The RHI support and compositor in QPlatformBackingStore were
tied to the surface format of the top level window owning
the backing store.
This meant that inserting an RHI-enabled widget (QRhiWidget,
QOpenGLWidget, QQuickWidget, QWebEngineView) into the widget
hierarchy required recreating the top level window with a
matching surface format that could support the RHI composition.
It also meant that we could not have two RHI enabled widgets
with different surface format requirements (Metal and OpenGL
for example) in the same top level widget hierarchy.
The recreation of the window had various visual side effects,
such as temporarily switching out of full screen state, or the
widget rendering a frame of black, as well as more serious
problems such as not correctly restoring the window geometry.
In addition, if client code had pulled out the winId() of the
window, and did not invalidate these references on window
destruction via QEvent::WinIdChange or QEvent::PlatformSurface,
the client would reference stale window handles. Although
this is a programming error (QWidget::winId() specifically
mentions this requirement), we should avoid recreation if
we can.
We were already supporting flushing the backingstore to
individual native child widgets, but always did so via a
single RHI managed by the platform backingstore. By
expanding QPlatformBackingStore to keep one set of RHI
support and compositor per surface format, we can refine
the logic in QWidget and QWidgetRepaintManager to not
require recreating the top level. Native child widgets
are then flushed independently, including any RHI textures
and raster content that overlaps with the widget.
We still assume that a single RHI support and compositor
can be be used for multiple windows, as long as those
windows have the same surface format. In the future, if
needed, we can refine this to use one set per surface
format e.g.
Fixes: QTBUG-119221
Fixes: QTBUG-121181
Fixes: QTBUG-120096
Task-number: QTBUG-115652
Task-number: QTBUG-108344
Task-number: QTBUG-113557
Task-number: QTBUG-119309
Change-Id: I2635ed3d20c2fb76eab3b8130007dd656a0b93e5
Reviewed-by: Laszlo Agocs <laszlo.agocs@qt.io>
As part of eacd58d4e7, a mechanism was
added to prevent posting redundant UpdateRequest events to the top
level widget, managed by QWidgetRepaintManager. The mechanism relied
on a boolean that was set when posting an update request event, and
reset when processing the event for the top level widget, as part
of QWidgetRepaintManager::sync().
However, for paint-on-screen widgets, we don't post an update request
to the top level, we post it to the paint-on-screen widget directly.
And when processing that event, we don't paint the widget though the
normal QWidgetRepaintManager machinery, but instead call the paint
event via QWidgetPrivate::paintOnScreen().
As a result, an update() on a paint-on-screen widget would result
in never receiving updates for non-paint-on-screen widgets, as
we assumed there was no reason to send further update requests.
We could fix this by clearing the updateRequestSent flag as part
of the paintOnScreen() code path, but that's incorrect as the flag
represents whether the top level QWidgetRepaintManager needs an
update request event or not, and would lead to missed updates
to normal widgets until the paint-on-screen widget finishes its
update.
Instead, we only set updateRequestSent if we're posting update
requests for non-paint-on-screen widgets, which in practice
means the top level widget.
The paint on screen logic in QWidgetRepaintManager::markDirty
still takes care of rate-limiting the update requests to the
paint-on-screen widget, by comparing the dirty area of the
widget.
There is definite room for improvement here, especially in the
direction of handling update requests via QWindow::requestUpdate
instead of manually posted events, but for now this will have to
do.
Fixes: QTBUG-80167
Pick-to: 6.7 6.6 6.5 6.2 5.15
Change-Id: Ib5685de7ca2fd7cd7883a25bb7bc0255ea242d30
Reviewed-by: Richard Moe Gustavsen <richard.gustavsen@qt.io>
This was a feature in Qt 4, but the only platform that carried it over
to Qt 5 was Windows, in ab76593f18, and
that's still the situation.
As a first step in possibly implementing this on more platforms, lets
replace the hard-coded check in QWidgetRepaintManager::hasStaticContents
with a proper QPlatformIntegration capability check in the only call site.
Pick-to: 6.7
Change-Id: I2067109f45116cd8c62facf7224cd748f19e845b
Reviewed-by: Axel Spoerl <axel.spoerl@qt.io>
QWidgetPrivate::setGeometry_sys wrongly triggers full update, which is
why the test was QEXPECT_FAIL, even on Windows, which does implement
static contents.
We can test the behavior without involving the broken QWidget behavior
by resizing the QWindow instead, which triggers a resize of the platform
window and will be fed back through QWSI, similar to when the user
resizes the window via the window frame.
The QEXPECT_FAIL message for the QWidget resize has been updated to
reflect why it's expected to fail.
tst_QWidget::optimizedResize_topLevel() has similar issues to this
test, and solves it by using Window specific platform code. For now
this test has been left alone, as there are a lot more going on in
that test in terms of workarounds and blacklistings, so it needs
further investigations.
Pick-to: 6.7
Change-Id: I28a55c0723b3e7f16b362d1ebc369e038d5a7ac4
Reviewed-by: Axel Spoerl <axel.spoerl@qt.io>
The scroll bars may show up in one of the grabs and not the other,
(as seen on macOS), which is interesting, and possibly a bug, but
is not what the test is checking for, so let's disable the scrollbars
to stabilize the test.
