Change-Id: I3b90ed63d4e4e5f2f5d1b7936f8622aa477a0ac8
Reviewed-by: Jørgen Lind <jorgen.lind@nokia.com>

Change-Id: I9b0bde8cc8d2575963db2e3165919b2d78213424
Reviewed-by: Jørgen Lind <jorgen.lind@nokia.com>

We need this to autodetect if we'r going to build WaylandSurfaceItem

Change-Id: Icdd5379d8334422ff4f780bf5192e9c01101e9db
Reviewed-by: Samuel Rødal <samuel.rodal@nokia.com>

Change-Id: I42c05384767989aa95b49abfadd875d0f8b157d4
Reviewed-by: Samuel Rødal <samuel.rodal@nokia.com>

Now we treat all buffers the same (shm or not).
We advance the queue when we send the frame signal

Change-Id: Id62b1037e03fbeae82048f5ab9078fa81ed8e45d
Reviewed-by: Samuel Rødal <samuel.rodal@nokia.com>

The wayland_sha1.txt in the root folder is the correct one.

Change-Id: Ic843dc3110e6d71fbb031691312848369e448e06
Reviewed-by: Robin Burchell <robin+qt@viroteck.net>

We can use the convenience API that has been added to wayland instead.

Change-Id: I9516256a14f62593b3c78064a0f6288bb496c583
Reviewed-by: Jørgen Lind <jorgen.lind@nokia.com>

Added way of controlling client side style hint ShowIsFullScreen from
compositor, and added support in plugin for the WindowFullScreen state.

Change-Id: I60efa5692b26e9a4079c3c2fb66eb72b67badd02
Reviewed-by: Laszlo Agocs <laszlo.p.agocs@nokia.com>

remove assignmets which are defaults or bogus

Change-Id: I6e4fcef11e905ffbbb7a32210608faac7aa111e5
Reviewed-by: Marius Storm-Olsen <marius.storm-olsen@nokia.com>

Change-Id: I2bbbdad7fdd0c726316a4192240a00326fd4d72e
Reviewed-by: Samuel Rødal <samuel.rodal@nokia.com>

The default state is Deactivate, so getting extra ones at startup
is not needed.

Change-Id: I9c6b4b5dab58651f1933900421a3d6d68e00522f
Reviewed-by: Jørgen Lind <jorgen.lind@nokia.com>

Initialize variables and send the event in the beginning so that
we can be sure both client and server are in sync.

Change-Id: Ibabd2fa3cfc2c7defd4804f3fcbdd184eba46652
Reviewed-by: Jørgen Lind <jorgen.lind@nokia.com>

After the recent refactoring, the frameFinished method was still called
where we should now be calling frameFinishedInternal

Change-Id: I6ac5c0467dda6bb7e62976e8da28dc9d19a986c8
Reviewed-by: Samuel Rødal <samuel.rodal@nokia.com>

Apparently some qtbase change triggers the need to include a bunch of
headers which were not properly included before. The extra mouse
button support in the platform plugin simply did not compile.

Change-Id: I7d518021fa81f2d5232277f0bce26e48effbeae3
Reviewed-by: Jørgen Lind <jorgen.lind@nokia.com>

Change-Id: I3d0ab79eb261de4d22c6aadf0672f5c84ff3533e
Reviewed-by: Samuel Rødal <samuel.rodal@nokia.com>

Change-Id: Ia0510e00ffc925cde98ec5345a8129d1979c478c
Reviewed-by: Jørgen Lind <jorgen.lind@nokia.com>

Change-Id: I85aa62c9deae308612ed2b13468e7f1e48f82acf
Reviewed-by: Jørgen Lind <jorgen.lind@nokia.com>

Update expands the mouse button assignment code block, which formerly
handled only values 272, 273, and 274. New version handles all mouse
mouse buttons shown in kernel's input.h header (decimal values 272
through 285).
Value 286 appears to be the first button in the range for joystick
devices, so I did not include any other kernel-enumerated values.
(Other values might be possible- but I don't know of a mouse which
is currently being sold with more than 15 non-wheel buttons. The
mouse with 15 buttons is the Razer Naga, and I don't own it.)

Task-number: QTBUG-22759

Change-Id: I77f16eb89e24d7d6410199badff28c0d48f7df93
Reviewed-by: Jørgen Lind <jorgen.lind@nokia.com>

Change-Id: I979e7d9ca7152d4db6c3e7adae8353935a05d6c5
Reviewed-by: Andy Nichols <andy.nichols@nokia.com>

Sometimes, the window's coordinate system is not bounded by
the coordinates of the QScreen (such as if the window uses a
landscape window orientation on a portrait screen).
QGuiApplication::topLevelAt() will fail to find this window, but since
we have the window already at the wayland plugin level we just use that.

We already did this for mouse events, so it's natural to do it here as
well.

Change-Id: Id086dd940aa1cf4eb4480117d243ab8a76da6cca
Reviewed-by: Laszlo Agocs <laszlo.p.agocs@nokia.com>

The new windowRotation property is especially useful for the compositor
to know how much the surface needs to be rotated in order to be
displayed correctly.

Change-Id: I81ee17880ceafd465c39b4439dee978f2cb80924
Reviewed-by: Jørgen Lind <jorgen.lind@nokia.com>

Change-Id: I86559add9925dad6aef530433dd92066f1708757
Reviewed-by: Laszlo Agocs <laszlo.p.agocs@nokia.com>

And moving Wayland::SurfaceBuffer into its own file.

Also remove the extra functions in Wayland::Surface which just relayed
down to extended surface. The pattern is that WaylandSurface is the
entry point so we check there is the Wayland::Surface has the extension
we'r looking for, then calling it directly from WaylandSurface.

