This commit backports the Vulkan server buffer and texture sharing code
from Qt 5.14 to Qt 5.12 as an opt-in feature. To enable, configure with
"-feature-wayland-client-texture-sharing-experimental
-feature-wayland-compositor-texture-sharing-experimental"
Contains code from the following commits:
Add server buffer integration based on Vulkan
(commit a998944b2035c80999db4c1d9f7666c87ba73a86)
Compressed texture support for vulkan server buffers
(commit 4f87adc442a318b9c05fc73ae1020d0df0dfded3)
Implement server-side toOpenGlTexture for Vulkan
(commit 19361e7259f04b08925b1e8e99faf9460770ee7b)
New texture sharing protocol and infrastructure
(commit b09e808bf222984196cebd4447880781e3265714)
Change-Id: I6c36ef7fddcd4db39e80d03a822d89f15eae3434
Reviewed-by: Lars Knoll <lars.knoll@qt.io>
Reviewed-by: Johan Helsing <johan.helsing@qt.io>
For shells that do not send configure events when exposed:
If setGeometry() was called before the window was initialized, Qt would
not mark the window as mapped, and the window would never become
visible.
Change-Id: Ic933cbbff20702424129c11264215181330cc7d1
Reviewed-by: Johan Helsing <johan.helsing@qt.io>
Implicit capture of 'this' in [=] is deprecated in C++20.
Fix by using explicit captures.
Change-Id: Ie3a94ec60d7c56b2856d201fa3d68d0670bdd7b9
Reviewed-by: Mårten Nordheim <marten.nordheim@qt.io>
If we hide a window whilst a compositor has a pending frame to show,
it's possible the compositor will not render the frame and not return
the callback.
If this happens on the next window expose we can be left with
mFrameCallbackTimedOut still true causing isExposed() to remain false
and us to not send the next buffer when we later show the window again.
Change-Id: I507410415d1a930fd5fa736412055e68fdf6c1d3
Reviewed-by: Johan Helsing <johan.helsing@qt.io>
When we did invokeMethod for handling the frame callbacks, we had to wait for
the GUI thread to finish whatever it's doing before we would stop blocking.
Fix it by clearing the frame callback timer and stop blocking immediately,
while delaying the rest of the work until it can be run on the other thread.
Fixes: QTBUG-76397
Change-Id: I343e4feac4838926b4fa2ccac2948988bc6c3bb7
Reviewed-by: Paul Olav Tvete <paul.tvete@qt.io>
Sometimes origin will be nullptr, triggering a crash.
[ChangeLog][QPA plugin] Fixed a crash that sometimes happened when
starting a drag-and-drop operation.
Fixes: QTBUG-76368
Change-Id: I8f4e6b05f073644834c3c72a8307dac5b897f626
Reviewed-by: Johan Helsing <johan.helsing@qt.io>
Neither Qt::ToolTip nor Qt::Popup are single bits in Qt::WindowFlags, and do in
fact include Qt::Window. This meant that when we or'ed them and did a bitwise
and with QWindow::type(), we would match more types than just Qt::Popup and
Qt::ToolTip. We would for instance get any Qt::Window as well, which meant the
main window would be added to activePopups, leading to strange things
happening, such as crashes and the main window closing unexpectedly.
[ChangeLog][QPA plugin] Fixed a crash when closing multiple popups at once.
Fixes: QTBUG-76124
Change-Id: I1a6a59e161a436604a7ac8ab824396481dc99a20
Reviewed-by: Paul Olav Tvete <paul.tvete@qt.io>
(cherry picked from commit 4a9a5e2f62935488b1eeb06cd5d69a168a5a9308)
Neither Qt::ToolTip nor Qt::Popup are single bits in Qt::WindowFlags, and do in
fact include Qt::Window. This meant that when we or'ed them and did a bitwise
and with QWindow::type(), we would match more types than just Qt::Popup and
Qt::ToolTip. We would for instance get any Qt::Window as well, which meant the
main window would be added to activePopups, leading to strange things
happening, such as crashes and the main window closing unexpectedly.
[ChangeLog][QPA plugin] Fixed a crash when closing multiple popups at once.
Fixes: QTBUG-76124
Change-Id: I1a6a59e161a436604a7ac8ab824396481dc99a20
Reviewed-by: Paul Olav Tvete <paul.tvete@qt.io>
The Wayland plugin now takes full control over delivering update request and
implement frame callbacks for both egl and shm.
[ChangeLog][QPA plugin] The non-blocking version of eglSwapBuffers is now used, if
supported. This fixed a bug where minimized windows would block the event loop.
