Merge lp:~robert-ancell/mir/shared-android-input into lp:~mir-team/mir/trunk

[Robert Ancell]

Revert android-input being a shared library:
 - It's not packaged
 - It doesn't have a globally unique name (it should be mirinput or similar to avoid colliding with names)
 - It's not versioned
 - There isn't a strong reason to have to maintain a separate library

Revert android-input being a shared library:
 - It's not packaged
 - It doesn't have a globally unique name (it should be mirinput or similar to avoid colliding with names)
 - It's not versioned
 - There isn't a strong reason to have to maintain a separate library

[PS Jenkins bot (ps-jenkins)]

PASSED: Continuous integration, rev:560
http://jenkins.qa.ubuntu.com/job/mir-ci/304/
Executed test runs:
    SUCCESS: http://jenkins.qa.ubuntu.com/job/mir-android-raring-i386-build/265
    SUCCESS: http://jenkins.qa.ubuntu.com/job/mir-clang-raring-amd64-build/148
    SUCCESS: http://jenkins.qa.ubuntu.com/job/mir-quantal-amd64-ci/309
        deb: http://jenkins.qa.ubuntu.com/job/mir-quantal-amd64-ci/309/artifact/work/output/*zip*/output.zip

Click here to trigger a rebuild:
http://s-jenkins:8080/job/mir-ci/304/rebuild

review: Approve (continuous-integration)

[Daniel van Vugt (vanvugt)]

This is a case where we explicitly want a STATIC library, and to not leave it up to CMake to decide (it might still use SHARED depending on CMake settings). So please change:
    add_library(android-input
to:
    add_library(android-input STATIC

review: Needs Fixing

[Robert Ancell (robert-ancell)]

> This is a case where we explicitly want a STATIC library, and to not leave it
> up to CMake to decide (it might still use SHARED depending on CMake settings).
> So please change:
> add_library(android-input
> to:
> add_library(android-input STATIC

Ah, I initially set it to STATIC but I noticed the old code didn't have this. Explicit is better than implicit here.

[Daniel van Vugt (vanvugt)]

Cool. If making it static and potentially duplicated across the client and server libraries increases code bloat, then that's fine. We can worry about reducing bloat later.

review: Approve

[PS Jenkins bot (ps-jenkins)]

PASSED: Continuous integration, rev:561
http://jenkins.qa.ubuntu.com/job/mir-ci/309/
Executed test runs:
    SUCCESS: http://jenkins.qa.ubuntu.com/job/mir-android-raring-i386-build/271
    SUCCESS: http://jenkins.qa.ubuntu.com/job/mir-clang-raring-amd64-build/154
    SUCCESS: http://jenkins.qa.ubuntu.com/job/mir-quantal-amd64-ci/314
        deb: http://jenkins.qa.ubuntu.com/job/mir-quantal-amd64-ci/314/artifact/work/output/*zip*/output.zip
    SUCCESS: http://jenkins.qa.ubuntu.com/job/mir-vm-ci-build/./distribution=precise,flavor=amd64/122

Click here to trigger a rebuild:
http://s-jenkins:8080/job/mir-ci/309/rebuild

review: Approve (continuous-integration)

[Robert Ancell (robert-ancell)]

Robert Carr - can you just check there's not a major side-effect of switching back to static? We had half a conversation on IRC and I think there weren't strong reasons for going shared.

[Alan Griffiths (alan-griffiths)]

This should be fine. We might plausibly want to split input out as a separate library at some point - but that ought to include mir::input along with the forked android code.

review: Approve

[PS Jenkins bot (ps-jenkins)]

FAILED: Autolanding.
More details in the following jenkins job:
http://jenkins.qa.ubuntu.com/job/mir-autolanding/119/
Executed test runs:
    SUCCESS: http://jenkins.qa.ubuntu.com/job/mir-android-raring-i386-build/277
    FAILURE: http://jenkins.qa.ubuntu.com/job/mir-clang-raring-amd64-build/160/console
    FAILURE: http://jenkins.qa.ubuntu.com/job/mir-quantal-amd64-autolanding/120/console
    SUCCESS: http://jenkins.qa.ubuntu.com/job/mir-vm-autolanding-build/./distribution=precise,flavor=amd64/16

review: Needs Fixing (continuous-integration)