Last,  the data of an extension is suppose to be members of the
extension, so I moved the properties to be members of ExtendedSurface

Change-Id: I5877671a17ee873b7644094f695fcd3e4677964a
Reviewed-by: Samuel Rødal <samuel.rodal@nokia.com>

Change-Id: Ic03d08ec936059344a1e092329a9def0b4ceb9ee
Reviewed-by: Jørgen Lind <jorgen.lind@nokia.com>

QOpenGLContext is always required even when QT_COMPOSITOR_WAYLAND_GL is
not defined.  Moving the include outside the ifdef makes compositors
build without problems.

Change-Id: I4b8903b06a5f58ad8fd2953dbde1476877683cb4
Reviewed-by: Laszlo Agocs <laszlo.p.agocs@nokia.com>

outside the ifdef

Change-Id: I6f9048cd7d5992ef46dcc0e9d05e86f2f7ca4033
Reviewed-by: Laszlo Agocs <laszlo.p.agocs@nokia.com>

Change-Id: I1b53c5f0b0f908490dcd0ab30e46ab681f885954
Reviewed-by: Jørgen Lind <jorgen.lind@nokia.com>

It maps to the protocol's touch_cancel pretty well. :)

Change-Id: Id417aac768106c2f6618b3e7ecb51d6929977c37
Reviewed-by: Samuel Rødal <samuel.rodal@nokia.com>

This hack was added to work around the problem of calling
wl_display_iterate in two separate threads, but now that we are
are enforceing that wl_display_iterate can only be called in one thread
it now causes a deadlock when we are waiting in eglSwapBuffers for the
gui thread to actually call wl_display_iterate.

Change-Id: Id3153b184c191494838fc363380f7cae621ac64e
Reviewed-by: Jørgen Lind <jorgen.lind@nokia.com>

Clearing the device list on resource destroy is wrong: it cleans away
data devices for all clients. Calling free when removing stale devices
for a client is also wrong.

Change-Id: I7bfcc928762dc4ca7dbf5abeebcd2489956e6828
Reviewed-by: Jørgen Lind <jorgen.lind@nokia.com>

The compositor can now be configured to tell the clients to generate
mouse events when receiving touch events. In touchscreen-only systems
this can be used to reduce the number of Wayland events.

Change-Id: I231a15cd4ed463ee81c510c082a270efa255a1f3
Reviewed-by: Jørgen Lind <jorgen.lind@nokia.com>

We must do something otherwise the next inputDevice->mouseFocus() call
will dereference an already deleted instance.

Change-Id: I3ac92a37715f63c6dc43191968a8da51befbccde
Sanity-Review: Qt Sanity Bot <qt_sanity_bot@ovi.com>
Reviewed-by: Jørgen Lind <jorgen.lind@nokia.com>

The pointer is out of scope by the time it's sent through wayland.

Change-Id: I00e9f050dfa47bf078921600d2bf63502f0b4897
Sanity-Review: Qt Sanity Bot <qt_sanity_bot@ovi.com>
Reviewed-by: Ian Monroe <ian.monroe@nokia.com>
Reviewed-by: Gunnar Sletta <gunnar.sletta@nokia.com>

The pointer focus must be set in pressed. We cannot rely on
the implicit mouse tracking behavior of QWindow for QML stuff
because the WaylandSurfaceItems will not get mouse moves outside
of a press-release sequence.

Change-Id: Ic2cf2a7497cac7397bcfa26440dd05cded6182dc
Sanity-Review: Qt Sanity Bot <qt_sanity_bot@ovi.com>
Reviewed-by: Jørgen Lind <jorgen.lind@nokia.com>

Change-Id: I6e182c048282f5edd30f49be19dcc0f020679b85
Sanity-Review: Qt Sanity Bot <qt_sanity_bot@ovi.com>
Reviewed-by: Laszlo Agocs <laszlo.p.agocs@nokia.com>
Reviewed-by: Jørgen Lind <jorgen.lind@nokia.com>

This reverts commit 42ed1265.

Weston has been fixed so attaching null buffer on hide is now a valid
operation.

Change-Id: I0fc6af7f87c97e07075da03ed1b54d2e3c1d45fd
Sanity-Review: Qt Sanity Bot <qt_sanity_bot@ovi.com>
Reviewed-by: Jørgen Lind <jorgen.lind@nokia.com>

Apparently Weston does not like this and causes instant crashes when
hiding a top-level widget.  The QtCompositor-side behavior is dubious
too although at least that does not crash. Nevertheless attaching a
null buffer seems unnecessary.

Change-Id: I1d7fe3e0b05e5d77773b661a71bc37311968eded
Sanity-Review: Qt Sanity Bot <qt_sanity_bot@ovi.com>
Reviewed-by: Jørgen Lind <jorgen.lind@nokia.com>

Otherwise shm based apps (widgets) will crash randomly.

Change-Id: I5345ca9419a6702170f93fd817bbab4d9ef2bc59
Sanity-Review: Qt Sanity Bot <qt_sanity_bot@ovi.com>
Reviewed-by: Jørgen Lind <jorgen.lind@nokia.com>

Change-Id: If720f7f4797e319ab03d02b6ecb9c88cb71356c2
Sanity-Review: Qt Sanity Bot <qt_sanity_bot@ovi.com>
Reviewed-by: Samuel Rødal <samuel.rodal@nokia.com>

As in the past, to avoid rewriting various autotests that contain
line-number information, an extra blank line has been inserted at the
end of the license text to ensure that this commit does not change the
total number of lines in the license header.

Change-Id: I08d8673382f0230d5c5a22a25efaf5c45cef539a
Sanity-Review: Qt Sanity Bot <qt_sanity_bot@ovi.com>
Reviewed-by: Rohan McGovern <rohan.mcgovern@nokia.com>