Mir

Merge lp://qastaging/~vanvugt/mir/manual-vsync-api-v2 into lp://qastaging/mir

manual-vsync-api-v2
Merge into development-branch

Proposed by Daniel van Vugt on 2017-04-27

Status:	Work in progress
Proposed branch:	lp://qastaging/~vanvugt/mir/manual-vsync-api-v2
Merge into:	lp://qastaging/mir
Diff against target:	233 lines (+103/-2) 9 files modified examples/eglapp.c (+15/-1) include/client/mir_toolkit/mir_window.h (+16/-0) src/client/frame_clock.cpp (+10/-0) src/client/frame_clock.h (+8/-0) src/client/mir_surface.cpp (+16/-0) src/client/mir_surface.h (+2/-0) src/client/mir_surface_api.cpp (+5/-0) src/client/symbols.map (+6/-1) tests/unit-tests/client/test_frame_clock.cpp (+25/-0)
To merge this branch:	bzr merge lp://qastaging/~vanvugt/mir/manual-vsync-api-v2
Related bugs:	Link a bug report

Reviewer	Review Type	Date Requested	Status
Mir CI Bot	continuous-integration		Approve on 2017-04-27
Mir development team		2017-04-27	Pending
Review via email: mp+323313@code.qastaging.launchpad.net

Commit message

Introducing a trivial API function that allows clients to throttle
themselves without ever blocking inside thirdparty code:
mir_window_get_microseconds_till_vblank

The intended methodology is for a client to set swap interval zero
and then use this new function to throttle its own swaps.

The intended use cases are:
* Toolkits/apps that have an existing event loop (particularly
   single threaded ones) and don't want anything that complicates
   that architecture, but do want synchronized/throttled rendering.
* Existing single-threaded users (Xmir only?) of the asynchronous
   mir_buffer_stream_swap_buffers(), can continue to have vsync and
   asynchronous buffer returns, and single-threaded simplicity,
   without putting a queuing/threading burden on libmirclient to
   throttle buffer returns at all. The existing client event loop can
   be used instead.

Revision history for this message

Mir CI Bot (mir-ci-bot) wrote on 2017-04-27:

Click here to trigger a rebuild:
https://mir-jenkins.ubuntu.com/job/mir-ci/3380/rebuild

review: Approve (continuous-integration)

Unmerged revisions

4167. By Daniel van Vugt on 2017-04-27: Rename the new unit test to match what it actually does
4166. By Daniel van Vugt on 2017-04-27: Use the new feature in examples and fix up C symbol exports.
4165. By Daniel van Vugt on 2017-04-27: And use the new function
4164. By Daniel van Vugt on 2017-04-27: Implement a new FrameClock method for the new API, which eases unit
testing and is crucially also right... so the test now passes.
4163. By Daniel van Vugt on 2017-04-27: Reintroduce and rewrite the unit test
4162. By Daniel van Vugt on 2017-04-27: Revert back to a simpler algorithm
4161. By Daniel van Vugt on 2017-04-27: A little progress
4160. By Daniel van Vugt on 2017-04-26: A new approach. Needs testing...
4159. By Daniel van Vugt on 2017-04-26: Drop the new unit test
4158. By Daniel van Vugt on 2017-04-26: Add a unit test for the expected behaviour from the new MirWindow
function... presently failing.