[06:32] <zzarr> duflu, thanks for your response about the Intel HD 3000
[06:39] <duflu> zzarr: tis OK. That's what we're here for
[06:41] <zzarr> I can hardly wait until I get the computer, it's an HP Probook 4330s with an upgraded CPU to a i7-2670QM and I will install a 480GB SSD + 8GB RAM :-D
[06:42] <zzarr> does XMir work out of the box or do I have to configure it myself?
[06:44] <duflu> zzarr: Xmir has been rewritten and is no longer designed as a full desktop replacement, but just a way to launch X apps in Mir. The Unity8 desktop is probably not ready for general usage but you can install it and switch between Unity7/8
[06:46] <duflu> zzarr: Excellent timing for a new computer, with 16.04 coming
[06:47] <zzarr> duflu, okey, it was the ability to run X-applications I wanted, but that should just work if I understood you correctly
[06:47] <zzarr> duflu, yes, it is :-D
[06:47] <zzarr> I will get it this week :-D
[06:50] <duflu> zzarr: Also worth noting, Libertine/Puritine (the launcher we're developing to transparently run X apps) is under development and I hear under redesign too
[06:51] <duflu> I don't know much about it. But am defacto maintainer of Xmir itself, which you can just keep running as an X server if you know how
[06:52] <zzarr> duflu, okey, what is the preformance like with Xmir?
[06:53] <duflu> zzarr: Quite reasonable, but not as good as native Unity7. On desktop it will default to high performance mode, but that is not stable. If in doubt use -sw to run it in stable software mode like we do on mobile
[06:53] <zzarr> duflu, okey
[06:53] <duflu> ... which will make it slower
[06:54] <zzarr> duflu, it will.... I assumed that
[06:54] <duflu> I'm working on performance enhancements that will benefit both Unity8 and Xmir right now. Although they're probably months away from release
[06:54] <zzarr> duflu, 16.10?
[06:54] <duflu> Yes
[06:54] <zzarr> nice
[06:56] <zzarr> duflu, could you help me later (when I got my new computer) to configure Xmir? (or is there a good tutorial?)
[06:56] <duflu> zzarr: Yes, it's easy
[06:57] <duflu> No official docs
[06:58] <zzarr> duflu, thanks :-D
[06:59] <zzarr> duflu, will Mir have a remote control possibility (like ssh -X ...)?
[06:59] <duflu> zzarr: Nothing officially planned, but never say never
[07:01] <zzarr> duflu, okey
[07:02] <zzarr> duflu, how does Mir handle multiple GPU's? (Some laptops have that)
[07:02] <RAOF> It doesn't, at the moment.
[07:02] <zzarr> okey
[07:02] <RAOF> It's on the near-term TODO list.
[07:02] <zzarr> the one I'm buying have only the Intel card
[07:02] <duflu> Fortunately RAOF is a sucker for punishment and working with that right now
[07:03] <RAOF> There's nothing fundamentally difficult about making it work (all the hard parts have already been done; mostly in the kernel) it just requires providing the right interfaces and mushing them together.
[07:04] <zzarr> RAOF, are you implementing it in such a way that one could plug'n'play a GPU? (like Thunderbolt 3 eGPU)
[07:05] <RAOF> That should be possible. I'm not sure if the kernel will play nicely with GPU hot plug, though :)
[07:06] <RAOF> For those playing at home, eglQueryDeviceStringEXT(devices[i], EGL_DRM_DEVICE_FILE_EXT) works better if you've actually filled in the devices array.
[07:06] <duflu> I've been thinking external GPUs are going to become a hot topic. Possibly the simplest path to a VR-capable machine for many people (assuming they have Thunderbolt 3)
[07:07] <zzarr> nice, I'm sure the kernel ether will play nicely or will in the future
[07:07] <duflu> Because you could theoretically match Vive/Oculus requirements with a laptop that way
[07:07] <RAOF> External GPU enclosures are surprisingly expensive.
