on the description15:59
didrockswould be nice if you can assign someone to it16:00
kgunnracarr: ricmm ^ can you guys please take a look into this bug ?16:00
didrocksthanks16:00
kgunndidrocks: so, if flashing dev-proposed will it be in that image ?16:00
didrockskgunn: just flash trusty16:01
kgunnack...thanks16:01
didrocksno need for -proposed16:01
didrockskgunn: maguro16:01
kgunndidrocks: ah...maguro16:01
didrocksyeah maguro (*sigh*)16:02
kgunndidrocks: maguro only good for sushi these days...right ?16:03
alf_kdub: so in KMS A plane respresents an image source that can be blended with or overlayed on top of a CRTC during the scanout process16:23
kdubso similar to android overlays16:23
kdubalf_, for what its worth, I think (to the compositor) we can model the hardware cursor as an overlay as well16:24
kdubso for gbm, it would be a kms plane or a hardware cursor, and for android, it would be a hwc surfcae16:24
alf_kdub: In KMS since this happens during scanout, the result is not really saved anywhere (not in user accessible buffer at least)16:24
kdubin android, i think its composited to a buffer, but we don't know exactly when that buffer is ready to see16:25
ogra_kgunn, well, maguro uses a PVR chipset ... we will soon have to support intel/android HW too ... that mostly comes with PVR (often enough even the same driver) ... so it makes sens to take care for it16:27
=== chihchun_afk is now known as chihchun
=== RoddieKieley1 is now known as RoddieKieley
=== chihchun is now known as chihchun_afk
alan_gkgunn: the only team member that I think might have a maguro/Galaxy Nexus is kdub16:32
kgunnalan_g: ...i thot racarr had one, but surely any help welcome16:33
alan_gkgunn: you may be right16:33
alf_kdub: @hardware cursor as overlay, probably as long as there is some kind of attribute that tells us that this is the cursor, since, at least with kms, there is a dedicated API for it so we need a special overlay implementation16:34
kdubalf_, right, but thats something the gbm platform can sort out so the compositor doesn't have to think about it16:34
kdubricmm, does maguro count on mir to turn the screen on and off? or is handled somehow with sysfs and powerd?16:43
greybackkdub: any way I could help you determine that? I've a gnexus here17:07
greybackkdub: powerd sets display state by talking to (unity-)mir over dbus.17:10
greybackwhich is setting the MirPowerMode on the display17:12
kdubgreyback, ok, i have an idea then17:12
ricmmkdub: it relies on mir, but afaik robert never plugged that path17:18
ricmmsomething might have changed that prevemts the hwc from restartong unledd there is a toich17:19
=== dandrader|lunch is now known as dandrader
ricmmpowerd does hpwever brong the system.down17:19
ricmmwhich results in the screen going off17:19
ricmmhaving lunch, can help in a bit17:19
kdubricmm, sure, should have something to try by then17:19
ricmmok17:20
ricmmtyping on phone is hard17:20
=== greyback is now known as greyback|away
* kdub sees the power button working intermittently on maguro17:57
kdubwas that the bug? that it doesn't work sometimes?17:57
=== alan_g is now known as alan_g|EOD
=== dandrader is now known as dandrader|afk
=== dandrader|afk is now known as dandrader
=== thomi_ is now known as thomi

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