Merge lp:~alan-griffiths/mir/fix-1358698 into lp:mir

[Alan Griffiths]

Possible fix for lp:1358698

Possible fix for lp:1358698

I've still not been able to reproduce locally, so this fix is speculative.

If this does pass CI, then we ought to resubmit it a few times to ensure the fix is real.

[PS Jenkins bot (ps-jenkins)]

FAILED: Continuous integration, rev:1853
http://jenkins.qa.ubuntu.com/job/mir-team-mir-development-branch-ci/2440/
Executed test runs:
    SUCCESS: http://jenkins.qa.ubuntu.com/job/mir-android-utopic-i386-build/1387
    SUCCESS: http://jenkins.qa.ubuntu.com/job/mir-clang-utopic-amd64-build/1393
    FAILURE: http://jenkins.qa.ubuntu.com/job/mir-mediumtests-utopic-touch/1370/console
    SUCCESS: http://jenkins.qa.ubuntu.com/job/mir-team-mir-development-branch-utopic-amd64-ci/962
        deb: http://jenkins.qa.ubuntu.com/job/mir-team-mir-development-branch-utopic-amd64-ci/962/artifact/work/output/*zip*/output.zip
    SUCCESS: http://jenkins.qa.ubuntu.com/job/mir-team-mir-development-branch-utopic-armhf-ci/963
        deb: http://jenkins.qa.ubuntu.com/job/mir-team-mir-development-branch-utopic-armhf-ci/963/artifact/work/output/*zip*/output.zip
    SUCCESS: http://jenkins.qa.ubuntu.com/job/mir-mediumtests-builder-utopic-armhf/307
        deb: http://jenkins.qa.ubuntu.com/job/mir-mediumtests-builder-utopic-armhf/307/artifact/work/output/*zip*/output.zip
    FAILURE: http://jenkins.qa.ubuntu.com/job/mir-mediumtests-runner-mako/2450/console
    SUCCESS: http://s-jenkins.ubuntu-ci:8080/job/touch-flash-device/11754

Click here to trigger a rebuild:
http://s-jenkins.ubuntu-ci:8080/job/mir-team-mir-development-branch-ci/2440/rebuild

review: Needs Fixing (continuous-integration)

[Daniel van Vugt (vanvugt)]

Jenkins says no. The bug persists :(

[Daniel van Vugt (vanvugt)]

I don't think "option optimize_for = LITE_RUNTIME;" is working, because we're not using protobuf-lite.

[Daniel van Vugt (vanvugt)]

P.S. I already tried the alternate workaround of using libmirprotobuf.a statically. It doesn't work in libmircommon.so because libmirprotobuf.a's contents are not relocatable (-fPIC). You would have to link all executables to libmirprotobuf.a directly, if not rebuild protobuf yourself.

[PS Jenkins bot (ps-jenkins)]

FAILED: Continuous integration, rev:1855
http://jenkins.qa.ubuntu.com/job/mir-team-mir-development-branch-ci/2446/
Executed test runs:
    SUCCESS: http://jenkins.qa.ubuntu.com/job/mir-android-utopic-i386-build/1396
    SUCCESS: http://jenkins.qa.ubuntu.com/job/mir-clang-utopic-amd64-build/1402
    FAILURE: http://jenkins.qa.ubuntu.com/job/mir-mediumtests-utopic-touch/1379/console
    SUCCESS: http://jenkins.qa.ubuntu.com/job/mir-team-mir-development-branch-utopic-amd64-ci/968
        deb: http://jenkins.qa.ubuntu.com/job/mir-team-mir-development-branch-utopic-amd64-ci/968/artifact/work/output/*zip*/output.zip
    SUCCESS: http://jenkins.qa.ubuntu.com/job/mir-mediumtests-builder-utopic-armhf/316
        deb: http://jenkins.qa.ubuntu.com/job/mir-mediumtests-builder-utopic-armhf/316/artifact/work/output/*zip*/output.zip
    FAILURE: http://jenkins.qa.ubuntu.com/job/mir-mediumtests-runner-mako/2459/console
    SUCCESS: http://s-jenkins.ubuntu-ci:8080/job/touch-flash-device/11791

