Author: Daniel van Vugt
Revision Date: 2013-07-02 08:53:34 UTC

Eliminate a duplicate interface; both DisplayBuffer and ViewableArea defined
their own pure virtual view_area(). It's more flexible for them to share a
single (pure virtual) interface: ViewableArea::view_area()

Author: Daniel van Vugt
Revision Date: 2013-10-17 09:08:11 UTC

Remember to mention enhancements (commits without bug numbers).

Also, correction: LP #1239160 is fixed in 0.0.15

Author: Daniel van Vugt
Revision Date: 2016-12-08 07:50:34 UTC

Emit adoption calls on set_streams

Author: Daniel van Vugt
Revision Date: 2014-11-14 04:00:20 UTC

Also use -rdynamic to stop the dynamic linker from hiding the symbols
at runtime.

Author: Daniel van Vugt
Revision Date: 2016-06-01 08:17:03 UTC

Proof of concept

Author: Daniel van Vugt
Revision Date: 2013-03-21 09:51:53 UTC

Remember puctuation.

Author: Daniel van Vugt
Revision Date: 2017-01-19 08:13:27 UTC

Merge latest trunk

Author: Daniel van Vugt
Revision Date: 2014-10-07 08:29:28 UTC

Merge latest development-branch and fix a conflict (untested)

Author: Daniel van Vugt
Revision Date: 2014-08-19 09:02:06 UTC

Tidy up

Author: Daniel van Vugt
Revision Date: 2013-08-28 10:11:46 UTC

Disable broken DRM/GBM tests, failing because the new drmModeAddFB2 is
not mocked, and possibly not mockable.

Author: Daniel van Vugt
Revision Date: 2016-04-07 04:00:26 UTC

ClientLatency tests: Clarify results to avoid reporting a latency of
zero, which is misleading.

The granularity of our measurements is only a frame counter, so we're
unable to show the subtleties of the exact timing. However it's more
accurate to report that any frame rendered in the 16ms prior to being
displayed has a latency of 1 frame and not 0. Particularly if you
consider that render times aim to be as low as possible, the exact
latency is really much closer to 1 than 0 frames.

Author: Daniel van Vugt
Revision Date: 2015-05-12 06:40:20 UTC

First attempt at bumping the client ABI to 9

Author: Daniel van Vugt
Revision Date: 2016-08-02 07:44:15 UTC

Prototype min/max render times for the client perf report.

Seems to be too low value to bother with right now though. The range between
min and max is so narrow in a one second sample window that the current
report of just the average is probably sufficient.

Author: Daniel van Vugt
Revision Date: 2017-05-02 07:46:01 UTC

More progress

Author: Daniel van Vugt
Revision Date: 2014-10-09 06:44:58 UTC

Bump the common ABI to 3. We broke it in r1968.

Author: Daniel van Vugt
Revision Date: 2014-11-28 07:34:52 UTC

Bump to the common ABI level 4. We broke the ABI with the removal of a
function in r2089.

Author: Daniel van Vugt
Revision Date: 2017-03-02 08:34:00 UTC

Just a proof of concept

Author: Daniel van Vugt
Revision Date: 2016-05-04 06:21:25 UTC

Try harder, Jenkins.

Author: Daniel van Vugt
Revision Date: 2014-05-05 03:17:56 UTC

Merge latest development-branch

Author: Daniel van Vugt
Revision Date: 2014-07-21 10:45:40 UTC

Work (hacking) in progress

Author: Daniel van Vugt
Revision Date: 2016-09-07 02:27:15 UTC

Merge latest trunk

Author: Daniel van Vugt
Revision Date: 2015-08-17 09:32:25 UTC

Prototype defered release

Author: Daniel van Vugt
Revision Date: 2016-11-04 05:34:22 UTC

Fix build failure

Author: Daniel van Vugt
Revision Date: 2013-11-11 06:55:25 UTC

Detect when multiple client buffers are requested simultaneously to
multiple compositor buffers (bypass mode). If you've only got 3 buffers,
doing so will cause blocking and delays. So throw an exception instead of
silently degrading performance (like we did in LP: #1249210).

