lp:~raof/mir/no-hidden-rpc-in-bufferstream

Branch merges

Propose for merging

No branches dependent on this one.

On hold for merging into lp:mir

Mir CI Bot: Needs Fixing (continuous-integration) on 2016-07-05

Kevin DuBois (community): Approve on 2016-06-23

Diff: 852 lines (+379/-72)

14 files modified

src/client/buffer_factory.cpp (+11/-1)
src/client/buffer_stream.cpp (+57/-21)
src/client/mir_connection.cpp (+22/-6)
src/client/mir_connection.h (+3/-0)
src/client/no_tls_future-inl.h (+76/-6)
src/client/rpc/mir_protobuf_rpc_channel.cpp (+1/-1)
tests/acceptance-tests/test_client_library.cpp (+2/-0)
tests/acceptance-tests/test_client_surfaces.cpp (+37/-0)
tests/include/mir/test/doubles/stub_buffer.h (+8/-6)
tests/integration-tests/test_surfaceloop.cpp (+44/-28)
tests/mir_test_doubles/stub_buffer.cpp (+5/-3)
tests/unit-tests/client/CMakeLists.txt (+1/-0)
tests/unit-tests/client/test_client_buffer_stream.cpp (+1/-0)
tests/unit-tests/client/test_no_tls_future.cpp (+111/-0)

api

Related source package recipes

No recipes using this branch. (?)

Create packaging recipe

Related snap packages

Related bugs

Related blueprints

3560. By Chris Halse Rogers on 2016-07-05

Wait for default BufferStream creation before calling surface created callback.

Punt completion of the operation into the delayed-processing queue, where it can block on
the creation MirWaitHandle of the default BufferStream.

3559. By Chris Halse Rogers on 2016-07-05

Be more verbose when receiving an unexpected buffer.

This *should* never happen, so being verbose in the error message is low cost and potentially
high reward.

3558. By Chris Halse Rogers on 2016-07-05

MirProtobufRpcChannel: Don't destroy exception information on error.

3557. By Chris Halse Rogers on 2016-07-05

Send valid buffers in SurfaceLoop integration test.

This requires making StubBuffer a bit stubbier and actually packing the IPC messages.

3556. By Chris Halse Rogers on 2016-06-24

Fix NoTLSFuture so that the move-setter is actually accessible.

Because NoTLSPromise only had a set_value(T val) (wat‽) method, PromiseState's set_value(T&& val)
method was unreachable. Which is apparently why g++ didn't notice that it was broken.

3555. By Chris Halse Rogers on 2016-06-23

BufferStream: Do not signal creation wait handle until fully ready.

With NewBufferSemantics we were signalling the creation wait handle after the construction of the mcl::BufferStream,
but at this point the BufferVault does not necessarily have a buffer yet - that comes later as an event in response
to the allocate_buffer requests made during construction.

Instead, wait until the current buffer future<> has resolved before signalling that the BufferStream
is ready.

3554. By Chris Halse Rogers on 2016-06-23

Add missing gmock header.

testing::Eq and EXPECT_THAT are defined in Google Mock, not Google Test.

This only builds because MIR_USE_PRECOMPILED_HEADERS means that all our tests effectively
get the same set of headers included.

3553. By Chris Halse Rogers on 2016-06-23

Remove synthesised-focuse-event unit tests on MirConnection.

These require an ungodly amount of setup, so much so that they're basically
acceptance tests. The change to making sure buffer-streams are completely
constructed before we inform the client that they're completely constructed
results in these tests requiring *even more* setup.

Since we've just added real acceptance tests for this behaviour, just delete
these “unit” tests.

3552. By Chris Halse Rogers on 2016-06-20

Add NoTLSPromise<T>::or_else()

3551. By Chris Halse Rogers on 2016-06-20

NoTLSFuture is a class, damnit.