lp:~raof/mir/dont-kill-the-poor-clients

Branch merges

Propose for merging

1 branch dependent on this one.

On hold for merging into lp:mir

PS Jenkins bot (community): Needs Fixing (continuous-integration) on 2014-09-12

Daniel van Vugt: Needs Fixing on 2014-09-12

Chris Halse Rogers: Abstain on 2014-09-10

Cemil Azizoglu (community): Approve on 2014-09-02

Alexandros Frantzis (community): Approve on 2014-09-02

Diff: 626 lines (+141/-172)

21 files modified

src/client/lifecycle_control.cpp (+13/-0)
src/client/lifecycle_control.h (+2/-0)
src/client/mir_connection.cpp (+11/-0)
src/client/rpc/mir_protobuf_rpc_channel.cpp (+0/-1)
tests/acceptance-tests/test_client_cursor_api.cpp (+0/-2)
tests/acceptance-tests/test_client_input.cpp (+0/-2)
tests/acceptance-tests/test_client_library.cpp (+1/-3)
tests/acceptance-tests/test_large_messages.cpp (+0/-2)
tests/acceptance-tests/test_nested_mir.cpp (+0/-2)
tests/acceptance-tests/test_prompt_session_client_api.cpp (+0/-2)
tests/acceptance-tests/test_protobuf.cpp (+0/-2)
tests/acceptance-tests/test_test_framework.cpp (+0/-2)
tests/include/mir_test_framework/basic_client_server_fixture.h (+0/-2)
tests/include/mir_test_framework/using_stub_client_platform.h (+0/-51)
tests/integration-tests/client/test_client_render.cpp (+0/-10)
tests/mir_test_framework/CMakeLists.txt (+0/-1)
tests/mir_test_framework/testing_process_manager.cpp (+0/-2)
tests/mir_test_framework/using_stub_client_platform.cpp (+0/-81)
tests/unit-tests/client/CMakeLists.txt (+1/-0)
tests/unit-tests/client/test_client_mir_surface.cpp (+2/-7)
tests/unit-tests/client/test_lifecycle_control.cpp (+111/-0)

api

Related source package recipes

No recipes using this branch. (?)

Create packaging recipe

Related snap packages

Related bugs

Related blueprints

1894. By Chris Halse Rogers on 2014-09-12

Don't force the completion of pending RPC calls on disconnect.

There's nothing sensible to be done there. If the client is simple, they'll be killed with a signal
(that awesomely might race with our force_completion() on the IO thread).

If the client is complex, they'll have a handler for the disconnect.

In neither case is it useful to call the RPC response handlers with a garbage non-reply.
Our response handlers are not generally capable of handling invalid protobuf::Message objects,
so this frequently results in calling a pure-virtual function and termination.

At best this would result in calling a client callback with an error object. But the client
already knows that the connection has died (or the process has been killed by a signal)
so that's not super-useful.

1893. By Chris Halse Rogers on 2014-09-11

Revert eglapp changes.

These will be proposed as a part of a wider cleanup of our examples

1892. By Chris Halse Rogers on 2014-09-11

Merge trunk, resolving conflicts

1891. By Chris Halse Rogers on 2014-09-08

Merge trunk, resolving conflicts

1890. By Chris Halse Rogers on 2014-09-05

Remove last vestages of UsingStubClientPlatform that crept back in

1889. By Chris Halse Rogers on 2014-09-05

Merge trunk, resolving conflicts

1888. By Chris Halse Rogers on 2014-09-05

Detect server disconnect in example EGL apps

1887. By Chris Halse Rogers on 2014-09-02

Remove some detritus

1886. By Chris Halse Rogers on 2014-09-02

Also remove unnecessary lifecycle setting in Android integration tests

1885. By Chris Halse Rogers on 2014-09-02

Remove UsingStubClientConnection.

This is now superfluous; it was only preventing the test code from being SIGHUPed when
it called mir_connection_release, which we no longer do.