| === chihchun_afk is now known as chihchun | ||
| kgunn | duflu: this one | 07:49 |
|---|---|---|
| kgunn | https://cvs.khronos.org/bugzilla/show_bug.cgi?id=12052 | 07:49 |
| RAOF | kgunn: Yeah, we need to work out what we want done there. | 07:52 |
| RAOF | Also, where did my comments go? | 07:52 |
| kgunn | RAOF: weird...my latest comment not there either ? | 08:03 |
| RAOF | EOD! | 08:11 |
| kgunn | alf__: for the non-block spike, duflu is wondering if we should "keep" the blocking eglswap as a potential path.... | 08:12 |
| kgunn | in fact, he thinks non-block is effectively "swapinterval0+sync" | 08:13 |
| kgunn | sync meaning it doesn't tear | 08:13 |
| duflu | As the default path at least. So clients and compositors don't regress/spin/etc | 08:13 |
| kgunn | alf__: also makes me wonder, are we trying too hard on the spike ? .... duflu says the swapinterval 0 today is non-tearing | 08:14 |
| duflu | eglSwapSleep(dpy, GL_TRUE) | 08:14 |
| kgunn | hmmm...maybe not on android tho | 08:14 |
| alf__ | kgunn: duflu: swapinterval 0 is not bound above by vsync, the spike is | 08:15 |
| alf__ | kgunn: duflu: that is, with just swap interval 0, clients will render as fast as they can | 08:16 |
| kgunn | alf__: ah....its in between | 08:19 |
| kgunn | or rather requires vsync...true... | 08:20 |
| kgunn | its like it wants some blocking (just not indeterminate blocking :) | 08:20 |
| alf__ | kgunn: at least some kind of throttling, which in the spike is a fake vsync at 60Hz | 08:21 |
| duflu | alf__: That's a bad idea. We fail the "zero wakeups" test during sleep. And that's something Mark mentioned this week too | 08:24 |
| alf__ | duflu: the buffer consuming thread only wakes up if triggered by a scene change, like the other compositing threads. It's just also enforces a fake vsync. So it's no worse than having another screen attached. | 08:26 |
| duflu | alf__: It's a basic power management requirement to avoid wakeups during sleep. Only intentional background threads should be the exception | 08:28 |
| duflu | A fake vsync would create a regression we'd just be forced to remove/fix eventually | 08:28 |
| alf__ | duflu: The fake vsync is only a throttling measure, it *doesn't* wake up the buffer consuming thread itself. The thread is only woken up if triggered by scene changes, and it consumes buffers at most at 60Hz, mimicking an output. | 08:31 |
| duflu | alf__: I know, but that's still poor power management, and a step backwards from Ubuntu as we know iyt | 08:32 |
| duflu | alf__: If nothing else, I think we need to keep defaulting to the existing behaviour. Then you can safely add optional behaviour for *anything* | 08:33 |
| alf__ | duflu: The existing behavior (eglSwapBuffers blocking indefinitely) 1. breaks Qt 2. Khronos told us that it's not how things should work | 08:35 |
| alf__ | kgunn: ^^ can you verify point 2, in case I misunderstood something | 08:35 |
| duflu | alf__: I know and I know. What I'm saying is don't make unblocking behaviour the default, but optional. We would break too many existing apps/compositors/toolkits | 08:36 |
| duflu | Plus you need buy-in from all the driver vendors, which will take time too | 08:36 |
| kgunn | alf__: accurate | 08:41 |
| kgunn | duflu: i agree, spec isn't clear & its not the greatest but side channels have to be used...which alot of OS's use (ios, android, qt)..so which toolkits don't have a side channel ? | 08:43 |
| kgunn | greyback: did you update the unity-mir branch for 0.1.8 ? (just wanted to make sure i wasn't mistakenly waiting) | 09:11 |
| greyback | kgunn: not yet, in meeting | 09:15 |
| kgunn | ack | 09:15 |
| === chihchun is now known as chihchun_afk | ||