Click here to trigger a rebuild:
http://s-jenkins.ubuntu-ci:8080/job/mir-team-mir-development-branch-ci/2446/rebuild

review: Needs Fixing (continuous-integration)

[PS Jenkins bot (ps-jenkins)]

FAILED: Continuous integration, rev:1857
http://jenkins.qa.ubuntu.com/job/mir-team-mir-development-branch-ci/2448/
Executed test runs:
    SUCCESS: http://jenkins.qa.ubuntu.com/job/mir-android-utopic-i386-build/1399
    SUCCESS: http://jenkins.qa.ubuntu.com/job/mir-clang-utopic-amd64-build/1405
    FAILURE: http://jenkins.qa.ubuntu.com/job/mir-mediumtests-utopic-touch/1382/console
    SUCCESS: http://jenkins.qa.ubuntu.com/job/mir-team-mir-development-branch-utopic-amd64-ci/970
        deb: http://jenkins.qa.ubuntu.com/job/mir-team-mir-development-branch-utopic-amd64-ci/970/artifact/work/output/*zip*/output.zip
    SUCCESS: http://jenkins.qa.ubuntu.com/job/mir-mediumtests-builder-utopic-armhf/319
        deb: http://jenkins.qa.ubuntu.com/job/mir-mediumtests-builder-utopic-armhf/319/artifact/work/output/*zip*/output.zip
    FAILURE: http://jenkins.qa.ubuntu.com/job/mir-mediumtests-runner-mako/2462/console
    SUCCESS: http://s-jenkins.ubuntu-ci:8080/job/touch-flash-device/11817

Click here to trigger a rebuild:
http://s-jenkins.ubuntu-ci:8080/job/mir-team-mir-development-branch-ci/2448/rebuild

review: Needs Fixing (continuous-integration)

[Alan Griffiths (alan-griffiths)]

What I think is happening is that when we dlopen("libmirplatformgraphics.so", ...) we *don't* load the libmirplatformgraphics.so platform from the build, but the one previously installed on the system.

Because we're using libmircommon.so.2 and the release libmirplatformgraphics.so has linked against libmircommon.so.1 we end up with two versions of mircommon registering with libprotobuf. Hence the error.

The current version of this MP should masks the problem by reverting the bump to libmircommon.so.2 - so we only get a single version loaded and CI works. It also removes some unnecessary linking against mircommon which (once that change is released) will reduce the likelihood of similar problems in the future.

The reason the problem started appearing was a combination of two factors:

1. prior to the latest release version it was an unversioned mirprotobuf that registered with libprotobuf, not a versioned mircommon
2. we (duflu thinks unnecessarily) bumped the mircommon version

As the latter landed before the former there was no CI error at that stage.

*This does not address the root cause*

The root cause is loading that we are dlopen()ing the incorrect libmirplatformgraphics.so - and I'm going to log that as a bug if this MP has the expected effect.

[PS Jenkins bot (ps-jenkins)]

FAILED: Continuous integration, rev:1861
http://jenkins.qa.ubuntu.com/job/mir-team-mir-development-branch-ci/2459/
Executed test runs:
    FAILURE: http://jenkins.qa.ubuntu.com/job/mir-android-utopic-i386-build/1411/console
    SUCCESS: http://jenkins.qa.ubuntu.com/job/mir-clang-utopic-amd64-build/1417
    FAILURE: http://jenkins.qa.ubuntu.com/job/mir-mediumtests-utopic-touch/1394/console
    SUCCESS: http://jenkins.qa.ubuntu.com/job/mir-team-mir-development-branch-utopic-amd64-ci/981
        deb: http://jenkins.qa.ubuntu.com/job/mir-team-mir-development-branch-utopic-amd64-ci/981/artifact/work/output/*zip*/output.zip
    FAILURE: http://jenkins.qa.ubuntu.com/job/mir-mediumtests-builder-utopic-armhf/331/console

Click here to trigger a rebuild:
http://s-jenkins.ubuntu-ci:8080/job/mir-team-mir-development-branch-ci/2459/rebuild

review: Needs Fixing (continuous-integration)

[PS Jenkins bot (ps-jenkins)]