[07:07] <zzarr> duflu, that's true
[07:08] <duflu> Expensive yes, but just a theoretical way to avoid large gaming rigs
[07:08] <zzarr> RAOF, yes they are, but I think they will be less expensive as the demand increases
[07:09]  * duflu doesn't see any reason why a cable and PCIe enclosure should stay expensive
[07:09] <duflu> And power supply
[07:09] <zzarr> I also think that there will be monitor/enclosure combos (all in one package)
[07:09] <RAOF> That would be pretty cool, actually.
[07:11] <duflu> Sadly requires prior investment in a very recent laptop. So nowhere near cheap
[07:12] <duflu> Although I guess micro desktops will follow with USB 3.1
[07:12] <zzarr> I thought of another thing one day, it would be awesome if Intel made a CPU with 2 M series cores (like the one in 12" Macbook) and a i5 or i7 quad core in one package in a 10" or 12" tablet
[07:13] <duflu> I'm sure they have considered it and have Apple design reasons why it doesn't exist... yet
[07:13] <RAOF> I don't think Intel fabs are set up for the low-power silicon process that would make that useful?
[07:14] <zzarr> and you could dock it with a keyboard/mousepad part which had a bigger battery and a stronger GPU
[07:14] <duflu> RAOF: The 12" Macbook is a reasonable proof of concept that it's catching up
[07:14] <RAOF> AIUI ARM's BIG.little architectures make sense because they use a different process/silicon for the little CPUs optimised for low leakage power rather than low active power?
[07:15] <zzarr> so when in tablet mode the M series core should be active and when in laptop mode the i5/i7
[07:15] <zzarr> RAOF, yes, that's true
[07:15] <RAOF> Why not just downclock/downvolt the i7 in tablet mode?
[07:15] <zzarr> RAOF, that works too :-)
[07:15] <duflu> That's a point. The larger cache of an i7 would still help
[07:16] <zzarr> perhaps turning of 2 cores?
[07:16] <RAOF> Race-to-idle is still a thing :)
[07:16] <zzarr> what is that?
[07:16] <RAOF> I think the current state of the art (for Intel, at least) is to run as fast as possible to process all the available work and then idle in an extremely low power state.
[07:17] <RAOF> Hence race-to-idle.
[07:17] <zzarr> ohh, I see
[07:17] <RAOF> Hence the counterintuitive result that limiting the CPU clock rate usually doesn't save you power.
[07:18] <RAOF> Because idle CPUs use much less power than even the lowest-power active state.
[07:18] <zzarr> RAOF, a valid point
[07:19] <zzarr> is it possible to suspend an application without suspending the hole system?
[07:19] <RAOF> Yes; Ubuntu Touch does it all the time.
[07:19] <zzarr> will the desktop do that too?
[07:19] <RAOF> (SIGSTOP is the relevant piece of posix)
[07:20] <RAOF> No, because it's got severe drawbacks (such as only your focused application being able to do anything at all) :)
[07:20] <zzarr> RAOF, yea, you're right
[07:21] <duflu> zzarr: You can do it any time:    'kill -STOP <pid>' and 'kill -CONT <pid>'
[07:21] <duflu> Just be careful what you stop
[07:21] <zzarr> RAOF, well the criteria for suspending/not suspending should be different for desktop
[07:22] <zzarr> duflu, very
[07:29] <zzarr> duflu, is it possible to hibernate an application too?
[07:29] <RAOF> What's the difference between that and suspending it?
[07:30] <duflu> zzarr: Not that I know of, but STOP will keep it frozen for the lifetime of the system (till you reboot)
[07:30] <RAOF> (CRIU might be the google-sauce you're looking for)
[07:30] <zzarr> suspending is stopping/hibernate is to save it to disk too
[07:31] <RAOF> zzarr: https://criu.org/Main_Page
[07:31] <duflu> SIGSTOP/ freezing all your processes since 1973
[07:32] <zzarr> thanks for the link, interesting
[07:32] <tjaalton> hi, anyone around to test mesa 11.2.0-rc4? I'd like to upload 11.2.0 to xenial today
[07:33] <tjaalton> there shouldn't be any surprises, at least the mir-egl patch didn't need updates for this
[07:37] <tjaalton> and if not, what would I need to do to test it myself?
[07:40] <RAOF> tjaalton: Install it, fire up a unity8 desktop session?
