This should provide a crisper cursor

Change-Id: Iaa3108cc2c6346e9db9755975ca2718875f470b7
Reviewed-by: David Edmundson <[email protected]>

Pick-to: 6.8
Change-Id: I4cec8d6c26076f0213dbeb9ab3d39585936d0c25
Reviewed-by: David Edmundson <[email protected]>

Wait until both the seat name and capabilities are known, to avoid
depending on the order of messages. On Weston, seat_name() is called
after seat_capabilities(), while it's the other way around on others.

Pass the seat name along in each place where the QPointingDevice ctor is
invoked.

This gets us closer to multi-seat support on Wayland: at least
qtdecl/examples/quick/pointerhandlers/singlePointHandlerProperties.qml
can see the seat name of the device provided with each QTouchEvent or
QTabletEvent. Followup to 5af836a and
dd8965b

Pick-to: 6.8
Task-number: QTBUG-72167
Task-number: QTBUG-115207
Change-Id: Ife25cfec0307893d949a2a50a9b49a8f560a60e0
Reviewed-by: David Edmundson <[email protected]>

Currently we don't set a cursor for tablet devices,
so we get either a generic fallback cursor (KWin)
or no cursor at all (GNOME).

This makes sure we get the same cursor we get for
mouse input.

The code is mostly identical to the mouse cursor
handling, so refactor things a bit to share

Pick-to: 6.8
Fixes: QTBUG-105843
Fixes: QTBUG-123776
Change-Id: Ie626ff978d9b66ec422804a103699eebec85e267
Reviewed-by: Shawn Rutledge <[email protected]>

When a fractional scaling value is sent, we expect the client DPR to
update, and a new frame sent to the compositor.

We can't be sure that our compositor thread won't already have the new
frame by the time we call surfaceCommitSpy.wait.

Change-Id: Iab7d57f07e854a79e69b97631fe1b6006a061b82
Reviewed-by: Shawn Rutledge <[email protected]>
Reviewed-by: David Redondo <[email protected]>

testMultiScreen explicitly sets clients on monitors, for this to work we
need autoEnter disabled.

e011b4b set this on startup, but due to
the way we reset the compositor this needs doing multiple times if
multiple tests are run.

Change-Id: Iaad3058b3ad8a12d01b3bd8272615fcd74ea250d
Reviewed-by: David Redondo <[email protected]>

When using OpenGLTexture in SharedMemoryBuffer, it is never freed.

Fixes: QTBUG-128350
Pick-to: 6.8 6.7 6.5
Change-Id: Ib277171c1b5409487dd236a97b54ce0e7547d6a1
Reviewed-by: Eskil Abrahamsen Blomfeldt <[email protected]>

There is an issue on Mali that keeps a pointer to the old wl_surface
which was current before, and when we make the new surface current it
tries to access it, even it have already destroyed it at that point.
The issue is reported here:

https://community.arm.com/support-forums/f/graphics-gaming-and-vr-forum/8146/r12p0-wayland-driver-odroid-xu3-frees-objects-too-early-leading-to-segm-fault/36417

Avoid the crash by calling eglMakeCurrent(EGL_NO_SURFACE) before
wl_surface destroy as the workaround for Mali GPU only.

Pick-to: 6.5 6.7 6.8
Fixes: QTBUG-127980
Change-Id: I900de9ce0f8ee32a7148f54c9aa53d36dc5d9bec
Reviewed-by: Laszlo Agocs <[email protected]>

Change-Id: I3adf020fa810bd6ddf5cf61e29a99abf181bf856
Reviewed-by: Qt Submodule Update Bot <[email protected]>

Change-Id: Ib3d49bc719dd2e41154baa250fab20dd515b3e39
Reviewed-by: Qt Submodule Update Bot <[email protected]>

This brings it more in-line with other platforms where the underlying
backing store has the lifespan of the window, implicitly fixing a
teardown issue with Qt Vulkan.

With this commit the wl_surface lifetime matches the QPlatformWindow, on
show and hide the shell surface is still recreated, with a null buffer
attached inbetween to get a blank slate as per the specification.

There is still some paths where a wl_surface is reset during the
platform window lifetime if the user ever converted a subsurface to a
toplevel or vice-versa, also for handling compositing restarts.

Task-number: QTBUG-125592
Change-Id: I449ec170a085c7c0fca504687556d702daddf79a
Reviewed-by: David Edmundson <[email protected]>

Change-Id: I15c405bf5bd49adc06e129910840c15e1e8ab2fb
Reviewed-by: Qt Submodule Update Bot <[email protected]>