Author: Daniel van Vugt
Revision Date: 2015-12-14 08:47:08 UTC

First quick and dirty prototype illustrating a fix for the pointer
acceleration issues introduced by libinput in Mir 0.18:
LP: #1524145 and LP: #1522295

This proof-of-concept just uses the native kernel/device motion data,
like Mir 0.17 and prior did. So the mouse cursor is smooth and
consistent again.

This is not the final answer though, as we would like a solution where
the acceleration is configurable. But it does suffice in resolving the
regressions that libinput brought in Mir 0.18 right now. So the mouse
experience is exactly as it was in Mir 0.17 and earlier.

Author: Daniel van Vugt
Revision Date: 2016-08-25 05:05:58 UTC

Merge latest trunk

Author: Daniel van Vugt
Revision Date: 2014-11-23 08:18:48 UTC

Even simpler

Author: Daniel van Vugt
Revision Date: 2015-08-19 09:21:15 UTC

Recursion!

Author: Daniel van Vugt
Revision Date: 2013-04-16 03:26:22 UTC

Remove unused "using namespace".

Author: Daniel van Vugt
Revision Date: 2016-02-15 08:25:54 UTC

Shrink the diff and add another handy test

Author: Daniel van Vugt
Revision Date: 2014-02-28 07:13:12 UTC

fingerpaint: Implement redraw on resize, because it's safe now!

Author: Daniel van Vugt
Revision Date: 2016-11-17 09:33:07 UTC

log_debug -> log_info

Author: Daniel van Vugt
Revision Date: 2015-03-10 06:27:09 UTC

Now buildable at least. But the excess of compiler workarounds means
there's no benefit to this now.

Author: Daniel van Vugt
Revision Date: 2015-01-20 09:10:02 UTC

Remember to place custom input rects relative to the client rect

Author: Daniel van Vugt
Revision Date: 2016-10-19 18:15:52 UTC

vanity: Filter out impossibly low latencies caused by interruptions
to the camera's view of the screen. The next time it thinks it sees
something after getting blinded it might record that as a spuriously
low measurement.

This isn't fullproof, it's computer vision after all. But this does
eliminate much of the errors you might encounter if the camera is
moved and loses sight of the screen.

Author: Daniel van Vugt
Revision Date: 2013-07-18 04:11:58 UTC

Use major and minor as the ABI version, and encode both in the soname.

This way most small ABI changes will only need the minor version to be
incremented. And most importantly, we won't have an ABI value which moves
separately to the version.

Author: Daniel van Vugt
Revision Date: 2014-07-21 09:22:30 UTC

Prototype synchronous input "finishing"

Author: Daniel van Vugt
Revision Date: 2016-08-03 09:59:33 UTC

Fix test failures

TODO: Fix 'Uninteresting mock function call' messages

Author: Daniel van Vugt
Revision Date: 2013-06-07 10:09:01 UTC

Don't throw a fatal exception when the display goes to sleep. Just wait
patiently for it to come back. (LP: #1188504)

This solves the case when the screen blanks during Mir execution. It does not
solve the case where the screen is blanked (and locked; EPERM) before the Mir
server starts. That one either should not be solved or requires more work
later.

Author: Daniel van Vugt
Revision Date: 2013-06-26 11:56:03 UTC

Fix a possible race in:
TEST_F(DefaultDisplayServerTestFixture, client_event_handler_is_thread_safe)

Author: Daniel van Vugt
Revision Date: 2013-09-27 08:02:12 UTC

Merge latest dev branch and resolve a conflict.