[ChangeLog][QPA plugin] Windows that don't get frame callbacks from the
compositor within 100 ms are now set as not exposed. This should stop most
clients from rendering unnecessary frames to minimized or hidden windows.
Also, when we relied on the QPA version of requestUpdate, we would sometimes
deliver one update request while we were waiting for a frame callback. When we
implement the fallback timer ourselves we can make sure we only deliver the
fallback if there are no pending frame callbacks.
QtQuick and other applications often depend on blocking swapBuffers to throttle
animations. If the context's surface format has a non-zero swapInterval, try to
emulate a blocking swap.
Fixes: QTBUG-69077
Change-Id: I3c6964f31a16e9aff70b8ec3c5340e640a30fef2
Reviewed-by: Paul Olav Tvete <paul.tvete@qt.io>
Because some shells don't allow attaching buffers before configure, we need to
not be exposed until we know that we don't want a shell surface.
Change-Id: Ida7101a99f953d02cf6401e4ea8d28cfabd6e102
Reviewed-by: Giulio Camuffo <giulio.camuffo@kdab.com>
Reviewed-by: David Edmundson <davidedmundson@kde.org>
The previous implementation had gaps in several places.
[ChangeLog][QPA plugin] Fixed a bug where the window decoration's damaged area
didn't cover the entire decoration. This meant some compositors would not
redraw those areas.
Change-Id: Ic72663dde301936635b2a1cfa90570a53227e8ea
Fixes: QTBUG-74341
Reviewed-by: Paul Olav Tvete <paul.tvete@qt.io>
Reviewed-by: Pier Luigi Fiorini <pierluigi.fiorini@liri.io>
The call to flushWindowSystemEvents() sometimes caused new popups to be shown
in the middle of hiding another. I.e. if multiple events show and hide
surfaces, they would be shown/hidden in opposite order of their corresponding
events. QMenus sometimes suffered from this (can be seen with the qopenglwidget
example from qtbase).
When the flush was added 5 years ago in 70c578cb5, it was to "reduce the
chances of seeing a bad frame". I don't see any rendering artifacts, though,
and I can't find any bug report on it. So let's hope it's safe to remove the
hack.
[ChangeLog][QPA plugin] Fixed a crash that sometimes happened when switching
popups.
Also adds more info to the workaround warning message that appears when a popup
grab is attempted and there already is another grabbing popup that is not the
parent.
Fixes: QTBUG-73524
Change-Id: Ibfcbb48c4bbe295c2be1a30add2d7e05cad398c5
Reviewed-by: Paul Olav Tvete <paul.tvete@qt.io>
This caused regressions because QtQuick depends on swapBuffers for throttling
animations. We probably need to emulate a blocking swapBuffers and continue
after a timeout, but until we have a patch for this, revert this to avoid
releasing a regression.
This brings back the bug with a frozen event loop when a surface is waiting for
a frame callback, but this is preferable to a regression.
This reverts commit 24ef6801e42b15be0f31860476199106c69e15d6.
Fixes: QTBUG-72578
Change-Id: If6435a947aae5e9fd775404649a392bfafe9130a
Reviewed-by: Paul Olav Tvete <paul.tvete@qt.io>
This adds a new client buffer integration: wayland-eglstream-controller,
which contains the EGLStream logic from wayland-egl, and additionally
uses NVIDIA's wayland-eglstream-controller protocol to avoid the issue
where the stream is not ready at the time of first buffer attach.
This is not enabled by default. Can be used like this:
QT_WAYLAND_CLIENT_BUFFER_INTEGRATION=wayland-eglstream-controller ./pure-qml
Fixes: QTBUG-71697
Change-Id: I73bb2a8fe9852afe1b5807cbb8c35dc4c7624dad
Reviewed-by: Johan Helsing <johan.helsing@qt.io>
[ChangeLog][QPA plugin] Fixed a bug where surface damage for window decorations
was outside the surface.
Fixes: QTBUG-72818
Change-Id: I32d00174f1c308952a98bdb55731eb77be54f331
Reviewed-by: Pier Luigi Fiorini <pierluigi.fiorini@liri.io>
Some compositors fail to send wl_pointer.enter events before button and axis
events. This led to nullptr dereference crashes.
[ChangeLog][QPA plugin] Fixed a crash caused by compositors sending incorrect
pointer events.
Fix it by adding nullptr checks before trying to access the pointer focus in
case the compositor didn't send enter, or if something deleted the surface
(it's a QPointer).
