[testsfail] NestedServerWithTwoDisplays.uses_passthrough_when_surface_size_is_appropriate

[Alan Griffiths]

https://mir-jenkins.ubuntu.com/job/build-2-binpkg-mir/arch=amd64,compiler=gcc,platform=mesa,release=xenial/4888/console

15:13:13 11: [ RUN ] NestedServerWithTwoDisplays.uses_passthrough_when_surface_size_is_appropriate ...
15:13:18 11: ==2589== Thread 6 Mir/Input Reade:
15:13:18 11: ==2589== Invalid read of size 8
15:13:18 11: ==2589== at 0x41FFA8A: std::__shared_ptr<mir_test_framework::FakeInputDeviceImpl::InputDevice, (__gnu_cxx::_Lock_policy)2>::operator->() const (shared_ptr_base.h:1055)
15:13:18 11: ==2589== by 0x41FB664: mir_test_framework::FakeInputDeviceImpl::emit_event(mir::input::synthesis::KeyParameters const&)::{lambda()#1}::operator()() const (fake_input_device_impl.cpp:68)

This looks like a race between the queue draining and deleting mtf::FakeInputDeviceImpl

[Alan Griffiths (alan-griffiths)]

Also seen here:

https://mir-jenkins.ubuntu.com/job/build-2-binpkg-mir/arch=i386,compiler=gcc,platform=mesa,release=xenial/4885/consoleFull

[Alan Griffiths (alan-griffiths)]

OK, in NestedMirRunner::wait_until_surface_ready() we send a pair of key-down/key-up events every 10ms until we see *any* event. We then run the test and teardown.

Other tests using FakeInputDevice ensure they have received the last event sent.

AFAICS There's nothing else to ensure the queue is drained.

[Mir CI Bot (mir-ci-bot)]

Fix committed into lp:mir at revision None, scheduled for release in mir, milestone 1.0.0