=== chihchun_afk is now known as chihchun === duflu_ is now known as duflu [06:37] Man, don't double-tap in gdb... [06:38] * RAOF 's console scrolls through the 2MB or so of symbol names [06:39] RAOF, verbosity for the win :) [06:40] Ahem. [06:40] Create the pipe *before* you fork() [06:40] And we win! [06:43] And with that, EOD! [06:45] Turns out that std::future is a quick way to large symbols. [06:47] RAOF: Not enough bytes in your life? [06:47] Later === ricmm_ is now known as ricmm === chihchun is now known as chihchun_afk [12:07] is anybody of you succesfully using sbuild to build mir? [12:08] for armhf [12:09] i just got weird cmake problems - i get cmake messages like successfully found pkg-config \n egl found and then on the next line pkg-config not found (found version 0.26) and glesv2 not found... [12:13] http://paste.ubuntu.com/6802564/ === alan_g is now known as alan_g|lunch [12:44] anpok: are you following this https://lists.launchpad.net/ubuntu-phone/msg05556.html [12:46] yes, but I also followed the configuration steps from here https://wiki.ubuntu.com/SimpleSbuild === dandrader is now known as dandrader|afk === alan_g|lunch is now known as alan_g === dandrader|afk is now known as dandrader [14:52] anpok: i tried week before last and i was actually able to build...but then failed the build's unit test run [14:53] hmmm, kinda looks like the chroot might not have set up correctly? [14:53] only a guess [15:00] Hmm, I just tried cross-building and got "/usr/lib/gcc-cross/arm-linux-gnueabihf/4.8/../../../../arm-linux-gnueabihf/bin/ld: cannot find /lib/arm-linux-gnueabihf/libpthread.so.0 inside" whatever that means [15:09] ... but zapping the "partial-armhf-chroot" fixes it [15:09] is that a known problem? https://jenkins.qa.ubuntu.com/job/mir-clang-trusty-amd64-build/713/console [15:10] "tests/acceptance-tests/test_protobuf.cpp:104:35: error: inheriting constructors are not supported" [15:11] dandrader: I thought that was supported now [15:11] * alan_g wonders which clang version this is [15:14] clang-3.2 [15:15] dandrader: not a known problem, but that C++ support is fairly recent in clang. 3.4 definitely supports it. [15:16] alan_g, that failed in https://code.launchpad.net/~dandrader/mir/SwitchingBundlePrint/+merge/202447. so the code is already there. maybe CI changed its clang version? [15:18] dandrader: maybe. Or maybe the node is out of date. I guess #ubuntu-ci-eng can tell us [15:18] alan_g, should I ask them or will you do it? [15:18] I can [15:18] alan_g, ok, thanks! [15:26] mterry: do you know which branches were involved in robert_ansell's failed attempt to get USC working without a filesystem endpoint? I'm suspecting the problem was lp:1271655 but can't find his code. [15:28] alan_g, hmm, let me dig. I remember they all had similar names [15:28] mterry: thanks === alan_g is now known as alan_g|tea [15:34] alan_g, hmm, let me dig. I remember they all had similar names [15:34] alan_g, at least lp:~robert-ancell/lightdm/private-mir-connection and lp:~robert-ancell/unity-system-compositor/private-mir-connection [15:34] alan_g, I think only those two... [15:34] alan_g, at the time I also had to grab mir/devel, but that's in trusty now [15:38] alan_g|tea, ^ sorry, had irc issue === alan_g|tea is now known as alan_g [15:42] mterry: thanks === dandrader is now known as dandrader|lunch [15:54] kgunn: yes .. just redoing the chroot resolved it.. === jono is now known as Guest95543 [16:28] alan_g: I pushed the branch with the problematic tests at lp:~afrantzis/mir/mir-output-capture-basic-client-api [16:29] alf_: I'll take a look - which test is the problem? [16:30] alan_g: bin/mir_acceptance_tests --gtest_filter=MirOutputCaptureTest.