lp:~raof/mir/+git/mir

Author: Chris Halse Rogers
Author Date: 2016-09-22 05:18:39 UTC

Add mir::ObserverRegistrar interface and mir::ObserverMultiplexer implementation.

Author: Chris Halse Rogers
Author Date: 2016-09-16 02:06:47 UTC

Test that configure_for_hardware_change() sends correct configuration on failure

Author: Chris Halse Rogers
Author Date: 2016-09-14 23:55:50 UTC

Correct argument-expectation in test.

We're testing that we apply the base configuration without restarting, so we should
be matching on the base configuration, not the applied configuration ☺

Author: Chris Halse Rogers
Author Date: 2016-08-23 04:24:00 UTC

UnresponsiveClient.does_not_hang_server: Extra paranoia in server test code.

Our test harness does not guarantee that session->default_surface() exists - running this test under load
I was able to crash it here.

Author: Chris Halse Rogers
Author Date: 2016-08-11 02:14:27 UTC

Add mir_output_get_{current,preferred}_mode_index API.

This is trivial to retreive in the client library, and I've found that I've been smattering ad-hoc,
worse, implementations of it around in my client code. So just expose it in the client API
and make everyone's lives easier.

Author: Daniel van Vugt
Author Date: 2016-08-19 03:16:09 UTC

Document early release operation (in case it is enabled) more accurately. Fixes: https://bugs.launchpad.net/bugs/1395421.

Approved by Cemil Azizoglu, mir-ci-bot.

Author: Chris Halse Rogers
Author Date: 2016-08-15 06:16:51 UTC

MirConnection: Don't double-close FDs on BufferStream error path.

Author: Chris Halse Rogers
Author Date: 2016-08-18 00:15:59 UTC

...and implement mir_connection_cancel_base_display_configuration_preview

Author: Chris Halse Rogers
Author Date: 2016-08-11 07:09:41 UTC

Improvements to mtd::FakeDisplay.

Throw an exception when writing to/reading from the configuration change event trigger
fails rather than silently continuing, blithely unaware that the call has failed.

Switch from a pipe to an eventfd; eventfd is the solution tailored for this use-case.

Presented in the forlorn hope that it makes a test that appears to be flaky on vivid+clang,
but does not appear that it's possible for it to be flaky, stops being flaky.
Either by consistently passing, or consistently failing!

Author: Chris Halse Rogers
Author Date: 2016-08-11 03:11:47 UTC

Merge remote-tracking branch 'origin/master' into mesa-read-subpixel-info

Author: Chris Halse Rogers
Author Date: 2016-08-09 06:35:21 UTC

Pipe MirSubpixelArrangement all the way through.

Author: Chris Halse Rogers
Author Date: 2016-03-10 06:18:17 UTC

Add missing ‘override’ specifier

Author: Chris Halse Rogers
Author Date: 2016-03-07 03:53:34 UTC

Some unittests for preview_base_configuration

Author: Chris Halse Rogers
Author Date: 2016-02-24 02:32:04 UTC

Document, and test, the things that are independent of orientation

Author: Chris Halse Rogers
Author Date: 2016-02-18 07:00:02 UTC

Add mir_output_get_form_factor

Author: Chris Halse Rogers
Author Date: 2016-02-18 00:38:32 UTC

Back mcl::DisplayConfiguration with the protobuf struct.

This makes it easier to extend what we send and receive as configuration.

Author: Chris Halse Rogers
Author Date: 2016-02-16 06:20:28 UTC

BAH

Author: Chris Halse Rogers
Author Date: 2016-02-15 06:17:27 UTC

Fix(?) racy tests

emit_configuration_change_event() doesn't ensure subsequent clients see that configuration.
It just mostly works.
Instead, wait until the configuration has taken.

