BUG:415998

When loading a file with a vector layer, the node manager will
activate and deactivate the default tool in the wrong order. This
cannot be worked around; and in fact, we get a LOT of activate
and deactivate calls because just creating the config widgets will
call those:

KoStrokeConfigWidget::KoStrokeConfigWidget DefaultToolTabbedWidget(0x562719ef0e00, name="default-tool-tabbed-widget")
KoFillConfigWidget::KoFillConfigWidget KoFillConfigWidget(0x562719f206a0)
KoFillConfigWidget::deactivate() KoFillConfigWidget(0x562719f206a0, name="KoFillConfigWidget")  locks: 0
KoFillConfigWidget::activate() KoFillConfigWidget(0x562719f206a0, name="KoFillConfigWidget") 2
KoStrokeConfigWidget::deActivate(): locks 0
KoFillConfigWidget::deactivate() KoFillConfigWidget(0x562719f206a0, name="KoFillConfigWidget")  locks: 0
KoFillConfigWidget::KoFillConfigWidget KoFillConfigWidget(0x56271a039190)
KoFillConfigWidget::deactivate() KoFillConfigWidget(0x56271a039190, name="KoFillConfigWidget")  locks: 0
KoFillConfigWidget::activate() KoFillConfigWidget(0x56271a039190, name="KoFillConfigWidget") 2
KoStrokeConfigWidget::activate(): locks 2
KoFillConfigWidget::activate() KoFillConfigWidget(0x562719f206a0, name="KoFillConfigWidget") 2
KoFillConfigWidget::deactivate() KoFillConfigWidget(0x56271a039190, name="KoFillConfigWidget")  locks: 0
KoStrokeConfigWidget::deActivate(): locks 0
KoFillConfigWidget::deactivate() KoFillConfigWidget(0x562719f206a0, name="KoFillConfigWidget")  locks: 0
KoFillConfigWidget::deactivate() KoFillConfigWidget(0x56271a039190, name="KoFillConfigWidget")  locks: 2
SAFE ASSERT (krita): "d->deactivationLocks.empty()" in file /home/boud/dev/krita/libs/ui/widgets/KoFillConfigWidget.cpp, line 358
KoStrokeConfigWidget::deActivate(): locks 2
SAFE ASSERT (krita): "d->deactivationLocks.empty()" in file /home/boud/dev/krita/libs/ui/widgets/KoStrokeConfigWidget.cpp, line 481

BUG:416000

by ef55574a and then the crash
happens again.

I have tested this change and it does not break PalleteDocker.

PS: KIS_ASSERT_RECOVER_RETURN_VALUE trigger an assert and krita
fails to start up. KisPaletteModel::index works if I use
KIS_SAFE_ASSERT_RECOVER_RETURN_VALUE and change
KoColorSet::getGroup to return null pointer when its name
parameter is an empty string, but KisSwatchGroup::rowCount
crashes, so I think it is simpler to return an default
QModelIndex as soon as possible in KisPaletteModel::index

CCBUG:414890

This reverts commit 69bc2897.

This doesn't properly fix bug 415625, but the assert here in addJob
shouldn't be fatal.

BUG:415625

BUG:414818

There was one missing code block delimiter (```),
inverting the following code blocks.

This change stores the full image layer data in exported ORA files, even when part of the layer is outside of the visible canvas.
(similar to how layer data is stored in KRA format) Properly adds the "x" and "y" attribute to the resulting layer stack to re assemble
 the image. This change address a previous issue where the layer data would become cropped when exporting to ORA format, resulting in
lost layer data in some cases.

BUG:415896

BUG:415935

Even though the color may not change e.g. on hue change when
saturation is zero, the other selector shape still needs to
update its background when hue changes.

Something about my specific development environment was causing a crash due
to QGuiApplication::screens() indexing throwing an Out of Range error.
It likely has something to do with my vertical-right oriented display I
just set up. I'll do more investigation of this issue, and whether its a
QT or GNOME bug. For now, preventing a hard crash for simple GUI
button size scaling is probably appropriate.

This message suggests to use transform mask instead of transform tool
on File layer or Clone Layer. Previous message misled users to assume
that transform operations can't be performed on File or Clone layer.

BUG:415656

State requires an active view in the canvas
to change widget status.

BUG:415033

Previously, the default for KineticScrollingGesture was
MiddleMouseButtonGesture. This commit changes the default to
TouchGesture on Android and leaves MiddleMouseButtonGesture on other
platforms.

This isn't perfect: if you're fast enough you can still start a
couple of Krita processes even if application process sharing is
enabled. That can't be helped, though.

BUG:415794

BUG:415810

This patch calculates the horizontal and vertical fade areas
differently to avoid hard edges fade is not symmetric.

Additionaly artifacts are reduces on dabs with spikes greater than 2

This commit makes the Add option of subbrush off when changing from
Copy Translate type to any other type. The option is toggled off
even for Copy Translate type of multibrush. Previously the Add
option stayed on and drawing with other tools was not possible unless
it was turned off.

BUG:415651

In case of conflict in i18n, keep the version of the branch "ours"
To resolve a particular conflict, "git checkout --ours path/to/file.desktop"

This commit makes the deletion and addition of palette
in palette docker show in colorpicker's tool options'
Add to palette comboBox. This also prevents a crash
due to a dangling pointer due to palette pointing to
deleted memory location.

BUG:413548

Reverts 97c03e0f

 "-1" is a QMap key to signal "root group"

Changed Q_ASSERT to KIS_ASSERT_RECOVER_RETURN_VALUE to
try and avoid crashing if Qt is compiled in debug mode.

BUG:414890

for macOS with case sensitive File Systems

BUG:415218

Before this commit, files created using copying and pasting
clone layers from one file to another would be corrupted
and crash Krita when the user tried to open them.
This commit fixes crashing on opening those files.
After this commit, all clone layers are replaced by
empty paint layers. (Empty, because all data is lost).

CCBUG:414699

it pointed to the page
https://www.davidrevoy.com/article193/compile-krita-from-source-code-on-linux-for-cats
which itself points to the more recent
https://docs.krita.org/en/untranslatable_pages/building_krita.html

When the value of a QDoubleSpinBox is changed it will be round down to the amount of decimals that are assigned to that spin box. The KisDoubleParseSpinBox would check the new unrounded value with the current rounded value, which resulted in an incorrect comparison.

Only in this mode will Qt respect the contents of the ComboBox popup. This is what Fusion uses.

BUG: 406595

Differential Revision: https://phabricator.kde.org/D25739

The crash happens only on systems that have asserts enabled, that
is, don't have NDEBUG defined.

Binding uninitialized '*m_savedTransformArgs' to a cont-reference
generated valid c++ code (a reference initialized with nullptr),
and given that finishStrokeImpl() didn't try to use/dereference
this reference, the code worked fine and didn't crash on systems
without asserts.

But on systems with asserts enabled (non NDEBUG), boost::optional
triggered a sanity check assert for dereferencing nullptr and crashed
the application.

The patch removes entire code for resetting m_savedTransformArgs in
cancelStrokeCallback(). It was actually an artifact of some initial
refactoring.

Thanks Fredrik Hansson for pointing out the problem!
kde/krita!197

BUG:414672