resources are unique within storage, filename and resourcetype.

- Linux: Work around Launchpad PPA using "krita" as the library prefix
- Windows: Pre-create the downloads dir at Krita build time

Mathias Wein authored Jan 29, 2021 and Halla Rempt committed Jan 29, 2021

If the labels are not supposed to influence the preferred widget size
(which determines some popup sizes), the solution should be to set
their size policy to "Ignored".
The layout will just shrink or grow them to fill the available space.

BUG:431825

Amy spark authored Jan 19, 2021 and Dmitry Kazakov committed Jan 29, 2021

- updates when the layer exists should not bypass the compressor
- fix possible first render to black

(Improved convertToPaintLayer)

And fixed compilation error for Launchpad build / docker instance.

Chris Venter authored Jan 27, 2021 and Halla Rempt committed Jan 28, 2021

BUG:432226

And add tooltips to the docker.

This would be hit if you would name your taskset resource something
like taskset1...

And remove resources.version.

This updates the database schema version.

CCMAIL:kimageshop@kde.org

It's not editable, but calculating the md5 sum needs this now, since
setting the md5 on loading isn't enough anymore.

...instead of protected.

... and not to the bundle they came from.

The reasons they were broken:

1) Md5 sum is now calculated by the storage, therefore the resource
   content cannot be empty.

2) Default name of the resource now coincides with the filename

3) There was a bug in new implementation of
   KisTagFilterResourceProxyModel::setResourceFilter

We shouldn't re-save the resource to get calculate MD5, it will
generate random md5sum depending on the version of the underlying
libraries (say, libpng). Instead, we should calculate MD5 directly
in the storage.

This patch also caches MD5 in the storage, so that it shouldn't be
recalculated every time resource is loaded into Krita.

Some dependent resources address the linked resources
by resource name, and in previous versions of Krita
we used to use filename for resources without explicit
name, e.g. PNG patterns.

There were two issues:

1) Instead of full file paths we now use file names only
   to distinguish resources.

2) In Krita4 we used filename as the default resource name
   (the the next patch in the patchset).

This reverts commit f2fee32e.

It fails compilation on Windows for some reason:

C:/dev/env-5/krita/libs/image/kis_layer_utils.cpp: In function 'void KisLayerUtils::mergeDown(KisImageSP, KisLayerSP, const KisMetaData::MergeStrategy*)':
C:/dev/env-5/krita/libs/image/kis_layer_utils.cpp:1317:72: error: call of overloaded 'RefreshDelayedUpdateLayers(KisLayerUtils::MergeDownInfoSP&)' is ambiguous
             applicator.applyCommand(new RefreshDelayedUpdateLayers(info), KisStrokeJobData::BARRIER);
                                                                        ^
C:/dev/env-5/krita/libs/image/kis_layer_utils.cpp:468:9: note: candidate: KisLayerUtils::RefreshDelayedUpdateLayers::RefreshDelayedUpdateLayers(KisLayerUtils::ConvertToPaintLayerInfoSP)
         RefreshDelayedUpdateLayers(ConvertToPaintLayerInfoSP info){
         ^~~~~~~~~~~~~~~~~~~~~~~~~~
C:/dev/env-5/krita/libs/image/kis_layer_utils.cpp:465:9: note: candidate: KisLayerUtils::RefreshDelayedUpdateLayers::RefreshDelayedUpdateLayers(KisLayerUtils::MergeDownInfoBaseSP)
         RefreshDelayedUpdateLayers(MergeDownInfoBaseSP info)
         ^~~~~~~~~~~~~~~~~~~~~~~~~~

Solves some conversion hedgecases where frame data would be lost or other layer data
would be lost in the conversion. Should also make it up to date with other layer
operations that require projection updates.

BUG:405741

Amy spark authored Jan 19, 2021 and Halla Rempt committed Jan 27, 2021

Turns out, GMic layer parsing is a three-stage process:

1. Image size is synced.
2. Layer count is synced; new layers are added if necessary (!!!)
3. Untouched layers are visited and updated with the data from GMic

While (3) is not much of an issue (after I added the name parsing code),
with (2) it means that some nodes will be overwritten, and its names
lost.

This commit fixes the issue by applying 1:1 the original name to the new
layer.

Amy spark authored Jan 18, 2021 and Halla Rempt committed Jan 27, 2021

CCBUG: 431151

Amy spark authored Dec 21, 2020 and Halla Rempt committed Jan 27, 2021

This implements the correct ordering for received layers and the
parameter specification as used by GMic.

BUG: 429851

Amy spark authored Dec 20, 2020 and Halla Rempt committed Jan 27, 2021

There are three layer modes that were removed in 2.8.2, and three more
that were unimplemented. The latter have been implemented in this
commit.