[09:22] <alan_g> alf_: your "Needs Fixing" has been addressed. OK now? https://code.launchpad.net/~andreas-pokorny/mir/libinput-package/+merge/269534
[09:23] <alf_> alan_g: yes, (top) approved
[09:28] <alan_g> ta
[09:43] <jibel> anyone would know what the message in dbus.log means in bug 1496773 ? Is it a problem with mir ?
[09:44] <jibel> it started suddenly this morning and makes the device very slow
[09:44] <alan_g> jibel: mir itself doesn't do dbus
[09:46] <jibel> alan_g, the dbus message refers to mir, any idea how I can find the source that triggers this message?
[09:48] <jibel> davmor2, ^ when did the problem start?
[09:49] <alan_g> jibel: well the Mir message says that an application is trying to connect to an invalid prompt session.
[09:49] <davmor2> jibel, alan_g: this morning  uptime is in the paste-bin in the bug report,  I'd not seen any slow down till this morning
[09:51] <alan_g> greyback: any idea why ^^ gets passed to dbus?
[09:54] <greyback> alan_g: not a clue. does the SSO service somehow return it as an error to the dbus daemon?
[09:57] <greyback> jibel: could you attach unity8.log to that bug please?
[09:57] <jibel> greyback, all the logs are in https://bugs.launchpad.net/ubuntu/+source/dbus/+bug/1496773/+attachment/4466701/+files/logs.tar.xz
[09:58] <greyback> jibel: ah nice
[09:59] <alan_g> jibel: From Mir's POV when a prompt helper (the SSO) creates a prompt session it must identify the application. This is failing because Mir doesn't recognise the application (maybe it just exited). So the prompt session is rejected. No idea why failing to create a prompt session causes a dbus message (presumably from the SSO service).
[10:00] <greyback> unity8 is indeed reporting online-account-service is starting and then stopping. No suggestion unity8 is interfering with it.
[10:02] <jibel> alan_g, okay, thanks for the explanation.
[11:50] <dandrader> anpok, ping
[11:52] <anpok> pong
[11:54] <dandrader> anpok, https://bugs.launchpad.net/mir/+bug/1496814
[11:54] <dandrader> anpok, is that easy to fix?
[12:27] <anpok> dandrader: yes I think so...
[13:05] <tsdgeos> alf_: you there? i'm investigating fd leaks in uniy8 process and i'd need some help
[13:05] <alf_> tsdgeos: sure, how can I help?
[13:05] <tsdgeos> alf_: so the fd leaks i'm seeing now are of the type anon_inode:dmabuf
[13:06] <tsdgeos> do you know where/if in mir one of those might be created?
[13:06] <tsdgeos> maybe DMABufTextureBinder ?
[13:07] <tsdgeos> alf_: the leaks now happen only when the process is killed because of memory pressure
[13:07] <tsdgeos> i have as many leaked anon_inode:dmabuf as processes where killed my the memory pressure
[13:08] <tsdgeos> unfortunately strace doesn't seem to "see" the calls that allocate those fds
[13:09] <alf_> tsdgeos: these are fds backing buffer objects
[13:10] <alf_> tsdgeos: that's on the desktop right?
[13:10] <tsdgeos> alf_: no, phone
[13:10] <alf_> tsdgeos: hmm, then I guess that's not the graphics buffers
[13:11] <alf_> tsdgeos: so, DMABufTextureBinder is even there on the phone, it's in the mesa platform plugin
[13:11] <tsdgeos> is -> isn't ?
[13:12] <alf_> tsdgeos: "is *not* event there"
[13:12] <alf_> tsdgeos: right
[13:12] <tsdgeos> ok
[13:12] <tsdgeos> so it's not that :/
[13:12] <tsdgeos> alf_: do you have any idea what may be using dmabuf inotes?
