=== chihchun_afk is now known as chihchun [02:30] kdub_: Hello? [03:12] duflu: he's actually US eastern standard time (he moved back to michigan)....so its 11pm there [03:13] just for future reference [03:13] Ah. Forgot [03:13] kgunn: How long for? [03:13] Oh "moved" [03:13] duflu: yeah...for the foreseeable future === chihchun is now known as chihchun_afk === chihchun_afk is now known as chihchun === chihchun is now known as chihchun_afk === chihchun_afk is now known as chihchun === renato is now known as Guest59288 [08:35] alf_: (non-urgent) does the latest version work for you? https://code.launchpad.net/~mir-team/mir/enable-late-release/+merge/227178 [08:44] alan_g: yes, although I think that if the failure is consistently reproducible we could have a test for this, checking the process exit status, like we do e.g. for ServerShutdown removes_endpoint_on_signal (but checking for a successful exit) [08:44] alan_g: not sure about any blocking details though [08:47] alf_: In principle, we could add a test that execs (say) the new test program. In practice, we need to be sure that it exists somewhere to be run. FWIW There are other tests like this I'd like to run but haven't built the infrastructure. [08:51] alan_g: what stops us from just using launch_client_process(client_config) instead of an external client? [08:53] alf_: I guess that might work, but remember it calls exit() in the teardown code instead of exiting from main() - a scenario with potentially significant differences [08:54] alan_g: ok (in any case, test not a blocker, but a nice-to-have) [08:56] * alan_g feels increasing motivation to write yet another test execution framework [09:15] RAOF: are your (blocking) concerns addressed? https://code.launchpad.net/~mir-team/mir/libmircommon/+merge/226704 [09:26] anpok: did you file a bug about Mir not showing the latest frame in some cases? [09:51] camako: are your concerns addressed? https://code.launchpad.net/~mir-team/mir/libmircommon/+merge/226704 [09:51] * camako checks [09:54] alan_g, yes.. approved [09:54] thanks [09:55] np [10:03] alf_: no I reflashed the device, another time and the issue was gone [10:03] -, [10:04] anpok: do you remember how to (try to) reproduce the issue? [10:04] alf: pin entry dialog - since there is no animation/blinking cursor [10:31] I am cross-compiling libmirserver23 from pristine trunk and copying the shared object to the phone, but unity8 crashes when using it (usc works normally) [10:31] Has anyone seen that? === dandrader is now known as dandrader|afk === dandrader|afk is now known as dandrader [11:18] alf_: Development branch is libmirserver24 :) [11:18] * duflu goes into hiding [12:11] anpok: ping [12:24] greyback: pong [12:24] greyback: thinking about injecting input? [12:25] anpok: slightly different topic, we're getting a crash with QtComp, wanted to ask your advice [12:25] anpok: crash happens running with autopilot, so apps are stopped & started very rapidly [12:26] and input events can be injected after the surface has been destroyed [12:26] anpok: http://pastebin.ubuntu.com/7835652/ is the stacktrace I get [12:28] could it be that if QtComp tries to send event to client which has stopped, but mir hasn't fully cleaned up after it, mir could call mir::AsioMainLoop::register_fd_handler to create a fd for that client again? [12:29] terminate called after throwing an instance of 'boost::exception_detail::clone_impl >' [12:29] what(): assign: File exists [12:29] is the exception thrown [12:29] greyback: hm not calling surface->consume? [12:30] anpok: we should be. BT doesn't seem to show that though [12:30] ok [12:30] compiler might have inlined it? [12:30] hm cool devirtualization [12:37] hm we are not creating a fd there.. [12:49] anpok: any ideas/theories? [13:01] hm i am following the call down to asio [13:02] the constructor seems to register the fd with the epoll fd [13:02] i would have assumed that happens some time later [13:03] EEXIST -> the supplied file descriptor fd is already registered with this epoll instance.. [13:04] do we have strace logs? [13:08] anpok: they can be got. I can also show you how to reproduce, whatever you like [13:09] oh then i take option 2 [13:10] ok. Install the qtcomp silo (https://wiki.ubuntu.com/Unity8/QtComp) [13:10] then we want to run an autopilot test suite [13:10] so [13:10] https://wiki.ubuntu.com/Unity8/QtComp [13:10] oops [13:10] phablet-config edges-intro --disable [13:11] adb shell powerd-cli display on bright& [13:11] phablet-test-run -p ubuntu-html5-ui-toolkit-autopilot ubuntu_html5_ui_toolkit [13:11] the last command should run the test. You'll see an app start & stop repeatedly [13:12] I often get the crash during the test suite run [13:13] and it's always after a stream of key events that it crashes [13:16] greyback, anpok, I got a bt with a tad more details http://paste.ubuntu.com/7836063/ === Guest59288 is now known as renatu === tvoss is now known as tvoss|errands [13:36] there is a race but not yet sure how to fix it [13:45] anpok, a race between who? [13:48] i guess between unregistering an fd and registering another one with the same fd number.. [13:51] ah no the fd is kept alive through the InputChannel even when the surface is gone === dandrader is now known as dandrader|afk === dandrader|afk is now known as dandrader [14:23] anpok, maybe the new fd type would help :) === alan_g is now known as alan_g|tea === alan_g|tea is now known as alan_g [14:44] Morning [14:45] greyback, dandrader: do we have ticket for that issue yet? [14:45] anpok: no [14:45] I can log one [14:47] ticket? [14:48] bug [14:49] dandrader: wording of my past... [14:49] that sometimes creeps into present [14:49] :) [14:51] hm trying to reproduce but I dont think that I get the error that I should get - [14:52] 'boost::exception_detail::clone_impl >' [14:52] what(): Failed to send message to server: Bad file descriptor [14:54] anpok: https://bugs.launchpad.net/mir/+bug/1346952 [14:54] anpok: hmm no, I reliably get a different error [14:54] s/reliably/always/ [14:59] you always get the other exception message from above? [15:02] I do [15:27] need one more review on https://code.launchpad.net/~mir-team/mir/touchspot-visualizer/+merge/224731 [15:27] cemil says he doesnt count as a proper review [15:27] :p [15:39] I'm trying to port UT in my device, device boots but unity-system-compositor is failing all the time. I've the similar issue as posted here https://bugs.launchpad.net/ubuntu/+source/ubuntu-touch-session/+bug/1283326/comments/3 [15:39] what is most recent qtcomp ppa/silo/thing I can install on my phone? [15:39] and here https://bugs.launchpad.net/ubuntu/+source/ubuntu-touch-session/+bug/1283326/comments/4 [15:39] trying to avoid rebuilding the whole stack while testing some things [15:40] what should I do to fix this ? [15:41] anpok, the crash is happening during a long and rapid stream of key events [15:41] anpok, as autopilot generates those key events synthetically [15:43] racarr_: https://wiki.ubuntu.com/Unity8/QtComp - silo6 [15:44] but I guess you already know this as you're able to reproduce it locally (event if it not yields the same crash) [15:44] Please, anyone can give me a solution? I've some logs if anyone wants to see :) [15:46] greyback: Ah thanks [15:46] greyback: Going to add touchspots [15:46] to qtmir/unity8 [15:47] racarr_: why? [15:47] unless you happen to be jumping with joy for new tasks [15:47] I figured I had some time to do it [15:47] thomi wants it [15:47] to assure quality [15:47] racarr_: I was expecting them to be partof USC [15:48] groot_: I'm guessing that unity-system-compositor didn't start up. You need to investigate if Mir works on your device. Get the "mir-demos" package and try out mir_demo_server_shell and some demo clients with it [15:48] I mean they could be....bar some other stuff that is taking a while [15:48] but I kind of feel like why not QML lol [15:48] because you probably want them to fade or pulse or burn on fire something [15:48] racarr_: MultiPointTouchArea is your friend then [15:48] not you personally [15:48] lol [15:49] I want them with particle effects [15:49] greyback: Oh thats one way... [15:49] I added a new [15:50] TouchspotVisualizer interface actually that juset gives raw pressure/point... [15:50] before they become input events or anything (down at input reader level) [15:50] greyback, thanks. Before I do that, can you take a look at some logs? Maybe this'll help you find the exact problem. [15:50] so plan was to add an interface for that to qtmir [15:50] sounds complicated. MulitPointTouchArea is very easy [15:51] hmm *goes to read* [15:52] groot_: show me the logs. I'm not much good to you if mir doesn't work, I'm not a core mir developer [15:53] greyback, no problem. here's syslog http://paste.ubuntu.com/7836697/ [15:53] lightdm.log http://paste.ubuntu.com/7836703/ [15:54] unity-system-compositor strace http://paste.ubuntu.com/7836711/ [15:54] greyback, i think handholding him through running the test apps might be a start (i sent groot_ here after checking his logs) [15:55] ogra_, you're right :) [15:56] groot_: seems mir fails with "Bad file descriptor" - suspect you'll need someone with more core mir knowledge than I. [15:57] groot_: We will need to try the demo servers like greyback described to get something more [15:57] groot_: but do install the mir-demos package, and try running the mir_demo_*_shell programs to see if they work [15:57] specific...the bad fd error is too weird...if you read [15:58] its a bad fd calling epoll_ctl on two fds which were just created, one an epoll, and one a timer [15:58] and...I dunno that seems reasonable lol. [15:58] so hard to get much out of the logs [15:59] greyback, as ogra said, you may have to help me with this demo run :| [15:59] racarr_, ^^ [16:00] AlbertA, hey, I found that if unity-system-compositor crashes, power button wake doesn't work any more [16:00] groot_: First you need mir-demos then you can run a mir instance [16:00] i.e. mir_demo_server_shell & [16:00] and then pick a client i.e. mir_demo_client_* [16:01] Saviq: this should help: https://code.launchpad.net/~albaguirre/unity-system-compositor/fix-1343919/+merge/227645 [16:01] alf_: I've updated https://code.launchpad.net/~albaguirre/unity-system-compositor/fix-1344101/+merge/227641 [16:02] racarr_, via ADB, right? I've to install mir-demos in desktop and run instance through adb ? [16:02] AlbertA: just looking at that... so the power menu has exactly the same issues the greeter (may not have enough time to render and stale frame) [16:03] groot_: No adb in and then install mir-demos on phone [16:05] AlbertA: perhaps we should apply the workaround you mentioned for the greeter (waiting a bit more to turn on the screen) for the power menu too. So if unity8 shows the menu after 2000ms perhaps we should turn on the screen after 2200ms (or something)? [16:05] alf_: so what are you seeing now? [16:06] racarr_, for that I need net connection, right? [16:06] racarr_, it's failing "E: Failed to fetch http://ports.ubuntu.com/ubuntu-ports/pool/universe/m/mir/mir-demos_0.4.1+14.10.20140714-0ubuntu1_armhf.deb Could not resolve 'ports.ubuntu.com'" :( [16:07] groot_: Yeah... [16:07] downnload on desktop [16:07] and adb push :) [16:07] should be fine [16:09] AlbertA: hmm, I thought I was seeing "power menu hasn't rendered yet" effect, but I now think it's the stale frame effect (I am seeing the greeter momentarily before the power menu shows up) [16:09] alf_: ahh, right [16:09] AlbertA: so I am not sure the extra timeout would help [16:10] alf_: well I think it would help [16:10] alf_: because the dialog doesn't start drawing itself [16:10] until that 2s alarm fires [16:14] alf_: Hey...something I've been meaning to ask. [16:15] Is there any generic way to write to GBM bos without using an opengl context? gbm_bo_write only works for