FAILED: Continuous integration, rev:1863
http://jenkins.qa.ubuntu.com/job/mir-team-mir-development-branch-ci/2463/
Executed test runs:
    FAILURE: http://jenkins.qa.ubuntu.com/job/mir-android-utopic-i386-build/1415/console
    SUCCESS: http://jenkins.qa.ubuntu.com/job/mir-clang-utopic-amd64-build/1421
    FAILURE: http://jenkins.qa.ubuntu.com/job/mir-mediumtests-utopic-touch/1400/console
    SUCCESS: http://jenkins.qa.ubuntu.com/job/mir-team-mir-development-branch-utopic-amd64-ci/985
        deb: http://jenkins.qa.ubuntu.com/job/mir-team-mir-development-branch-utopic-amd64-ci/985/artifact/work/output/*zip*/output.zip
    SUCCESS: http://jenkins.qa.ubuntu.com/job/mir-mediumtests-builder-utopic-armhf/337
        deb: http://jenkins.qa.ubuntu.com/job/mir-mediumtests-builder-utopic-armhf/337/artifact/work/output/*zip*/output.zip
    FAILURE: http://jenkins.qa.ubuntu.com/job/mir-mediumtests-runner-mako/2479/console
    SUCCESS: http://s-jenkins.ubuntu-ci:8080/job/touch-flash-device/11930

Click here to trigger a rebuild:
http://s-jenkins.ubuntu-ci:8080/job/mir-team-mir-development-branch-ci/2463/rebuild

review: Needs Fixing (continuous-integration)

[Daniel van Vugt (vanvugt)]

It looks like the branch is to blame but I can't tell why...

/usr/lib/gcc-cross/arm-linux-gnueabihf/4.9/../../../../arm-linux-gnueabihf/bin/ld: cannot find -landroid-properties
collect2: error: ld returned 1 exit status
src/platform/graphics/android/CMakeFiles/mirplatformgraphicsandroid.dir/build.make:117: recipe for target 'lib/android/libmirplatformgraphics.so' failed
make[2]: *** [lib/android/libmirplatformgraphics.so] Error 1

[Chris Halse Rogers (raof)]

Oh! You know why it's looking in /usr/ for dlopen("libmirplatformgraphics.so")? Because CMake likes to set ALL THE RPATHS. This also means that I think we're not necessarily testing the right client code, too.

I think I know how to get CMake to be less rpath-happy, and see whether that resolves this...

[Chris Halse Rogers (raof)]

I'm confident that https://code.launchpad.net/~raof/mir/buildsystems-are-the-worst/+merge/231840 should fix the root cause of this for us; it fixes it on my N10 at least, on which I was able to reproduce the bug.

[Alan Griffiths (alan-griffiths)]

> It looks like the branch is to blame but I can't tell why...
>
> /usr/lib/gcc-cross/arm-linux-gnueabihf/4.9/../../../../arm-linux-
> gnueabihf/bin/ld: cannot find -landroid-properties
> collect2: error: ld returned 1 exit status
> src/platform/graphics/android/CMakeFiles/mirplatformgraphicsandroid.dir/build.
> make:117: recipe for target 'lib/android/libmirplatformgraphics.so' failed
> make[2]: *** [lib/android/libmirplatformgraphics.so] Error 1

Yes, the branch is to blame - my fault for not testing a clean build.

[PS Jenkins bot (ps-jenkins)]

FAILED: Continuous integration, rev:1864
http://jenkins.qa.ubuntu.com/job/mir-team-mir-development-branch-ci/2470/
Executed test runs:
    SUCCESS: http://jenkins.qa.ubuntu.com/job/mir-android-utopic-i386-build/1422
    SUCCESS: http://jenkins.qa.ubuntu.com/job/mir-clang-utopic-amd64-build/1428
    FAILURE: http://jenkins.qa.ubuntu.com/job/mir-mediumtests-utopic-touch/1407/console
    SUCCESS: http://jenkins.qa.ubuntu.com/job/mir-team-mir-development-branch-utopic-amd64-ci/992
        deb: http://jenkins.qa.ubuntu.com/job/mir-team-mir-development-branch-utopic-amd64-ci/992/artifact/work/output/*zip*/output.zip
    SUCCESS: http://jenkins.qa.ubuntu.com/job/mir-mediumtests-builder-utopic-armhf/344
        deb: http://jenkins.qa.ubuntu.com/job/mir-mediumtests-builder-utopic-armhf/344/artifact/work/output/*zip*/output.zip
    FAILURE: http://jenkins.qa.ubuntu.com/job/mir-mediumtests-runner-mako/2485/console
    SUCCESS: http://s-jenkins.ubuntu-ci:8080/job/touch-flash-device/11954