[13:16] <alf_> tsdgeos: not really, it's a general cross-device/driver buffer sharing mechanism, I am only familiar with its use for graphics buffers
[13:16] <tsdgeos> ok, thanks
[13:16] <tsdgeos> i'll keep trying to find out who's creating it
[13:17] <alf_> tsdgeos: did you use strace -f , perhaps a subprocess is involved
[13:17] <tsdgeos> alf_: but a subprocess could not create fds on the unity8 process, no?
[13:18] <alf_> tsdgeos: right, thinko...
[13:21] <alf_> tsdgeos: which calls are you tracking with strace (or all calls)?
[13:21] <tsdgeos> alf_: all of them
[13:21] <tsdgeos> the thing is
[13:21] <tsdgeos> strace doesn't seem to be showing me the calls that create fds of dmabuf type
[13:22] <tsdgeos> since for example i have fd 240 shown in proc
[13:22] <tsdgeos> but the last time 240 shows up in the strace output
[13:22] <tsdgeos> it was because it was a pipe
[13:22] <tsdgeos> and was correctly closed
[13:24] <alf_> perhaps they are sent to unity8 with sendmsg, does strace show fds sent this way?
[13:24] <alf_> tsdgeos: ^^
[13:25] <tsdgeos> i do not know
[13:25] <tsdgeos> i'll check that thanks :)
[13:27] <alf_> tsdgeos: also, now I am reading that android has started using dmabuf internally, so perhaps we can't rule out these being part of graphics buffers
[13:28] <alf_> tsdgeos: (although at a lower level, which we don't access directly)
[13:28] <tsdgeos> right
[13:39] <alf_> tsdgeos: also, sometimes fd leaks are associated with memory leaks (classes holding fd resources internally, especially if the fd handling in the app is robust with RAII etc), so that's something to look out for in case the strace approach doesn't lead anywhere
[13:39] <tsdgeos> totally
[13:53] <kgunn> alf_: i rebooted about the time you were talking about android gfx possibly using dmabuf...curious if i missed anything
[13:55] <alf_> kgunn: nothing new since then
[14:18] <tsdgeos> so it's not sendmsg/receivemsg because it doesn't happen while starting apps
[14:32] <kgunn> tsdgeos: on this idea of closing apps, any chance you've looked at associated screen shots in spread
[14:32] <tsdgeos> kgunn: yeah getting there
[16:18] <alan_g> bschaefer: your vote is "blocking" this. Do you want to reconsider? https://code.launchpad.net/~mir-team/mir/attestable-timestamps-server/+merge/270457
[16:19] <bschaefer> alan_g, well i want to go through and fix the small things mentioned
[16:19] <bschaefer> o wait you pushed some changes :)
[16:20] <alan_g> bschaefer: just some small things
[16:20] <bschaefer> yeah those changes look good
[16:24] <bschaefer> anpok, braces of lambda functor? Should they be line or same line?
[16:24]  * bschaefer reading through mir style
[16:29] <bschaefer> alan_g, everything looks good to me, anything small can be fixed in later MPs
[16:29]  * bschaefer will have to re-do the client code but shouldnt be to hard
[16:29] <bschaefer> and will most likley need some changing around :)
[16:33]  * alan_g thinks bschaefer will need to change "override_the_cookie_factory(mir::cookie::Secret const& secret);" to "override_the_cookie_factory(Builder<> const& builder);" to capture the secret in the builder.
[16:34] <bschaefer> alan_g, well from what i understand you would store the secret before calling override
[16:35] <bschaefer> at lease thats how it was done before when we kept the secret around in tests/default_server_config
[16:35] <alan_g> Where would you get the first secret from?
[16:35] <bschaefer> yes that'll be different
[16:36] <alan_g> I think using a builder will solve it, and be more consistent with the rest of the API
[16:36] <alan_g> ymmv
[16:36] <bschaefer> right, just need to figure out what the builder is :)
[16:36] <bschaefer> yeah we are trying to predict the future, which is fine just not 100% sure how RAOF imagined using it
[16:37] <bschaefer> though this is a bit easier to predict how it *should* be used
[16:38] <anpok> alan_g: there is one thing in the device settings MP.. that I still ponder about.. what would be the best way to indicate whether a setting is applicable for a device..