the dumb bos... [16:15] which only work for the cursor bo. [16:15] which has the whole no DRI image problem so you can't EGLCreateImage (as you encountered doing software I guess) [16:16] The thing is. I can't actually use the cursor buffer for the cursors or touchspots [16:16] AlbertA: YMMV, 2200 doesn't make much of a difference, but 2500 is better: the "not rendered yet" effect is gone, but we occasionally get a stale greeter frame [16:16] because software wont work on it [16:16] err [16:16] not software [16:16] GL [16:16] lol === tvoss|errands is now known as tvoss [16:18] I could just use software buffers but then [16:18] AlbertA, oh cool [16:19] no overlay on HWC for touchspots [16:19] racarr_, installed mir-demos from here https://launchpad.net/ubuntu/utopic/armhf/mir-demos/0.4.1+14.10.20140714-0ubuntu1 [16:19] alf_: yeah it looks better with 500ms... [16:19] alf_: so would this suffice? http://paste.ubuntu.com/7836986/ [16:19] groot_: Ok seems about right! So are you able to run mir_demo_server_shell [16:19] which command should I run now? [16:20] alf_: or just run another alarm? [16:20] racarr_, ran it but got "segmentation fault (core dumped)" [16:22] groot_: Can you grab a backtrace with gdb? May have to use the same trick to install it [16:22] AlbertA: Why not make the timeout itself 2500? I don't like too many things running concurrently :) [16:22] alf_: because if the screen is on... [16:22] alf_: and the dialog shows up, you can let go of the button (so power key up) [16:23] before the 2.5 second alarm goes on [16:23] i.e. in between 2 and 2.5s [16:23] so basically that would turn off the screen [16:23] AlbertA: ahh, right, and the screen would go off [16:24] racarr_, mir_demo_server_shell& gives no error. gdb command is available in phone. What command should I run? [16:24] AlbertA: well, I am OK with the thread if we know that screen_state_handler->set_screen_power_mode() is thread safe [16:25] groot_: gdb mir_demo_server_shell [16:25] then run [16:25] then bt [16:26] racarr_: @"Is there any generic way to write to GBM bos without using an opengl context", no that I know of [16:28] racarr_, it's giving "No stack" output. [16:28] racarr_: do you want to avoid rendering with GL because of the hassle or because there is some problem with it? [16:32] AlbertA: EOD, so preapproving MP to unblock [16:32] alf_: Mostly because of the hassle and in particular the hassle of [16:32] AlbertA: do what you think is best :) [16:32] using opengl outside of the renderer [16:34] alf_: yeah it's thread safe...but I think I will just setup another alarm [16:36] groot_: just to confirm, you ran "gdb mir_demo_server_shell" and after a brief loading period, ended up at a command line. You entered "run", hit enter, and you should have seen some output about threads, then it stopped with a error? Did you enter "bt" then? [16:38] racarr_, my bad sorry. Here's the output http://paste.ubuntu.com/7837084/ [16:38] greyback, thanks for pointing out [16:38] groot_: Hmm [16:38] something with your driver is unhappy ;) [16:38] surprise! [16:39] kdub_: ^ does that mean anything to you? [16:39] the stack trace [16:39] groot_: good luck with it, you've gone into alien territory for me ;) [16:41] the segfault being at 0x0 kind of makes me think its deferencing one of the android device functions...hard to tell without debug info in libmirplatformgraphics [16:42] I can't find...our debug symbol packages [16:42] greyback, don't say that. I've already spent a month trying to port ;) [16:42] racarr_: http://ddebs.ubuntu.com/pool/main/m/mir/ [16:43] groot_: nah, hopefully these guys can help you. I just know nothing about working with the android drivers [16:45] greyback: Thanks :) [16:45] groot: If you install the package for libmirplatformgraphics-android from http://ddebs.ubuntu.com/pool/main/m/mir/ [16:45] that corresponds to the version you see from [16:45] apt-cache policy libmirplatformgraphics-android (on the