**** BEGIN LOGGING AT Wed May 16 03:00:13 2018 May 16 05:17:37 Herrie: bumping LXC shouldn't help here, the problem is more between LXC and the kernel May 16 05:19:48 narf, bpaste.net's certificate is out-of-date May 16 05:21:12 Tofe: if you get the container up and running and things still crash that is most likely because the android.img provided on build.anbox.io has the UIDs shifted to 100000, if you're running with --privileged you need to unshift everything May 16 05:40:36 morphis__: OK that means we'd need to re-build the andoid.img right with lower UID? May 16 05:40:55 morphis__: well the issue seems to be introduced with a specific LXC version May 16 05:42:30 morphis__: Seems there's something on launchpad about it: https://bugs.launchpad.net/canonical-devices-system-image/+bug/1516971 May 16 05:42:44 I'm not having it on my 3.18 kernel so it's likely the above May 16 05:43:02 Just I'm not that good with launchpad and couldn't find how they actually patched/fixed it May 16 06:34:26 Herrie: git clone github.com/lxc/lxc and then search through the history for preserve_ns **** BEGIN LOGGING AT Wed May 16 06:34:47 2018 May 16 07:23:39 it failed to build a bit for me over night: May 16 07:23:40 log.qemux86:| /OE/build/owpb/webos-ports/tmp-glibc/work/i586-webos-linux/anbox/3.0+gitrAUTOINC+4a5cb84231-r0/git/external/android-emugl/host/include/OpenGLESDispatch/gles_functions.h:17:10: fatal error: OpenGLESDispatch/gles_common_functions.h: No such file or directory May 16 07:23:44 log.qemux86-64:| /OE/build/owpb/webos-ports/tmp-glibc/work/core2-64-webos-linux/anbox/3.0+gitrAUTOINC+4a5cb84231-r0/recipe-sysroot/usr/include/core/dbus/asio/executor.h:29:7: error: using typedef-name 'boost::asio::io_service' after 'class' May 16 07:23:48 log.qemux86-64:| /OE/build/owpb/webos-ports/tmp-glibc/work/core2-64-webos-linux/anbox/3.0+gitrAUTOINC+4a5cb84231-r0/recipe-sysroot/usr/include/core/dbus/asio/executor.h:29:7: error: using typedef-name 'boost::asio::io_service' after 'class' May 16 07:23:52 log.raspberrypi3:| /OE/build/owpb/webos-ports/tmp-glibc/work/cortexa7hf-neon-vfpv4-webos-linux-gnueabi/anbox/3.0+gitrAUTOINC+4a5cb84231-r0/git/external/android-emugl/host/include/OpenGLESDispatch/gles_functions.h:18:10: fatal error: OpenGLESDispatch/gles_extensions_functions.h: No such file or directory May 16 07:23:57 log.raspberrypi3-64:| /OE/build/owpb/webos-ports/tmp-glibc/work/aarch64-webos-linux/anbox/3.0+gitrAUTOINC+4a5cb84231-r0/git/external/android-emugl/host/include/OpenGLESDispatch/gles_functions.h:17:10: fatal error: OpenGLESDispatch/gles_common_functions.h: No such May 16 07:24:05 the io_service seems to be related to https://github.com/anbox/anbox/issues/637 May 16 07:45:18 JaMa: We don't have those with pyro May 16 07:45:26 So I guess it's due to sumo changes in boost or something May 16 07:48:56 JaMa: it could be gles_common_functions.h is meant to be auto-generated, but I'll need to check that May 16 07:50:36 JaMa: yes it was this commit https://github.com/anbox/anbox/commit/067201babafddf50466c47cc5ad33aec3290d85d May 16 07:50:43 let me see what could have gone wrong May 16 07:52:31 JaMa: you don't have any error related to scripts/gen-emugl-entries.py ? May 16 07:53:05 Tofe: For the preserve_ns issue, seems we might need to backport setns functionality, though not 100% sure. There's a patchset in https://chromium.googlesource.com/chromiumos/third_party/kernel/+/b2cb3912d4219a2a65481d784a5dc99f1b919f2e May 16 07:53:38 It's coming from one of the Ubuntu devs who did a lot of the lxc commits as well May 16 07:54:23 I can try that yes May 16 07:54:58 But we shouldn't try too hard to make anbox run on 3.4 kernels either, and first try to have it running on 3.10+ kernels ilke mido May 16 07:55:15 boost/dbus-cpp issue fixed by patching dbus-cpp May 16 07:58:48 Tofe: I see only execution of shell script scripts/gen-emugl-headers.sh May 16 07:59:37 trying cleansstate May 16 07:59:39 ok then it's weird it didn't generate the header May 16 08:00:04 Tofe: I agree 3.10+ should be the primary focus, but it would be nice if it would run on 3.4 of course May 16 08:00:10 And if a simple patchset can achieve that May 16 08:00:14 JaMa: it's a script started by a CMake command, do