Author: Chris Halse Rogers
Author Date: 2016-02-10 02:09:20 UTC

Mark AgingBuffer overrides as override

Author: Chris Halse Rogers
Author Date: 2016-02-09 22:32:47 UTC

Client symbols: Specify a local: *; glob in exactly one stanza.

Author: Chris Halse Rogers
Author Date: 2016-01-21 23:52:26 UTC

Remove superfluous mt::Signal from test

Author: Daniel van Vugt
Author Date: 2016-01-18 20:22:55 UTC

Add environment variable MIR_CLIENT_FORCE_SWAP_INTERVAL to forcefully
override the swap interval used by a client.

This is very useful for testing ideas without needing to rebuild your
apps/toolkits/shells. In particular, some existing users of libmirclient
are probably capable and ready to use a swap interval of zero already,
which would significantly reduce lag. However we need a way of testing
that without having to rebuild anything.
.

Approved by Alan Griffiths, PS Jenkins bot, mir-ci-bot.

Author: Chris Halse Rogers
Author Date: 2015-12-17 22:35:20 UTC

Fix the testcase-test for clang

Author: Chris Halse Rogers
Author Date: 2015-12-16 01:23:56 UTC

Hide LifetimeGuard::WaitableCount in implementation file.

Also resolve TSan's (probably?) false-positive about destroying the std::condition_variable

Author: Chris Halse Rogers
Author Date: 2015-12-17 01:14:14 UTC

TimeoutApplicationNotRespondingDetector: Consider applications that respond to input as responsive.

Author: Brandon Schaefer
Author Date: 2015-12-04 02:31:22 UTC

* Move to using a unit8_t* for the mac + a uint32_t mac_size to pass a variable length mac hash
  through the fd

Author: Chris Halse Rogers
Author Date: 2015-11-04 09:15:04 UTC

Add mir-screencast-gst, a mirascast screencasting source

Author: Alberto Aguirre
Author Date: 2015-10-30 01:16:32 UTC

Revert -r3066

The server cannot start throwing std::bad_weak_ptr exception.

Author: Chris Halse Rogers
Author Date: 2015-10-26 04:25:58 UTC

Expose the_broadcasting_input_send_observer()

Author: Kevin DuBois
Author Date: 2015-10-23 21:15:50 UTC

compositor: improve the nbs system enough to pass the BufferScheduling test for updating the queue around visibility events.

Approved by PS Jenkins bot, Alexandros Frantzis.

Author: Chris Halse Rogers
Author Date: 2015-10-21 10:22:59 UTC

Optimise the no-contention case for ThreadSafeList mutation

Author: Chris Halse Rogers
Author Date: 2015-10-21 02:02:14 UTC

Add mircookie-dev dependency to mirclient-dev.

We've got a Requires.private on it in mirclient.pc and use its headers.

Author: Chris Halse Rogers
Author Date: 2015-10-01 01:27:17 UTC

DisplayConfigurationController: Delete CopyAssign

Author: Chris Halse Rogers
Author Date: 2015-09-11 07:34:25 UTC

BufferingMessageSender: Split uncork() and drain().

We actually want to send messages *out of order*; the create_surface reply first,
*then* the buffered messages (which possibly reference that surface).

To do this we need to first uncork(), so that the reply can be sent, *then* drain(),
sending any events that were buffered up.

Author: Chris Halse Rogers
Author Date: 2015-09-11 11:05:54 UTC

Pass a non-null sender in ApplicationSessionSurfaceOutput tests.

This lets them pass, rather than SEGV when they try to dereference it...

Author: Chris Halse Rogers
Author Date: 2015-09-04 04:17:07 UTC

client: Add mir_surface_spec_set_event_handler.

This allows the client to set the event handler for a surface before it is
constructed, (half-)closing the race between the server sending an event and
the client installing an event handler for the surface.

Author: Chris Halse Rogers
Author Date: 2015-08-25 07:05:55 UTC

Wrap display in a unique_ptr, ensuring it's destroyed *after* the other globals