Click here to trigger a rebuild:
http://s-jenkins.ubuntu-ci:8080/job/mir-team-mir-development-branch-ci/2470/rebuild

review: Needs Fixing (continuous-integration)

[Alan Griffiths (alan-griffiths)]

> [==========] Running 1 test from 1 test case.
> [----------] Global test environment set-up.
> [----------] 1 test from GLMark2Test
> [ RUN ] GLMark2Test.benchmark_fullscreen_default
>
> Segmentation fault (core dumped)
> + mir_rc=-1

WTF?

(At least this failure is a step forwards.)

[PS Jenkins bot (ps-jenkins)]

FAILED: Continuous integration, rev:1864
http://jenkins.qa.ubuntu.com/job/mir-team-mir-development-branch-ci/2473/
Executed test runs:
    SUCCESS: http://jenkins.qa.ubuntu.com/job/mir-android-utopic-i386-build/1425
    SUCCESS: http://jenkins.qa.ubuntu.com/job/mir-clang-utopic-amd64-build/1431
    FAILURE: http://jenkins.qa.ubuntu.com/job/mir-mediumtests-utopic-touch/1410/console
    SUCCESS: http://jenkins.qa.ubuntu.com/job/mir-team-mir-development-branch-utopic-amd64-ci/995
        deb: http://jenkins.qa.ubuntu.com/job/mir-team-mir-development-branch-utopic-amd64-ci/995/artifact/work/output/*zip*/output.zip
    SUCCESS: http://jenkins.qa.ubuntu.com/job/mir-mediumtests-builder-utopic-armhf/347
        deb: http://jenkins.qa.ubuntu.com/job/mir-mediumtests-builder-utopic-armhf/347/artifact/work/output/*zip*/output.zip
    FAILURE: http://jenkins.qa.ubuntu.com/job/mir-mediumtests-runner-mako/2488/console
    SUCCESS: http://s-jenkins.ubuntu-ci:8080/job/touch-flash-device/11963

Click here to trigger a rebuild:
http://s-jenkins.ubuntu-ci:8080/job/mir-team-mir-development-branch-ci/2473/rebuild

review: Needs Fixing (continuous-integration)

[Kevin DuBois (kdub)]

The performance test was disabled to unblock everything. No one could reproduce failure on device. retriggering, hopefully it works this time...

[PS Jenkins bot (ps-jenkins)]

review: Approve (continuous-integration)

=== modified file 'cmake/MirCommon.cmake'
--- cmake/MirCommon.cmake	2014-04-15 05:31:19 +0000
+++ cmake/MirCommon.cmake	2014-08-22 08:42:25 +0000
@@ -54,6 +54,14 @@
       ${EXECUTABLE_OUTPUT_PATH}/${EXECUTABLE} --gtest_list_tests > /dev/null
       WORKING_DIRECTORY ${CMAKE_CURRENT_BINARY_DIR}
       COMMENT "Check that discovering Tests in ${EXECUTABLE} works")
+      
+    if (MIR_BUILD_PLATFORM_ANDROID)
+      add_dependencies(${CHECK_TEST_DISCOVERY_TARGET_NAME} mirplatformgraphicsandroid)
+    endif()
+    
+    if (MIR_BUILD_PLATFORM_MESA)
+      add_dependencies(${CHECK_TEST_DISCOVERY_TARGET_NAME} mirplatformgraphicsmesa)
+    endif()
 
     if (${ARGC} GREATER 1)
       foreach (env ${ARGN})

