lp:~raof/mir/fix-client-module-error-reporting

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

Related blueprints

3529. By Chris Halse Rogers on 2016-05-31

mcl::ProbingClientPlatformFactory: handle exceptions in create_client_platform.

Previously the exception thrown from the client platform module would outlive the
mir::SharedLibrary it came from, meaning that if create_client_platform threw
the client would SIGSEGV trying to access the now-unloaded exception message.

3528. By Chris Halse Rogers on 2016-05-27

Add missing override specifier to fix the Clang build

3527. By Chris Halse Rogers on 2016-05-27

Fix the android client platform build

3526. By Chris Halse Rogers on 2016-05-27

Update the final test for “dummy”→“mir:stub-graphics” change

3525. By Chris Halse Rogers on 2016-05-27

Fix longstanding TODO in client platform detection.

Inspect, and match on, the server's MirModuleProperties directly rather than hope we can
distinguish platforms by shape of the MirPlatformPackage they send.

The EGLStream MirPlatformPackage will look the same as the Android platform package.

3524. By Chris Halse Rogers on 2016-05-26

Expose populate_graphics_module to mcl::ClientContext

This lets the client modules actually check that there's a graphics driver they support
on the other end of the connection.

3523. By Daniel van Vugt on 2016-05-25

Fix failing acceptance test:
ClientLatency.triple_buffered_client_has_less_than_two_frames_latency
(LP: #1576690)

Technically we expect exactly two frames latency when triple buffering
and no early release. We see that on regular machines, but CI is often
too loaded to stay at two frames latency and yields ~2.6 instead. So
accommodate slow machines without weakening the test too much. Fixes: https://bugs.launchpad.net/bugs/1576690.

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

3522. By Chris Halse Rogers on 2016-05-25

DRMModeResources: Add iterator-implementing Connectors, Encoders, and CRTCs.

Quite a lot of the users of the connector-, encoder-, and CRTC-enumerating functions want to early-return, which is awkward with the current for_each_* API.

Providing an iterator-based API makes the code using it more natural, and lets us use <algorithm> where we want.

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

3521. By Chris Halse Rogers on 2016-05-25

Pull some KMS code out into a platform helper.

We'll have at least two consumers of this; mesa-kms and eglstream-kms.

Approved by mir-ci-bot, Cemil Azizoglu.

3520. By Kevin DuBois on 2016-05-25

Fix LP: #1579076 by correcting a situation where an incorrectly-sized
buffer didn't end up getting freed, resulting in performance
degradation.

Consolidates how set_size and set_scale works under NBS as a bonus, now
both will immediately free buffers that can be freed, and allocate the
buffers back once a wait situation is encountered.