* hangs in MirOutputCaptureTest.gets_valid_egl_native_window [16:30] alan_g: but MirOutputCaptureTest.gets_valid_egl_native_window by itself runs [16:31] alf_: ack. Building [16:31] alan_g: the last two tests MirOutputCaptureTest.create_and_release_contact_server and MirOutputCaptureTest.gets_valid_egl_native_window are similar [16:31] alan_g: you can get the same effect if you run either twice (with --gtest_repeat) [16:32] alan_g: Removing "delete capture;" from src/client/mir_output_capture_api.cpp alleviates the problem (but causes a leak, of course) [16:34] alf_: I can reproduce the hang (in first incantation above) [16:36] alan_g: The apparrent cause of the hang is the display server not stopping when told to do so, i.e., it seems that the main loop just ignores(?) our stop() call. Pressing ctrl-c unblocks it somehow [16:36] s/apparrent/apparent/ [16:37] alf_: OK - just looking at where the threads are at [16:41] Hmm. exiting the debugger seems to unblock too [16:43] alan_g: it's really weird... it's even more weird that removing the "delete capture" as mentioned above "fixes" the problem... I can't see the connection (unless it is a timing issue?) [16:44] alf_: weird just means there's a fact we're missing [16:50] alan_g: wanna have a chat now about the potential release process changes [17:01] dandrader_: fginther & cjwatson figured out what happened. === dandrader_ is now known as dandrader [17:15] alan_g, great! [17:17] alan_g, with ~vanvugt/mir/fix-1271853, would you be okay with me top-approving? [17:18] kdub: yes, you have the right. [17:19] alan_g, :D was just checking if that needs-info was a blocker [17:19] kdub: it would be nice to know === jono is now known as Guest52764 [17:37] Saviq: i've mixed a bug in cmake, and mir & unity8 cross-building works once again out of the box. [17:37] s/mixed/fixed/ [17:38] dont mix the fixes :) [17:47] kdub, I'm running mir/devel on my mako with USC. USC is crashing with "error binding buffer to texture". Any clues? [17:50] mterry, new to me [17:54] kdub, it happens because eglCreateImageKHR returns EGL_NO_IMAGE_KHR in mga::Buffer::bind_to_texture() [17:54] kdub, this doesn't happen with trusty mir [17:56] mterry, let me try something === alan_g is now known as alan_g|EOD [18:08] xnox: i thought it work again but now I got that message: make[1]: Leaving directory `/«PKGBUILDDIR»/obj-arm-linux-gnueabihf' [18:08] dh_install -O--parallel -O--fail-missing [18:09] dh_install: usr/lib/arm-linux-gnueabihf/libmirserver.so.13 exists in debian/tmp but is not installed to anywhere [18:09] dh_install: missing files, aborting [18:09] make: *** [binary] Error 255 [18:09] dpkg-buildpackage: error: fakeroot debian/rules binary gave error exit status 2 [18:09] anpok: with mir or unity8? [18:10] anpok: i've compiled mir_0.1.3+14.04.20140108-0ubuntu1_armhf.build as in the archive [18:10] anpok: http://paste.ubuntu.com/6804190/ [18:10] anpok: sounds like a packaging change is needed of whichever revision shows that bug above. [18:11] ah [18:11] building a mir branch [18:12] i think i have mixture of abi verison 13 and 14 [18:15] mterry, i see... [18:15] : Unsupported buffer format: 3. [18:15] this is probably a regression of some sort [18:15] i havent been tinkering in this area lately, not sure what it could be at the moment [18:15] Yar. Is this from that alpha patch? [18:16] it might be, lets try backing it out and seeing if that helps [18:18] mterry, i can try in about... [18:18] 30 min [18:18] kdub, I'll give it a go [18:19] mterry, at any rate, i think that robotfuel has some scripts for firing up a server and checking a client against it [18:19] we should improve those to fire a server, nested server, and client [18:19] robotfuel, ^ pretty please :) [18:20] * robotfuel looks at backlog [18:20] robotfuel, especially since we're about to enable nested mode [18:20] by default [18:20] because I don't think we have a full test for nested [18:20] in the CI process [18:20] what are you trying to run? [18:20] mterry: what are you trying to run? [18:21] ./mir_demo_server -f /tmp/socket1 & ./mir_demo_server --host-socket /tmp/socket1 -f /tmp/socket2 & ./mir_demo_client_egltriangle -m /tmp/socket2 [18:21] robotfuel, so the problem I hit is using nested mode. I'm trying to run unity-system-compositor (USC) and it crashes with "error binding buffer to texture". But kdub was just saying that we don't have good test coverage for nested mode in general (which allows such regressions to slip in) [18:21] robotfuel, so apparently you have a nice script for testing that may benefit from adding a nested server [18:22] robotfuel, i'm