lp:~vanvugt/mir/improve-1570698

Branch merges

Propose for merging

No branches dependent on this one.

Related source package recipes

No recipes using this branch. (?)

Create packaging recipe

Related snap packages

Related bugs

Bug #1570698: CI failure in TestClientInput.receives_one_touch_event_per_frame

Medium

Fix Released

Link a bug report

Related blueprints

3713. By Daniel van Vugt on 2016-09-22

Work in progress.

It works, but is somehow less reliable under Valgrind now.

3712. By Chris Halse Rogers on 2016-09-21

Prevent unnecessary reinitialisation of the compositor on display configuration (round 1)

We currently need to tear down and reinitialise the compositor on display changes because (a) we have no way of adding or removing compositors short of a full teardown/rebuild cycle and, worse (b) because the compositor takes a non-owning reference to a DisplayBuffer and display configuration will frequently destroy all existing DisplayBuffers.

Add an interface to the platform's Display to check when a display configuration change will *not* destroy existing DisplayBuffers. If it will not, that avoids problem (b), so check if any new outputs have been enabled. If neither of those are the case we can safely apply the display configuration without reinitialising the compositor.

Follow up work will enable this in the various graphics platforms.

Fixes (part one): https://bugs.launchpad.net/mir/+bug/1556142.

Approved by mir-ci-bot, Cemil Azizoglu, Alan Griffiths.

3711. By Daniel van Vugt on 2016-09-20

Fix protobuf3 syntax warnings (LP: #1624971)
. Fixes: https://bugs.launchpad.net/bugs/1624971.

Approved by Kevin DuBois, Cemil Azizoglu, mir-ci-bot.

3710. By Daniel van Vugt on 2016-09-19

Rename 'post_renderables_if_optimizable' to something less vague
and less wordy. Also don't mention 'post_update' which no longer
exists.

Approved by Cemil Azizoglu, Kevin DuBois, mir-ci-bot.

3709. By Kevin DuBois on 2016-09-19

nested: define the NativeBuffer type for the nested platform. Its largely a wrapper around the mir_buffer.h API.

Approved by mir-ci-bot, Daniel van Vugt, Cemil Azizoglu.

3708. By Daniel van Vugt on 2016-09-19

Generalise CompositorTest into "SystemPerformanceTest" so we can reuse
it for non-compositor system performance testing.

Approved by mir-ci-bot, Cemil Azizoglu.

3707. By Daniel van Vugt on 2016-09-16

Don't log [E]GL_EXTENSIONS on start-up. They're too noisy (which will
only get worse with NVIDIA support), and hinder readability of other
server log messages.

Besides, if extensions matter to us at all they should be seen only in
error messages when one is missing (like you already see from
eglstream).

Approved by mir-ci-bot, Kevin DuBois.

3706. By Andreas Pokorny on 2016-09-16

Solve two issues with the nested mir tests - and a bandaid for the third

This adds another server status to listen to - server being ready to receive user input. That is the case as soon as udev tells us about input devices found via the libinput fd, on in a nested setup as soon we have received the device information from the host server - thus after the surfaces of the nested server have been focused on the host server. This new status is used in the tests to avoid the unbounded sleeps.

The other issues in the tests are related to test setup and the broken way per surface cursor images (when backed by buffer streams) are implemented. The problem with the cursor is only observed when the feature is used together with a SoftwareCursor on the host compositor. We do not ship that a setup at the moment. So this MP weakens the success criteria. There are now FIXMEs explaining the problem. So only a bandaid for now.

Finally the second fixed issue: Segmentation faults because mock actions might be executed due to the problem described above while being removed. Fixes: https://bugs.launchpad.net/bugs/1523621.

Approved by mir-ci-bot, Alan Griffiths.

3705. By Daniel van Vugt on 2016-09-16

Document a feature of mir_wait_for*() that we seemingly rely on in our
swap buffers implementation but had never explicitly documented.

Approved by mir-ci-bot, Daniel van Vugt, Cemil Azizoglu, Andreas Pokorny.

3704. By Daniel van Vugt on 2016-09-16

report_vsync with real output IDs instead of arbitrarily chosen IDs.
(LP: #1621352)

So now you can match up the report output with the display config. Fixes: https://bugs.launchpad.net/bugs/1621352.

Approved by Andreas Pokorny, mir-ci-bot, Cemil Azizoglu.