Signed-off-by: Nikolay Sivov <nsivov@codeweavers.com>

Also we don't care about the content of the buffer so using
HeapRealloc() makes no sense.

Otherwise it's impossible to know which iteration failed.

Wine-Bug: https://bugs.winehq.org/show_bug.cgi?id=55777

This mostly matters for Mac as segment CS is handled as a real LDT
entry.
Signed-off-by: Eric Pouech <epouech@codeweavers.com>

Signed-off-by: Eric Pouech <epouech@codeweavers.com>

Fixes cursor not being released when alt-tabbing out of an unresponsive
foreground window.

If the window is already fullscreen and its size is compatible with what
the compositor is requesting, don't force a resize, since some applications
are very insistent on a particular fullscreen size (which may not match
the monitor size).

Rename the function wayland_surface_configure_is_compatible to
wayland_surface_config_is_compatible to match the associated struct name.

The maximized and fullscreen states disallow surface contents that are
larger than the configured size. However, if we have a larger surface,
and only part of the surface contains useful client data, we can inform
the compositor using xdg_surface_set_geometry, so that only the
specified region is considered for compositor-side positioning and
state constraints.

We use the geometry to support the presentation of windows that insist
on particular positions and sizes which may not match the compositor's
requested size (e.g., fullscreen but may overshoot a few pixels outside
the monitor boundary).

Detect when an window wants to become fullscreen and request the Wayland
fullscreen state from the compositor. The fullscreen state for windows
is inferred by their size and position relative to the monitor that
contains their largest region.

Since there is no explicit fullscreen state for windows, compositor
initiated fullscreen requests may lead to a size change, but will not
directly affect other aspects of the window state.

Instead of querying the window config for a wayland_surface on demand,
query it and store it only when it is applied through SetWindowPos. In
addition to being more efficient, this prevents deadlocks by avoiding
calling NtUser functions that require the user lock, while holding
the window_surface lock.

Tiled states don't place strict constraints on the surface size, but
they indicate a strong size preference, so try to respect it.

If we are processing a config request by the compositor, consider that
config as authoritative for the window. Otherwise use the window state
to determine the new Wayland state to apply.

It leaves the client in a bad state, causing test failures on recent Windows versions.

This avoids a crash when Wine is compiled using --without-krb5.
This makes test_ticket_cache() more consistent with test_kerberos().

Since self.cursorFrames is a copy property, its pointer value will only
equal frames when both are nil. Also return early from the function if
the array contents are equivalent.

Wine-Bug: https://bugs.winehq.org/show_bug.cgi?id=55713

The functions are in gdi32 nowadays.