| === chihchun_afk is now known as chihchun | ||
| greyback | kgunn: https://code.launchpad.net/~alan-griffiths/unity-mir/compatibility-with-mir-0.1.8/+merge/214610 | 09:58 |
| kgunn | greyback: thanks...sorry about that...my fault | 10:00 |
| Saviq | duflu, https://bugs.launchpad.net/ubuntu/+source/unity8/+bug/1304959 | 10:25 |
| ubot5 | Ubuntu bug 1304959 in unity8 (Ubuntu) "unity8 crashed with SIGABRT in qt_message_fatal()" [Medium,New] | 10:25 |
| duflu | Saviq: That's the current (stable) Mir. I can't see any issue with Mir as yet | 10:34 |
| kgunn | racarr__: here's some fun... | 14:03 |
| kgunn | https://bugs.launchpad.net/mir/+bug/1305080 | 14:03 |
| ubot5 | Ubuntu bug 1305080 in Mir "libmirclientplatform-android isn't loaded upon libmirserver install" [Undecided,New] | 14:03 |
| kgunn | alf__ might have an opinion | 14:03 |
| mterry | kgunn, I tested your Mir 0.1.8 silo, seemed to work for me | 14:18 |
| mterry | kgunn, in related news, is there a silo open for split greeter again? I think I can avoid the Mir troubles I was having last silo by just basing off 0.1.8 and patching it for the one fix I need (which doesn't break ABI) | 14:19 |
| PreSSion | hello! I want start to develoment a tool to be used in multiplataform linux smartphones, tablets and pc, I going to choose Qt, but I need the develoment of the MIR protocol is seriously and ubuntu will abandon this. | 14:21 |
| PreSSion | and sry for my "engrish" | 14:21 |
| kgunn | PreSSion: hi, Mir is serious and won't be abandoned | 14:30 |
| kgunn | PreSSion: consider the fact that Mir is _the_ display server being used by default in our ubuntu on phones & tablets | 14:31 |
| PreSSion | nice | 14:31 |
| kgunn | and we are working toward eventually making it the default display server on the desktop as well | 14:31 |
| PreSSion | because I want develoment a "yakuake" but not a command line, a tool to make a human script software, where you can say script in human language, for example: Delete files from /home/user that contain the word hello, and a voice input and i will try to add google voice | 14:33 |
| PreSSion | I think woulkd be fine, I have got money and I will have got full time then summer | 14:33 |
| PreSSion | I will write in Qt | 14:34 |
| PreSSion | and kgunn, just the last question, MIR will be use in the future in PC's¿? | 14:35 |
| kgunn | PreSSion: yes, absolutely, it will be used in our desktop / PC distributions | 14:37 |
| PreSSion | thanks! good bye | 14:37 |
| kgunn | PreSSion: you're welcome | 14:37 |
| === jono is now known as Guest1478 | ||
| === olli__ is now known as olli | ||
| fginther | kgunn, is josharenson the right contact for the benchmarking work? | 16:26 |
| kgunn | fginther: he is!...i think he was able to connect with robotfuel but dunno | 16:28 |
| kgunn | he might have more questions | 16:28 |
| robotfuel | fginther: is there a new way to stop unity8 that you know of? | 16:29 |
| robotfuel | fginther: stop unity8 as the phablet user used to work, but now I get stop: Unable to connect to Upstart: Empty address '' | 16:30 |
| fginther | robotfuel, I don't have any specific knowledge of this, try asking in ubuntu-ci-eng | 16:31 |
| fginther | kgunn, thanks | 16:32 |
| kgunn | fginther: no thank you! | 16:32 |
| kgunn | Saviq: ^ any ideas on robotfuel's ? | 16:33 |
| Saviq | robotfuel, upstart died for you | 16:38 |
| Saviq | robotfuel, there's a known crash | 16:39 |
| Saviq | robotfuel, restart lightdm as root | 16:39 |
| === chihchun is now known as chihchun_afk | ||
| === chihchun_afk is now known as chihchun | ||
| === om26er_ is now known as om26er | ||
Generated by irclog2html.py 2.7 by Marius Gedminas - find it at mg.pov.lt!