This means we are now silently ignoring events for misbehaving compositors
(which is probably not optimal), but I don't see how we can easily distinguish
between a missing enter and a deleted QWaylandSurface without adding more
bookkeeping overhead.
Ditto for wl_keyboard.
Fixes: QTBUG-72235
Change-Id: I4bfaa3da4ebae37b059e1e450975566559edf40b
Reviewed-by: Pier Luigi Fiorini <pierluigi.fiorini@liri.io>
[ChangeLog][QPA plugin] Fixed a protocol error that used to happen when closing
a menu with an active tooltip.
Fixes: QTBUG-71734
Change-Id: I784fef08494fabaa4debea11f51116cf9de1f86e
Reviewed-by: David Edmundson <davidedmundson@kde.org>
Reviewed-by: Pier Luigi Fiorini <pierluigi.fiorini@liri.io>
The Wayland plugin now takes full control over delivering update request and
implement frame callbacks for both egl and shm.
Fixes two bugs:
[ChangeLog][Client] The non-blocking version of eglSwapBuffers is now used.
This fixed a bug where minimized windows would block the event loop.
Also, when we relied on the QPA version of requestUpdate, we would sometimes
deliver one update request while we were waiting for a frame callback. When we
implement the fallback timer ourselves we can make sure we only deliver the
fallback if there are no pending frame callbacks.
Fixes: QTBUG-69077
Change-Id: I2d3a6896c32e63d8520b57448a3601a817816a91
Reviewed-by: Paul Olav Tvete <paul.tvete@qt.io>
The xkbcommon configure logic was refactored in
qtbase/c3a963da1f9e7b1d37e63eedded61da4fbdaaf9a.
For more details see the relevant commit.
Change-Id: Ic1aa26846ab8266c589f6e92dc8b81aba36df58a
Reviewed-by: Johan Helsing <johan.helsing@qt.io>
QList is not included when built with QT_NO_QOBJECT defined.
Change-Id: Iaf8874583e81b08a49ec246a5de0b62803c8ed7e
Reviewed-by: Sami Nurmenniemi <sami.nurmenniemi@qt.io>
Reviewed-by: Liang Qi <liang.qi@qt.io>
GCC 4.8 doesn't like QPointers in signal connections.
Also, Default Member Initializers for aggregates
are C++14, so we need to write a constructor to make
a class with DMIs work in C++11.
Change-Id: I567c60a8081fa2f268517ec747bd6534674bcce5
Reviewed-by: Simon Hausmann <simon.hausmann@qt.io>
This is used when creating new grabbed popups to ensure they are only
created on the topmost grabbing popup as per the specification.
Non-grabbing popups don't have this restriction.
Change-Id: I8e2b8cd0f091688c847f81f2e5d33ce2f0df96a1
Reviewed-by: Johan Helsing <johan.helsing@qt.io>
It's valid to create a non-grabbing XDG-popup, such as a tooltip, before an input event.
The current guard should only apply for grabbing popups, otherwise tooltips created early
are turned into XDG-toplevels resulting in incorrect positioning.
Change-Id: I5bb59a68964ef2375c81e8b4e1a73361b9d2cbcc
Reviewed-by: Johan Helsing <johan.helsing@qt.io>
Avoids printing deprecation warnings for shells that are available by
default but not actually used. In particular, avoids warning about
wl-shell when using ivi-shell.
Change-Id: Id075e97572e6ec876ad24f369808c61a479e9b07
Reviewed-by: Johan Helsing <johan.helsing@qt.io>
Before 46aece51330b6f9b5b4bfd68a5b3b183dc5b1c38 we would always call
QWaylandWindow::applyConfigure which in turn sends an expose event.
Without that clients do not commit their initial buffer.
This adds back the expose event on the first received configure.
Fixes: QTBUG-71509
Change-Id: Ica39bb23245957948e1e1f75caf4f16f7b086ef7
Reviewed-by: Johan Helsing <johan.helsing@qt.io>
The code for isNormal was wrong, so set_toplevel (which clears fullscreen
state) would be sent unless the window was both maximized and fullscreen. Fixed
it and made it a little bit more readable.
Change-Id: Icaa52d3a058df064a0bd1164119af3f511cb4220
Fixes: QTBUG-71350
Reviewed-by: Paul Olav Tvete <paul.tvete@qt.io>
Reviewed-by: David Edmundson <davidedmundson@kde.org>
Only send expose event when the geometry actually changed.