=== modified file 'debian/control'
--- debian/control	2014-08-14 05:59:02 +0000
+++ debian/control	2014-08-22 08:42:25 +0000
@@ -108,7 +108,7 @@
 Architecture: i386 amd64 armhf arm64
 Multi-Arch: same
 Pre-Depends: ${misc:Pre-Depends}
-Depends: libmircommon2 (= ${binary:Version}), 
+Depends: libmircommon1 (= ${binary:Version}), 
          libprotobuf-dev (>= 2.4.1),
          ${misc:Depends},
 Breaks: mircommon-dev (<< 0.6)
@@ -252,7 +252,7 @@
  .
  Contains a tool for stress testing the Mir display server
 
-Package: libmircommon2
+Package: libmircommon1
 Section: libs
 Architecture: i386 amd64 armhf arm64
 Multi-Arch: same

=== renamed file 'debian/libmircommon2.install' => 'debian/libmircommon1.install'
--- debian/libmircommon2.install	2014-08-14 05:59:02 +0000
+++ debian/libmircommon1.install	2014-08-22 08:42:25 +0000
@@ -1,1 +1,1 @@
-usr/lib/*/libmircommon.so.2
+usr/lib/*/libmircommon.so.1

=== modified file 'src/platform/graphics/android/CMakeLists.txt'
--- src/platform/graphics/android/CMakeLists.txt	2014-08-04 08:58:40 +0000
+++ src/platform/graphics/android/CMakeLists.txt	2014-08-22 08:42:25 +0000
@@ -51,14 +51,12 @@
   mirplatformgraphicsandroid
 
   mirsharedandroid
-  mirplatform
-  mircommon
 
   ${Boost_PROGRAM_OPTIONS_LIBRARY}
   ${LIBHARDWARE_LIBRARIES}
   ${EGL_LDFLAGS} ${EGL_LIBRARIES}
   ${GLESv2_LDFLAGS} ${GLESv2_LIBRARIES}
-  ${ANDROID_PROPERTIES_LIBRARIES}
+  ${ANDROID_PROPERTIES_LDFLAGS}
 )
 
 install(TARGETS mirplatformgraphicsandroid LIBRARY DESTINATION ${CMAKE_INSTALL_LIBDIR}/mir/platformgraphics/android)

=== modified file 'src/platform/graphics/mesa/CMakeLists.txt'
--- src/platform/graphics/mesa/CMakeLists.txt	2014-08-04 08:58:40 +0000
+++ src/platform/graphics/mesa/CMakeLists.txt	2014-08-22 08:42:25 +0000
@@ -47,7 +47,6 @@
 )
 
 target_link_libraries(mirplatformgraphicsmesa
-  mirplatform
 
   ${Boost_PROGRAM_OPTIONS_LIBRARY}
   ${DRM_LDFLAGS} ${DRM_LIBRARIES}

=== modified file 'src/shared/CMakeLists.txt'
--- src/shared/CMakeLists.txt	2014-08-14 05:59:02 +0000
+++ src/shared/CMakeLists.txt	2014-08-22 08:42:25 +0000
@@ -40,7 +40,7 @@
 )
 
 # TODO we need a place to manage ABI and related versioning but use this as placeholder
-set(MIRCOMMON_ABI 2)
+set(MIRCOMMON_ABI 1)
 
 set_target_properties(mircommon
 

=== modified file 'tests/integration-tests/CMakeLists.txt'
--- tests/integration-tests/CMakeLists.txt	2014-07-31 13:20:18 +0000
+++ tests/integration-tests/CMakeLists.txt	2014-08-22 08:42:25 +0000
@@ -74,7 +74,7 @@
 if (MIR_TEST_PLATFORM STREQUAL "android")
 target_link_libraries(mir_integration_tests
   mirsharedandroid
-  ${ANDROID_PROPERTIES_LIBRARIES}
+  ${ANDROID_PROPERTIES_LDFLAGS}
 )
 endif()
 

=== modified file 'tests/unit-tests/CMakeLists.txt'
--- tests/unit-tests/CMakeLists.txt	2014-08-15 13:42:19 +0000
+++ tests/unit-tests/CMakeLists.txt	2014-08-22 08:42:25 +0000
@@ -75,7 +75,7 @@
 if (MIR_TEST_PLATFORM STREQUAL "android")
 target_link_libraries(mir_unit_tests 
   mirsharedandroid
-  ${ANDROID_PROPERTIES_LIBRARIES}
+  ${ANDROID_PROPERTIES_LDFLAGS}
 )
 endif()