GitLab

Summary:
MSVC complained about using these template definitions without including
the class definitions explicitly.

Reviewers: rempt

Differential Revision: https://phabricator.kde.org/D255

Summary: This was hiding in a lump of things I did to make my original MSVC build.

Reviewers: rempt

Maniphest Tasks: T360

Differential Revision: https://phabricator.kde.org/D252

Summary: This is for the Intel graphics card check. Part of T360

Reviewers: rempt

Differential Revision: https://phabricator.kde.org/D251

Summary:
Not sure whether this requires strict adherence to a header in Qt5.

This is based on the Frameworks branch. I can rebase it to 2.9 but I don't have a build configured there.

Reviewers: dkazakov, rempt

Differential Revision: https://phabricator.kde.org/D246

Conflicts:
	krita/plugins/formats/xcf/3rdparty/xcftools/config.h
	krita/ui/CMakeLists.txt
	krita/ui/KisDocumentEntry.h
	krita/ui/KisFilterEntry.h
	krita/ui/canvas/kis_grid_manager.cpp
	krita/ui/input/wintab/kis_tablet_support_x11.h
	libs/db/expression.cpp

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"

And warn the user that we're downscaling their 16 bit image to
8 bit. A 2007 todo done!

The result looks wrong, though... Inverted. This also makes
the filter obey the batch mode setting, so it won't show a
warning message box when saving lab or ycbcr to jpeg.

The GUI uses the emitter's threads to do the color space conversion so
we shouldn't emit in bigger patches.

BUG:351195

This cleans up the pattern option a bit and makes the
embedded pattern manager work much harder to find out
whether the embedded pattern exists. And if it really
doesn't exist, we do add it to the resource manager,
but do not save it to disk.

1) The active field should be saved correctly
2) The outline cache should be invalidated when loading

Fixes T568

You can now merge them using usual Ctrl+E on one layer, or just
select multiple selection masks and press Ctrl+E again.

CC:kimageshop@kde.org

From now on active selection masks are saved through a merge operation
and are added to a new merged layer as inactive ones. You can activate
them later manually.

Next we will implement merging of selection masks, then the user will be
able to merge selection masks as well.

Fixes T508
CC:kimageshop@kde.org

Added optional flags to KisDocument::openUrl() and made "File Layer" not add its file to the recent files list.

BUG:345560

REVIEW:124732

Masks' extent should not be included into the extent of a passthrough
layer, because they are explicitly disabled in this mode.

BUG:351224
Fixes T567

Patch by Alexey Elnatanov. Thanks!
CCMAIL: invalpoint@gmail.com

Form layout instead of grid layout, set enabled on reading config
and the random opacity checkbox now has space for its label.

Now all types of 8 bit and 16 bit PSD files should load fine, including
multilayered ones. Saving should work as well.

STILL NEEDS TESTING:
We need to test if Photoshop will load 16bit multilayered files saved
in Krita. The point is we still save them differently from what PS is
used to.

CC:kimageshop@kde.org

fixup for cf69d1e294ceb63cfbadbff95836433c2314cf21

The testing files themselves are stored in external unittests data
folder.

See
https://community.kde.org/Krita/Docs/Unittests_in_Krita

It also fixes a set of loading/saving bugs we had

This is exactly what PS does. Indeed, to show that your image projection
has transparency, according to PSD spec you should set the number of layers
to -1, and you cannot do that if you don't have a layer info block at all :)

When you finished adding layers to the group just lock it and the
new layers will not be added to it anymore. They will be added as
sibling nodes.

Fixes T475
CC:kimageshop@kde.org

When transforming a group recursively, some of the layers may be invisible
so the user may just forget about them. Now we transform the layers even if
they are invisible. The user still can avoid this behavior by
setting the layer as "locked".

Thanks Roman Ilin for suggestion!

Ref T475
CC:kimageshop@kde.org

The default bounds associated with a mask need to know about the
image. When we load a dropped mask, we create it with a temporary
image to reconstruct the hierarchy, and then we need to "internalize"
the nodes to the target image.

That was only done for layers, because we said

// Only layers know about the image

that was a lie...

Cache the paint devices we use for rendering the color selectors
for as long as possible.

QColor accepts only integers in the constructor.

psd_layer_effects_shadow_base is a class not a struct.

Commenting out unused code.