Change-Id: Ic06986ce5d11e0ff7a842303f57093b8ff25b9f6
Reviewed-by: Paul Olav Tvete <paul.tvete@qt.io>
Reviewed-by: David Edmundson <davidedmundson@kde.org>
QBackingStore::flush is sometimes called with an unxeposed window, in that
case, don't attach the buffer to the wl_surface immediately, as that causes
protocol errors with xdg_shell.
Flushed buffers are instead stored until we get the first configure event.
[ChangeLog][QPA plugin][xdg-shell] Fixed a bug where buffers were sometimes
attached and committed before the first configure event, causing protocol
errors.
Fixes: QTBUG-71345
Change-Id: If9409d97bd25f6b13940c56141920a664c349c8e
Reviewed-by: Paul Olav Tvete <paul.tvete@qt.io>
Reviewed-by: David Edmundson <davidedmundson@kde.org>
kwin and perhaps other compositors sometimes send questionable configure events
with state set to maximized and size == {0,0}.
Previously, we would then restore the m_normalSize, which would give us a
maximized window with the size of a windowed (normal) one.
A size of zero usually means that it's up to the client to decide, which makes
sense for non-maximized windows, but not so much for maximized ones. This is
what the protocol spec says about the maximized state:
The surface is maximized. The window geometry specified in the configure
event must be obeyed by the client.
It's fixed in the dev version of kwin, but until then, let's be on the safe
side and only resize the window if things make sense.
[ChangeLog][QPA plugin] Fixed a bug where maximized windows would resize to
their unmaximized size if the compositor sent an invalid configure event.
Change-Id: I2371b29c82426ac48cd1c18c14c3dd0fe4f2250e
Reviewed-by: David Edmundson <davidedmundson@kde.org>
Reviewed-by: Paul Olav Tvete <paul.tvete@qt.io>
cursorSize was a static which was multiplied with the dpr each time
loadCursorSize was called resulting in a bigger and bigger cursor size.
Fix it by making the static constant.
Change-Id: Ie23cf0cc0d6c13721feecebe214d7c794be35077
Reviewed-by: Paul Olav Tvete <paul.tvete@qt.io>
Reviewed-by: David Edmundson <davidedmundson@kde.org>
Whilst wayland can't actually move the window in a QWindow::setGeometry
call the previous behavior was to keep it so that it appeared to the
client that it had moved.
The new behavior is non-consistent, QWindow::setPosition will update
QWindow::position until the first configure call is received at which
point it will move back to 0,0. This then generates a QMoveEvent for the
window that hasn't moved.
This patch keeps the QWindow::position consistent with the user request
and doesn't generate move events.
Change-Id: Ifb1433b3902d44c1b2e43036bc1805a6e00128fb
Reviewed-by: Johan Helsing <johan.helsing@qt.io>
Some compositors don't advertise wl_data_device_manager support, and the client
has no control over this, so the assertion doesn't make sense. Returning
nullptr in that case should be fine.
Change-Id: If0b145326d074ea4ebf8f5a12654962d56012a2d
Reviewed-by: Paul Olav Tvete <paul.tvete@qt.io>
When we switch to only including core wayland headers, wl_resource will be an
opaque type. Use the getters and setter functions instead.
Task-number: QTBUG-70553
Change-Id: I7d84d48a4ee3586f231a331cd15716686dcee775
Reviewed-by: Paul Olav Tvete <paul.tvete@qt.io>
The old xdg-shell-unstable-v5.xml was malformed, and was missing some
<description> tags.
The updated version passes wayland-scanner run with --strict. Hopefully we can
start to use it soon.
Otherwise, it's just documentation fixes.
Change-Id: Id28ad7196643156cd22407e12ed7ff1e7fb34ce6
Reviewed-by: Pier Luigi Fiorini <pierluigi.fiorini@liri.io>
Since we're using the deprecated ways of including wayland (wayland-server.h
instead of wayland-server-core.h), we can't upgrade wayland.xml to a version
that's newer than the system libwayland. Our minimum supported version is 1.6.
This reverts commit 1d51503daead2e09cc2526a0114bedd4669c1d3f.
Fixes: QTBUG-70528
Change-Id: I102cb215a40a1aaaeba6f463200eefc3be653c52
Reviewed-by: Paul Olav Tvete <paul.tvete@qt.io>
It requires libwayland 1.8 and we're supposed to support versions as old as 1.6
Task-number: QTBUG-70528
Change-Id: Iff4c2542d103d53aba8db99ecceddfccb79cfcb6
Reviewed-by: Paul Olav Tvete <paul.tvete@qt.io>
