GitLab

It was dropped when implementing RTL support for layers docker.

Before this commit, color labels combobox would look exactly like
in the filter label for layers, but stretched horizontally.
It means the view-filter icon when no filtering is used,
and a small circle with colors when there are colors selected.
This commit changes the icon to "tag", which seems more appropriate,
and the visuals of the combobox to be in simple rectangles instead of
a circle, so it fills the combobox better.

We should use KisSequentialIterator whenever is possible, it is
twice faster than normal hline-iterator.

They are not a part of kritaui, they are built explicitly into
the unittest object file.

(cherry picked from commit ae0e16e242be231c795f67c0110b546fd0293aec)

CCBUG:415117
(cherry picked from commit dd1f4a13f60b8958d960120c89f52e89ee186e3b)

Because otherwise it's impossible to figure out which profile is
which.

BUG:415117
(cherry picked from commit 53df9ca12274841579bf6481f1b4523c332fe7e7)

BUG:421080
(cherry picked from commit 423c474205c0091b127aa90340b4c5e29f072e92)

BUG:421081
(cherry picked from commit abac7729eed0abf50da65a01575c2f937c3cabc0)

Before this commit, while loop would first check the value of the
array in a specific index and only later if the index is in range.
That shouldn't lead to any problems per se unless the array is very
short, because it always starts from the other end of the array and
the loop ends before the index can fall out of range.
However it seems like it triggered some statis analysis tool and it
might potentially deceive a future programmer if some of the external
conditions (like the starting index) changes.
This commit reverses the order of the conditions so the index is checked
first.

BUG:421079

Before this commit, KoZoomAction class didn't have any tests for itself.
This commit adds a very small, easy test.
(I `just needed it to check something, it should be expanded to be a real
test, for now it checks very basic functionality).

Before this commit, there were two asserts that were checked
only after both of the variables were used. This means that instead of
asserting, Krita would crash. It probaby doesn't change much in case of
Q_ASSERTs (it would be different with asserts that allows recovering)
but nonetheless, it makes no sense to assert after using a variable.
This commit moves the asserts above the usage.
BUG:421076

Before this commit, one of the newly added strings in Fill Tool
lacked context. This commit adds context to that string.

Before this commit, color label combobox in Fill Tool would be
enabled in some specific cases, for example after disabling the
Fast Mode, even though the sampling method would be Current Layer
or All Layers. This commit fixes that behaviour by making sure
that the combobox is enabled only if all three are true:
- Fill Tool doesn't use Fast Mode
- Fill Tool doesn't fill the entire selection
- Fill Tool uses the Color Labeled Layers

Before this commit, translators could be confused what is the string
about. This commit adds contexts to every translated string so it should
be clear how to translate it into other languages.

Otherwise you will have to wait until the even loop kicks back
in before the images are really deleted, and that only happens
after the script is done running.

CCBUG:412740
(cherry picked from commit 543565e9cb5d1ddd8fd46bbd65e98e46d059a507)

When creating a document from libkis' Krita class, we don't want
the Notifier to go off and create spurious and pretty useless
documents.

CCBUG:412740
(cherry picked from commit 7f973f64007637a2d4d175de5a4a2fab33bbc13c)

(cherry picked from commit bc6005f668a7aeca4157c9de4a2f29293f700852)

And clean up the code: comparing the translated strings in a
combobox with untranslated English strings to figure out what
option was chosen is, of course, a big no-no.

BUG:411482
(cherry picked from commit 9c8454f3090013b03306f7281a9110ffe75c5f8a)

(cherry picked from commit b3e6596d27bc112cf1591f321ea29a25eedace8d)

(cherry picked from commit e11edf31e43ce0c35f805162a600e9169340f3e8)

KisAssistantTool::removeAllAssistants function wasn't making correct use
of EditAssistantsCommand. EditAssistantsCommand also contained a flawed and
unnecessary assertion.

Thanks for the help Eoin.

BUG:419189

Add the first of any selection masks that are children of the projection
to the parent at the end of adding all the other child nodes. Otherwise
it won't be seen as the global selection.