just suggesting once we get the problem we're seeing fixed, we should improve the CI to protect the nested server from regressions [18:22] mterry: yes we could easily add a test for that. let me give you a link to the current one. [18:22] which me or another mir team member could help with :) [18:22] robotfuel, thanks [18:28] mterry: there is this https://bazaar.launchpad.net/~chris.gagnon/+junk/mir-medium-test-runner-for-jenkins/view/head:/mir-mediumtest-runner.sh and this https://bazaar.launchpad.net/~chris.gagnon/+junk/mir-demo-runner/view/head:/bin/mir-demo-tester [18:40] mir-demo-tester is in a all the way finish because I thought shell bindings for subunit were going to land, but they haven't yet. https://code.launchpad.net/~thomir/subunit/trunk-add-cmd-line-script [18:40] s/is in/isn't [18:49] robotfuel, I'm less interested in running the script myself than just making sure a nested server gets added to it to avoid regressions in the future. Something like kdub's line above: [18:49] ./mir_demo_server -f /tmp/socket1 & ./mir_demo_server --host-socket /tmp/socket1 -f /tmp/socket2 & ./mir_demo_client_egltriangle -m /tmp/socket2 [18:50] mterry: does that automatically stop the server when it's finished? [18:51] I doubt it [18:57] mterry, filed a bug https://bugs.launchpad.net/mir/+bug/1272041 [18:57] Ubuntu bug 1272041 in Mir "mir nested server failure: what(): error binding buffer to texture" [Critical,New] [18:58] kdub, awesome. I'm compiling a version of mir just before the alpha bits landed for testing. But compiling takes a while [19:00] mterry, okay :) (also, you know of cross-compile-chroot for mir?) [19:00] kdub, I've cross compiled with pbuilder before, but it's not much faster [19:00] not much use for packages, but if you're working in a mir-only world for some problems, its a real time saver [19:01] well, thats an emulated cross compile, the script does a native cross compile [19:01] eg, a toolchain that runs on x86 that produces armhf libs [19:01] hmm. I'm so wedded to the packaging workflow, but that sounds tempting [19:03] mterry, yeah, i just thought I'd throw it out there [19:13] mterry, lp:mir/devel rev 1328 is the culprit [19:14] kdub, aha [19:15] kdub, yup, that's the alpha patch I was going to test [19:15] anpok, ^ [19:16] bug 1272041 [19:16] bug 1272041 in Mir "mir nested server failure: what(): error binding buffer to texture" [Critical,New] https://launchpad.net/bugs/1272041 [19:29] re [19:30] xnox: yes thanks for parsing the error message [19:36] mterry: that happened on n10? [19:37] anpok, mako (n4) [19:47] * mterry goes to gym for a bit [21:25] anpok, any luck? Any testing I can help with? [21:42] kdub: so it seems mir is working mostly fine with flo (nexus 7 - 2013), I have just one critical issue atm [21:42] kdub: after mir/powerd turns the screen off, nothing can turn the screen on again [21:43] so the rendering keeps working, but screen is constantly off [21:43] had to disable powerd to have unity8/mir working with it [21:43] and this also happens when running mir_demo_server_basic [21:44] it works fine first time you run it, but it seems it turns the screen off automatically when you stop it [21:44] so next time you run the screen will be completely off, as it fails to turn on the screen again [21:45] kdub: I'm just uploading the image I got here, will have the link ready in a few [21:45] rsalveti, cool, its encouraging mir works [21:50] rsalveti, taking a late lunch, be back shortly [21:52] kdub: ok [22:08] mterry: kdub already has a solution which i can only verify to work on none n4 devices - lp:~andreas-pokorny/mir/fix-1272041 [22:10] anpok, ooh, will test [22:16] kdub: flash http://people.canonical.com/~rsalveti/aosp/flo/, then http://cdimage.ubuntu.com/ubuntu-touch/daily-preinstalled/20140123.1/trusty-preinstalled-touch-armhf.zip via recovery [22:17] kdub: the android container takes ~1 min to load, still investigating why, and current mir will be broken because of hwcomposer 1.2, so if you want a working unity8 just use my ppa: https://launchpad.net/~rsalveti/+archive/ppa [22:17] and make sure to disable powerd as well [23:04] rsalveti thanks