CCBUG:438305

This patch fixes the problem in the inplace mode. The patch for legacy
mode is still pending

We should use projectionPlane()->tightUserVisibleBounds() instead of
extent() for that.

Two issues:

1) We need to explicitly ask shape layers to regenerate their projection
   and wait for that using a barrier.

2) It should happen inside the blocked-setdirty-calls block to make
   sure no intermediate results are visible to the user.

CCBUG:437886

This patch creates flickering on applying a transform to
a shape layer.

This reverts commit d67d9292.

CCBUG:437886

We only revert sip and pyqt, python update is good t have as it adds more
stability on arm builds. The issues that need to be resolved before
updating sip and pyqt:

- On arm pyqt causes
t.qpa.openglcontext: Could not create NSOpenGLContext with shared context, falling back to unshared context.
UNSUPPORTED (log once): POSSIBLE ISSUE: unit 0 GLD_TEXTURE_INDEX_2D is unloadable and bound to sampler type (Float) - using zero texture because texture unloadable
- pyqt-builder fails on detecting sip as dependency on setuptools build
(We fixed this by removing deps on pyqt-builder package)
- Also for some reason sip env fails on krita build phase breaking the
nighties.

This reverts commit 8b25f3a2.

If the user cancels a continued transform in IP mode, then the "new"
updates were not added to the update list. Therefore, only the initial
area was updated.

We used incorrect update rect for the undo commands in continued
mode. Undo and redo should usethe update rect that was present
**before** the previous (continued) state, not just what we had
before the current action.

If one of these masks was transformation mask, it could cause a crash.
Now if the layer has a transform mask **and** some other mask, the
root of the transformation switches to this transform mask.

BUG:439729

Amy spark authored Jul 13, 2021 and Dmitry Kazakov committed Jul 14, 2021

- MSVC warns that the constructor taking an instance of the superclass
is infinitely recursive (C4717). Fix it by calling the superclass's
instead.

- boost::orable, at least on 1.71, requires a definition of the |= operator.

CCBUG: 438398

This was required to publish on Google Play, now with App Bundles all
APKs will share the same versionCode.

A step necessary to move to App Bundles as per Google's Play Store
requirements.

Transform should now properly be initialized to the center of device
when creating an AnimatedTransformMaskParams instance.

BUG:438343

This changes KoCompositeOp.cpp to read the category names from
KoCompositeOpRegistry, effectively consolidating all blending mode
category name i18n calls in KoCompositeOpRegistry.cpp to avoid
duplication of these translation messages.

This basically moves all blending mode names to be translated only in
KoCompositeOpRegistry.cpp, using i18nc with context string. The
exceptions are COMPOSITE_IN and COMPOSITE_OUT which are not registered
in KoCompositeOpRegistry and had only been using i18n without context
string.

Reason for regression: it was assumed that the color of a "No Fill"
Stroke is Qt::transparent. But since this assumption no longer holds, we
need to check if the color is valid.

BUG:439725

Agata Cacko authored Jul 08, 2021 and Alvin Wong committed Jul 13, 2021

Before this commit, the translation context string would have
the same base no matter which resource or which storage it is.
However some storages and some resources require special treatment.

1) resources stored in memory storage don't have a file,
so the translation context string will be
memory/[resource_type]/[resource_filename] to fit other
strings.
2) myb presets are stored in a different place than most resources
(in the source code) so they have different paths, which means
different context strings.
3) resources from bundles need to be somewhat distinguishible as well,
so they have
[path_to_bundle]:[resource_path (usually resource type)]/[resource_filename]
context string.

Agata Cacko authored Jul 07, 2021 and Alvin Wong committed Jul 13, 2021

Before this commit, Seexpr scripts didn't have translatable names.
This commit ensures they have. Since they don't have a name inside
the .kse file, we use parsed filename to extract a nice name.

Agata Cacko authored Jul 07, 2021 and Alvin Wong committed Jul 13, 2021

Before this commit, resources from the bundles
didn't have translations. This commit ensures that the resources
in the bundles can have their names translated.

To make sure the resources of the same type have the same code
for reading both from the folder and from the bundle,
functions readGeneric() and readAllLinesGeneric() were written
to allow just calling the function that will take care of
the specific syntax of reading from the file.
Note that readGeneric() takes a different hash for the zip
handle than readGenericAllLines (explanation below).

Note:
$ziph->read() didn't work correctly (always returning empty string)
so I used a hack with reading the whole file using getline().
Possibly I was using read() incorrectly; but this method works.
If anyone in the future wants to improve this to use read()
directly, without the hack, feel free to.

Agata Cacko authored Jul 06, 2021 and Alvin Wong committed Jul 13, 2021

Before this commit, the extracted comments weren't really
helpful, the translator could not know what to do with a specific
string. This commit introduces some explanations like
what the path in the context string mean etc.

Agata Cacko authored Jul 01, 2021 and Alvin Wong committed Jul 13, 2021

Before this commit, the file was only closed for particular
filenames. This commit ensures that the file is closed
in all cases.

Agata Cacko authored Jul 01, 2021 and Alvin Wong committed Jul 13, 2021

Before this commit, only gradients, palettes and potentially
brush tips could be translated.
This commit adds some other files types like .kpp presets,
.kwl window layouts, .kws workspaces, .kgm gamutmasks,
.png patterns, .myb presets and .svg symbol libraries
and brush tips.

CCMAIL:kde-i18n-doc@kde.org
CCMAIL:kimageshop@kde.org

Qt::Tool can't receive keyboard events