BUG:408963
(cherry picked from commit 23a48da3c40102198eadf53a5e28e578841a68e5)

BUG:420935

It shouldn't be possible, but bug 420626 shows that the settings
object can be zero under some circumstances.

BUG:420626
(cherry picked from commit 3ca93708da557b034b706340d2b5bf5b2e4c3be0)

This merge request is implementing self-update capabilities for
AppImages inside Krita. Apart from integration of the tools into Krita's
codebase, there are also updated scripts for AppImage building and icons
for update channels.

There are multiple parts to this patch:

1. C++ code for integration of AppImageUpdate and changes to previous
update notifications Both of those options are called Updaters. There
is:
  * The Manual Updater, which is in fact a fancy name for checking for
new version on the website and displaying a message to the user together
with a link to the release notes. This is the original update
notification code, I have slightly modified version checking and moved
the code into it's own class.
  * The AppImage Updater, which is used if
Krita detects that it runs from an AppImage. This updater can actually
download new version of Krita, if it is available.

2. updated scripts for building appimages I have added code to bundle
AppImageUpdate into the AppImage being built. Also, the build scripts
infer the update channel (stable, Beta, Plus or Next) from a combination
of Krita version and git information. This info is used for setting up
proper update channel and selecting the right branding option. I have
also added helper scripts for signing the already built AppImage
(sign_appimage.sh, generate_zsync.sh; also strip_appimage_signature.sh and
validate_appimage_signature.sh for debugging purposes) and rewriting
update information of the built AppImage (update_updinfo.sh).

3. icons for update channels I have added icons I made for myself to
make sense of all my AppImages. The icons are in krita/pics/branding,
together with a script for generating all icon sizes from svg
(generate_icons.sh).

Changes to building - there are two new cmake flags:

  * ENABLE_UPDATERS (bool, can be ON or OFF) - this flag toggles all
    updater code, both the appimage updating and update notifications
  * BRANDING (string, can be default, Beta, Plus, or Next) - this flag
    controls which app icons will be installed

Perceptual quantization (PQ) is just not useful for a
color selector.

(cherry picked from commit 1e87724bb676357c79ecf82860cd09f61df4bec2)

When the color space of a document or current layer changes,
the color selector did not react to this.

After a lot of brain racking I came to the conclusion that the
most sane solution is to make color space changes explicit
for KisVisualColorSelector, and pass it the original
foreground color again.
Changing color space was added to KisColorSelectorInterface too.

More implicit behavior and assumptions about expected outcome
seem counterproductive at this point.

(cherry picked from commit 7c897c282e3524db8e75ec9f276852c8583e3c1b)

The ITUR_2100_PQ_FULL.ICC has no colorants and revealed the bugs.
Iteration order was inverted so normalization didn't work.
Also only the first colorant was ever written, and
consequently the luma coefficients were also incomplete.

Additionally, RGB vs. BGR channel ordering was not respected,
and the sum of luma coefficients needs to be 1.0 for HSY
color space calculations.

And while at it, calculate a few more points so the graph in
the color space browser looks at least somewhat like a curve.

(cherry picked from commit fe47c0619d97e0d727afc9a48c95bf18c4475ddd)

Without any kind of barrier between left and right half,
it was not really possible to select the min or max value
as you could in a bar control.
With this change, the cursor coordinate will be limited
to the side where mouse dragging started.

(cherry picked from commit 418ebf10)

Constructing a QVector is expensive, so expensive that it was
actually slower than the non-fast versions.
It ranked pretty high in perf and removing it improved HSY'->RGB
conversion times quite a bit.

(cherry picked from commit f83ba6560e62be58502b085a281ec8b4dc670a7d)

The RGB<=>HSY conversion really only works properly in linear RGB.
Gamma needs to be applied to Y', and is now taken from Advanced Color Selector
configuration.

Luma coefficients are now fetched on reconfiguration instead of every single
color conversion.

Caveat: Some color profiles like "ITUR_2100_PQ_FULL.ICC" define no colorants
and hence they cannot be linearized with the current implementation.
A substitution of color spaces like the Small Color Selector does seems
unavoidable in the long run...

(cherry picked from commit 6a660c25f68396d7fdb9fcce69048b5ee8d4637a)