Summary: Found via `codespell -q 3 -S *.xpm,./3rdparty -L ba,fo,lod,pard,uint,vertexes`

Reviewers: rempt, #krita, woltherav

Reviewed By: #krita, woltherav

Subscribers: woltherav

Tags: #krita

Differential Revision: https://phabricator.kde.org/D24680

despite conditioning tput message keeps running in the
factory and sends an error message breaking the process

Now we have difference between KisFilter::factoryConfguration() and
KisFilter::defaultConfiguration():

* KisFilter::factoryConfguration() creates an empty configuration
  object of a proper C++ type. This object should be filled by fromXML()
  right after this call.

* KisFilter::defaultConfiguration() calls factoryConfiguration() and
  fill it with default values at the state how they should look in
  filter dialog.

This split is needed, because new versions of Krita may introduce
options not available in older saves configs. In such cases, Krita
should be able to recognize if the config is the "old" one, and
doesn't have new parameters.

BUG:412982

BUG: 413033

And restore the current composition after export.
BUG:412953

These signals only ended up in used parts of the color selector
code.

Whe shouldn't change value to fix new saturation position. Human
eye is more precise in lightness, so clamping value will be really
visible to a human eye.

Tha patch also changes the way how 'L' slider is applied. Now it also
does some compensation for 'S' to keep saturation unchanged during the
operation. Otherwise, the image loses some contrast in saturated area.

BUG:412966

Only signal the terminal if we have one. Nightly env does
not have an TERM flag needed for tput to work.

isIdentity will only resolve to true if the corners are placed
in the corners and the middle points are all x = y

BUG:412491

Some filter configurations (e.g. HSV-filter) expect that older configs
don't have newly introduced keys (e.g. "compatibilityMode"). Therefore
we shouldn't use default filter configuration as a source when loading
the layer. Instead, we should create a new one, and only then load.

CCBUG:411863

Basically, the transformations are substituted with their chroma-based
counterparts, with clever sliding on the edges of the triangle.

The new filters behave a bit differently, therefore the patch adds
a "Compatibility Mode" switch for the filter. It enables old behavior.

BUG:411863

* KoCreatePathTool is inherited by KisToolPath, no factory needed
* KoPencilToolFactory is inherited by KisToolPencil, no factory needed
* KoCreateShapeTool is not used anywhere anymore

All convolution filters should have neededRect adn changedRect. They
worked previously only due to a happy mathematical coincidence :)

BUG:411159

When we make a clone of a shape layer, we must ensure that no updates
are initiated after cloning the image. We used to block updates at
the level of KisShapeLayerCanvas, but it works only when we save/clone
the image from the GUI thread. When we save it from the worker thread,
KoShapeManager queues the event into the GUI events queue, so an update
comes asynchronously.

To resolve this issue, the patch moves locking from KisShapeLayerCanvas
to KoShapeManager. We must ensure that there is no compressors nor queues
between the adding code and blocking code.

BUG:412835

KisSelectionDefaultBounds should not care about the image, it should
take all the values from the parent device's default bounds object.

BUG:412622

Don't assume response was free of error, and print
the response from notarize run for debugging. 

We shouldn't use style's need rect when they are disabled

Background color should be converted using the monitor profile
instead of display converter.

CCBUG:406698

On HDR displays before this commit transparency checkers
looked white because it was just a SDR image showed in
full HDR glory. Hence both light and darker light grey
checkers looked just bright white, no different from each
other, even though in fact their colors were different.
This commit fixes this behaviour by converting the colors
properly from SDR/sRGB to Rec2020-pq (or other display format)
as it should be done.

BUG:407238
CCBUG:406698

of tiles

CCBUG: 412804

This actually wasn't documented like this, and it was broken, too:
the flattening happened after the separation layers were added.

BUG: 412804

The appimage builds are always made from the script in master, not
the one in the release. So our linux releases always show safe asserts,
and that isn't what we want.

* This removes the code for saving the separations to files. This
has been disabled since 2006
* Fixes adding the layers (this was broken because the progress
reporter was set to 100 before adding the layers, which means that
iterrupted() was true, so the layers were never added)
* Fixes unlocking the image (which happened inside the block that
added nodes, which wasn't executed because of see above)
* And makes it possible to only activate the relevant channel when
separating to color

BUG:336694,412624