Pick-to: 6.7
Change-Id: Iad97ca5b22783bb1bee51b560957c5461b7d9923
Reviewed-by: Richard Moe Gustavsen <richard.gustavsen@qt.io>
According to QUIP-18 [1], all tests file should be
LicenseRef-Qt-Commercial OR GPL-3.0-only
[1]: https://contribute.qt-project.org/quips/18
Pick-to: 6.7
Task-number: QTBUG-121787
Change-Id: I9657df5d660820e56c96d511ea49d321c54682e8
Reviewed-by: Christian Ehrlicher <ch.ehrlicher@gmx.de>
This reverts commit 4096667d66.
Reason for revert: <Screen shot workaround is flaky.>
Change-Id: I6e01cc584c094d0d0b8c1544b9daf72d648dd002
Reviewed-by: Tor Arne Vestbø <tor.arne.vestbo@qt.io>
Due to an XCB library change, QXcbBackingStore::toImage() cannot be
safely assumed to return an image identical to QWidget::grab().
The test functions fastMove(), moveAccross() and moveInOutOverlapped()
relied on QXcbBackingStore::toImage() and failed.
They were backlisted on Linux/XCB.
This patch obtains a screen shot instead of an image from the backing
store on XCB platforms. It processes events until the screen shot
matches QWidget::grab(). It makes the test fail only if the comparison
times out.
The patch also removes the BLACKLIST file, containing only the test
functions mentioned above.
Fixes: QTBUG-109036
Pick-to: 6.5 6.4
Change-Id: I26dd5b89dc62b313db066a285f6ad7d4d92baaf2
Reviewed-by: Volker Hilsheimer <volker.hilsheimer@qt.io>
When a widget's palette has different brushes in Active / Inactive
color groups, or it is pixmap based, multiple paint events are
triggered. This leads to unpredictable, double entries in
QWidgetRepaintManager::dirtyWidgetList().
This patch overrides event() of all test widgets and ignores
activation / deactivation events in order to make the entries of
QWidgetRepaintManager::dirtyWidgetList() predictable.
Pick-to: 6.4
Change-Id: I164d7ab4148551590ac3c50fcc3b9f98c5ac0535
Reviewed-by: Volker Hilsheimer <volker.hilsheimer@qt.io>
The helper function TestWidget::waitForPainted returned after the test
widget had consumed the first paint event. In case of multiple paint
events, QRegion paintedRegions did not match the entire region that
was supposed to be painted. The test function children() failed/flaked
due to that.
This patch extends the helper function. After consumption of the first
paint event, it processes events until the painted regions no longer
change.
Pick-to: 6.4
Fixes: QTBUG-108764
Change-Id: I54e14bb07725dd1f602cc93085da13836e3b7494
Reviewed-by: Volker Hilsheimer <volker.hilsheimer@qt.io>
Replace the current license disclaimer in files by
a SPDX-License-Identifier.
Files that have to be modified by hand are modified.
License files are organized under LICENSES directory.
Task-number: QTBUG-67283
Change-Id: Id880c92784c40f3bbde861c0d93f58151c18b9f1
Reviewed-by: Qt CI Bot <qt_ci_bot@qt-project.org>
Reviewed-by: Lars Knoll <lars.knoll@qt.io>
Reviewed-by: Jörg Bornemann <joerg.bornemann@qt.io>
Expose QWidgetRepaintManager's data structures so that we can write
unit tests, and verify that they are correct after moving opaque
widgets (which triggers the accelerated move code path).
Improve the compareWidget logic to not rely on screen grabbing
(which requires permissions), but instead use QPlatformBackingStore's
toImage function, which is faster and more reliable, and also doesn't
require us to show the UI we want to grab full screen in order to
avoid issues with overlapping windows etc.
Change-Id: Iff2ea419f03a390ab6baca26814fef6ff45f7470
Reviewed-by: Tor Arne Vestbø <tor.arne.vestbo@qt.io>
Added to QWidgetRepaintManager test case, which is the only place where
the function is used.
Includes a helper that creates a complex scene with opaque children,
which will be used in additional unit tests.
Change-Id: I0e0188dd560923a552a8967d8e992dc17cc849d6
Reviewed-by: Tor Arne Vestbø <tor.arne.vestbo@qt.io>
In paintAndFlush, QWidgetRepaintManager subtracts opaque children if
the target isn't overlapped and isMoved is set to true. So in moveRect,
set isMoved to true after the blitting of movable areas, and reset it to
false if we have overlapped sibling or child regions. Otherwise, moving
so far that sourceRect is invalid (none of the original pixels are
visible after the move) we end up in a code path that sets isMoved to
true even with overlapping children or siblings, which then breaks
paintAndFlush's assumptions.
Reuse the test case written by Sergiy Korobov <tiamatenko@gmail.com> in
earlier attempts to fix this bug.
Fixes: QTBUG-26269
Pick-to: 6.2
Change-Id: If7443863f5eee79a80220cd587522122f42a21e4
Reviewed-by: Qt CI Bot <qt_ci_bot@qt-project.org>
Reviewed-by: Allan Sandfeld Jensen <allan.jensen@qt.io>
At the current state, Qt::WA_StaticContents is not respected on any
platform, and optimizations for opaque widgets is not implemented on
Cocoa.
Task-number: QTBUG-73117
Change-Id: I48b7b0c4436da5a609fdbe51bb1c9789a2a059bd
Reviewed-by: Qt CI Bot <qt_ci_bot@qt-project.org>
Reviewed-by: Tor Arne Vestbø <tor.arne.vestbo@qt.io>
Tor Arne Vestbø
Lucie Gérard
Axel Spoerl
Volker Hilsheimer