[Kevin DuBois (kdub)]

not sure what jenkins is complaining about, but looks sensible to me as well.

review: Approve

[PS Jenkins bot (ps-jenkins)]

FAILED: Continuous integration, rev:562
http://jenkins.qa.ubuntu.com/job/mir-ci/314/
Executed test runs:
    SUCCESS: http://jenkins.qa.ubuntu.com/job/mir-android-raring-i386-build/282
    FAILURE: http://jenkins.qa.ubuntu.com/job/mir-clang-raring-amd64-build/165/console
    FAILURE: http://jenkins.qa.ubuntu.com/job/mir-quantal-amd64-ci/319/console
    SUCCESS: http://jenkins.qa.ubuntu.com/job/mir-vm-ci-build/./distribution=precise,flavor=amd64/127

Click here to trigger a rebuild:
http://s-jenkins:8080/job/mir-ci/314/rebuild

review: Needs Fixing (continuous-integration)

[PS Jenkins bot (ps-jenkins)]

FAILED: Autolanding.
More details in the following jenkins job:
http://jenkins.qa.ubuntu.com/job/mir-autolanding/121/
Executed test runs:
    SUCCESS: http://jenkins.qa.ubuntu.com/job/mir-android-raring-i386-build/283
    FAILURE: http://jenkins.qa.ubuntu.com/job/mir-clang-raring-amd64-build/166/console
    FAILURE: http://jenkins.qa.ubuntu.com/job/mir-quantal-amd64-autolanding/122/console
    SUCCESS: http://jenkins.qa.ubuntu.com/job/mir-vm-autolanding-build/./distribution=precise,flavor=amd64/18

review: Needs Fixing (continuous-integration)

[Robert Ancell (robert-ancell)]

There is some crash going on in the acceptance tests, investigating.

[Daniel van Vugt (vanvugt)]

[ 53%] Check that discovering Tests in acceptance-tests works
cd /home/dan/bzr/mir/tmp.ra/build/tests/acceptance-tests && /home/dan/bzr/mir/tmp.ra/build/bin/acceptance-tests --gtest_list_tests > /dev/null
Segmentation fault (core dumped)

The crash only happens to *-tests binaries. It does not happen if you run the partially built bin/mir and a real client.

So only crashing in *-tests and being triggered by a switch to STATIC, means it's almost certainly a symbol mix-up. This will happen in *-tests because those binaries link to both libmirclient and libmirserver. And that's now dangerous because those shared libraries will have identical symbols they both pull in from libandroid-input.

This is not to say that libandroid-input needs to always remain dynamic. We just have to tweak the dynamic loader and/or avoid "static" symbols in static/archive libraries.

[Daniel van Vugt (vanvugt)]

The crash in both unit-tests and acceptance-tests looks like it's caused by this static:
    android::KeyCharacterMap::sEmpty

It's being unref'd and deleted twice. Because there is a copy in both libmirclient and libmirserver.

[Daniel van Vugt (vanvugt)]

Full fix proposed here:
  https://code.launchpad.net/~vanvugt/mir/fix-1164253/+merge/157285

[Daniel van Vugt (vanvugt)]

Wow, launchpad got confused.

This is not merged. It is rejected. Because the branch that actually merged is:
  lp:~vanvugt/mir/fix-1164253

=== modified file '3rd_party/android-input/android/CMakeLists.txt'
--- 3rd_party/android-input/android/CMakeLists.txt	2013-03-13 04:54:15 +0000
+++ 3rd_party/android-input/android/CMakeLists.txt	2013-04-04 20:13:23 +0000
@@ -49,7 +49,7 @@
 
 
 add_library(
-  android-input SHARED
+  android-input STATIC
   # The stuff that we want
   frameworks/base/services/input/EventHub.cpp
   frameworks/base/services/input/InputApplication.cpp
@@ -97,7 +97,3 @@
   PROPERTIES
   COMPILE_FLAGS ${ANDROID_INPUT_COMPILE_FLAGS}
 )
-
-install(
-  TARGETS android-input
-  LIBRARY DESTINATION lib)