Author: Daniel van Vugt
Revision Date: 2013-10-16 09:40:27 UTC

Also remove unused typedef

Author: Daniel van Vugt
Revision Date: 2014-04-17 09:30:34 UTC

Merge latest development-branch

Author: Daniel van Vugt
Revision Date: 2014-04-17 08:36:53 UTC

Fix warnings arising from unexpected (uninteresting) mock_gl calls
(LP: #1308905)

Author: Daniel van Vugt
Revision Date: 2014-05-05 04:01:50 UTC

Merge latest development-branch

Author: Daniel van Vugt
Revision Date: 2014-07-09 06:21:09 UTC

Revert r1687 as it is producing visibly incorrect colours when compared
to r1686. (LP: #1339471)

Author: Daniel van Vugt
Revision Date: 2014-07-24 03:10:47 UTC

Merge latest 0.5 branch

Author: Daniel van Vugt
Revision Date: 2014-09-25 08:19:39 UTC

Add a changelog entry and version bump too

Author: Daniel van Vugt
Revision Date: 2014-10-10 03:54:06 UTC

Fix faulty integration tests which assumed 5 frames could be scheduled.
Actually we only need a maximum of 3 (or 2 arguably).

Author: Daniel van Vugt
Revision Date: 2015-10-21 07:20:00 UTC

Backport a simplified version of RAOF's fix for LP: #1499134 so that
lp:mir/0.15 can build on wily again (with the new Mesa 11.0).

Author: Daniel van Vugt
Revision Date: 2015-12-04 09:21:50 UTC

Avoid running timing-sensitive performance tests under Valgrind, as
they are likely to fail (and we don't care). (LP: #1522031)

Author: Daniel van Vugt
Revision Date: 2015-12-04 09:02:53 UTC

And protect Alt+mousewheel from magnitude changes too.

Author: Daniel van Vugt
Revision Date: 2016-01-20 15:59:43 UTC

Merge latest trunk and fix a conflict.

Author: Daniel van Vugt
Revision Date: 2016-01-20 20:34:29 UTC

Merge lp:~andreas-pokorny/mir/fix-fix-1527449-make_module_ptr

Author: Daniel van Vugt
Revision Date: 2016-10-07 08:50:27 UTC

Experimental very high max buffer count of 20.

Turns out it never actually needs more than 5 or 6, but this does
indicate we need to speed up buffer returns from the server to avoid
blocking the client.

Author: Daniel van Vugt
Revision Date: 2016-09-22 10:09:50 UTC

Work in progress.

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

Author: Daniel van Vugt
Revision Date: 2017-01-12 10:49:24 UTC

Add a mutex too (because LP: #1194384)

Author: Daniel van Vugt
Revision Date: 2017-03-01 03:50:01 UTC

Tidy up the test

Author: Daniel van Vugt
Revision Date: 2016-02-08 10:15:28 UTC

Tidy up and fix iterator madness.

Author: Daniel van Vugt
Revision Date: 2014-09-09 03:44:04 UTC

Merge latest development-branch and fix conflicts

Author: Daniel van Vugt
Revision Date: 2014-05-06 08:25:22 UTC

More work in progress

Author: Daniel van Vugt
Revision Date: 2014-11-21 09:12:00 UTC

Use the logger

Author: Daniel van Vugt
Revision Date: 2017-02-24 08:53:16 UTC

Proof of concept (bypass doesn't work though)

Author: Daniel van Vugt
Revision Date: 2014-09-17 04:02:38 UTC

Move to a simpler, safer, more conservative algorithm

Author: Daniel van Vugt
Revision Date: 2014-04-15 04:05:04 UTC

Fix more (soft) conflicts from recent changes

Author: Daniel van Vugt
Revision Date: 2016-03-02 10:20:35 UTC

Prototype dropping of custom loggers from libmirclient

Author: Daniel van Vugt
Revision Date: 2017-03-02 08:43:24 UTC

Move the throttle back into the compositor. A bit slower, but more
elegant and can be improved on still...

Author: Daniel van Vugt
Revision Date: 2015-07-29 06:30:18 UTC

Name mesa-x11 using the existing naming pattern.

Author: Daniel van Vugt
Revision Date: 2016-11-14 10:26:04 UTC

Add a default error handler (which logs)

Author: Daniel van Vugt
Revision Date: 2015-08-17 08:51:25 UTC

Move more

Author: Daniel van Vugt
Revision Date: 2015-08-28 07:13:07 UTC

Port from the original handoff prototype that was based on the
faulty earlier-release(-1) branch

Author: Daniel van Vugt
Revision Date: 2015-06-03 09:06:19 UTC

Avoid premature release() and drop frame rate limit to 100Hz

Author: Daniel van Vugt
Revision Date: 2016-09-22 10:09:50 UTC

Work in progress.