phone) [16:45] then repeat the gdb experiment [16:45] can get a little more info [16:46] android drivers are not my expertise either, but maybe we can at least distill it down in to a bug report with exactly what is going wrong or some such [16:48] racarr_, thanks. just because you're looking at it gives me hope :). I'll try and let you know after a while. [16:48] groot_: Cheers :D [16:48] what kind of device is it? [16:48] (just curious) [16:49] Sony Xperia U. First android series from sony ;) [16:51] Ah cool :) that shouldnt be too old or anything [16:59] groot_, reading through now... [17:00] lunchtime is over :) [17:03] kdub_, thanks :). I'm installing libmirplatformgraphics now [17:03] great [17:03] so it looks like "bad file descriptor" is thrown from mir [17:04] yes [17:04] it could just be having the mesa platform installed [17:04] which I assume you're installing the android one [17:04] kdub_: Look at the segfault from mir_demo_server_shell though [17:05] http://paste.ubuntu.com/7837084/ [17:05] hmm, yes, more tricky [17:06] was it run as root? sometimes we have to change udev permissions === alan_g is now known as alan_g|EOD [17:07] racarr_, here's the output http://paste.ubuntu.com/7837213/ [17:07] kdub_, ^^ [17:09] well, thats interesting [17:10] I'll be back in 5 mins. Thanks for looking at it guys :) [17:13] I guess it means the hwc_composer_device struct [17:13] is bbad [17:13] or at leat this callback is [17:13] 0x0 [17:13] or something...lol [17:16] racarr_, yeah, thats what I think too.. probably some workaround though [17:20] I'm back. So, it can be fixed right ? :) [17:21] groot_, maybe :D are you able to compile/test patches? [17:22] kdub_, I'll do anything if you walk me through it :P [17:24] groot_, its slightly involved, but basically, if you download lp:mir/0.5 [17:25] and then run cross-compile-chroot.sh, that should try to cross compile for the device [17:25] (bzr branch lp:mir/0.5) [17:25] and at that point, we can just cherry-pick a library file and push it to the device (not much abi breakage) [17:28] downloading right now. my net is very slow today :( [17:30] well, the script will go and download a bunch of arm packages and put them in a folder, so not a good day for slow net :) [17:31] kdub_, is there a way to directly download the source without using bzr ? [17:31] not that I know of [17:32] and groot_ what's the resolution of the device? [17:32] i'll leave a patch for you to try if I'm not around [17:32] with my first idea [17:32] 480 x 854 [17:32] cool [17:34] out for lunch will be back soon though [17:35] racarr_, yup come soon :p [17:35] dl speed is now 0 !! :( [17:55] groot_, well, my first shot at figuring out what is wrong is here: lp:~kdub/mir/sony-experiment [17:55] (just hardcoding some stuff to see if that gets a sane mir environment) [17:55] kdub_, looks like it'll take a while to download the source :(. Will you be available ? [17:56] a few more hours [17:57] hopefully it'll be complete by then [17:59] very dirty hack :D [18:17] kdub_, is this the correct source https://launchpad.net/mir/0.5/0.5.0/+download/mir-0.5.0.tar.bz2 [18:18] ah, good to know [18:18] downloaded :) [18:19] kdub_, what should I do now ? [18:20] install the arm cross compiler (g++-4.9-arm-linux-gnueabihf) and try that script (cross-compile-chroot.sh) in the root dir [18:21] first modify framebuffers.cpp like yours? [18:23] kdub_, ^^ [18:24] yeah [18:28] kdub_, I download this one http://packages.ubuntu.com/utopic/amd64/g++-4.9-arm-linux-gnueabihf/download [18:28] but got error "dependency problems - leaving unconfigured" [18:29] just use apt-get [18:30] kdub_, E: Couldn't find any package by regex 'g++-4.9-arm-linux-gnueabihf' [18:30] guess it's not in the source.list [18:30] oh, /me is on utopic [18:31] you could probably try 4.8 (provided the ubuntu touch image is trusty too) [18:32] kdub_, I'm using utopic image [18:33] the point here :) is that the toolchain version of the