[16:38] <zzarr> hello! I have a plan, to build a tablet/laptop 2-in-1 unit, I wish to use a strong SBC, I have looked att Hummingbird H88, wish have a A80 CPU
[16:38] <alan_g> [&]() -> std::shared_ptr<CookeFactory> { ... }
[16:38] <zzarr> do anyone know how "good" they are?
[16:39] <anpok> that A80 is an allwinner?
[16:39] <bschaefer> alan_g, cool, i see a lot of other examples in the code as well
[16:39] <zzarr> anpok, yes AllWinner A80 is correct
[16:39] <bschaefer> but mainly we pass in the way we want to build it (i guess?) Ill have to get a better understand once i attempt to implement it :)
[16:40] <anpok> zzarr: where do you get your drivers from? and do those drivers provide access to the hardware overlays?
[16:42] <alan_g> anpok: just pondering... Maybe you should be using a struct of optional fields?
[16:42] <zzarr> anpok, on the page it says that it supports Android 4.4.2 and Linux 3.4.39
[16:42] <anpok> alan_g: but struct might be more problematic for abi compilance?
[16:43] <zzarr> anpok, but that sounds old in my ears
[16:43] <alan_g> if we just add fields at the end it isn't much different to adding to vtab
[16:43] <alan_g> Well, if we disable creating in user code
[16:46] <zzarr> as it seams the source code for the SoC is open and some specifications http://linux-sunxi.org/images/1/10/A80_Datasheet_Revision_1.0_0404.pdf
[16:50] <anpok> zzarr: driver quality is essential...
[16:50] <anpok> i would look out for other projects that try to use it, and look for experiences..
[16:50] <zzarr> I know, I might have to use another SoC
[16:50] <zzarr> yepp
[16:51] <zzarr> I'll do that
[16:52] <anpok> zzarr: if you want to use mir you would have to use android drivers I guess..
[16:53] <anpok> i guess the stock imagination tec winsys drivers wont let .. hmm ..
[16:53] <anpok> that could be an interesting task.. not sure if the image tec winsys interface would allow you implement a mir platform ..
[16:54] <anpok> alan_g: hm ok I will give it a try .. it is kind of annoying that this seems to become a growing struct.. (or structs i could split them into devices classes..)
[16:55] <alan_g> anpok: I've not given it deep thought. Just been musing
[16:57] <zzarr> anpok, yes, but is it a problem that it's an old version of Android?
[17:53] <greyback_> bregma: can I have this added to your "sooner rather than later" list: https://bugs.launchpad.net/mir/+bug/1496814
[17:54] <bregma> greyback_, sure
[17:54] <greyback_> anpok: opinion would be appreciated, as that bug means our cursor handling code fails with trackpads
[18:07] <anpok> nope oh I should add my thoughts on what causes this..
[18:07] <anpok> -nope
[19:14] <mhall119> camako: ping
[19:15] <camako> mhall119, hi
[19:15] <mhall119> camako: hey, I just watched https://www.youtube.com/watch?v=QCsZ2jYUTqk very cool stuff
[19:15] <mhall119> I was wondering if qmlscene and QML apps can be run in a Mir on X window that way
[19:16] <mhall119> and if so, how to tell qmlscene to use Mir instead of X11
[19:17] <camako> mhall119, I haven't tried myself... Have you seen this? ---> https://plus.google.com/+MarcoTrevisan/posts/ZLWYfRMZWNG
[19:17] <mhall119> yes, I did, and I asked Trevinho about it already, trying to get mir_demo_server working on my laptop
[19:17] <mhall119> but I just want to run a single app, not all of Unity
[19:20] <camako> mhall119, TBH, I haven't looked into it yet.
[19:53] <Trevinho> mhall119: ah, if you only need an app, I think what's in wily should be enough to get that working
[19:53] <Trevinho> Since last Mir is needed only for unity
[21:21] <kgunn> Trevinho: that post is kick ass
[21:21] <Trevinho> :)