you think it's possible to get its output? May 16 08:00:25 can you pastebin your log.do_compile May 16 08:00:27 Herrie|Laptop: yup :) May 16 08:00:45 JaMa: not before tonight, unfortunately May 16 08:01:12 on, np, just wanted to check if scripts/gen-emugl-entries.py is shown in your log.do_compile May 16 08:01:28 but thanks for hint, I'll check in this direction May 16 08:02:01 ah sorry now I see May 16 08:02:12 scripts/gen-emugl-headers.sh is what calls scripts/gen-emugl-entries.py inside May 16 08:03:49 ... issue with python (host-side) somehow ? May 16 08:04:24 JaMa: This could be python 2 vs 3 May 16 08:04:33 There's a PR for that pending in Anbox May 16 08:05:35 it built ok, when executed run.do_compile manually May 16 08:06:42 JaMa: https://github.com/anbox/anbox/pull/512/commits/1df0964af5a31bfeed380caca302d019daf398f2 May 16 08:06:51 Seems script is python2 so you need to force it May 16 08:06:55 Not sure that's in our build May 16 08:07:00 ok, it's a bit fragile build it seems May 16 08:07:43 interesting, when do_compile finishes ok and you execute it again, then it only regenerates the header, if you execute it again, then it rebuilds bunch of stuff because of regenerated headers and if you trigger it again, then just headers again and so on May 16 08:08:06 meh, dependency hell May 16 08:09:21 Herrie|Laptop: python defaults to python2 here (when executing do_compile from devshell) but will try to change shebang and build directly with bitbake May 16 08:10:03 Tofe: Well Sumo is quite newer compared to Pyro May 16 08:10:08 So quite some things got updated May 16 08:10:13 Some issues were to be expected May 16 08:10:54 and this is a bit newer than Sumo :) May 16 08:12:20 JaMa: cmake has also been update ? May 16 08:12:22 updated* May 16 08:13:14 not between Sumo and master IIRC May 16 08:13:23 but I think it was upgraded in Sumo or Rocko May 16 08:14:52 it's race-condition for sure, [21/220] builds GLEScontext.cpp which includes OpenGLESDispatch/gles_extensions_functions.h and scripts/gen-emugl-headers.sh successfully generates this header as [24/220] May 16 08:18:25 ok, so it's a missing dependency actually May 16 08:21:14 dependency declaration seems to be there https://github.com/anbox/anbox/blob/master/external/android-emugl/host/libs/libOpenGLESDispatch/CMakeLists.txt#L7 but should it be *before* add_library ? is cmake that dumb ? May 16 08:22:24 Tofe: it shouldn't May 16 08:24:14 there is a DEPENDS missing in https://github.com/anbox/anbox/blob/master/external/android-emugl/host/libs/CMakeLists.txt#L16 May 16 08:25:49 there are only 3 explicit dependencies on it: May 16 08:25:50 external/android-emugl/host/libs/GLESv1_dec/CMakeLists.txt:add_dependencies(GLESv1_dec GLHeaders) May 16 08:25:53 external/android-emugl/host/libs/GLESv2_dec/CMakeLists.txt:add_dependencies(GLESv2_dec GLHeaders) May 16 08:25:56 external/android-emugl/host/libs/libOpenGLESDispatch/CMakeLists.txt:add_dependencies(OpenGLESDispatch GLHeaders) May 16 08:26:51 Tofe, JaMa : https://paste.ubuntu.com/p/2xjdDC7wWs/ should fix it May 16 08:30:47 I'm testing https://github.com/webOS-ports/meta-webos-ports/commit/8d5f0910751e54f6a554fde26f38c29e5894d956 May 16 08:33:06 built ok with my change for qemux86-64 now, will try 3 more MACHINEs May 16 09:38:16 morphis__: Is there an easy way to unshift the UID? May 16 09:38:25 Or I need to rebuild the Android image for this? May 16 09:40:15 all MACHINEs built ok with my fix May 16 09:42:52 JaMa: Nice 1 line fix then :D May 16 10:07:33 morphis__: do you want me to send PR or can you take it directly? **** BEGIN LOGGING AT Wed May 16 15:51:01 2018 May 16 16:14:58 Tofe: anbox session-manager in my build with latest Yocto: https://paste.pound-python.org/show/lbYCyc57dDLm5UyHGolH/ May 16 16:29:04 Herrie: i was doing uid shift by building squashfs-tools and then packing/unpacking anbox image in emulator May 16 16:44:00 JaMa: that's a bit better :) May 16 16:44:08 JaMa: you needed to start dbus ? May 16 16:45:34 "Failed to create context: error=0x3005" ... meaning EGL_BAD_CONFIG May 16 16:46:36 Tofe: dbus was running, but the variable wasn't exported in the terminal for some reason May 16 16:48:20 ok **** ENDING LOGGING AT Thu May 17 03:00:02 2018