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

Author: Daniel van Vugt
Revision Date: 2013-12-16 10:19:00 UTC

Separate kernel-to-server and server-to-client latencies

Author: Daniel van Vugt
Revision Date: 2014-11-12 07:23:03 UTC

Merge latest trunk

Author: Daniel van Vugt
Revision Date: 2013-04-18 11:11:01 UTC

Correct send_response for test_file_descriptors

Author: Daniel van Vugt
Revision Date: 2014-05-30 06:54:39 UTC

Merge latest development-branch

Author: Daniel van Vugt
Revision Date: 2016-02-01 09:53:09 UTC

Try again Jenkins

Author: Daniel van Vugt
Revision Date: 2016-12-22 10:07:00 UTC

Prototype

Author: Daniel van Vugt
Revision Date: 2016-04-01 06:48:25 UTC

Make new stanza names less redundant. No need to mention the library name
in every public symbol.

Author: Daniel van Vugt
Revision Date: 2016-08-09 07:15:58 UTC

Bump the libmirprotobuf ABI in one last clumsy effort to get
mir_test_reload_protobuf passing on vivid.

Author: Daniel van Vugt
Revision Date: 2015-09-23 10:04:55 UTC

Add a log warning message about clients that die unexpectedly.

We do have the session mediator report doing the same thing, but most of
the time that's not enabled.

Author: Daniel van Vugt
Revision Date: 2015-09-11 06:51:19 UTC

Use the std::string log_info to get over the fixed buffer size of the
formatting log_info.

Author: Daniel van Vugt
Revision Date: 2015-09-11 06:07:49 UTC

anonymouse!

Author: Daniel van Vugt
Revision Date: 2015-01-29 03:25:24 UTC

Merge latest trunk and fix the same conflict, again.

Author: Daniel van Vugt
Revision Date: 2014-11-24 08:11:31 UTC

Convert to std::chrono, which is probably what the fans want.
Also fixes start_time accuracy.

Author: Daniel van Vugt
Revision Date: 2016-11-09 09:04:33 UTC

Try again

Author: Daniel van Vugt
Revision Date: 2016-11-11 05:57:14 UTC

Restore Android's fallback from n/0 to 0

Author: Daniel van Vugt
Revision Date: 2015-01-27 03:09:33 UTC

Merge latest trunk.

Author: Daniel van Vugt
Revision Date: 2017-05-01 08:28:10 UTC

Revert/drop the prereq branch

Author: Daniel van Vugt
Revision Date: 2017-04-27 08:36:14 UTC

Rename the new unit test to match what it actually does

Author: Daniel van Vugt
Revision Date: 2013-04-19 09:18:15 UTC

Avoid passing method names as strings in every single invocation. You can
use their unique integer indices instead.

Author: Daniel van Vugt
Revision Date: 2016-11-18 10:14:12 UTC

Public headers should look different

Author: Daniel van Vugt
Revision Date: 2015-07-27 06:59:05 UTC

Begin attempt #2.

Author: Daniel van Vugt
Revision Date: 2015-01-19 03:43:34 UTC

Changed syntax per Cemil's comments.

Author: Daniel van Vugt
Revision Date: 2015-08-10 08:38:56 UTC

Rewrite/simplify? the new overlapping compositors test

Author: Daniel van Vugt
Revision Date: 2016-04-21 03:50:13 UTC

Minor progress

Author: Daniel van Vugt
Revision Date: 2014-12-03 02:20:41 UTC

Merge latest trunk

Author: Daniel van Vugt
Revision Date: 2015-01-22 09:41:39 UTC

Merge latest trunk and fix conflicts.

Author: Daniel van Vugt
Revision Date: 2014-03-18 09:26:35 UTC

cppguide.xml: Remove section "Function Parameter Ordering".

No justification was provided for that style point and I would contend that
reading text left-to-right makes inputs-before-outputs more logical anyway.
But I don't feel strongly enough to want to enforce either style. So drop
the whole section.