host cross compiler should match the version of the phone device image [18:38] kdub_, it says to add the mirror to sources.list. http://packages.ubuntu.com/utopic/amd64/gcc-4.9-arm-linux-gnueabihf/download [18:38] Maybe then I can use apt-get [18:38] I'll try [18:38] sounds like a good suggestion [18:46] gmock incantations [18:48] kdub_, still got unmet dependency error :( [18:49] groot_, maybe apt-get -f install? otherwise, out of suggestions [18:52] kdub_, did that, no change. I just started the cross-compile-chroot.sh. I'll let you know what happens [18:52] if you dont have the cross compiler, it'll fail [18:53] back [18:55] kdub_, it's retrieving many things from net. How can I check if I've any cross compileer installed? [18:55] racarr_, welcome back :) [19:01] kdub_, my desktop has 14.04 but phone's got 14.10. So 4.8 or 4.9 compiler? [19:02] 4.9 i suppose [19:09] kdub_, looks like 4.9 is for 14.10 :(. I'll try with 4.8. What should I expect if all goes smoothly ? [19:10] something on the screen if running that demo shell [19:10] kdub_, I meant after compilation. What files should I push back to phone ? [19:11] the other scripts have an enumeration [19:14] kdub_, sorry, I don't understand. [19:14] the deploy-and-test script === tvoss is now known as tvoss|test === tvoss|test is now known as tvoss [19:37] kdub_, I've to go. Will you be available tommorrow at the same time ? [19:38] yep [19:39] kdub_, then I'll try to compile within this time. Thanks for everything you've done :) [19:39] I'll check back tommorrow [19:40] cool [19:41] OK, bye. [20:15] ummm, why would an alarm not fire [20:15] when coming out of deep suspend... [20:15] uhoh [20:15] so I get the power key down event....which schedules an alarm [20:15] that never fires...but only when out of deep suspend... [20:16] (wild guess) hardware timer got suspended? [20:17] sorry so its like [20:17] you set alarm then [20:17] suspend then come back [20:17] and alarm doesnt fire, or what? [20:17] you suspend... [20:17] screen goes off [20:17] then you press and hold the power key [20:17] so I get a power key down event [20:18] so it should be out of suspend... [20:18] which then schedules an alarm [20:18] that doesn't trigger [20:18] how are you getting power key down events while its suspended [20:18] you mean early suspend [20:18] ? [20:18] kernel [20:18] usually hooks up the power key [20:18] to wake it up [20:18] and take it out of suspend [20:19] oh ok lol [20:19] hmm [20:19] maybe an early resume issue [20:20] like kdub_ said...maybe the hw timers are not brought up yet [20:21] also something else is waking the screen up before USC actually does [20:22] hw timers not up yet is weird youd think] [20:22] someone could give you an error [20:22] my only other guess is maybe some funny signals flying around that mess things up... [20:23] ...cant think of anything specific [20:23] hmm [20:40] so apparently it is an early resume thing.... [20:41] if I wait for powerd to broadcast that it's completely out of suspend before scheduling the alarms...then the alarms fire just fine [21:01] AlbertA: Very weird...wonder what the rules are [21:07] nevermind...it wasn't going to deep suspend that time [21:08] so alarm still doesn't fire...weird... [22:18] oooh.... [22:18] why do I get [22:18] a power key down and up at the same time? [22:18] after coming out of suspend === renato is now known as Guest49276 [22:19] that's what was cancelling my alarm... [22:19] don't know what to do about it though... [22:20] AlbertA: maybe its the power key up from before [22:20] you suspend? [22:20] I guess [22:20] you suspend on up [22:20] nvm [22:21] right it suspends after power key up is received [22:55] umm I don't remember this happening with powerd... which used a more recent version of the android input stack... [23:52] anyone want any reviews or anything before I end my core day? [23:52] the active reviewsd is actually kind of benign right now wow [23:59] :)