[07:41] <tjaalton> ah, too easy then :)
[07:41] <tjaalton> might even have that installed somewhere..
[07:59] <tjaalton> I see a bunch of segfaults due to banfdaemon, and a black screen
[08:00] <tjaalton> oh and upstart taking most of cpu
[08:20] <tjaalton> so banfdaemon and ibus-ui-gtk3 are segfaulting on unity8 startup here
[08:29] <seb128> tjaalton, you mean bamf instead of banf? unsure why that would start under unity8, it's quite x11 centric I think and unity8 doesn't need a matching service that's built in u8/mir
[08:29] <tjaalton> seb128: oh right, bamf
[08:30] <tjaalton> I just installed unity8-desktop-session-mir and tried to start it
[08:30] <tjaalton> but doesn't run
[08:32] <seb128> not likely due to bamf or ibus
[08:35] <tjaalton> so how to test it without hosing another laptop? does it run on qemu?
[08:36] <tjaalton> lighdm looks all weird on this machine now
[08:36] <tjaalton> better after reboot, but still can't select unity from it
[08:37] <tjaalton> err, unity8
[08:37] <tjaalton> huh, it got removed
[08:38] <tjaalton> ok now I got something
[08:38] <tjaalton> another password prompt?
[08:38] <tjaalton> with a big circle
[08:40] <tjaalton> ok it works, just kbd layout was wrong
[08:41] <tjaalton> works as in the session runs, apps don't start
[08:42] <tjaalton> how do I tell it should be this broken or due to something I did?
[08:42] <tjaalton> there's the scopes window open, clicking browser or system settings opens a window which immediately closes
[08:49] <duflu> tjaalton: Yeah the immediate closing is a known bug (ted owns it)
[08:50] <duflu> At least one of them anyway
[08:50] <duflu> It got mostly fixed in the past month or so tho
[08:50] <duflu> ?
[08:51] <tjaalton> ok, this is fresh xenial
[08:51] <seb128> tjaalton, cgmanager needed to be active at some point, unsure if that's still true
[08:51] <seb128> try to systemctl start it
[08:51] <seb128> if it's not
[08:51] <seb128> to see if that makes a difference
[08:52] <tjaalton> wasn't running, doesn't help though
[08:52] <seb128> :-/
[08:52] <seb128> do you get anything in .cache/upstart/<job>.log
[08:52] <duflu> Oh that's a cool test. A non-nested server has the same latency as a nested server if you just raise the buffer queue from 3 to 5
[08:56] <tjaalton> seb128: which job?
[08:57] <tjaalton> no log for cgmanager
[08:57] <seb128> tjaalton, whatever you try to start
[08:57] <seb128> ubuntu-system-settings for example
[08:58] <seb128> webbrowser is known to not work atm I think
[08:58] <tjaalton> nope
[08:58] <seb128> rm .cache/upstart/*
[08:58] <seb128> try to start settings
[08:58] <seb128> see what gets written?
[08:58] <seb128> also #ubuntu-unity is likely a better channel
[08:58] <seb128> those issues are not mir ones
[08:59] <seb128> you might get more people about the help in the unity team
[08:59] <tjaalton> sure
[08:59] <tjaalton> guess mesa is working fine so I'm fine :)
[08:59] <tjaalton> er, happy
[08:59] <seb128> :-)
[09:03] <duflu> There are pixels. But they do not do as instructed.
[09:03] <duflu> Pixels have been sufficiently herded though
[20:34] <josharenson> I'm trying to get the greeter (running under mir) to launch a unity7 session (under X). It wasn't working before because of a lightdm bug, but I think I've fixed it, and now I get this exception in u-s-c's log http://pastebin.ubuntu.com/15637751/
[23:32] <RAOF> josharenson: Hm. Looks like usc is giving up DRM master and then trying to change display configuration.
[23:33] <josharenson> RAOF: Should it just stop when X is started?
[23:33] <RAOF> If that's the plan, yes.
[23:33] <RAOF> It should stop *before* X is started :)
[23:33] <josharenson> RAOF: ok, so its likely a lightdm issue then I suppose (since lightdm is orchestrating all of this)
[23:36] <RAOF> Possibly. USC also shouldn't try to change display configuration after being paused.